Mercurial > vim
annotate src/spell.c @ 2770:25672ad7f377 v7.3.161
updated for version 7.3.161
Problem: Items on the stack may be too big.
Solution: Make items static or allocate them.
| author | Bram Moolenaar <bram@vim.org> |
|---|---|
| date | Mon, 11 Apr 2011 21:35:11 +0200 |
| parents | c5e47b752f07 |
| children | 6cd0d6413a2e |
| rev | line source |
|---|---|
| 223 | 1 /* vi:set ts=8 sts=4 sw=4: |
| 2 * | |
| 3 * VIM - Vi IMproved by Bram Moolenaar | |
| 4 * | |
| 5 * Do ":help uganda" in Vim to read copying and usage conditions. | |
| 6 * Do ":help credits" in Vim to see a list of people who contributed. | |
| 7 * See README.txt for an overview of the Vim source code. | |
| 8 */ | |
| 9 | |
| 10 /* | |
| 11 * spell.c: code for spell checking | |
| 226 | 12 * |
| 300 | 13 * The spell checking mechanism uses a tree (aka trie). Each node in the tree |
| 14 * has a list of bytes that can appear (siblings). For each byte there is a | |
| 15 * pointer to the node with the byte that follows in the word (child). | |
| 324 | 16 * |
| 17 * A NUL byte is used where the word may end. The bytes are sorted, so that | |
| 18 * binary searching can be used and the NUL bytes are at the start. The | |
| 19 * number of possible bytes is stored before the list of bytes. | |
| 20 * | |
| 21 * The tree uses two arrays: "byts" stores the characters, "idxs" stores | |
| 22 * either the next index or flags. The tree starts at index 0. For example, | |
| 23 * to lookup "vi" this sequence is followed: | |
| 24 * i = 0 | |
| 25 * len = byts[i] | |
| 26 * n = where "v" appears in byts[i + 1] to byts[i + len] | |
| 27 * i = idxs[n] | |
| 28 * len = byts[i] | |
| 29 * n = where "i" appears in byts[i + 1] to byts[i + len] | |
| 30 * i = idxs[n] | |
| 31 * len = byts[i] | |
| 32 * find that byts[i + 1] is 0, idxs[i + 1] has flags for "vi". | |
| 300 | 33 * |
| 339 | 34 * There are two word trees: one with case-folded words and one with words in |
| 300 | 35 * original case. The second one is only used for keep-case words and is |
| 36 * usually small. | |
| 37 * | |
| 481 | 38 * There is one additional tree for when not all prefixes are applied when |
| 339 | 39 * generating the .spl file. This tree stores all the possible prefixes, as |
| 40 * if they were words. At each word (prefix) end the prefix nr is stored, the | |
| 41 * following word must support this prefix nr. And the condition nr is | |
| 42 * stored, used to lookup the condition that the word must match with. | |
| 43 * | |
| 300 | 44 * Thanks to Olaf Seibert for providing an example implementation of this tree |
| 45 * and the compression mechanism. | |
| 625 | 46 * LZ trie ideas: |
| 47 * http://www.irb.hr/hr/home/ristov/papers/RistovLZtrieRevision1.pdf | |
| 48 * More papers: http://www-igm.univ-mlv.fr/~laporte/publi_en.html | |
| 243 | 49 * |
| 50 * Matching involves checking the caps type: Onecap ALLCAP KeepCap. | |
| 51 * | |
| 236 | 52 * Why doesn't Vim use aspell/ispell/myspell/etc.? |
| 53 * See ":help develop-spell". | |
| 54 */ | |
| 55 | |
| 469 | 56 /* Use SPELL_PRINTTREE for debugging: dump the word tree after adding a word. |
| 482 | 57 * Only use it for small word lists! */ |
| 469 | 58 #if 0 |
| 59 # define SPELL_PRINTTREE | |
| 482 | 60 #endif |
| 61 | |
| 633 | 62 /* Use DEBUG_TRIEWALK to print the changes made in suggest_trie_walk() for a |
| 63 * specific word. */ | |
| 625 | 64 #if 0 |
| 65 # define DEBUG_TRIEWALK | |
| 66 #endif | |
| 67 | |
| 300 | 68 /* |
| 324 | 69 * Use this to adjust the score after finding suggestions, based on the |
| 70 * suggested word sounding like the bad word. This is much faster than doing | |
| 71 * it for every possible suggestion. | |
| 625 | 72 * Disadvantage: When "the" is typed as "hte" it sounds quite different ("@" |
| 73 * vs "ht") and goes down in the list. | |
| 344 | 74 * Used when 'spellsuggest' is set to "best". |
| 75 */ | |
| 76 #define RESCORE(word_score, sound_score) ((3 * word_score + sound_score) / 4) | |
| 77 | |
| 78 /* | |
| 625 | 79 * Do the opposite: based on a maximum end score and a known sound score, |
| 1705 | 80 * compute the maximum word score that can be used. |
| 625 | 81 */ |
| 82 #define MAXSCORE(word_score, sound_score) ((4 * word_score - sound_score) / 3) | |
| 83 | |
| 84 /* | |
| 339 | 85 * Vim spell file format: <HEADER> |
| 483 | 86 * <SECTIONS> |
| 339 | 87 * <LWORDTREE> |
| 88 * <KWORDTREE> | |
| 89 * <PREFIXTREE> | |
| 300 | 90 * |
| 483 | 91 * <HEADER>: <fileID> <versionnr> |
| 92 * | |
| 93 * <fileID> 8 bytes "VIMspell" | |
| 94 * <versionnr> 1 byte VIMSPELLVERSION | |
| 95 * | |
| 96 * | |
| 97 * Sections make it possible to add information to the .spl file without | |
| 98 * making it incompatible with previous versions. There are two kinds of | |
| 99 * sections: | |
| 100 * 1. Not essential for correct spell checking. E.g. for making suggestions. | |
| 101 * These are skipped when not supported. | |
| 102 * 2. Optional information, but essential for spell checking when present. | |
| 103 * E.g. conditions for affixes. When this section is present but not | |
| 104 * supported an error message is given. | |
| 300 | 105 * |
| 483 | 106 * <SECTIONS>: <section> ... <sectionend> |
| 107 * | |
| 108 * <section>: <sectionID> <sectionflags> <sectionlen> (section contents) | |
| 109 * | |
| 110 * <sectionID> 1 byte number from 0 to 254 identifying the section | |
| 111 * | |
| 112 * <sectionflags> 1 byte SNF_REQUIRED: this section is required for correct | |
| 113 * spell checking | |
| 114 * | |
| 115 * <sectionlen> 4 bytes length of section contents, MSB first | |
| 116 * | |
| 117 * <sectionend> 1 byte SN_END | |
| 118 * | |
| 119 * | |
| 714 | 120 * sectionID == SN_INFO: <infotext> |
| 121 * <infotext> N bytes free format text with spell file info (version, | |
| 122 * website, etc) | |
| 123 * | |
| 483 | 124 * sectionID == SN_REGION: <regionname> ... |
| 125 * <regionname> 2 bytes Up to 8 region names: ca, au, etc. Lower case. | |
| 300 | 126 * First <regionname> is region 1. |
| 127 * | |
| 483 | 128 * sectionID == SN_CHARFLAGS: <charflagslen> <charflags> |
| 129 * <folcharslen> <folchars> | |
| 300 | 130 * <charflagslen> 1 byte Number of bytes in <charflags> (should be 128). |
| 131 * <charflags> N bytes List of flags (first one is for character 128): | |
| 324 | 132 * 0x01 word character CF_WORD |
| 133 * 0x02 upper-case character CF_UPPER | |
| 483 | 134 * <folcharslen> 2 bytes Number of bytes in <folchars>. |
| 135 * <folchars> N bytes Folded characters, first one is for character 128. | |
| 300 | 136 * |
| 483 | 137 * sectionID == SN_MIDWORD: <midword> |
| 138 * <midword> N bytes Characters that are word characters only when used | |
| 366 | 139 * in the middle of a word. |
| 140 * | |
| 483 | 141 * sectionID == SN_PREFCOND: <prefcondcnt> <prefcond> ... |
| 339 | 142 * <prefcondcnt> 2 bytes Number of <prefcond> items following. |
| 143 * <prefcond> : <condlen> <condstr> | |
| 144 * <condlen> 1 byte Length of <condstr>. | |
| 145 * <condstr> N bytes Condition for the prefix. | |
| 146 * | |
| 483 | 147 * sectionID == SN_REP: <repcount> <rep> ... |
| 148 * <repcount> 2 bytes number of <rep> items, MSB first. | |
| 323 | 149 * <rep> : <repfromlen> <repfrom> <reptolen> <repto> |
| 483 | 150 * <repfromlen> 1 byte length of <repfrom> |
| 151 * <repfrom> N bytes "from" part of replacement | |
| 152 * <reptolen> 1 byte length of <repto> | |
| 153 * <repto> N bytes "to" part of replacement | |
| 323 | 154 * |
| 625 | 155 * sectionID == SN_REPSAL: <repcount> <rep> ... |
| 156 * just like SN_REP but for soundfolded words | |
| 157 * | |
| 483 | 158 * sectionID == SN_SAL: <salflags> <salcount> <sal> ... |
| 159 * <salflags> 1 byte flags for soundsalike conversion: | |
| 323 | 160 * SAL_F0LLOWUP |
| 161 * SAL_COLLAPSE | |
| 162 * SAL_REM_ACCENTS | |
| 483 | 163 * <salcount> 2 bytes number of <sal> items following |
| 323 | 164 * <sal> : <salfromlen> <salfrom> <saltolen> <salto> |
| 483 | 165 * <salfromlen> 1 byte length of <salfrom> |
| 166 * <salfrom> N bytes "from" part of soundsalike | |
| 167 * <saltolen> 1 byte length of <salto> | |
| 168 * <salto> N bytes "to" part of soundsalike | |
| 323 | 169 * |
| 483 | 170 * sectionID == SN_SOFO: <sofofromlen> <sofofrom> <sofotolen> <sofoto> |
| 171 * <sofofromlen> 2 bytes length of <sofofrom> | |
| 172 * <sofofrom> N bytes "from" part of soundfold | |
| 173 * <sofotolen> 2 bytes length of <sofoto> | |
| 174 * <sofoto> N bytes "to" part of soundfold | |
| 323 | 175 * |
| 625 | 176 * sectionID == SN_SUGFILE: <timestamp> |
| 177 * <timestamp> 8 bytes time in seconds that must match with .sug file | |
| 178 * | |
| 698 | 179 * sectionID == SN_NOSPLITSUGS: nothing |
| 180 * | |
| 625 | 181 * sectionID == SN_WORDS: <word> ... |
| 182 * <word> N bytes NUL terminated common word | |
| 183 * | |
| 483 | 184 * sectionID == SN_MAP: <mapstr> |
| 185 * <mapstr> N bytes String with sequences of similar characters, | |
| 323 | 186 * separated by slashes. |
| 300 | 187 * |
| 809 | 188 * sectionID == SN_COMPOUND: <compmax> <compminlen> <compsylmax> <compoptions> |
| 189 * <comppatcount> <comppattern> ... <compflags> | |
| 483 | 190 * <compmax> 1 byte Maximum nr of words in compound word. |
| 191 * <compminlen> 1 byte Minimal word length for compounding. | |
| 192 * <compsylmax> 1 byte Maximum nr of syllables in compound word. | |
| 809 | 193 * <compoptions> 2 bytes COMP_ flags. |
| 194 * <comppatcount> 2 bytes number of <comppattern> following | |
| 714 | 195 * <compflags> N bytes Flags from COMPOUNDRULE items, separated by |
| 483 | 196 * slashes. |
| 197 * | |
| 809 | 198 * <comppattern>: <comppatlen> <comppattext> |
| 199 * <comppatlen> 1 byte length of <comppattext> | |
| 200 * <comppattext> N bytes end or begin chars from CHECKCOMPOUNDPATTERN | |
| 201 * | |
| 202 * sectionID == SN_NOBREAK: (empty, its presence is what matters) | |
| 492 | 203 * |
| 483 | 204 * sectionID == SN_SYLLABLE: <syllable> |
| 205 * <syllable> N bytes String from SYLLABLE item. | |
| 300 | 206 * |
| 207 * <LWORDTREE>: <wordtree> | |
| 208 * | |
| 339 | 209 * <KWORDTREE>: <wordtree> |
| 210 * | |
| 211 * <PREFIXTREE>: <wordtree> | |
| 212 * | |
| 213 * | |
| 300 | 214 * <wordtree>: <nodecount> <nodedata> ... |
| 215 * | |
| 216 * <nodecount> 4 bytes Number of nodes following. MSB first. | |
| 217 * | |
| 218 * <nodedata>: <siblingcount> <sibling> ... | |
| 219 * | |
| 220 * <siblingcount> 1 byte Number of siblings in this node. The siblings | |
| 221 * follow in sorted order. | |
| 222 * | |
| 339 | 223 * <sibling>: <byte> [ <nodeidx> <xbyte> |
| 481 | 224 * | <flags> [<flags2>] [<region>] [<affixID>] |
| 225 * | [<pflags>] <affixID> <prefcondnr> ] | |
| 300 | 226 * |
| 227 * <byte> 1 byte Byte value of the sibling. Special cases: | |
| 228 * BY_NOFLAGS: End of word without flags and for all | |
| 229 * regions. | |
| 481 | 230 * For PREFIXTREE <affixID> and |
| 366 | 231 * <prefcondnr> follow. |
| 390 | 232 * BY_FLAGS: End of word, <flags> follow. |
| 481 | 233 * For PREFIXTREE <pflags>, <affixID> |
| 455 | 234 * and <prefcondnr> follow. |
| 390 | 235 * BY_FLAGS2: End of word, <flags> and <flags2> |
| 236 * follow. Not used in PREFIXTREE. | |
| 237 * BY_INDEX: Child of sibling is shared, <nodeidx> | |
| 300 | 238 * and <xbyte> follow. |
| 239 * | |
| 240 * <nodeidx> 3 bytes Index of child for this sibling, MSB first. | |
| 241 * | |
| 242 * <xbyte> 1 byte byte value of the sibling. | |
| 243 * | |
| 244 * <flags> 1 byte bitmask of: | |
| 245 * WF_ALLCAP word must have only capitals | |
| 246 * WF_ONECAP first char of word must be capital | |
| 389 | 247 * WF_KEEPCAP keep-case word |
| 248 * WF_FIXCAP keep-case word, all caps not allowed | |
| 300 | 249 * WF_RARE rare word |
| 389 | 250 * WF_BANNED bad word |
| 300 | 251 * WF_REGION <region> follows |
| 481 | 252 * WF_AFX <affixID> follows |
| 300 | 253 * |
| 500 | 254 * <flags2> 1 byte Bitmask of: |
| 390 | 255 * WF_HAS_AFF >> 8 word includes affix |
| 500 | 256 * WF_NEEDCOMP >> 8 word only valid in compound |
| 698 | 257 * WF_NOSUGGEST >> 8 word not used for suggestions |
| 809 | 258 * WF_COMPROOT >> 8 word already a compound |
| 819 | 259 * WF_NOCOMPBEF >> 8 no compounding before this word |
| 260 * WF_NOCOMPAFT >> 8 no compounding after this word | |
| 390 | 261 * |
| 455 | 262 * <pflags> 1 byte bitmask of: |
| 263 * WFP_RARE rare prefix | |
| 264 * WFP_NC non-combining prefix | |
| 265 * WFP_UP letter after prefix made upper case | |
| 266 * | |
| 300 | 267 * <region> 1 byte Bitmask for regions in which word is valid. When |
| 268 * omitted it's valid in all regions. | |
| 269 * Lowest bit is for region 1. | |
| 270 * | |
| 481 | 271 * <affixID> 1 byte ID of affix that can be used with this word. In |
| 339 | 272 * PREFIXTREE used for the required prefix ID. |
| 273 * | |
| 274 * <prefcondnr> 2 bytes Prefix condition number, index in <prefcond> list | |
| 275 * from HEADER. | |
| 300 | 276 * |
| 277 * All text characters are in 'encoding', but stored as single bytes. | |
| 278 */ | |
| 279 | |
| 625 | 280 /* |
| 281 * Vim .sug file format: <SUGHEADER> | |
| 282 * <SUGWORDTREE> | |
| 283 * <SUGTABLE> | |
| 284 * | |
| 285 * <SUGHEADER>: <fileID> <versionnr> <timestamp> | |
| 286 * | |
| 287 * <fileID> 6 bytes "VIMsug" | |
| 288 * <versionnr> 1 byte VIMSUGVERSION | |
| 289 * <timestamp> 8 bytes timestamp that must match with .spl file | |
| 290 * | |
| 291 * | |
| 292 * <SUGWORDTREE>: <wordtree> (see above, no flags or region used) | |
| 293 * | |
| 294 * | |
| 295 * <SUGTABLE>: <sugwcount> <sugline> ... | |
| 296 * | |
| 297 * <sugwcount> 4 bytes number of <sugline> following | |
| 298 * | |
| 299 * <sugline>: <sugnr> ... NUL | |
| 300 * | |
| 301 * <sugnr>: X bytes word number that results in this soundfolded word, | |
| 302 * stored as an offset to the previous number in as | |
| 303 * few bytes as possible, see offset2bytes()) | |
| 304 */ | |
| 305 | |
| 223 | 306 #include "vim.h" |
| 307 | |
| 737 | 308 #if defined(FEAT_SPELL) || defined(PROTO) |
| 223 | 309 |
| 625 | 310 #ifndef UNIX /* it's in os_unix.h for Unix */ |
| 311 # include <time.h> /* for time_t */ | |
| 312 #endif | |
| 313 | |
| 323 | 314 #define MAXWLEN 250 /* Assume max. word len is this many bytes. |
| 315 Some places assume a word length fits in a | |
| 316 byte, thus it can't be above 255. */ | |
| 226 | 317 |
| 490 | 318 /* Type used for indexes in the word tree need to be at least 4 bytes. If int |
| 324 | 319 * is 8 bytes we could use something smaller, but what? */ |
| 490 | 320 #if SIZEOF_INT > 3 |
| 324 | 321 typedef int idx_T; |
| 322 #else | |
| 323 typedef long idx_T; | |
| 324 #endif | |
| 325 | |
| 2660 | 326 #ifdef VMS |
| 327 # define SPL_FNAME_TMPL "%s_%s.spl" | |
| 328 # define SPL_FNAME_ADD "_add." | |
| 329 # define SPL_FNAME_ASCII "_ascii." | |
| 330 #else | |
| 331 # define SPL_FNAME_TMPL "%s.%s.spl" | |
| 332 # define SPL_FNAME_ADD ".add." | |
| 333 # define SPL_FNAME_ASCII ".ascii." | |
| 334 #endif | |
| 335 | |
| 324 | 336 /* Flags used for a word. Only the lowest byte can be used, the region byte |
| 337 * comes above it. */ | |
| 300 | 338 #define WF_REGION 0x01 /* region byte follows */ |
| 339 #define WF_ONECAP 0x02 /* word with one capital (or all capitals) */ | |
| 340 #define WF_ALLCAP 0x04 /* word must be all capitals */ | |
| 341 #define WF_RARE 0x08 /* rare word */ | |
| 307 | 342 #define WF_BANNED 0x10 /* bad word */ |
| 481 | 343 #define WF_AFX 0x20 /* affix ID follows */ |
| 389 | 344 #define WF_FIXCAP 0x40 /* keep-case word, allcap not allowed */ |
| 323 | 345 #define WF_KEEPCAP 0x80 /* keep-case word */ |
| 346 | |
| 390 | 347 /* for <flags2>, shifted up one byte to be used in wn_flags */ |
| 348 #define WF_HAS_AFF 0x0100 /* word includes affix */ | |
| 500 | 349 #define WF_NEEDCOMP 0x0200 /* word only valid in compound */ |
| 698 | 350 #define WF_NOSUGGEST 0x0400 /* word not to be suggested */ |
| 809 | 351 #define WF_COMPROOT 0x0800 /* already compounded word, COMPOUNDROOT */ |
| 819 | 352 #define WF_NOCOMPBEF 0x1000 /* no compounding before this word */ |
| 353 #define WF_NOCOMPAFT 0x2000 /* no compounding after this word */ | |
| 390 | 354 |
| 633 | 355 /* only used for su_badflags */ |
| 356 #define WF_MIXCAP 0x20 /* mix of upper and lower case: macaRONI */ | |
| 357 | |
| 389 | 358 #define WF_CAPMASK (WF_ONECAP | WF_ALLCAP | WF_KEEPCAP | WF_FIXCAP) |
| 300 | 359 |
| 455 | 360 /* flags for <pflags> */ |
| 822 | 361 #define WFP_RARE 0x01 /* rare prefix */ |
| 362 #define WFP_NC 0x02 /* prefix is not combining */ | |
| 363 #define WFP_UP 0x04 /* to-upper prefix */ | |
| 364 #define WFP_COMPPERMIT 0x08 /* prefix with COMPOUNDPERMITFLAG */ | |
| 365 #define WFP_COMPFORBID 0x10 /* prefix with COMPOUNDFORBIDFLAG */ | |
| 366 | |
| 367 /* Flags for postponed prefixes in "sl_pidxs". Must be above affixID (one | |
| 368 * byte) and prefcondnr (two bytes). */ | |
| 369 #define WF_RAREPFX (WFP_RARE << 24) /* rare postponed prefix */ | |
| 370 #define WF_PFX_NC (WFP_NC << 24) /* non-combining postponed prefix */ | |
| 371 #define WF_PFX_UP (WFP_UP << 24) /* to-upper postponed prefix */ | |
| 372 #define WF_PFX_COMPPERMIT (WFP_COMPPERMIT << 24) /* postponed prefix with | |
| 373 * COMPOUNDPERMITFLAG */ | |
| 374 #define WF_PFX_COMPFORBID (WFP_COMPFORBID << 24) /* postponed prefix with | |
| 375 * COMPOUNDFORBIDFLAG */ | |
| 376 | |
| 390 | 377 |
| 809 | 378 /* flags for <compoptions> */ |
| 379 #define COMP_CHECKDUP 1 /* CHECKCOMPOUNDDUP */ | |
| 380 #define COMP_CHECKREP 2 /* CHECKCOMPOUNDREP */ | |
| 381 #define COMP_CHECKCASE 4 /* CHECKCOMPOUNDCASE */ | |
| 382 #define COMP_CHECKTRIPLE 8 /* CHECKCOMPOUNDTRIPLE */ | |
| 383 | |
| 390 | 384 /* Special byte values for <byte>. Some are only used in the tree for |
| 385 * postponed prefixes, some only in the other trees. This is a bit messy... */ | |
| 386 #define BY_NOFLAGS 0 /* end of word without flags or region; for | |
| 455 | 387 * postponed prefix: no <pflags> */ |
| 388 #define BY_INDEX 1 /* child is shared, index follows */ | |
| 389 #define BY_FLAGS 2 /* end of word, <flags> byte follows; for | |
| 390 * postponed prefix: <pflags> follows */ | |
| 391 #define BY_FLAGS2 3 /* end of word, <flags> and <flags2> bytes | |
| 392 * follow; never used in prefix tree */ | |
| 393 #define BY_SPECIAL BY_FLAGS2 /* highest special byte value */ | |
| 236 | 394 |
| 625 | 395 /* Info from "REP", "REPSAL" and "SAL" entries in ".aff" file used in si_rep, |
| 396 * si_repsal, sl_rep, and si_sal. Not for sl_sal! | |
| 323 | 397 * One replacement: from "ft_from" to "ft_to". */ |
| 398 typedef struct fromto_S | |
| 236 | 399 { |
| 323 | 400 char_u *ft_from; |
| 401 char_u *ft_to; | |
| 402 } fromto_T; | |
| 236 | 403 |
| 344 | 404 /* Info from "SAL" entries in ".aff" file used in sl_sal. |
| 405 * The info is split for quick processing by spell_soundfold(). | |
| 406 * Note that "sm_oneof" and "sm_rules" point into sm_lead. */ | |
| 407 typedef struct salitem_S | |
| 408 { | |
| 409 char_u *sm_lead; /* leading letters */ | |
| 410 int sm_leadlen; /* length of "sm_lead" */ | |
| 375 | 411 char_u *sm_oneof; /* letters from () or NULL */ |
| 344 | 412 char_u *sm_rules; /* rules like ^, $, priority */ |
| 413 char_u *sm_to; /* replacement. */ | |
| 372 | 414 #ifdef FEAT_MBYTE |
| 415 int *sm_lead_w; /* wide character copy of "sm_lead" */ | |
| 375 | 416 int *sm_oneof_w; /* wide character copy of "sm_oneof" */ |
| 372 | 417 int *sm_to_w; /* wide character copy of "sm_to" */ |
| 418 #endif | |
| 344 | 419 } salitem_T; |
| 420 | |
| 375 | 421 #ifdef FEAT_MBYTE |
| 422 typedef int salfirst_T; | |
| 423 #else | |
| 424 typedef short salfirst_T; | |
| 425 #endif | |
| 426 | |
| 483 | 427 /* Values for SP_*ERROR are negative, positive values are used by |
| 428 * read_cnt_string(). */ | |
| 429 #define SP_TRUNCERROR -1 /* spell file truncated error */ | |
| 430 #define SP_FORMERROR -2 /* format error in spell file */ | |
| 495 | 431 #define SP_OTHERERROR -3 /* other error while reading spell file */ |
| 483 | 432 |
| 236 | 433 /* |
| 243 | 434 * Structure used to store words and other info for one language, loaded from |
| 435 * a .spl file. | |
| 300 | 436 * The main access is through the tree in "sl_fbyts/sl_fidxs", storing the |
| 437 * case-folded words. "sl_kbyts/sl_kidxs" is for keep-case words. | |
| 438 * | |
| 439 * The "byts" array stores the possible bytes in each tree node, preceded by | |
| 440 * the number of possible bytes, sorted on byte value: | |
| 441 * <len> <byte1> <byte2> ... | |
| 442 * The "idxs" array stores the index of the child node corresponding to the | |
| 443 * byte in "byts". | |
| 444 * Exception: when the byte is zero, the word may end here and "idxs" holds | |
| 481 | 445 * the flags, region mask and affixID for the word. There may be several |
| 446 * zeros in sequence for alternative flag/region/affixID combinations. | |
| 236 | 447 */ |
| 448 typedef struct slang_S slang_T; | |
| 449 struct slang_S | |
| 450 { | |
| 451 slang_T *sl_next; /* next language */ | |
| 452 char_u *sl_name; /* language name "en", "en.rare", "nl", etc. */ | |
| 310 | 453 char_u *sl_fname; /* name of .spl file */ |
| 323 | 454 int sl_add; /* TRUE if it's a .add file. */ |
| 339 | 455 |
| 300 | 456 char_u *sl_fbyts; /* case-folded word bytes */ |
| 324 | 457 idx_T *sl_fidxs; /* case-folded word indexes */ |
| 300 | 458 char_u *sl_kbyts; /* keep-case word bytes */ |
| 324 | 459 idx_T *sl_kidxs; /* keep-case word indexes */ |
| 339 | 460 char_u *sl_pbyts; /* prefix tree word bytes */ |
| 461 idx_T *sl_pidxs; /* prefix tree word indexes */ | |
| 462 | |
| 714 | 463 char_u *sl_info; /* infotext string or NULL */ |
| 464 | |
| 236 | 465 char_u sl_regions[17]; /* table with up to 8 region names plus NUL */ |
| 323 | 466 |
| 376 | 467 char_u *sl_midword; /* MIDWORD string or NULL */ |
| 468 | |
| 625 | 469 hashtab_T sl_wordcount; /* hashtable with word count, wordcount_T */ |
| 470 | |
| 809 | 471 int sl_compmax; /* COMPOUNDWORDMAX (default: MAXWLEN) */ |
| 501 | 472 int sl_compminlen; /* COMPOUNDMIN (default: 0) */ |
| 483 | 473 int sl_compsylmax; /* COMPOUNDSYLMAX (default: MAXWLEN) */ |
| 809 | 474 int sl_compoptions; /* COMP_* flags */ |
| 475 garray_T sl_comppat; /* CHECKCOMPOUNDPATTERN items */ | |
| 714 | 476 regprog_T *sl_compprog; /* COMPOUNDRULE turned into a regexp progrm |
| 483 | 477 * (NULL when no compounding) */ |
| 1762 | 478 char_u *sl_comprules; /* all COMPOUNDRULE concatenated (or NULL) */ |
| 483 | 479 char_u *sl_compstartflags; /* flags for first compound word */ |
| 485 | 480 char_u *sl_compallflags; /* all flags for compound words */ |
| 492 | 481 char_u sl_nobreak; /* When TRUE: no spaces between words */ |
| 483 | 482 char_u *sl_syllable; /* SYLLABLE repeatable chars or NULL */ |
| 483 garray_T sl_syl_items; /* syllable items */ | |
| 481 | 484 |
| 339 | 485 int sl_prefixcnt; /* number of items in "sl_prefprog" */ |
| 486 regprog_T **sl_prefprog; /* table with regprogs for prefixes */ | |
| 487 | |
| 323 | 488 garray_T sl_rep; /* list of fromto_T entries from REP lines */ |
| 489 short sl_rep_first[256]; /* indexes where byte first appears, -1 if | |
| 490 there is none */ | |
| 344 | 491 garray_T sl_sal; /* list of salitem_T entries from SAL lines */ |
| 375 | 492 salfirst_T sl_sal_first[256]; /* indexes where byte first appears, -1 if |
| 323 | 493 there is none */ |
| 625 | 494 int sl_followup; /* SAL followup */ |
| 495 int sl_collapse; /* SAL collapse_result */ | |
| 496 int sl_rem_accents; /* SAL remove_accents */ | |
| 375 | 497 int sl_sofo; /* SOFOFROM and SOFOTO instead of SAL items: |
| 498 * "sl_sal_first" maps chars, when has_mbyte | |
| 499 * "sl_sal" is a list of wide char lists. */ | |
| 625 | 500 garray_T sl_repsal; /* list of fromto_T entries from REPSAL lines */ |
| 501 short sl_repsal_first[256]; /* sl_rep_first for REPSAL lines */ | |
| 698 | 502 int sl_nosplitsugs; /* don't suggest splitting a word */ |
| 625 | 503 |
| 504 /* Info from the .sug file. Loaded on demand. */ | |
| 505 time_t sl_sugtime; /* timestamp for .sug file */ | |
| 506 char_u *sl_sbyts; /* soundfolded word bytes */ | |
| 507 idx_T *sl_sidxs; /* soundfolded word indexes */ | |
| 508 buf_T *sl_sugbuf; /* buffer with word number table */ | |
| 509 int sl_sugloaded; /* TRUE when .sug file was loaded or failed to | |
| 510 load */ | |
| 511 | |
| 330 | 512 int sl_has_map; /* TRUE if there is a MAP line */ |
| 513 #ifdef FEAT_MBYTE | |
| 514 hashtab_T sl_map_hash; /* MAP for multi-byte chars */ | |
| 515 int sl_map_array[256]; /* MAP for first 256 chars */ | |
| 516 #else | |
| 517 char_u sl_map_array[256]; /* MAP for first 256 chars */ | |
| 518 #endif | |
| 625 | 519 hashtab_T sl_sounddone; /* table with soundfolded words that have |
| 520 handled, see add_sound_suggest() */ | |
| 236 | 521 }; |
| 522 | |
| 243 | 523 /* First language that is loaded, start of the linked list of loaded |
| 524 * languages. */ | |
| 236 | 525 static slang_T *first_lang = NULL; |
| 526 | |
| 323 | 527 /* Flags used in .spl file for soundsalike flags. */ |
| 528 #define SAL_F0LLOWUP 1 | |
| 529 #define SAL_COLLAPSE 2 | |
| 530 #define SAL_REM_ACCENTS 4 | |
| 531 | |
| 236 | 532 /* |
| 533 * Structure used in "b_langp", filled from 'spelllang'. | |
| 534 */ | |
| 535 typedef struct langp_S | |
| 536 { | |
| 503 | 537 slang_T *lp_slang; /* info for this language */ |
| 538 slang_T *lp_sallang; /* language used for sound folding or NULL */ | |
| 539 slang_T *lp_replang; /* language used for REP items or NULL */ | |
| 236 | 540 int lp_region; /* bitmask for region or REGION_ALL */ |
| 541 } langp_T; | |
| 542 | |
| 543 #define LANGP_ENTRY(ga, i) (((langp_T *)(ga).ga_data) + (i)) | |
| 544 | |
| 307 | 545 #define REGION_ALL 0xff /* word valid in all regions */ |
| 546 | |
| 483 | 547 #define VIMSPELLMAGIC "VIMspell" /* string at start of Vim spell file */ |
| 548 #define VIMSPELLMAGICL 8 | |
| 549 #define VIMSPELLVERSION 50 | |
| 550 | |
| 625 | 551 #define VIMSUGMAGIC "VIMsug" /* string at start of Vim .sug file */ |
| 552 #define VIMSUGMAGICL 6 | |
| 553 #define VIMSUGVERSION 1 | |
| 554 | |
| 483 | 555 /* Section IDs. Only renumber them when VIMSPELLVERSION changes! */ |
| 556 #define SN_REGION 0 /* <regionname> section */ | |
| 557 #define SN_CHARFLAGS 1 /* charflags section */ | |
| 558 #define SN_MIDWORD 2 /* <midword> section */ | |
| 559 #define SN_PREFCOND 3 /* <prefcond> section */ | |
| 560 #define SN_REP 4 /* REP items section */ | |
| 561 #define SN_SAL 5 /* SAL items section */ | |
| 562 #define SN_SOFO 6 /* soundfolding section */ | |
| 563 #define SN_MAP 7 /* MAP items section */ | |
| 564 #define SN_COMPOUND 8 /* compound words section */ | |
| 565 #define SN_SYLLABLE 9 /* syllable section */ | |
| 492 | 566 #define SN_NOBREAK 10 /* NOBREAK section */ |
| 625 | 567 #define SN_SUGFILE 11 /* timestamp for .sug file */ |
| 568 #define SN_REPSAL 12 /* REPSAL items section */ | |
| 569 #define SN_WORDS 13 /* common words */ | |
| 698 | 570 #define SN_NOSPLITSUGS 14 /* don't split word for suggestions */ |
| 714 | 571 #define SN_INFO 15 /* info section */ |
| 483 | 572 #define SN_END 255 /* end of sections */ |
| 573 | |
| 574 #define SNF_REQUIRED 1 /* <sectionflags>: required section */ | |
| 575 | |
| 307 | 576 /* Result values. Lower number is accepted over higher one. */ |
| 577 #define SP_BANNED -1 | |
| 236 | 578 #define SP_OK 0 |
| 307 | 579 #define SP_RARE 1 |
| 580 #define SP_LOCAL 2 | |
| 581 #define SP_BAD 3 | |
| 236 | 582 |
| 381 | 583 /* file used for "zG" and "zW" */ |
| 385 | 584 static char_u *int_wordlist = NULL; |
| 381 | 585 |
| 625 | 586 typedef struct wordcount_S |
| 587 { | |
| 588 short_u wc_count; /* nr of times word was seen */ | |
| 589 char_u wc_word[1]; /* word, actually longer */ | |
| 590 } wordcount_T; | |
| 591 | |
| 592 static wordcount_T dumwc; | |
| 835 | 593 #define WC_KEY_OFF (unsigned)(dumwc.wc_word - (char_u *)&dumwc) |
| 625 | 594 #define HI2WC(hi) ((wordcount_T *)((hi)->hi_key - WC_KEY_OFF)) |
| 595 #define MAXWORDCOUNT 0xffff | |
| 596 | |
| 236 | 597 /* |
| 323 | 598 * Information used when looking for suggestions. |
| 599 */ | |
| 600 typedef struct suginfo_S | |
| 601 { | |
| 602 garray_T su_ga; /* suggestions, contains "suggest_T" */ | |
| 344 | 603 int su_maxcount; /* max. number of suggestions displayed */ |
| 323 | 604 int su_maxscore; /* maximum score for adding to su_ga */ |
| 625 | 605 int su_sfmaxscore; /* idem, for when doing soundfold words */ |
| 344 | 606 garray_T su_sga; /* like su_ga, sound-folded scoring */ |
| 323 | 607 char_u *su_badptr; /* start of bad word in line */ |
| 608 int su_badlen; /* length of detected bad word in line */ | |
| 346 | 609 int su_badflags; /* caps flags for bad word */ |
| 323 | 610 char_u su_badword[MAXWLEN]; /* bad word truncated at su_badlen */ |
| 611 char_u su_fbadword[MAXWLEN]; /* su_badword case-folded */ | |
| 534 | 612 char_u su_sal_badword[MAXWLEN]; /* su_badword soundfolded */ |
| 323 | 613 hashtab_T su_banned; /* table with banned words */ |
| 503 | 614 slang_T *su_sallang; /* default language for sound folding */ |
| 323 | 615 } suginfo_T; |
| 616 | |
| 617 /* One word suggestion. Used in "si_ga". */ | |
| 618 typedef struct suggest_S | |
| 619 { | |
| 620 char_u *st_word; /* suggested word, allocated string */ | |
| 625 | 621 int st_wordlen; /* STRLEN(st_word) */ |
| 323 | 622 int st_orglen; /* length of replaced text */ |
| 623 int st_score; /* lower is better */ | |
| 344 | 624 int st_altscore; /* used when st_score compares equal */ |
| 625 int st_salscore; /* st_score is for soundalike */ | |
| 324 | 626 int st_had_bonus; /* bonus already included in score */ |
| 503 | 627 slang_T *st_slang; /* language used for sound folding */ |
| 323 | 628 } suggest_T; |
| 629 | |
| 344 | 630 #define SUG(ga, i) (((suggest_T *)(ga).ga_data)[i]) |
| 323 | 631 |
| 625 | 632 /* TRUE if a word appears in the list of banned words. */ |
| 633 #define WAS_BANNED(su, word) (!HASHITEM_EMPTY(hash_find(&su->su_banned, word))) | |
| 634 | |
| 1705 | 635 /* Number of suggestions kept when cleaning up. We need to keep more than |
| 625 | 636 * what is displayed, because when rescore_suggestions() is called the score |
| 637 * may change and wrong suggestions may be removed later. */ | |
| 638 #define SUG_CLEAN_COUNT(su) ((su)->su_maxcount < 130 ? 150 : (su)->su_maxcount + 20) | |
| 324 | 639 |
| 640 /* Threshold for sorting and cleaning up suggestions. Don't want to keep lots | |
| 641 * of suggestions that are not going to be displayed. */ | |
| 625 | 642 #define SUG_MAX_COUNT(su) (SUG_CLEAN_COUNT(su) + 50) |
| 323 | 643 |
| 644 /* score for various changes */ | |
| 344 | 645 #define SCORE_SPLIT 149 /* split bad word */ |
| 698 | 646 #define SCORE_SPLIT_NO 249 /* split bad word with NOSPLITSUGS */ |
| 323 | 647 #define SCORE_ICASE 52 /* slightly different case */ |
| 483 | 648 #define SCORE_REGION 200 /* word is for different region */ |
| 323 | 649 #define SCORE_RARE 180 /* rare word */ |
| 625 | 650 #define SCORE_SWAP 75 /* swap two characters */ |
| 323 | 651 #define SCORE_SWAP3 110 /* swap two characters in three */ |
| 532 | 652 #define SCORE_REP 65 /* REP replacement */ |
| 323 | 653 #define SCORE_SUBST 93 /* substitute a character */ |
| 654 #define SCORE_SIMILAR 33 /* substitute a similar character */ | |
| 479 | 655 #define SCORE_SUBCOMP 33 /* substitute a composing character */ |
| 324 | 656 #define SCORE_DEL 94 /* delete a character */ |
| 532 | 657 #define SCORE_DELDUP 66 /* delete a duplicated character */ |
| 479 | 658 #define SCORE_DELCOMP 28 /* delete a composing character */ |
| 324 | 659 #define SCORE_INS 96 /* insert a character */ |
| 532 | 660 #define SCORE_INSDUP 67 /* insert a duplicate character */ |
| 475 | 661 #define SCORE_INSCOMP 30 /* insert a composing character */ |
| 366 | 662 #define SCORE_NONWORD 103 /* change non-word to word char */ |
| 323 | 663 |
| 372 | 664 #define SCORE_FILE 30 /* suggestion from a file */ |
| 323 | 665 #define SCORE_MAXINIT 350 /* Initial maximum score: higher == slower. |
| 666 * 350 allows for about three changes. */ | |
| 344 | 667 |
| 625 | 668 #define SCORE_COMMON1 30 /* subtracted for words seen before */ |
| 669 #define SCORE_COMMON2 40 /* subtracted for words often seen */ | |
| 670 #define SCORE_COMMON3 50 /* subtracted for words very often seen */ | |
| 671 #define SCORE_THRES2 10 /* word count threshold for COMMON2 */ | |
| 672 #define SCORE_THRES3 100 /* word count threshold for COMMON3 */ | |
| 673 | |
| 674 /* When trying changed soundfold words it becomes slow when trying more than | |
| 675 * two changes. With less then two changes it's slightly faster but we miss a | |
| 676 * few good suggestions. In rare cases we need to try three of four changes. | |
| 677 */ | |
| 678 #define SCORE_SFMAX1 200 /* maximum score for first try */ | |
| 679 #define SCORE_SFMAX2 300 /* maximum score for second try */ | |
| 680 #define SCORE_SFMAX3 400 /* maximum score for third try */ | |
| 681 | |
| 344 | 682 #define SCORE_BIG SCORE_INS * 3 /* big difference */ |
| 625 | 683 #define SCORE_MAXMAX 999999 /* accept any score */ |
| 684 #define SCORE_LIMITMAX 350 /* for spell_edit_score_limit() */ | |
| 685 | |
| 686 /* for spell_edit_score_limit() we need to know the minimum value of | |
| 687 * SCORE_ICASE, SCORE_SWAP, SCORE_DEL, SCORE_SIMILAR and SCORE_INS */ | |
| 688 #define SCORE_EDIT_MIN SCORE_SIMILAR | |
| 323 | 689 |
| 690 /* | |
| 236 | 691 * Structure to store info for word matching. |
| 692 */ | |
| 693 typedef struct matchinf_S | |
| 694 { | |
| 695 langp_T *mi_lp; /* info for language and region */ | |
| 243 | 696 |
| 697 /* pointers to original text to be checked */ | |
| 236 | 698 char_u *mi_word; /* start of word being checked */ |
| 339 | 699 char_u *mi_end; /* end of matching word so far */ |
| 243 | 700 char_u *mi_fend; /* next char to be added to mi_fword */ |
| 300 | 701 char_u *mi_cend; /* char after what was used for |
| 702 mi_capflags */ | |
| 243 | 703 |
| 704 /* case-folded text */ | |
| 705 char_u mi_fword[MAXWLEN + 1]; /* mi_word case-folded */ | |
| 300 | 706 int mi_fwordlen; /* nr of valid bytes in mi_fword */ |
| 243 | 707 |
| 339 | 708 /* for when checking word after a prefix */ |
| 709 int mi_prefarridx; /* index in sl_pidxs with list of | |
| 481 | 710 affixID/condition */ |
| 339 | 711 int mi_prefcnt; /* number of entries at mi_prefarridx */ |
| 712 int mi_prefixlen; /* byte length of prefix */ | |
| 455 | 713 #ifdef FEAT_MBYTE |
| 714 int mi_cprefixlen; /* byte length of prefix in original | |
| 715 case */ | |
| 716 #else | |
| 717 # define mi_cprefixlen mi_prefixlen /* it's the same value */ | |
| 718 #endif | |
| 339 | 719 |
| 481 | 720 /* for when checking a compound word */ |
| 721 int mi_compoff; /* start of following word offset */ | |
| 483 | 722 char_u mi_compflags[MAXWLEN]; /* flags for compound words used */ |
| 723 int mi_complen; /* nr of compound words used */ | |
| 809 | 724 int mi_compextra; /* nr of COMPOUNDROOT words */ |
| 481 | 725 |
| 243 | 726 /* others */ |
| 236 | 727 int mi_result; /* result so far: SP_BAD, SP_OK, etc. */ |
| 300 | 728 int mi_capflags; /* WF_ONECAP WF_ALLCAP WF_KEEPCAP */ |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
729 win_T *mi_win; /* buffer being checked */ |
| 492 | 730 |
| 731 /* for NOBREAK */ | |
| 732 int mi_result2; /* "mi_resul" without following word */ | |
| 733 char_u *mi_end2; /* "mi_end" without following word */ | |
| 236 | 734 } matchinf_T; |
| 735 | |
| 307 | 736 /* |
| 737 * The tables used for recognizing word characters according to spelling. | |
| 738 * These are only used for the first 256 characters of 'encoding'. | |
| 739 */ | |
| 740 typedef struct spelltab_S | |
| 741 { | |
| 742 char_u st_isw[256]; /* flags: is word char */ | |
| 743 char_u st_isu[256]; /* flags: is uppercase char */ | |
| 744 char_u st_fold[256]; /* chars: folded case */ | |
| 324 | 745 char_u st_upper[256]; /* chars: upper case */ |
| 307 | 746 } spelltab_T; |
| 747 | |
| 748 static spelltab_T spelltab; | |
| 749 static int did_set_spelltab; | |
| 750 | |
| 324 | 751 #define CF_WORD 0x01 |
| 752 #define CF_UPPER 0x02 | |
| 307 | 753 |
| 754 static void clear_spell_chartab __ARGS((spelltab_T *sp)); | |
| 755 static int set_spell_finish __ARGS((spelltab_T *new_st)); | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
756 static int spell_iswordp __ARGS((char_u *p, win_T *wp)); |
| 376 | 757 static int spell_iswordp_nmw __ARGS((char_u *p)); |
| 758 #ifdef FEAT_MBYTE | |
| 1580 | 759 static int spell_mb_isword_class __ARGS((int cl)); |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
760 static int spell_iswordp_w __ARGS((int *p, win_T *wp)); |
| 376 | 761 #endif |
| 483 | 762 static int write_spell_prefcond __ARGS((FILE *fd, garray_T *gap)); |
| 307 | 763 |
| 764 /* | |
| 344 | 765 * For finding suggestions: At each node in the tree these states are tried: |
| 330 | 766 */ |
| 767 typedef enum | |
| 768 { | |
| 344 | 769 STATE_START = 0, /* At start of node check for NUL bytes (goodword |
| 770 * ends); if badword ends there is a match, otherwise | |
| 771 * try splitting word. */ | |
| 375 | 772 STATE_NOPREFIX, /* try without prefix */ |
| 344 | 773 STATE_SPLITUNDO, /* Undo splitting. */ |
| 330 | 774 STATE_ENDNUL, /* Past NUL bytes at start of the node. */ |
| 775 STATE_PLAIN, /* Use each byte of the node. */ | |
| 776 STATE_DEL, /* Delete a byte from the bad word. */ | |
| 625 | 777 STATE_INS_PREP, /* Prepare for inserting bytes. */ |
| 330 | 778 STATE_INS, /* Insert a byte in the bad word. */ |
| 779 STATE_SWAP, /* Swap two bytes. */ | |
| 344 | 780 STATE_UNSWAP, /* Undo swap two characters. */ |
| 781 STATE_SWAP3, /* Swap two characters over three. */ | |
| 782 STATE_UNSWAP3, /* Undo Swap two characters over three. */ | |
| 783 STATE_UNROT3L, /* Undo rotate three characters left */ | |
| 784 STATE_UNROT3R, /* Undo rotate three characters right */ | |
| 330 | 785 STATE_REP_INI, /* Prepare for using REP items. */ |
| 786 STATE_REP, /* Use matching REP items from the .aff file. */ | |
| 787 STATE_REP_UNDO, /* Undo a REP item replacement. */ | |
| 788 STATE_FINAL /* End of this node. */ | |
| 789 } state_T; | |
| 790 | |
| 791 /* | |
| 346 | 792 * Struct to keep the state at each level in suggest_try_change(). |
| 323 | 793 */ |
| 794 typedef struct trystate_S | |
| 795 { | |
| 330 | 796 state_T ts_state; /* state at this level, STATE_ */ |
| 323 | 797 int ts_score; /* score */ |
| 344 | 798 idx_T ts_arridx; /* index in tree array, start of node */ |
| 330 | 799 short ts_curi; /* index in list of child nodes */ |
| 800 char_u ts_fidx; /* index in fword[], case-folded bad word */ | |
| 801 char_u ts_fidxtry; /* ts_fidx at which bytes may be changed */ | |
| 802 char_u ts_twordlen; /* valid length of tword[] */ | |
| 482 | 803 char_u ts_prefixdepth; /* stack depth for end of prefix or |
| 485 | 804 * PFD_PREFIXTREE or PFD_NOPREFIX */ |
| 805 char_u ts_flags; /* TSF_ flags */ | |
| 330 | 806 #ifdef FEAT_MBYTE |
| 807 char_u ts_tcharlen; /* number of bytes in tword character */ | |
| 808 char_u ts_tcharidx; /* current byte index in tword character */ | |
| 809 char_u ts_isdiff; /* DIFF_ values */ | |
| 810 char_u ts_fcharstart; /* index in fword where badword char started */ | |
| 811 #endif | |
| 483 | 812 char_u ts_prewordlen; /* length of word in "preword[]" */ |
| 813 char_u ts_splitoff; /* index in "tword" after last split */ | |
| 492 | 814 char_u ts_splitfidx; /* "ts_fidx" at word split */ |
| 483 | 815 char_u ts_complen; /* nr of compound words used */ |
| 485 | 816 char_u ts_compsplit; /* index for "compflags" where word was spit */ |
| 346 | 817 char_u ts_save_badflags; /* su_badflags saved here */ |
| 625 | 818 char_u ts_delidx; /* index in fword for char that was deleted, |
| 819 valid when "ts_flags" has TSF_DIDDEL */ | |
| 323 | 820 } trystate_T; |
| 821 | |
| 330 | 822 /* values for ts_isdiff */ |
| 823 #define DIFF_NONE 0 /* no different byte (yet) */ | |
| 824 #define DIFF_YES 1 /* different byte found */ | |
| 825 #define DIFF_INSERT 2 /* inserting character */ | |
| 826 | |
| 485 | 827 /* values for ts_flags */ |
| 828 #define TSF_PREFIXOK 1 /* already checked that prefix is OK */ | |
| 829 #define TSF_DIDSPLIT 2 /* tried split at this point */ | |
| 625 | 830 #define TSF_DIDDEL 4 /* did a delete, "ts_delidx" has index */ |
| 485 | 831 |
| 375 | 832 /* special values ts_prefixdepth */ |
| 485 | 833 #define PFD_NOPREFIX 0xff /* not using prefixes */ |
| 482 | 834 #define PFD_PREFIXTREE 0xfe /* walking through the prefix tree */ |
| 625 | 835 #define PFD_NOTSPECIAL 0xfd /* highest value that's not special */ |
| 375 | 836 |
| 339 | 837 /* mode values for find_word */ |
| 481 | 838 #define FIND_FOLDWORD 0 /* find word case-folded */ |
| 839 #define FIND_KEEPWORD 1 /* find keep-case word */ | |
| 840 #define FIND_PREFIX 2 /* find word after prefix */ | |
| 841 #define FIND_COMPOUND 3 /* find case-folded compound word */ | |
| 842 #define FIND_KEEPCOMPOUND 4 /* find keep-case compound word */ | |
| 339 | 843 |
| 236 | 844 static slang_T *slang_alloc __ARGS((char_u *lang)); |
| 845 static void slang_free __ARGS((slang_T *lp)); | |
| 310 | 846 static void slang_clear __ARGS((slang_T *lp)); |
| 625 | 847 static void slang_clear_sug __ARGS((slang_T *lp)); |
| 339 | 848 static void find_word __ARGS((matchinf_T *mip, int mode)); |
| 1762 | 849 static int match_checkcompoundpattern __ARGS((char_u *ptr, int wlen, garray_T *gap)); |
| 483 | 850 static int can_compound __ARGS((slang_T *slang, char_u *word, char_u *flags)); |
| 1762 | 851 static int can_be_compound __ARGS((trystate_T *sp, slang_T *slang, char_u *compflags, int flag)); |
| 852 static int match_compoundrule __ARGS((slang_T *slang, char_u *compflags)); | |
| 455 | 853 static int valid_word_prefix __ARGS((int totprefcnt, int arridx, int flags, char_u *word, slang_T *slang, int cond_req)); |
| 485 | 854 static void find_prefix __ARGS((matchinf_T *mip, int mode)); |
| 339 | 855 static int fold_more __ARGS((matchinf_T *mip)); |
| 389 | 856 static int spell_valid_case __ARGS((int wordflags, int treeflags)); |
| 498 | 857 static int no_spell_checking __ARGS((win_T *wp)); |
| 307 | 858 static void spell_load_lang __ARGS((char_u *lang)); |
| 310 | 859 static char_u *spell_enc __ARGS((void)); |
| 385 | 860 static void int_wordlist_spl __ARGS((char_u *fname)); |
| 310 | 861 static void spell_load_cb __ARGS((char_u *fname, void *cookie)); |
| 323 | 862 static slang_T *spell_load_file __ARGS((char_u *fname, char_u *lang, slang_T *old_lp, int silent)); |
| 389 | 863 static char_u *read_cnt_string __ARGS((FILE *fd, int cnt_bytes, int *lenp)); |
| 483 | 864 static int read_region_section __ARGS((FILE *fd, slang_T *slang, int len)); |
| 865 static int read_charflags_section __ARGS((FILE *fd)); | |
| 866 static int read_prefcond_section __ARGS((FILE *fd, slang_T *lp)); | |
| 625 | 867 static int read_rep_section __ARGS((FILE *fd, garray_T *gap, short *first)); |
| 483 | 868 static int read_sal_section __ARGS((FILE *fd, slang_T *slang)); |
| 625 | 869 static int read_words_section __ARGS((FILE *fd, slang_T *lp, int len)); |
| 870 static void count_common_word __ARGS((slang_T *lp, char_u *word, int len, int count)); | |
| 871 static int score_wordcount_adj __ARGS((slang_T *slang, int score, char_u *word, int split)); | |
| 483 | 872 static int read_sofo_section __ARGS((FILE *fd, slang_T *slang)); |
| 873 static int read_compound __ARGS((FILE *fd, slang_T *slang, int len)); | |
| 495 | 874 static int byte_in_str __ARGS((char_u *str, int byte)); |
| 483 | 875 static int init_syl_tab __ARGS((slang_T *slang)); |
| 876 static int count_syllables __ARGS((slang_T *slang, char_u *word)); | |
| 381 | 877 static int set_sofo __ARGS((slang_T *lp, char_u *from, char_u *to)); |
| 878 static void set_sal_first __ARGS((slang_T *lp)); | |
| 372 | 879 #ifdef FEAT_MBYTE |
| 880 static int *mb_str2wide __ARGS((char_u *s)); | |
| 881 #endif | |
| 625 | 882 static int spell_read_tree __ARGS((FILE *fd, char_u **bytsp, idx_T **idxsp, int prefixtree, int prefixcnt)); |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
883 static idx_T read_tree_node __ARGS((FILE *fd, char_u *byts, idx_T *idxs, int maxidx, idx_T startidx, int prefixtree, int maxprefcondnr)); |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
884 static void clear_midword __ARGS((win_T *buf)); |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
885 static void use_midword __ARGS((slang_T *lp, win_T *buf)); |
| 236 | 886 static int find_region __ARGS((char_u *rp, char_u *region)); |
| 887 static int captype __ARGS((char_u *word, char_u *end)); | |
| 474 | 888 static int badword_captype __ARGS((char_u *word, char_u *end)); |
| 323 | 889 static void spell_reload_one __ARGS((char_u *fname, int added_word)); |
| 483 | 890 static void set_spell_charflags __ARGS((char_u *flags, int cnt, char_u *upp)); |
| 307 | 891 static int set_spell_chartab __ARGS((char_u *fol, char_u *low, char_u *upp)); |
| 892 static int spell_casefold __ARGS((char_u *p, int len, char_u *buf, int buflen)); | |
| 475 | 893 static int check_need_cap __ARGS((linenr_T lnum, colnr_T col)); |
| 638 | 894 static void spell_find_suggest __ARGS((char_u *badptr, int badlen, suginfo_T *su, int maxcount, int banbadword, int need_cap, int interactive)); |
| 372 | 895 #ifdef FEAT_EVAL |
| 896 static void spell_suggest_expr __ARGS((suginfo_T *su, char_u *expr)); | |
| 897 #endif | |
| 898 static void spell_suggest_file __ARGS((suginfo_T *su, char_u *fname)); | |
| 625 | 899 static void spell_suggest_intern __ARGS((suginfo_T *su, int interactive)); |
| 900 static void suggest_load_files __ARGS((void)); | |
| 901 static void tree_count_words __ARGS((char_u *byts, idx_T *idxs)); | |
| 344 | 902 static void spell_find_cleanup __ARGS((suginfo_T *su)); |
| 324 | 903 static void onecap_copy __ARGS((char_u *word, char_u *wcopy, int upper)); |
| 344 | 904 static void allcap_copy __ARGS((char_u *word, char_u *wcopy)); |
| 346 | 905 static void suggest_try_special __ARGS((suginfo_T *su)); |
| 906 static void suggest_try_change __ARGS((suginfo_T *su)); | |
| 625 | 907 static void suggest_trie_walk __ARGS((suginfo_T *su, langp_T *lp, char_u *fword, int soundfold)); |
| 908 static void go_deeper __ARGS((trystate_T *stack, int depth, int score_add)); | |
| 455 | 909 #ifdef FEAT_MBYTE |
| 910 static int nofold_len __ARGS((char_u *fword, int flen, char_u *word)); | |
| 911 #endif | |
| 323 | 912 static void find_keepcap_word __ARGS((slang_T *slang, char_u *fword, char_u *kword)); |
| 344 | 913 static void score_comp_sal __ARGS((suginfo_T *su)); |
| 914 static void score_combine __ARGS((suginfo_T *su)); | |
| 351 | 915 static int stp_sal_score __ARGS((suggest_T *stp, suginfo_T *su, slang_T *slang, char_u *badsound)); |
| 625 | 916 static void suggest_try_soundalike_prep __ARGS((void)); |
| 346 | 917 static void suggest_try_soundalike __ARGS((suginfo_T *su)); |
| 625 | 918 static void suggest_try_soundalike_finish __ARGS((void)); |
| 919 static void add_sound_suggest __ARGS((suginfo_T *su, char_u *goodword, int score, langp_T *lp)); | |
| 920 static int soundfold_find __ARGS((slang_T *slang, char_u *word)); | |
| 323 | 921 static void make_case_word __ARGS((char_u *fword, char_u *cword, int flags)); |
| 330 | 922 static void set_map_str __ARGS((slang_T *lp, char_u *map)); |
| 323 | 923 static int similar_chars __ARGS((slang_T *slang, int c1, int c2)); |
| 625 | 924 static void add_suggestion __ARGS((suginfo_T *su, garray_T *gap, char_u *goodword, int badlen, int score, int altscore, int had_bonus, slang_T *slang, int maxsf)); |
| 925 static void check_suggestions __ARGS((suginfo_T *su, garray_T *gap)); | |
| 323 | 926 static void add_banned __ARGS((suginfo_T *su, char_u *word)); |
| 324 | 927 static void rescore_suggestions __ARGS((suginfo_T *su)); |
| 534 | 928 static void rescore_one __ARGS((suginfo_T *su, suggest_T *stp)); |
| 344 | 929 static int cleanup_suggestions __ARGS((garray_T *gap, int maxscore, int keep)); |
| 375 | 930 static void spell_soundfold __ARGS((slang_T *slang, char_u *inword, int folded, char_u *res)); |
| 931 static void spell_soundfold_sofo __ARGS((slang_T *slang, char_u *inword, char_u *res)); | |
| 932 static void spell_soundfold_sal __ARGS((slang_T *slang, char_u *inword, char_u *res)); | |
| 372 | 933 #ifdef FEAT_MBYTE |
| 375 | 934 static void spell_soundfold_wsal __ARGS((slang_T *slang, char_u *inword, char_u *res)); |
| 372 | 935 #endif |
| 344 | 936 static int soundalike_score __ARGS((char_u *goodsound, char_u *badsound)); |
| 625 | 937 static int spell_edit_score __ARGS((slang_T *slang, char_u *badword, char_u *goodword)); |
| 938 static int spell_edit_score_limit __ARGS((slang_T *slang, char_u *badword, char_u *goodword, int limit)); | |
| 939 #ifdef FEAT_MBYTE | |
| 940 static int spell_edit_score_limit_w __ARGS((slang_T *slang, char_u *badword, char_u *goodword, int limit)); | |
| 941 #endif | |
| 701 | 942 static void dump_word __ARGS((slang_T *slang, char_u *word, char_u *pat, int *dir, int round, int flags, linenr_T lnum)); |
| 943 static linenr_T dump_prefixes __ARGS((slang_T *slang, char_u *word, char_u *pat, int *dir, int round, int flags, linenr_T startlnum)); | |
| 625 | 944 static buf_T *open_spellbuf __ARGS((void)); |
| 945 static void close_spellbuf __ARGS((buf_T *buf)); | |
| 323 | 946 |
| 324 | 947 /* |
| 948 * Use our own character-case definitions, because the current locale may | |
| 949 * differ from what the .spl file uses. | |
| 950 * These must not be called with negative number! | |
| 951 */ | |
| 952 #ifndef FEAT_MBYTE | |
| 953 /* Non-multi-byte implementation. */ | |
| 1883 | 954 # define SPELL_TOFOLD(c) ((c) < 256 ? (int)spelltab.st_fold[c] : (c)) |
| 955 # define SPELL_TOUPPER(c) ((c) < 256 ? (int)spelltab.st_upper[c] : (c)) | |
| 324 | 956 # define SPELL_ISUPPER(c) ((c) < 256 ? spelltab.st_isu[c] : FALSE) |
| 957 #else | |
| 447 | 958 # if defined(HAVE_WCHAR_H) |
| 959 # include <wchar.h> /* for towupper() and towlower() */ | |
| 960 # endif | |
| 324 | 961 /* Multi-byte implementation. For Unicode we can call utf_*(), but don't do |
| 962 * that for ASCII, because we don't want to use 'casemap' here. Otherwise use | |
| 963 * the "w" library function for characters above 255 if available. */ | |
| 964 # ifdef HAVE_TOWLOWER | |
| 965 # define SPELL_TOFOLD(c) (enc_utf8 && (c) >= 128 ? utf_fold(c) \ | |
| 1883 | 966 : (c) < 256 ? (int)spelltab.st_fold[c] : (int)towlower(c)) |
| 324 | 967 # else |
| 968 # define SPELL_TOFOLD(c) (enc_utf8 && (c) >= 128 ? utf_fold(c) \ | |
| 1883 | 969 : (c) < 256 ? (int)spelltab.st_fold[c] : (c)) |
| 324 | 970 # endif |
| 971 | |
| 972 # ifdef HAVE_TOWUPPER | |
| 973 # define SPELL_TOUPPER(c) (enc_utf8 && (c) >= 128 ? utf_toupper(c) \ | |
| 1883 | 974 : (c) < 256 ? (int)spelltab.st_upper[c] : (int)towupper(c)) |
| 324 | 975 # else |
| 976 # define SPELL_TOUPPER(c) (enc_utf8 && (c) >= 128 ? utf_toupper(c) \ | |
| 1883 | 977 : (c) < 256 ? (int)spelltab.st_upper[c] : (c)) |
| 324 | 978 # endif |
| 979 | |
| 980 # ifdef HAVE_ISWUPPER | |
| 981 # define SPELL_ISUPPER(c) (enc_utf8 && (c) >= 128 ? utf_isupper(c) \ | |
| 982 : (c) < 256 ? spelltab.st_isu[c] : iswupper(c)) | |
| 983 # else | |
| 984 # define SPELL_ISUPPER(c) (enc_utf8 && (c) >= 128 ? utf_isupper(c) \ | |
| 474 | 985 : (c) < 256 ? spelltab.st_isu[c] : (FALSE)) |
| 324 | 986 # endif |
| 987 #endif | |
| 988 | |
| 307 | 989 |
| 990 static char *e_format = N_("E759: Format error in spell file"); | |
| 381 | 991 static char *e_spell_trunc = N_("E758: Truncated spell file"); |
| 482 | 992 static char *e_afftrailing = N_("Trailing text in %s line %d: %s"); |
| 495 | 993 static char *e_affname = N_("Affix name too long in %s line %d: %s"); |
| 994 static char *e_affform = N_("E761: Format error in affix file FOL, LOW or UPP"); | |
| 995 static char *e_affrange = N_("E762: Character in FOL, LOW or UPP is out of range"); | |
| 469 | 996 static char *msg_compressing = N_("Compressing word tree..."); |
| 236 | 997 |
| 626 | 998 /* Remember what "z?" replaced. */ |
| 999 static char_u *repl_from = NULL; | |
| 1000 static char_u *repl_to = NULL; | |
| 1001 | |
| 236 | 1002 /* |
| 1003 * Main spell-checking function. | |
| 300 | 1004 * "ptr" points to a character that could be the start of a word. |
| 534 | 1005 * "*attrp" is set to the highlight index for a badly spelled word. For a |
| 1006 * non-word or when it's OK it remains unchanged. | |
| 236 | 1007 * This must only be called when 'spelllang' is not empty. |
| 323 | 1008 * |
| 385 | 1009 * "capcol" is used to check for a Capitalised word after the end of a |
| 1010 * sentence. If it's zero then perform the check. Return the column where to | |
| 1011 * check next, or -1 when no sentence end was found. If it's NULL then don't | |
| 1012 * worry. | |
| 323 | 1013 * |
| 236 | 1014 * Returns the length of the word in bytes, also when it's OK, so that the |
| 1015 * caller can skip over the word. | |
| 1016 */ | |
| 1017 int | |
| 625 | 1018 spell_check(wp, ptr, attrp, capcol, docount) |
| 236 | 1019 win_T *wp; /* current window */ |
| 1020 char_u *ptr; | |
| 534 | 1021 hlf_T *attrp; |
| 385 | 1022 int *capcol; /* column to check for Capital */ |
| 625 | 1023 int docount; /* count good words */ |
| 236 | 1024 { |
| 1025 matchinf_T mi; /* Most things are put in "mi" so that it can | |
| 1026 be passed to functions quickly. */ | |
| 344 | 1027 int nrlen = 0; /* found a number first */ |
| 385 | 1028 int c; |
| 483 | 1029 int wrongcaplen = 0; |
| 500 | 1030 int lpi; |
| 625 | 1031 int count_word = docount; |
| 236 | 1032 |
| 307 | 1033 /* A word never starts at a space or a control character. Return quickly |
| 1034 * then, skipping over the character. */ | |
| 1035 if (*ptr <= ' ') | |
| 1036 return 1; | |
| 690 | 1037 |
| 1038 /* Return here when loading language files failed. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1039 if (wp->w_s->b_langp.ga_len == 0) |
| 690 | 1040 return 1; |
| 1041 | |
| 483 | 1042 vim_memset(&mi, 0, sizeof(matchinf_T)); |
| 236 | 1043 |
| 344 | 1044 /* A number is always OK. Also skip hexadecimal numbers 0xFF99 and |
| 584 | 1045 * 0X99FF. But always do check spelling to find "3GPP" and "11 |
| 1046 * julifeest". */ | |
| 300 | 1047 if (*ptr >= '0' && *ptr <= '9') |
| 1048 { | |
| 316 | 1049 if (*ptr == '0' && (ptr[1] == 'x' || ptr[1] == 'X')) |
| 1050 mi.mi_end = skiphex(ptr + 2); | |
| 300 | 1051 else |
| 344 | 1052 mi.mi_end = skipdigits(ptr); |
| 835 | 1053 nrlen = (int)(mi.mi_end - ptr); |
| 584 | 1054 } |
| 346 | 1055 |
| 1056 /* Find the normal end of the word (until the next non-word character). */ | |
| 344 | 1057 mi.mi_word = ptr; |
| 584 | 1058 mi.mi_fend = ptr; |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1059 if (spell_iswordp(mi.mi_fend, wp)) |
| 344 | 1060 { |
| 1061 do | |
| 1062 { | |
| 307 | 1063 mb_ptr_adv(mi.mi_fend); |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1064 } while (*mi.mi_fend != NUL && spell_iswordp(mi.mi_fend, wp)); |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1065 |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1066 if (capcol != NULL && *capcol == 0 && wp->w_s->b_cap_prog != NULL) |
| 385 | 1067 { |
| 1068 /* Check word starting with capital letter. */ | |
| 455 | 1069 c = PTR2CHAR(ptr); |
| 385 | 1070 if (!SPELL_ISUPPER(c)) |
| 483 | 1071 wrongcaplen = (int)(mi.mi_fend - ptr); |
| 385 | 1072 } |
| 1073 } | |
| 1074 if (capcol != NULL) | |
| 1075 *capcol = -1; | |
| 344 | 1076 |
| 1077 /* We always use the characters up to the next non-word character, | |
| 1078 * also for bad words. */ | |
| 1079 mi.mi_end = mi.mi_fend; | |
| 1080 | |
| 1081 /* Check caps type later. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1082 mi.mi_capflags = 0; |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1083 mi.mi_cend = NULL; |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1084 mi.mi_win = wp; |
| 344 | 1085 |
| 483 | 1086 /* case-fold the word with one non-word character, so that we can check |
| 1087 * for the word end. */ | |
| 344 | 1088 if (*mi.mi_fend != NUL) |
| 1089 mb_ptr_adv(mi.mi_fend); | |
| 1090 | |
| 1091 (void)spell_casefold(ptr, (int)(mi.mi_fend - ptr), mi.mi_fword, | |
| 1092 MAXWLEN + 1); | |
| 835 | 1093 mi.mi_fwordlen = (int)STRLEN(mi.mi_fword); |
| 344 | 1094 |
| 1095 /* The word is bad unless we recognize it. */ | |
| 1096 mi.mi_result = SP_BAD; | |
| 492 | 1097 mi.mi_result2 = SP_BAD; |
| 344 | 1098 |
| 1099 /* | |
| 1100 * Loop over the languages specified in 'spelllang'. | |
| 625 | 1101 * We check them all, because a word may be matched longer in another |
| 1102 * language. | |
| 344 | 1103 */ |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1104 for (lpi = 0; lpi < wp->w_s->b_langp.ga_len; ++lpi) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1105 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1106 mi.mi_lp = LANGP_ENTRY(wp->w_s->b_langp, lpi); |
| 500 | 1107 |
| 1108 /* If reloading fails the language is still in the list but everything | |
| 1109 * has been cleared. */ | |
| 1110 if (mi.mi_lp->lp_slang->sl_fidxs == NULL) | |
| 1111 continue; | |
| 1112 | |
| 344 | 1113 /* Check for a matching word in case-folded words. */ |
| 1114 find_word(&mi, FIND_FOLDWORD); | |
| 1115 | |
| 1116 /* Check for a matching word in keep-case words. */ | |
| 1117 find_word(&mi, FIND_KEEPWORD); | |
| 1118 | |
| 1119 /* Check for matching prefixes. */ | |
| 485 | 1120 find_prefix(&mi, FIND_FOLDWORD); |
| 492 | 1121 |
| 1122 /* For a NOBREAK language, may want to use a word without a following | |
| 1123 * word as a backup. */ | |
| 1124 if (mi.mi_lp->lp_slang->sl_nobreak && mi.mi_result == SP_BAD | |
| 1125 && mi.mi_result2 != SP_BAD) | |
| 1126 { | |
| 1127 mi.mi_result = mi.mi_result2; | |
| 1128 mi.mi_end = mi.mi_end2; | |
| 1129 } | |
| 625 | 1130 |
| 1131 /* Count the word in the first language where it's found to be OK. */ | |
| 1132 if (count_word && mi.mi_result == SP_OK) | |
| 1133 { | |
| 1134 count_common_word(mi.mi_lp->lp_slang, ptr, | |
| 1135 (int)(mi.mi_end - ptr), 1); | |
| 1136 count_word = FALSE; | |
| 1137 } | |
| 344 | 1138 } |
| 1139 | |
| 1140 if (mi.mi_result != SP_OK) | |
| 1141 { | |
| 346 | 1142 /* If we found a number skip over it. Allows for "42nd". Do flag |
| 1143 * rare and local words, e.g., "3GPP". */ | |
| 344 | 1144 if (nrlen > 0) |
| 346 | 1145 { |
| 1146 if (mi.mi_result == SP_BAD || mi.mi_result == SP_BANNED) | |
| 1147 return nrlen; | |
| 1148 } | |
| 344 | 1149 |
| 1150 /* When we are at a non-word character there is no error, just | |
| 1151 * skip over the character (try looking for a word after it). */ | |
| 390 | 1152 else if (!spell_iswordp_nmw(ptr)) |
| 243 | 1153 { |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1154 if (capcol != NULL && wp->w_s->b_cap_prog != NULL) |
| 385 | 1155 { |
| 1156 regmatch_T regmatch; | |
| 1157 | |
| 1158 /* Check for end of sentence. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1159 regmatch.regprog = wp->w_s->b_cap_prog; |
| 385 | 1160 regmatch.rm_ic = FALSE; |
| 1161 if (vim_regexec(®match, ptr, 0)) | |
| 1162 *capcol = (int)(regmatch.endp[0] - ptr); | |
| 1163 } | |
| 1164 | |
| 344 | 1165 #ifdef FEAT_MBYTE |
| 1166 if (has_mbyte) | |
| 474 | 1167 return (*mb_ptr2len)(ptr); |
| 344 | 1168 #endif |
| 1169 return 1; | |
| 300 | 1170 } |
| 483 | 1171 else if (mi.mi_end == ptr) |
| 1172 /* Always include at least one character. Required for when there | |
| 1173 * is a mixup in "midword". */ | |
| 1174 mb_ptr_adv(mi.mi_end); | |
| 492 | 1175 else if (mi.mi_result == SP_BAD |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1176 && LANGP_ENTRY(wp->w_s->b_langp, 0)->lp_slang->sl_nobreak) |
| 492 | 1177 { |
| 1178 char_u *p, *fp; | |
| 1179 int save_result = mi.mi_result; | |
| 1180 | |
| 1181 /* First language in 'spelllang' is NOBREAK. Find first position | |
| 1182 * at which any word would be valid. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1183 mi.mi_lp = LANGP_ENTRY(wp->w_s->b_langp, 0); |
| 500 | 1184 if (mi.mi_lp->lp_slang->sl_fidxs != NULL) |
| 1185 { | |
| 1186 p = mi.mi_word; | |
| 1187 fp = mi.mi_fword; | |
| 1188 for (;;) | |
| 1189 { | |
| 1190 mb_ptr_adv(p); | |
| 1191 mb_ptr_adv(fp); | |
| 1192 if (p >= mi.mi_end) | |
| 1193 break; | |
| 835 | 1194 mi.mi_compoff = (int)(fp - mi.mi_fword); |
| 500 | 1195 find_word(&mi, FIND_COMPOUND); |
| 1196 if (mi.mi_result != SP_BAD) | |
| 1197 { | |
| 1198 mi.mi_end = p; | |
| 1199 break; | |
| 1200 } | |
| 1201 } | |
| 1202 mi.mi_result = save_result; | |
| 1203 } | |
| 492 | 1204 } |
| 243 | 1205 |
| 344 | 1206 if (mi.mi_result == SP_BAD || mi.mi_result == SP_BANNED) |
| 534 | 1207 *attrp = HLF_SPB; |
| 344 | 1208 else if (mi.mi_result == SP_RARE) |
| 534 | 1209 *attrp = HLF_SPR; |
| 344 | 1210 else |
| 534 | 1211 *attrp = HLF_SPL; |
| 243 | 1212 } |
| 1213 | |
| 483 | 1214 if (wrongcaplen > 0 && (mi.mi_result == SP_OK || mi.mi_result == SP_RARE)) |
| 1215 { | |
| 1216 /* Report SpellCap only when the word isn't badly spelled. */ | |
| 534 | 1217 *attrp = HLF_SPC; |
| 483 | 1218 return wrongcaplen; |
| 1219 } | |
| 1220 | |
| 300 | 1221 return (int)(mi.mi_end - ptr); |
| 236 | 1222 } |
| 1223 | |
| 1224 /* | |
| 300 | 1225 * Check if the word at "mip->mi_word" is in the tree. |
| 339 | 1226 * When "mode" is FIND_FOLDWORD check in fold-case word tree. |
| 1227 * When "mode" is FIND_KEEPWORD check in keep-case word tree. | |
| 1228 * When "mode" is FIND_PREFIX check for word after prefix in fold-case word | |
| 1229 * tree. | |
| 300 | 1230 * |
| 1231 * For a match mip->mi_result is updated. | |
| 243 | 1232 */ |
| 1233 static void | |
| 339 | 1234 find_word(mip, mode) |
| 243 | 1235 matchinf_T *mip; |
| 339 | 1236 int mode; |
| 243 | 1237 { |
| 324 | 1238 idx_T arridx = 0; |
| 300 | 1239 int endlen[MAXWLEN]; /* length at possible word endings */ |
| 324 | 1240 idx_T endidx[MAXWLEN]; /* possible word endings */ |
| 300 | 1241 int endidxcnt = 0; |
| 1242 int len; | |
| 1243 int wlen = 0; | |
| 1244 int flen; | |
| 1245 int c; | |
| 1246 char_u *ptr; | |
| 324 | 1247 idx_T lo, hi, m; |
| 243 | 1248 #ifdef FEAT_MBYTE |
| 300 | 1249 char_u *s; |
| 490 | 1250 #endif |
| 339 | 1251 char_u *p; |
| 307 | 1252 int res = SP_BAD; |
| 300 | 1253 slang_T *slang = mip->mi_lp->lp_slang; |
| 1254 unsigned flags; | |
| 1255 char_u *byts; | |
| 324 | 1256 idx_T *idxs; |
| 481 | 1257 int word_ends; |
| 485 | 1258 int prefix_found; |
| 492 | 1259 int nobreak_result; |
| 481 | 1260 |
| 1261 if (mode == FIND_KEEPWORD || mode == FIND_KEEPCOMPOUND) | |
| 236 | 1262 { |
| 300 | 1263 /* Check for word with matching case in keep-case tree. */ |
| 1264 ptr = mip->mi_word; | |
| 1265 flen = 9999; /* no case folding, always enough bytes */ | |
| 1266 byts = slang->sl_kbyts; | |
| 1267 idxs = slang->sl_kidxs; | |
| 481 | 1268 |
| 1269 if (mode == FIND_KEEPCOMPOUND) | |
| 1270 /* Skip over the previously found word(s). */ | |
| 1271 wlen += mip->mi_compoff; | |
| 236 | 1272 } |
| 1273 else | |
| 1274 { | |
| 300 | 1275 /* Check for case-folded in case-folded tree. */ |
| 1276 ptr = mip->mi_fword; | |
| 1277 flen = mip->mi_fwordlen; /* available case-folded bytes */ | |
| 1278 byts = slang->sl_fbyts; | |
| 1279 idxs = slang->sl_fidxs; | |
| 339 | 1280 |
| 1281 if (mode == FIND_PREFIX) | |
| 1282 { | |
| 1283 /* Skip over the prefix. */ | |
| 1284 wlen = mip->mi_prefixlen; | |
| 1285 flen -= mip->mi_prefixlen; | |
| 1286 } | |
| 481 | 1287 else if (mode == FIND_COMPOUND) |
| 1288 { | |
| 1289 /* Skip over the previously found word(s). */ | |
| 1290 wlen = mip->mi_compoff; | |
| 1291 flen -= mip->mi_compoff; | |
| 1292 } | |
| 1293 | |
| 243 | 1294 } |
| 1295 | |
| 300 | 1296 if (byts == NULL) |
| 1297 return; /* array is empty */ | |
| 236 | 1298 |
| 1299 /* | |
| 307 | 1300 * Repeat advancing in the tree until: |
| 1301 * - there is a byte that doesn't match, | |
| 1302 * - we reach the end of the tree, | |
| 1303 * - or we reach the end of the line. | |
| 236 | 1304 */ |
| 300 | 1305 for (;;) |
| 236 | 1306 { |
| 346 | 1307 if (flen <= 0 && *mip->mi_fend != NUL) |
| 339 | 1308 flen = fold_more(mip); |
| 300 | 1309 |
| 1310 len = byts[arridx++]; | |
| 1311 | |
| 1312 /* If the first possible byte is a zero the word could end here. | |
| 1313 * Remember this index, we first check for the longest word. */ | |
| 1314 if (byts[arridx] == 0) | |
| 1315 { | |
| 307 | 1316 if (endidxcnt == MAXWLEN) |
| 1317 { | |
| 1318 /* Must be a corrupted spell file. */ | |
| 1319 EMSG(_(e_format)); | |
| 1320 return; | |
| 1321 } | |
| 300 | 1322 endlen[endidxcnt] = wlen; |
| 1323 endidx[endidxcnt++] = arridx++; | |
| 1324 --len; | |
| 1325 | |
| 1326 /* Skip over the zeros, there can be several flag/region | |
| 1327 * combinations. */ | |
| 1328 while (len > 0 && byts[arridx] == 0) | |
| 1329 { | |
| 1330 ++arridx; | |
| 1331 --len; | |
| 1332 } | |
| 1333 if (len == 0) | |
| 1334 break; /* no children, word must end here */ | |
| 1335 } | |
| 1336 | |
| 1337 /* Stop looking at end of the line. */ | |
| 1338 if (ptr[wlen] == NUL) | |
| 1339 break; | |
| 1340 | |
| 1341 /* Perform a binary search in the list of accepted bytes. */ | |
| 1342 c = ptr[wlen]; | |
| 346 | 1343 if (c == TAB) /* <Tab> is handled like <Space> */ |
| 1344 c = ' '; | |
| 300 | 1345 lo = arridx; |
| 1346 hi = arridx + len - 1; | |
| 1347 while (lo < hi) | |
| 1348 { | |
| 1349 m = (lo + hi) / 2; | |
| 1350 if (byts[m] > c) | |
| 1351 hi = m - 1; | |
| 1352 else if (byts[m] < c) | |
| 1353 lo = m + 1; | |
| 1354 else | |
| 1355 { | |
| 1356 lo = hi = m; | |
| 1357 break; | |
| 236 | 1358 } |
| 1359 } | |
| 300 | 1360 |
| 1361 /* Stop if there is no matching byte. */ | |
| 1362 if (hi < lo || byts[lo] != c) | |
| 1363 break; | |
| 1364 | |
| 1365 /* Continue at the child (if there is one). */ | |
| 1366 arridx = idxs[lo]; | |
| 1367 ++wlen; | |
| 1368 --flen; | |
| 346 | 1369 |
| 1370 /* One space in the good word may stand for several spaces in the | |
| 1371 * checked word. */ | |
| 1372 if (c == ' ') | |
| 1373 { | |
| 1374 for (;;) | |
| 1375 { | |
| 1376 if (flen <= 0 && *mip->mi_fend != NUL) | |
| 1377 flen = fold_more(mip); | |
| 1378 if (ptr[wlen] != ' ' && ptr[wlen] != TAB) | |
| 1379 break; | |
| 1380 ++wlen; | |
| 1381 --flen; | |
| 1382 } | |
| 1383 } | |
| 236 | 1384 } |
| 1385 | |
| 300 | 1386 /* |
| 1387 * Verify that one of the possible endings is valid. Try the longest | |
| 1388 * first. | |
| 1389 */ | |
| 1390 while (endidxcnt > 0) | |
| 1391 { | |
| 1392 --endidxcnt; | |
| 1393 arridx = endidx[endidxcnt]; | |
| 1394 wlen = endlen[endidxcnt]; | |
| 236 | 1395 |
| 300 | 1396 #ifdef FEAT_MBYTE |
| 1397 if ((*mb_head_off)(ptr, ptr + wlen) > 0) | |
| 1398 continue; /* not at first byte of character */ | |
| 1399 #endif | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1400 if (spell_iswordp(ptr + wlen, mip->mi_win)) |
| 481 | 1401 { |
| 492 | 1402 if (slang->sl_compprog == NULL && !slang->sl_nobreak) |
| 481 | 1403 continue; /* next char is a word character */ |
| 1404 word_ends = FALSE; | |
| 1405 } | |
| 1406 else | |
| 1407 word_ends = TRUE; | |
| 485 | 1408 /* The prefix flag is before compound flags. Once a valid prefix flag |
| 1409 * has been found we try compound flags. */ | |
| 1410 prefix_found = FALSE; | |
| 300 | 1411 |
| 1412 #ifdef FEAT_MBYTE | |
| 339 | 1413 if (mode != FIND_KEEPWORD && has_mbyte) |
| 300 | 1414 { |
| 1415 /* Compute byte length in original word, length may change | |
| 339 | 1416 * when folding case. This can be slow, take a shortcut when the |
| 1417 * case-folded word is equal to the keep-case word. */ | |
| 300 | 1418 p = mip->mi_word; |
| 339 | 1419 if (STRNCMP(ptr, p, wlen) != 0) |
| 1420 { | |
| 1421 for (s = ptr; s < ptr + wlen; mb_ptr_adv(s)) | |
| 1422 mb_ptr_adv(p); | |
| 835 | 1423 wlen = (int)(p - mip->mi_word); |
| 339 | 1424 } |
| 300 | 1425 } |
| 1426 #endif | |
| 236 | 1427 |
| 339 | 1428 /* Check flags and region. For FIND_PREFIX check the condition and |
| 1429 * prefix ID. | |
| 1430 * Repeat this if there are more flags/region alternatives until there | |
| 1431 * is a match. */ | |
| 1432 res = SP_BAD; | |
| 1433 for (len = byts[arridx - 1]; len > 0 && byts[arridx] == 0; | |
| 1434 --len, ++arridx) | |
| 300 | 1435 { |
| 1436 flags = idxs[arridx]; | |
| 324 | 1437 |
| 339 | 1438 /* For the fold-case tree check that the case of the checked word |
| 1439 * matches with what the word in the tree requires. | |
| 1440 * For keep-case tree the case is always right. For prefixes we | |
| 1441 * don't bother to check. */ | |
| 1442 if (mode == FIND_FOLDWORD) | |
| 300 | 1443 { |
| 1444 if (mip->mi_cend != mip->mi_word + wlen) | |
| 1445 { | |
| 323 | 1446 /* mi_capflags was set for a different word length, need |
| 1447 * to do it again. */ | |
| 300 | 1448 mip->mi_cend = mip->mi_word + wlen; |
| 323 | 1449 mip->mi_capflags = captype(mip->mi_word, mip->mi_cend); |
| 300 | 1450 } |
| 1451 | |
| 346 | 1452 if (mip->mi_capflags == WF_KEEPCAP |
| 1453 || !spell_valid_case(mip->mi_capflags, flags)) | |
| 339 | 1454 continue; |
| 300 | 1455 } |
| 236 | 1456 |
| 339 | 1457 /* When mode is FIND_PREFIX the word must support the prefix: |
| 1458 * check the prefix ID and the condition. Do that for the list at | |
| 366 | 1459 * mip->mi_prefarridx that find_prefix() filled. */ |
| 485 | 1460 else if (mode == FIND_PREFIX && !prefix_found) |
| 481 | 1461 { |
| 366 | 1462 c = valid_word_prefix(mip->mi_prefcnt, mip->mi_prefarridx, |
| 390 | 1463 flags, |
| 455 | 1464 mip->mi_word + mip->mi_cprefixlen, slang, |
| 1465 FALSE); | |
| 366 | 1466 if (c == 0) |
| 339 | 1467 continue; |
| 366 | 1468 |
| 1469 /* Use the WF_RARE flag for a rare prefix. */ | |
| 1470 if (c & WF_RAREPFX) | |
| 1471 flags |= WF_RARE; | |
| 485 | 1472 prefix_found = TRUE; |
| 339 | 1473 } |
| 1474 | |
| 492 | 1475 if (slang->sl_nobreak) |
| 1476 { | |
| 1477 if ((mode == FIND_COMPOUND || mode == FIND_KEEPCOMPOUND) | |
| 1478 && (flags & WF_BANNED) == 0) | |
| 1479 { | |
| 1480 /* NOBREAK: found a valid following word. That's all we | |
| 1481 * need to know, so return. */ | |
| 1482 mip->mi_result = SP_OK; | |
| 1483 break; | |
| 1484 } | |
| 1485 } | |
| 1486 | |
| 1487 else if ((mode == FIND_COMPOUND || mode == FIND_KEEPCOMPOUND | |
| 1488 || !word_ends)) | |
| 481 | 1489 { |
| 974 | 1490 /* If there is no compound flag or the word is shorter than |
| 483 | 1491 * COMPOUNDMIN reject it quickly. |
| 1492 * Makes you wonder why someone puts a compound flag on a word | |
| 481 | 1493 * that's too short... Myspell compatibility requires this |
| 1494 * anyway. */ | |
| 490 | 1495 if (((unsigned)flags >> 24) == 0 |
| 1496 || wlen - mip->mi_compoff < slang->sl_compminlen) | |
| 483 | 1497 continue; |
| 500 | 1498 #ifdef FEAT_MBYTE |
| 1499 /* For multi-byte chars check character length against | |
| 1500 * COMPOUNDMIN. */ | |
| 1501 if (has_mbyte | |
| 501 | 1502 && slang->sl_compminlen > 0 |
| 500 | 1503 && mb_charlen_len(mip->mi_word + mip->mi_compoff, |
| 1504 wlen - mip->mi_compoff) < slang->sl_compminlen) | |
| 1505 continue; | |
| 1506 #endif | |
| 483 | 1507 |
| 809 | 1508 /* Limit the number of compound words to COMPOUNDWORDMAX if no |
| 490 | 1509 * maximum for syllables is specified. */ |
| 809 | 1510 if (!word_ends && mip->mi_complen + mip->mi_compextra + 2 |
| 1511 > slang->sl_compmax | |
| 490 | 1512 && slang->sl_compsylmax == MAXWLEN) |
| 483 | 1513 continue; |
| 1514 | |
| 819 | 1515 /* Don't allow compounding on a side where an affix was added, |
| 1516 * unless COMPOUNDPERMITFLAG was used. */ | |
| 1517 if (mip->mi_complen > 0 && (flags & WF_NOCOMPBEF)) | |
| 1518 continue; | |
| 1519 if (!word_ends && (flags & WF_NOCOMPAFT)) | |
| 1520 continue; | |
| 1521 | |
| 485 | 1522 /* Quickly check if compounding is possible with this flag. */ |
| 495 | 1523 if (!byte_in_str(mip->mi_complen == 0 |
| 485 | 1524 ? slang->sl_compstartflags |
| 1525 : slang->sl_compallflags, | |
| 495 | 1526 ((unsigned)flags >> 24))) |
| 481 | 1527 continue; |
| 1528 | |
| 1762 | 1529 /* If there is a match with a CHECKCOMPOUNDPATTERN rule |
| 1530 * discard the compound word. */ | |
| 1531 if (match_checkcompoundpattern(ptr, wlen, &slang->sl_comppat)) | |
| 1532 continue; | |
| 1533 | |
| 490 | 1534 if (mode == FIND_COMPOUND) |
| 1535 { | |
| 1536 int capflags; | |
| 1537 | |
| 1538 /* Need to check the caps type of the appended compound | |
| 1539 * word. */ | |
| 1540 #ifdef FEAT_MBYTE | |
| 1541 if (has_mbyte && STRNCMP(ptr, mip->mi_word, | |
| 1542 mip->mi_compoff) != 0) | |
| 1543 { | |
| 1544 /* case folding may have changed the length */ | |
| 1545 p = mip->mi_word; | |
| 1546 for (s = ptr; s < ptr + mip->mi_compoff; mb_ptr_adv(s)) | |
| 1547 mb_ptr_adv(p); | |
| 1548 } | |
| 1549 else | |
| 1550 #endif | |
| 1551 p = mip->mi_word + mip->mi_compoff; | |
| 1552 capflags = captype(p, mip->mi_word + wlen); | |
| 1553 if (capflags == WF_KEEPCAP || (capflags == WF_ALLCAP | |
| 1554 && (flags & WF_FIXCAP) != 0)) | |
| 1555 continue; | |
| 1556 | |
| 1557 if (capflags != WF_ALLCAP) | |
| 1558 { | |
| 1559 /* When the character before the word is a word | |
| 1560 * character we do not accept a Onecap word. We do | |
| 1561 * accept a no-caps word, even when the dictionary | |
| 1562 * word specifies ONECAP. */ | |
| 1563 mb_ptr_back(mip->mi_word, p); | |
| 1564 if (spell_iswordp_nmw(p) | |
| 1565 ? capflags == WF_ONECAP | |
| 1566 : (flags & WF_ONECAP) != 0 | |
| 1567 && capflags != WF_ONECAP) | |
| 1568 continue; | |
| 1569 } | |
| 1570 } | |
| 1571 | |
| 483 | 1572 /* If the word ends the sequence of compound flags of the |
| 714 | 1573 * words must match with one of the COMPOUNDRULE items and |
| 483 | 1574 * the number of syllables must not be too large. */ |
| 1575 mip->mi_compflags[mip->mi_complen] = ((unsigned)flags >> 24); | |
| 1576 mip->mi_compflags[mip->mi_complen + 1] = NUL; | |
| 1577 if (word_ends) | |
| 1578 { | |
| 1579 char_u fword[MAXWLEN]; | |
| 1580 | |
| 1581 if (slang->sl_compsylmax < MAXWLEN) | |
| 1582 { | |
| 1583 /* "fword" is only needed for checking syllables. */ | |
| 1584 if (ptr == mip->mi_word) | |
| 1585 (void)spell_casefold(ptr, wlen, fword, MAXWLEN); | |
| 1586 else | |
| 1587 vim_strncpy(fword, ptr, endlen[endidxcnt]); | |
| 1588 } | |
| 1589 if (!can_compound(slang, fword, mip->mi_compflags)) | |
| 1590 continue; | |
| 1591 } | |
| 1762 | 1592 else if (slang->sl_comprules != NULL |
| 1593 && !match_compoundrule(slang, mip->mi_compflags)) | |
| 1594 /* The compound flags collected so far do not match any | |
| 1595 * COMPOUNDRULE, discard the compounded word. */ | |
| 1596 continue; | |
| 481 | 1597 } |
| 1598 | |
| 500 | 1599 /* Check NEEDCOMPOUND: can't use word without compounding. */ |
| 1600 else if (flags & WF_NEEDCOMP) | |
| 1601 continue; | |
| 1602 | |
| 492 | 1603 nobreak_result = SP_OK; |
| 1604 | |
| 481 | 1605 if (!word_ends) |
| 1606 { | |
| 492 | 1607 int save_result = mip->mi_result; |
| 1608 char_u *save_end = mip->mi_end; | |
| 501 | 1609 langp_T *save_lp = mip->mi_lp; |
| 1610 int lpi; | |
| 492 | 1611 |
| 1612 /* Check that a valid word follows. If there is one and we | |
| 1613 * are compounding, it will set "mi_result", thus we are | |
| 1614 * always finished here. For NOBREAK we only check that a | |
| 1615 * valid word follows. | |
| 481 | 1616 * Recursive! */ |
| 492 | 1617 if (slang->sl_nobreak) |
| 1618 mip->mi_result = SP_BAD; | |
| 481 | 1619 |
| 1620 /* Find following word in case-folded tree. */ | |
| 1621 mip->mi_compoff = endlen[endidxcnt]; | |
| 1622 #ifdef FEAT_MBYTE | |
| 1623 if (has_mbyte && mode == FIND_KEEPWORD) | |
| 1624 { | |
| 1625 /* Compute byte length in case-folded word from "wlen": | |
| 1626 * byte length in keep-case word. Length may change when | |
| 1627 * folding case. This can be slow, take a shortcut when | |
| 1628 * the case-folded word is equal to the keep-case word. */ | |
| 1629 p = mip->mi_fword; | |
| 1630 if (STRNCMP(ptr, p, wlen) != 0) | |
| 1631 { | |
| 1632 for (s = ptr; s < ptr + wlen; mb_ptr_adv(s)) | |
| 1633 mb_ptr_adv(p); | |
| 835 | 1634 mip->mi_compoff = (int)(p - mip->mi_fword); |
| 481 | 1635 } |
| 1636 } | |
| 1637 #endif | |
| 485 | 1638 c = mip->mi_compoff; |
| 483 | 1639 ++mip->mi_complen; |
| 809 | 1640 if (flags & WF_COMPROOT) |
| 1641 ++mip->mi_compextra; | |
| 501 | 1642 |
| 1643 /* For NOBREAK we need to try all NOBREAK languages, at least | |
| 1644 * to find the ".add" file(s). */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1645 for (lpi = 0; lpi < mip->mi_win->w_s->b_langp.ga_len; ++lpi) |
| 501 | 1646 { |
| 1647 if (slang->sl_nobreak) | |
| 1648 { | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
1649 mip->mi_lp = LANGP_ENTRY(mip->mi_win->w_s->b_langp, lpi); |
| 501 | 1650 if (mip->mi_lp->lp_slang->sl_fidxs == NULL |
| 1651 || !mip->mi_lp->lp_slang->sl_nobreak) | |
| 1652 continue; | |
| 1653 } | |
| 1654 | |
| 1655 find_word(mip, FIND_COMPOUND); | |
| 1656 | |
| 1657 /* When NOBREAK any word that matches is OK. Otherwise we | |
| 1658 * need to find the longest match, thus try with keep-case | |
| 1659 * and prefix too. */ | |
| 492 | 1660 if (!slang->sl_nobreak || mip->mi_result == SP_BAD) |
| 1661 { | |
| 501 | 1662 /* Find following word in keep-case tree. */ |
| 1663 mip->mi_compoff = wlen; | |
| 1664 find_word(mip, FIND_KEEPCOMPOUND); | |
| 1665 | |
| 819 | 1666 #if 0 /* Disabled, a prefix must not appear halfway a compound word, |
| 1667 unless the COMPOUNDPERMITFLAG is used and then it can't be a | |
| 1668 postponed prefix. */ | |
| 501 | 1669 if (!slang->sl_nobreak || mip->mi_result == SP_BAD) |
| 1670 { | |
| 1671 /* Check for following word with prefix. */ | |
| 1672 mip->mi_compoff = c; | |
| 1673 find_prefix(mip, FIND_COMPOUND); | |
| 1674 } | |
| 819 | 1675 #endif |
| 492 | 1676 } |
| 501 | 1677 |
| 1678 if (!slang->sl_nobreak) | |
| 1679 break; | |
| 492 | 1680 } |
| 483 | 1681 --mip->mi_complen; |
| 809 | 1682 if (flags & WF_COMPROOT) |
| 1683 --mip->mi_compextra; | |
| 501 | 1684 mip->mi_lp = save_lp; |
| 485 | 1685 |
| 492 | 1686 if (slang->sl_nobreak) |
| 1687 { | |
| 1688 nobreak_result = mip->mi_result; | |
| 1689 mip->mi_result = save_result; | |
| 1690 mip->mi_end = save_end; | |
| 1691 } | |
| 1692 else | |
| 1693 { | |
| 1694 if (mip->mi_result == SP_OK) | |
| 1695 break; | |
| 1696 continue; | |
| 1697 } | |
| 481 | 1698 } |
| 1699 | |
| 339 | 1700 if (flags & WF_BANNED) |
| 1701 res = SP_BANNED; | |
| 1702 else if (flags & WF_REGION) | |
| 1703 { | |
| 1704 /* Check region. */ | |
| 390 | 1705 if ((mip->mi_lp->lp_region & (flags >> 16)) != 0) |
| 300 | 1706 res = SP_OK; |
| 339 | 1707 else |
| 1708 res = SP_LOCAL; | |
| 300 | 1709 } |
| 339 | 1710 else if (flags & WF_RARE) |
| 1711 res = SP_RARE; | |
| 307 | 1712 else |
| 339 | 1713 res = SP_OK; |
| 1714 | |
| 492 | 1715 /* Always use the longest match and the best result. For NOBREAK |
| 1716 * we separately keep the longest match without a following good | |
| 1717 * word as a fall-back. */ | |
| 1718 if (nobreak_result == SP_BAD) | |
| 1719 { | |
| 1720 if (mip->mi_result2 > res) | |
| 1721 { | |
| 1722 mip->mi_result2 = res; | |
| 1723 mip->mi_end2 = mip->mi_word + wlen; | |
| 1724 } | |
| 1725 else if (mip->mi_result2 == res | |
| 1726 && mip->mi_end2 < mip->mi_word + wlen) | |
| 1727 mip->mi_end2 = mip->mi_word + wlen; | |
| 1728 } | |
| 1729 else if (mip->mi_result > res) | |
| 339 | 1730 { |
| 1731 mip->mi_result = res; | |
| 1732 mip->mi_end = mip->mi_word + wlen; | |
| 1733 } | |
| 351 | 1734 else if (mip->mi_result == res && mip->mi_end < mip->mi_word + wlen) |
| 339 | 1735 mip->mi_end = mip->mi_word + wlen; |
| 1736 | |
| 492 | 1737 if (mip->mi_result == SP_OK) |
| 339 | 1738 break; |
| 300 | 1739 } |
| 1740 | |
| 492 | 1741 if (mip->mi_result == SP_OK) |
| 300 | 1742 break; |
| 1743 } | |
| 236 | 1744 } |
| 1745 | |
| 323 | 1746 /* |
| 1762 | 1747 * Return TRUE if there is a match between the word ptr[wlen] and |
| 1748 * CHECKCOMPOUNDPATTERN rules, assuming that we will concatenate with another | |
| 1749 * word. | |
| 1750 * A match means that the first part of CHECKCOMPOUNDPATTERN matches at the | |
| 1751 * end of ptr[wlen] and the second part matches after it. | |
| 1752 */ | |
| 1753 static int | |
| 1754 match_checkcompoundpattern(ptr, wlen, gap) | |
| 1755 char_u *ptr; | |
| 1756 int wlen; | |
| 1757 garray_T *gap; /* &sl_comppat */ | |
| 1758 { | |
| 1759 int i; | |
| 1760 char_u *p; | |
| 1761 int len; | |
| 1762 | |
| 1763 for (i = 0; i + 1 < gap->ga_len; i += 2) | |
| 1764 { | |
| 1765 p = ((char_u **)gap->ga_data)[i + 1]; | |
| 1766 if (STRNCMP(ptr + wlen, p, STRLEN(p)) == 0) | |
| 1767 { | |
| 1768 /* Second part matches at start of following compound word, now | |
| 1769 * check if first part matches at end of previous word. */ | |
| 1770 p = ((char_u **)gap->ga_data)[i]; | |
| 1771 | 1771 len = (int)STRLEN(p); |
| 1762 | 1772 if (len <= wlen && STRNCMP(ptr + wlen - len, p, len) == 0) |
| 1773 return TRUE; | |
| 1774 } | |
| 1775 } | |
| 1776 return FALSE; | |
| 1777 } | |
| 1778 | |
| 1779 /* | |
| 626 | 1780 * Return TRUE if "flags" is a valid sequence of compound flags and "word" |
| 1781 * does not have too many syllables. | |
| 482 | 1782 */ |
| 1783 static int | |
| 483 | 1784 can_compound(slang, word, flags) |
| 482 | 1785 slang_T *slang; |
| 483 | 1786 char_u *word; |
| 1787 char_u *flags; | |
| 1788 { | |
| 1789 regmatch_T regmatch; | |
| 495 | 1790 #ifdef FEAT_MBYTE |
| 1791 char_u uflags[MAXWLEN * 2]; | |
| 1792 int i; | |
| 1793 #endif | |
| 1794 char_u *p; | |
| 483 | 1795 |
| 1796 if (slang->sl_compprog == NULL) | |
| 1797 return FALSE; | |
| 495 | 1798 #ifdef FEAT_MBYTE |
| 1799 if (enc_utf8) | |
| 1800 { | |
| 1801 /* Need to convert the single byte flags to utf8 characters. */ | |
| 1802 p = uflags; | |
| 1803 for (i = 0; flags[i] != NUL; ++i) | |
| 1804 p += mb_char2bytes(flags[i], p); | |
| 1805 *p = NUL; | |
| 1806 p = uflags; | |
| 1807 } | |
| 1808 else | |
| 1809 #endif | |
| 1810 p = flags; | |
| 483 | 1811 regmatch.regprog = slang->sl_compprog; |
| 1812 regmatch.rm_ic = FALSE; | |
| 495 | 1813 if (!vim_regexec(®match, p, 0)) |
| 483 | 1814 return FALSE; |
| 1815 | |
| 490 | 1816 /* Count the number of syllables. This may be slow, do it last. If there |
| 1817 * are too many syllables AND the number of compound words is above | |
| 809 | 1818 * COMPOUNDWORDMAX then compounding is not allowed. */ |
| 483 | 1819 if (slang->sl_compsylmax < MAXWLEN |
| 1820 && count_syllables(slang, word) > slang->sl_compsylmax) | |
| 495 | 1821 return (int)STRLEN(flags) < slang->sl_compmax; |
| 483 | 1822 return TRUE; |
| 482 | 1823 } |
| 1824 | |
| 1825 /* | |
| 1762 | 1826 * Return TRUE when the sequence of flags in "compflags" plus "flag" can |
| 1827 * possibly form a valid compounded word. This also checks the COMPOUNDRULE | |
| 1828 * lines if they don't contain wildcards. | |
| 1829 */ | |
| 1830 static int | |
| 1831 can_be_compound(sp, slang, compflags, flag) | |
| 1832 trystate_T *sp; | |
| 1833 slang_T *slang; | |
| 1834 char_u *compflags; | |
| 1835 int flag; | |
| 1836 { | |
| 1837 /* If the flag doesn't appear in sl_compstartflags or sl_compallflags | |
| 1838 * then it can't possibly compound. */ | |
| 1839 if (!byte_in_str(sp->ts_complen == sp->ts_compsplit | |
| 1840 ? slang->sl_compstartflags : slang->sl_compallflags, flag)) | |
| 1841 return FALSE; | |
| 1842 | |
| 1843 /* If there are no wildcards, we can check if the flags collected so far | |
| 1844 * possibly can form a match with COMPOUNDRULE patterns. This only | |
| 1845 * makes sense when we have two or more words. */ | |
| 1846 if (slang->sl_comprules != NULL && sp->ts_complen > sp->ts_compsplit) | |
| 1847 { | |
| 1848 int v; | |
| 1849 | |
| 1850 compflags[sp->ts_complen] = flag; | |
| 1851 compflags[sp->ts_complen + 1] = NUL; | |
| 1852 v = match_compoundrule(slang, compflags + sp->ts_compsplit); | |
| 1853 compflags[sp->ts_complen] = NUL; | |
| 1854 return v; | |
| 1855 } | |
| 1856 | |
| 1857 return TRUE; | |
| 1858 } | |
| 1859 | |
| 1860 | |
| 1861 /* | |
| 1862 * Return TRUE if the compound flags in compflags[] match the start of any | |
| 1863 * compound rule. This is used to stop trying a compound if the flags | |
| 1864 * collected so far can't possibly match any compound rule. | |
| 1865 * Caller must check that slang->sl_comprules is not NULL. | |
| 1866 */ | |
| 1867 static int | |
| 1868 match_compoundrule(slang, compflags) | |
| 1869 slang_T *slang; | |
| 1870 char_u *compflags; | |
| 1871 { | |
| 1872 char_u *p; | |
| 1873 int i; | |
| 1874 int c; | |
| 1875 | |
| 1876 /* loop over all the COMPOUNDRULE entries */ | |
| 1877 for (p = slang->sl_comprules; *p != NUL; ++p) | |
| 1878 { | |
| 1879 /* loop over the flags in the compound word we have made, match | |
| 1880 * them against the current rule entry */ | |
| 1881 for (i = 0; ; ++i) | |
| 1882 { | |
| 1883 c = compflags[i]; | |
| 1884 if (c == NUL) | |
| 1885 /* found a rule that matches for the flags we have so far */ | |
| 1886 return TRUE; | |
| 1887 if (*p == '/' || *p == NUL) | |
| 1888 break; /* end of rule, it's too short */ | |
| 1889 if (*p == '[') | |
| 1890 { | |
| 1891 int match = FALSE; | |
| 1892 | |
| 1893 /* compare against all the flags in [] */ | |
| 1894 ++p; | |
| 1895 while (*p != ']' && *p != NUL) | |
| 1896 if (*p++ == c) | |
| 1897 match = TRUE; | |
| 1898 if (!match) | |
| 1899 break; /* none matches */ | |
| 1900 } | |
| 1901 else if (*p != c) | |
| 1902 break; /* flag of word doesn't match flag in pattern */ | |
| 1903 ++p; | |
| 1904 } | |
| 1905 | |
| 1906 /* Skip to the next "/", where the next pattern starts. */ | |
| 1907 p = vim_strchr(p, '/'); | |
| 1908 if (p == NULL) | |
| 1909 break; | |
| 1910 } | |
| 1911 | |
| 1912 /* Checked all the rules and none of them match the flags, so there | |
| 1913 * can't possibly be a compound starting with these flags. */ | |
| 1914 return FALSE; | |
| 1915 } | |
| 1916 | |
| 1917 /* | |
| 390 | 1918 * Return non-zero if the prefix indicated by "arridx" matches with the prefix |
| 1919 * ID in "flags" for the word "word". | |
| 366 | 1920 * The WF_RAREPFX flag is included in the return value for a rare prefix. |
| 351 | 1921 */ |
| 1922 static int | |
| 455 | 1923 valid_word_prefix(totprefcnt, arridx, flags, word, slang, cond_req) |
| 351 | 1924 int totprefcnt; /* nr of prefix IDs */ |
| 1925 int arridx; /* idx in sl_pidxs[] */ | |
| 390 | 1926 int flags; |
| 351 | 1927 char_u *word; |
| 1928 slang_T *slang; | |
| 455 | 1929 int cond_req; /* only use prefixes with a condition */ |
| 351 | 1930 { |
| 1931 int prefcnt; | |
| 1932 int pidx; | |
| 1933 regprog_T *rp; | |
| 1934 regmatch_T regmatch; | |
| 390 | 1935 int prefid; |
| 1936 | |
| 1937 prefid = (unsigned)flags >> 24; | |
| 351 | 1938 for (prefcnt = totprefcnt - 1; prefcnt >= 0; --prefcnt) |
| 1939 { | |
| 1940 pidx = slang->sl_pidxs[arridx + prefcnt]; | |
| 1941 | |
| 1942 /* Check the prefix ID. */ | |
| 1943 if (prefid != (pidx & 0xff)) | |
| 1944 continue; | |
| 1945 | |
| 390 | 1946 /* Check if the prefix doesn't combine and the word already has a |
| 1947 * suffix. */ | |
| 1948 if ((flags & WF_HAS_AFF) && (pidx & WF_PFX_NC)) | |
| 1949 continue; | |
| 1950 | |
| 351 | 1951 /* Check the condition, if there is one. The condition index is |
| 366 | 1952 * stored in the two bytes above the prefix ID byte. */ |
| 1953 rp = slang->sl_prefprog[((unsigned)pidx >> 8) & 0xffff]; | |
| 351 | 1954 if (rp != NULL) |
| 1955 { | |
| 1956 regmatch.regprog = rp; | |
| 1957 regmatch.rm_ic = FALSE; | |
| 1958 if (!vim_regexec(®match, word, 0)) | |
| 1959 continue; | |
| 1960 } | |
| 455 | 1961 else if (cond_req) |
| 1962 continue; | |
| 1963 | |
| 1964 /* It's a match! Return the WF_ flags. */ | |
| 366 | 1965 return pidx; |
| 1966 } | |
| 1967 return 0; | |
| 351 | 1968 } |
| 1969 | |
| 1970 /* | |
| 339 | 1971 * Check if the word at "mip->mi_word" has a matching prefix. |
| 1972 * If it does, then check the following word. | |
| 1973 * | |
| 485 | 1974 * If "mode" is "FIND_COMPOUND" then do the same after another word, find a |
| 1975 * prefix in a compound word. | |
| 1976 * | |
| 339 | 1977 * For a match mip->mi_result is updated. |
| 1978 */ | |
| 1979 static void | |
| 485 | 1980 find_prefix(mip, mode) |
| 339 | 1981 matchinf_T *mip; |
| 485 | 1982 int mode; |
| 339 | 1983 { |
| 1984 idx_T arridx = 0; | |
| 1985 int len; | |
| 1986 int wlen = 0; | |
| 1987 int flen; | |
| 1988 int c; | |
| 1989 char_u *ptr; | |
| 1990 idx_T lo, hi, m; | |
| 1991 slang_T *slang = mip->mi_lp->lp_slang; | |
| 1992 char_u *byts; | |
| 1993 idx_T *idxs; | |
| 1994 | |
| 375 | 1995 byts = slang->sl_pbyts; |
| 1996 if (byts == NULL) | |
| 1997 return; /* array is empty */ | |
| 1998 | |
| 339 | 1999 /* We use the case-folded word here, since prefixes are always |
| 2000 * case-folded. */ | |
| 2001 ptr = mip->mi_fword; | |
| 2002 flen = mip->mi_fwordlen; /* available case-folded bytes */ | |
| 485 | 2003 if (mode == FIND_COMPOUND) |
| 2004 { | |
| 2005 /* Skip over the previously found word(s). */ | |
| 2006 ptr += mip->mi_compoff; | |
| 2007 flen -= mip->mi_compoff; | |
| 2008 } | |
| 339 | 2009 idxs = slang->sl_pidxs; |
| 2010 | |
| 2011 /* | |
| 2012 * Repeat advancing in the tree until: | |
| 2013 * - there is a byte that doesn't match, | |
| 2014 * - we reach the end of the tree, | |
| 2015 * - or we reach the end of the line. | |
| 2016 */ | |
| 2017 for (;;) | |
| 2018 { | |
| 2019 if (flen == 0 && *mip->mi_fend != NUL) | |
| 2020 flen = fold_more(mip); | |
| 2021 | |
| 2022 len = byts[arridx++]; | |
| 2023 | |
| 2024 /* If the first possible byte is a zero the prefix could end here. | |
| 2025 * Check if the following word matches and supports the prefix. */ | |
| 2026 if (byts[arridx] == 0) | |
| 2027 { | |
| 2028 /* There can be several prefixes with different conditions. We | |
| 2029 * try them all, since we don't know which one will give the | |
| 2030 * longest match. The word is the same each time, pass the list | |
| 2031 * of possible prefixes to find_word(). */ | |
| 2032 mip->mi_prefarridx = arridx; | |
| 2033 mip->mi_prefcnt = len; | |
| 2034 while (len > 0 && byts[arridx] == 0) | |
| 2035 { | |
| 2036 ++arridx; | |
| 2037 --len; | |
| 2038 } | |
| 2039 mip->mi_prefcnt -= len; | |
| 2040 | |
| 2041 /* Find the word that comes after the prefix. */ | |
| 2042 mip->mi_prefixlen = wlen; | |
| 485 | 2043 if (mode == FIND_COMPOUND) |
| 2044 /* Skip over the previously found word(s). */ | |
| 2045 mip->mi_prefixlen += mip->mi_compoff; | |
| 2046 | |
| 455 | 2047 #ifdef FEAT_MBYTE |
| 2048 if (has_mbyte) | |
| 2049 { | |
| 2050 /* Case-folded length may differ from original length. */ | |
| 485 | 2051 mip->mi_cprefixlen = nofold_len(mip->mi_fword, |
| 2052 mip->mi_prefixlen, mip->mi_word); | |
| 455 | 2053 } |
| 2054 else | |
| 485 | 2055 mip->mi_cprefixlen = mip->mi_prefixlen; |
| 455 | 2056 #endif |
| 339 | 2057 find_word(mip, FIND_PREFIX); |
| 2058 | |
| 2059 | |
| 2060 if (len == 0) | |
| 2061 break; /* no children, word must end here */ | |
| 2062 } | |
| 2063 | |
| 2064 /* Stop looking at end of the line. */ | |
| 2065 if (ptr[wlen] == NUL) | |
| 2066 break; | |
| 2067 | |
| 2068 /* Perform a binary search in the list of accepted bytes. */ | |
| 2069 c = ptr[wlen]; | |
| 2070 lo = arridx; | |
| 2071 hi = arridx + len - 1; | |
| 2072 while (lo < hi) | |
| 2073 { | |
| 2074 m = (lo + hi) / 2; | |
| 2075 if (byts[m] > c) | |
| 2076 hi = m - 1; | |
| 2077 else if (byts[m] < c) | |
| 2078 lo = m + 1; | |
| 2079 else | |
| 2080 { | |
| 2081 lo = hi = m; | |
| 2082 break; | |
| 2083 } | |
| 2084 } | |
| 2085 | |
| 2086 /* Stop if there is no matching byte. */ | |
| 2087 if (hi < lo || byts[lo] != c) | |
| 2088 break; | |
| 2089 | |
| 2090 /* Continue at the child (if there is one). */ | |
| 2091 arridx = idxs[lo]; | |
| 2092 ++wlen; | |
| 2093 --flen; | |
| 2094 } | |
| 2095 } | |
| 2096 | |
| 2097 /* | |
| 2098 * Need to fold at least one more character. Do until next non-word character | |
| 626 | 2099 * for efficiency. Include the non-word character too. |
| 339 | 2100 * Return the length of the folded chars in bytes. |
| 2101 */ | |
| 2102 static int | |
| 2103 fold_more(mip) | |
| 2104 matchinf_T *mip; | |
| 2105 { | |
| 2106 int flen; | |
| 2107 char_u *p; | |
| 2108 | |
| 2109 p = mip->mi_fend; | |
| 2110 do | |
| 2111 { | |
| 2112 mb_ptr_adv(mip->mi_fend); | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
2113 } while (*mip->mi_fend != NUL && spell_iswordp(mip->mi_fend, mip->mi_win)); |
| 339 | 2114 |
| 626 | 2115 /* Include the non-word character so that we can check for the word end. */ |
| 339 | 2116 if (*mip->mi_fend != NUL) |
| 2117 mb_ptr_adv(mip->mi_fend); | |
| 2118 | |
| 2119 (void)spell_casefold(p, (int)(mip->mi_fend - p), | |
| 2120 mip->mi_fword + mip->mi_fwordlen, | |
| 2121 MAXWLEN - mip->mi_fwordlen); | |
| 835 | 2122 flen = (int)STRLEN(mip->mi_fword + mip->mi_fwordlen); |
| 339 | 2123 mip->mi_fwordlen += flen; |
| 2124 return flen; | |
| 2125 } | |
| 2126 | |
| 2127 /* | |
| 323 | 2128 * Check case flags for a word. Return TRUE if the word has the requested |
| 2129 * case. | |
| 2130 */ | |
| 2131 static int | |
| 389 | 2132 spell_valid_case(wordflags, treeflags) |
| 2133 int wordflags; /* flags for the checked word. */ | |
| 323 | 2134 int treeflags; /* flags for the word in the spell tree */ |
| 2135 { | |
| 389 | 2136 return ((wordflags == WF_ALLCAP && (treeflags & WF_FIXCAP) == 0) |
| 323 | 2137 || ((treeflags & (WF_ALLCAP | WF_KEEPCAP)) == 0 |
| 474 | 2138 && ((treeflags & WF_ONECAP) == 0 |
| 2139 || (wordflags & WF_ONECAP) != 0))); | |
| 323 | 2140 } |
| 2141 | |
| 351 | 2142 /* |
| 2143 * Return TRUE if spell checking is not enabled. | |
| 2144 */ | |
| 2145 static int | |
| 498 | 2146 no_spell_checking(wp) |
| 2147 win_T *wp; | |
| 2148 { | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
2149 if (!wp->w_p_spell || *wp->w_s->b_p_spl == NUL |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
2150 || wp->w_s->b_langp.ga_len == 0) |
| 351 | 2151 { |
| 2152 EMSG(_("E756: Spell checking is not enabled")); | |
| 2153 return TRUE; | |
| 2154 } | |
| 2155 return FALSE; | |
| 2156 } | |
| 300 | 2157 |
| 236 | 2158 /* |
| 2159 * Move to next spell error. | |
| 500 | 2160 * "curline" is FALSE for "[s", "]s", "[S" and "]S". |
| 2161 * "curline" is TRUE to find word under/after cursor in the same line. | |
| 483 | 2162 * For Insert mode completion "dir" is BACKWARD and "curline" is TRUE: move |
| 2163 * to after badly spelled word before the cursor. | |
| 495 | 2164 * Return 0 if not found, length of the badly spelled word otherwise. |
| 236 | 2165 */ |
| 2166 int | |
| 498 | 2167 spell_move_to(wp, dir, allwords, curline, attrp) |
| 2168 win_T *wp; | |
| 236 | 2169 int dir; /* FORWARD or BACKWARD */ |
| 500 | 2170 int allwords; /* TRUE for "[s"/"]s", FALSE for "[S"/"]S" */ |
| 323 | 2171 int curline; |
| 534 | 2172 hlf_T *attrp; /* return: attributes of bad word or NULL |
| 2173 (only when "dir" is FORWARD) */ | |
| 236 | 2174 { |
| 249 | 2175 linenr_T lnum; |
| 2176 pos_T found_pos; | |
| 495 | 2177 int found_len = 0; |
| 236 | 2178 char_u *line; |
| 2179 char_u *p; | |
| 346 | 2180 char_u *endp; |
| 534 | 2181 hlf_T attr; |
| 236 | 2182 int len; |
| 737 | 2183 # ifdef FEAT_SYN_HL |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
2184 int has_syntax = syntax_present(wp); |
| 944 | 2185 # endif |
| 249 | 2186 int col; |
| 2187 int can_spell; | |
| 346 | 2188 char_u *buf = NULL; |
| 2189 int buflen = 0; | |
| 2190 int skip = 0; | |
| 385 | 2191 int capcol = -1; |
| 500 | 2192 int found_one = FALSE; |
| 2193 int wrapped = FALSE; | |
| 236 | 2194 |
| 498 | 2195 if (no_spell_checking(wp)) |
| 495 | 2196 return 0; |
| 236 | 2197 |
| 249 | 2198 /* |
| 2199 * Start looking for bad word at the start of the line, because we can't | |
| 817 | 2200 * start halfway a word, we don't know where it starts or ends. |
| 249 | 2201 * |
| 2202 * When searching backwards, we continue in the line to find the last | |
| 2203 * bad word (in the cursor line: before the cursor). | |
| 346 | 2204 * |
| 2205 * We concatenate the start of the next line, so that wrapped words work | |
| 2206 * (e.g. "et<line-break>cetera"). Doesn't work when searching backwards | |
| 2207 * though... | |
| 249 | 2208 */ |
| 498 | 2209 lnum = wp->w_cursor.lnum; |
| 698 | 2210 clearpos(&found_pos); |
| 236 | 2211 |
| 2212 while (!got_int) | |
| 2213 { | |
| 498 | 2214 line = ml_get_buf(wp->w_buffer, lnum, FALSE); |
| 346 | 2215 |
| 835 | 2216 len = (int)STRLEN(line); |
| 346 | 2217 if (buflen < len + MAXWLEN + 2) |
| 2218 { | |
| 2219 vim_free(buf); | |
| 2220 buflen = len + MAXWLEN + 2; | |
| 2221 buf = alloc(buflen); | |
| 2222 if (buf == NULL) | |
| 2223 break; | |
| 2224 } | |
| 2225 | |
| 385 | 2226 /* In first line check first word for Capital. */ |
| 2227 if (lnum == 1) | |
| 2228 capcol = 0; | |
| 2229 | |
| 2230 /* For checking first word with a capital skip white space. */ | |
| 2231 if (capcol == 0) | |
| 835 | 2232 capcol = (int)(skipwhite(line) - line); |
| 2233 else if (curline && wp == curwin) | |
| 2234 { | |
| 2235 /* For spellbadword(): check if first word needs a capital. */ | |
| 944 | 2236 col = (int)(skipwhite(line) - line); |
| 835 | 2237 if (check_need_cap(lnum, col)) |
| 2238 capcol = col; | |
| 2239 | |
| 2240 /* Need to get the line again, may have looked at the previous | |
| 2241 * one. */ | |
| 2242 line = ml_get_buf(wp->w_buffer, lnum, FALSE); | |
| 2243 } | |
| 385 | 2244 |
| 346 | 2245 /* Copy the line into "buf" and append the start of the next line if |
| 2246 * possible. */ | |
| 2247 STRCPY(buf, line); | |
| 498 | 2248 if (lnum < wp->w_buffer->b_ml.ml_line_count) |
| 883 | 2249 spell_cat_line(buf + STRLEN(buf), |
| 2250 ml_get_buf(wp->w_buffer, lnum + 1, FALSE), MAXWLEN); | |
| 346 | 2251 |
| 2252 p = buf + skip; | |
| 2253 endp = buf + len; | |
| 2254 while (p < endp) | |
| 236 | 2255 { |
| 500 | 2256 /* When searching backward don't search after the cursor. Unless |
| 2257 * we wrapped around the end of the buffer. */ | |
| 300 | 2258 if (dir == BACKWARD |
| 498 | 2259 && lnum == wp->w_cursor.lnum |
| 500 | 2260 && !wrapped |
| 498 | 2261 && (colnr_T)(p - buf) >= wp->w_cursor.col) |
| 300 | 2262 break; |
| 249 | 2263 |
| 300 | 2264 /* start of word */ |
| 534 | 2265 attr = HLF_COUNT; |
| 625 | 2266 len = spell_check(wp, p, &attr, &capcol, FALSE); |
| 249 | 2267 |
| 534 | 2268 if (attr != HLF_COUNT) |
| 300 | 2269 { |
| 2270 /* We found a bad word. Check the attribute. */ | |
| 534 | 2271 if (allwords || attr == HLF_SPB) |
| 236 | 2272 { |
| 300 | 2273 /* When searching forward only accept a bad word after |
| 2274 * the cursor. */ | |
| 2275 if (dir == BACKWARD | |
| 500 | 2276 || lnum != wp->w_cursor.lnum |
| 498 | 2277 || (lnum == wp->w_cursor.lnum |
| 500 | 2278 && (wrapped |
| 2279 || (colnr_T)(curline ? p - buf + len | |
| 346 | 2280 : p - buf) |
| 500 | 2281 > wp->w_cursor.col))) |
| 236 | 2282 { |
| 737 | 2283 # ifdef FEAT_SYN_HL |
| 300 | 2284 if (has_syntax) |
| 249 | 2285 { |
| 835 | 2286 col = (int)(p - buf); |
| 498 | 2287 (void)syn_get_id(wp, lnum, (colnr_T)col, |
| 1504 | 2288 FALSE, &can_spell, FALSE); |
| 857 | 2289 if (!can_spell) |
| 2290 attr = HLF_COUNT; | |
| 300 | 2291 } |
| 2292 else | |
| 737 | 2293 #endif |
| 300 | 2294 can_spell = TRUE; |
| 249 | 2295 |
| 300 | 2296 if (can_spell) |
| 2297 { | |
| 857 | 2298 found_one = TRUE; |
| 300 | 2299 found_pos.lnum = lnum; |
| 835 | 2300 found_pos.col = (int)(p - buf); |
| 249 | 2301 #ifdef FEAT_VIRTUALEDIT |
| 300 | 2302 found_pos.coladd = 0; |
| 249 | 2303 #endif |
| 300 | 2304 if (dir == FORWARD) |
| 2305 { | |
| 2306 /* No need to search further. */ | |
| 498 | 2307 wp->w_cursor = found_pos; |
| 346 | 2308 vim_free(buf); |
| 498 | 2309 if (attrp != NULL) |
| 2310 *attrp = attr; | |
| 495 | 2311 return len; |
| 249 | 2312 } |
| 483 | 2313 else if (curline) |
| 2314 /* Insert mode completion: put cursor after | |
| 2315 * the bad word. */ | |
| 2316 found_pos.col += len; | |
| 495 | 2317 found_len = len; |
| 249 | 2318 } |
| 236 | 2319 } |
| 857 | 2320 else |
| 2321 found_one = TRUE; | |
| 236 | 2322 } |
| 2323 } | |
| 2324 | |
| 300 | 2325 /* advance to character after the word */ |
| 2326 p += len; | |
| 385 | 2327 capcol -= len; |
| 236 | 2328 } |
| 2329 | |
| 483 | 2330 if (dir == BACKWARD && found_pos.lnum != 0) |
| 2331 { | |
| 500 | 2332 /* Use the last match in the line (before the cursor). */ |
| 498 | 2333 wp->w_cursor = found_pos; |
| 483 | 2334 vim_free(buf); |
| 495 | 2335 return found_len; |
| 483 | 2336 } |
| 2337 | |
| 323 | 2338 if (curline) |
| 346 | 2339 break; /* only check cursor line */ |
| 323 | 2340 |
| 236 | 2341 /* Advance to next line. */ |
| 249 | 2342 if (dir == BACKWARD) |
| 2343 { | |
| 500 | 2344 /* If we are back at the starting line and searched it again there |
| 2345 * is no match, give up. */ | |
| 2346 if (lnum == wp->w_cursor.lnum && wrapped) | |
| 346 | 2347 break; |
| 500 | 2348 |
| 2349 if (lnum > 1) | |
| 2350 --lnum; | |
| 2351 else if (!p_ws) | |
| 2352 break; /* at first line and 'nowrapscan' */ | |
| 2353 else | |
| 2354 { | |
| 2355 /* Wrap around to the end of the buffer. May search the | |
| 2356 * starting line again and accept the last match. */ | |
| 2357 lnum = wp->w_buffer->b_ml.ml_line_count; | |
| 2358 wrapped = TRUE; | |
| 503 | 2359 if (!shortmess(SHM_SEARCH)) |
| 2360 give_warning((char_u *)_(top_bot_msg), TRUE); | |
| 500 | 2361 } |
| 385 | 2362 capcol = -1; |
| 249 | 2363 } |
| 2364 else | |
| 2365 { | |
| 500 | 2366 if (lnum < wp->w_buffer->b_ml.ml_line_count) |
| 2367 ++lnum; | |
| 2368 else if (!p_ws) | |
| 2369 break; /* at first line and 'nowrapscan' */ | |
| 2370 else | |
| 2371 { | |
| 2372 /* Wrap around to the start of the buffer. May search the | |
| 2373 * starting line again and accept the first match. */ | |
| 2374 lnum = 1; | |
| 2375 wrapped = TRUE; | |
| 503 | 2376 if (!shortmess(SHM_SEARCH)) |
| 2377 give_warning((char_u *)_(bot_top_msg), TRUE); | |
| 500 | 2378 } |
| 2379 | |
| 2380 /* If we are back at the starting line and there is no match then | |
| 2381 * give up. */ | |
| 1808 | 2382 if (lnum == wp->w_cursor.lnum && (!found_one || wrapped)) |
| 346 | 2383 break; |
| 2384 | |
| 2385 /* Skip the characters at the start of the next line that were | |
| 2386 * included in a match crossing line boundaries. */ | |
| 534 | 2387 if (attr == HLF_COUNT) |
| 835 | 2388 skip = (int)(p - endp); |
| 346 | 2389 else |
| 2390 skip = 0; | |
| 385 | 2391 |
| 835 | 2392 /* Capcol skips over the inserted space. */ |
| 385 | 2393 --capcol; |
| 2394 | |
| 2395 /* But after empty line check first word in next line */ | |
| 2396 if (*skipwhite(line) == NUL) | |
| 2397 capcol = 0; | |
| 249 | 2398 } |
| 236 | 2399 |
| 2400 line_breakcheck(); | |
| 2401 } | |
| 2402 | |
| 346 | 2403 vim_free(buf); |
| 495 | 2404 return 0; |
| 346 | 2405 } |
| 2406 | |
| 2407 /* | |
| 2408 * For spell checking: concatenate the start of the following line "line" into | |
| 2409 * "buf", blanking-out special characters. Copy less then "maxlen" bytes. | |
| 1577 | 2410 * Keep the blanks at the start of the next line, this is used in win_line() |
| 2411 * to skip those bytes if the word was OK. | |
| 346 | 2412 */ |
| 2413 void | |
| 2414 spell_cat_line(buf, line, maxlen) | |
| 2415 char_u *buf; | |
| 2416 char_u *line; | |
| 2417 int maxlen; | |
| 2418 { | |
| 2419 char_u *p; | |
| 2420 int n; | |
| 2421 | |
| 2422 p = skipwhite(line); | |
| 2423 while (vim_strchr((char_u *)"*#/\"\t", *p) != NULL) | |
| 2424 p = skipwhite(p + 1); | |
| 2425 | |
| 2426 if (*p != NUL) | |
| 2427 { | |
| 1577 | 2428 /* Only worth concatenating if there is something else than spaces to |
| 2429 * concatenate. */ | |
| 2430 n = (int)(p - line) + 1; | |
| 2431 if (n < maxlen - 1) | |
| 2432 { | |
| 2433 vim_memset(buf, ' ', n); | |
| 2434 vim_strncpy(buf + n, p, maxlen - 1 - n); | |
| 2435 } | |
| 346 | 2436 } |
| 236 | 2437 } |
| 2438 | |
| 626 | 2439 /* |
| 2440 * Structure used for the cookie argument of do_in_runtimepath(). | |
| 2441 */ | |
| 501 | 2442 typedef struct spelload_S |
| 2443 { | |
| 2444 char_u sl_lang[MAXWLEN + 1]; /* language name */ | |
| 2445 slang_T *sl_slang; /* resulting slang_T struct */ | |
| 2446 int sl_nobreak; /* NOBREAK language found */ | |
| 2447 } spelload_T; | |
| 2448 | |
| 236 | 2449 /* |
| 307 | 2450 * Load word list(s) for "lang" from Vim spell file(s). |
| 310 | 2451 * "lang" must be the language without the region: e.g., "en". |
| 236 | 2452 */ |
| 307 | 2453 static void |
| 236 | 2454 spell_load_lang(lang) |
| 2455 char_u *lang; | |
| 2456 { | |
| 310 | 2457 char_u fname_enc[85]; |
| 236 | 2458 int r; |
| 501 | 2459 spelload_T sl; |
| 649 | 2460 #ifdef FEAT_AUTOCMD |
| 2461 int round; | |
| 2462 #endif | |
| 307 | 2463 |
| 310 | 2464 /* Copy the language name to pass it to spell_load_cb() as a cookie. |
| 307 | 2465 * It's truncated when an error is detected. */ |
| 501 | 2466 STRCPY(sl.sl_lang, lang); |
| 2467 sl.sl_slang = NULL; | |
| 2468 sl.sl_nobreak = FALSE; | |
| 307 | 2469 |
| 649 | 2470 #ifdef FEAT_AUTOCMD |
| 2471 /* We may retry when no spell file is found for the language, an | |
| 2472 * autocommand may load it then. */ | |
| 2473 for (round = 1; round <= 2; ++round) | |
| 2474 #endif | |
| 2475 { | |
| 2476 /* | |
| 2477 * Find the first spell file for "lang" in 'runtimepath' and load it. | |
| 2478 */ | |
| 2479 vim_snprintf((char *)fname_enc, sizeof(fname_enc) - 5, | |
| 2660 | 2480 #ifdef VMS |
| 2481 "spell/%s_%s.spl", | |
| 2482 #else | |
| 2483 "spell/%s.%s.spl", | |
| 2484 #endif | |
| 2485 lang, spell_enc()); | |
| 649 | 2486 r = do_in_runtimepath(fname_enc, FALSE, spell_load_cb, &sl); |
| 2487 | |
| 2488 if (r == FAIL && *sl.sl_lang != NUL) | |
| 2489 { | |
| 2490 /* Try loading the ASCII version. */ | |
| 2491 vim_snprintf((char *)fname_enc, sizeof(fname_enc) - 5, | |
| 2660 | 2492 #ifdef VMS |
| 2493 "spell/%s_ascii.spl", | |
| 2494 #else | |
| 2495 "spell/%s.ascii.spl", | |
| 2496 #endif | |
| 2497 lang); | |
| 649 | 2498 r = do_in_runtimepath(fname_enc, FALSE, spell_load_cb, &sl); |
| 2499 | |
| 2500 #ifdef FEAT_AUTOCMD | |
| 2501 if (r == FAIL && *sl.sl_lang != NUL && round == 1 | |
| 2502 && apply_autocmds(EVENT_SPELLFILEMISSING, lang, | |
| 2503 curbuf->b_fname, FALSE, curbuf)) | |
| 2504 continue; | |
| 2505 break; | |
| 2506 #endif | |
| 2507 } | |
| 714 | 2508 #ifdef FEAT_AUTOCMD |
| 2509 break; | |
| 2510 #endif | |
| 307 | 2511 } |
| 2512 | |
| 2513 if (r == FAIL) | |
| 649 | 2514 { |
| 2660 | 2515 smsg((char_u *) |
| 2516 #ifdef VMS | |
| 2517 _("Warning: Cannot find word list \"%s_%s.spl\" or \"%s_ascii.spl\""), | |
| 2518 #else | |
| 2519 _("Warning: Cannot find word list \"%s.%s.spl\" or \"%s.ascii.spl\""), | |
| 2520 #endif | |
| 483 | 2521 lang, spell_enc(), lang); |
| 649 | 2522 } |
| 501 | 2523 else if (sl.sl_slang != NULL) |
| 2524 { | |
| 626 | 2525 /* At least one file was loaded, now load ALL the additions. */ |
| 310 | 2526 STRCPY(fname_enc + STRLEN(fname_enc) - 3, "add.spl"); |
| 501 | 2527 do_in_runtimepath(fname_enc, TRUE, spell_load_cb, &sl); |
| 310 | 2528 } |
| 2529 } | |
| 2530 | |
| 2531 /* | |
| 2532 * Return the encoding used for spell checking: Use 'encoding', except that we | |
| 2533 * use "latin1" for "latin9". And limit to 60 characters (just in case). | |
| 2534 */ | |
| 2535 static char_u * | |
| 2536 spell_enc() | |
| 2537 { | |
| 2538 | |
| 2539 #ifdef FEAT_MBYTE | |
| 2540 if (STRLEN(p_enc) < 60 && STRCMP(p_enc, "iso-8859-15") != 0) | |
| 2541 return p_enc; | |
| 2542 #endif | |
| 2543 return (char_u *)"latin1"; | |
| 236 | 2544 } |
| 2545 | |
| 2546 /* | |
| 385 | 2547 * Get the name of the .spl file for the internal wordlist into |
| 2548 * "fname[MAXPATHL]". | |
| 2549 */ | |
| 2550 static void | |
| 2551 int_wordlist_spl(fname) | |
| 2552 char_u *fname; | |
| 2553 { | |
| 2660 | 2554 vim_snprintf((char *)fname, MAXPATHL, SPL_FNAME_TMPL, |
| 385 | 2555 int_wordlist, spell_enc()); |
| 2556 } | |
| 2557 | |
| 2558 /* | |
| 625 | 2559 * Allocate a new slang_T for language "lang". "lang" can be NULL. |
| 236 | 2560 * Caller must fill "sl_next". |
| 2561 */ | |
| 2562 static slang_T * | |
| 2563 slang_alloc(lang) | |
| 2564 char_u *lang; | |
| 2565 { | |
| 2566 slang_T *lp; | |
| 2567 | |
| 300 | 2568 lp = (slang_T *)alloc_clear(sizeof(slang_T)); |
| 236 | 2569 if (lp != NULL) |
| 2570 { | |
| 625 | 2571 if (lang != NULL) |
| 2572 lp->sl_name = vim_strsave(lang); | |
| 323 | 2573 ga_init2(&lp->sl_rep, sizeof(fromto_T), 10); |
| 625 | 2574 ga_init2(&lp->sl_repsal, sizeof(fromto_T), 10); |
| 483 | 2575 lp->sl_compmax = MAXWLEN; |
| 2576 lp->sl_compsylmax = MAXWLEN; | |
| 625 | 2577 hash_init(&lp->sl_wordcount); |
| 2578 } | |
| 2579 | |
| 236 | 2580 return lp; |
| 2581 } | |
| 2582 | |
| 2583 /* | |
| 2584 * Free the contents of an slang_T and the structure itself. | |
| 2585 */ | |
| 2586 static void | |
| 2587 slang_free(lp) | |
| 2588 slang_T *lp; | |
| 2589 { | |
| 2590 vim_free(lp->sl_name); | |
| 310 | 2591 vim_free(lp->sl_fname); |
| 2592 slang_clear(lp); | |
| 2593 vim_free(lp); | |
| 2594 } | |
| 2595 | |
| 2596 /* | |
| 2597 * Clear an slang_T so that the file can be reloaded. | |
| 2598 */ | |
| 2599 static void | |
| 2600 slang_clear(lp) | |
| 2601 slang_T *lp; | |
| 2602 { | |
| 339 | 2603 garray_T *gap; |
| 2604 fromto_T *ftp; | |
| 344 | 2605 salitem_T *smp; |
| 339 | 2606 int i; |
| 625 | 2607 int round; |
| 323 | 2608 |
| 300 | 2609 vim_free(lp->sl_fbyts); |
| 310 | 2610 lp->sl_fbyts = NULL; |
| 300 | 2611 vim_free(lp->sl_kbyts); |
| 310 | 2612 lp->sl_kbyts = NULL; |
| 339 | 2613 vim_free(lp->sl_pbyts); |
| 2614 lp->sl_pbyts = NULL; | |
| 2615 | |
| 300 | 2616 vim_free(lp->sl_fidxs); |
| 310 | 2617 lp->sl_fidxs = NULL; |
| 300 | 2618 vim_free(lp->sl_kidxs); |
| 310 | 2619 lp->sl_kidxs = NULL; |
| 339 | 2620 vim_free(lp->sl_pidxs); |
| 2621 lp->sl_pidxs = NULL; | |
| 323 | 2622 |
| 625 | 2623 for (round = 1; round <= 2; ++round) |
| 2624 { | |
| 2625 gap = round == 1 ? &lp->sl_rep : &lp->sl_repsal; | |
| 2626 while (gap->ga_len > 0) | |
| 2627 { | |
| 2628 ftp = &((fromto_T *)gap->ga_data)[--gap->ga_len]; | |
| 2629 vim_free(ftp->ft_from); | |
| 2630 vim_free(ftp->ft_to); | |
| 2631 } | |
| 2632 ga_clear(gap); | |
| 2633 } | |
| 344 | 2634 |
| 2635 gap = &lp->sl_sal; | |
| 375 | 2636 if (lp->sl_sofo) |
| 376 | 2637 { |
| 2638 /* "ga_len" is set to 1 without adding an item for latin1 */ | |
| 2639 if (gap->ga_data != NULL) | |
| 2640 /* SOFOFROM and SOFOTO items: free lists of wide characters. */ | |
| 2641 for (i = 0; i < gap->ga_len; ++i) | |
| 2642 vim_free(((int **)gap->ga_data)[i]); | |
| 2643 } | |
| 375 | 2644 else |
| 2645 /* SAL items: free salitem_T items */ | |
| 2646 while (gap->ga_len > 0) | |
| 2647 { | |
| 2648 smp = &((salitem_T *)gap->ga_data)[--gap->ga_len]; | |
| 2649 vim_free(smp->sm_lead); | |
| 2650 /* Don't free sm_oneof and sm_rules, they point into sm_lead. */ | |
| 2651 vim_free(smp->sm_to); | |
| 2652 #ifdef FEAT_MBYTE | |
| 2653 vim_free(smp->sm_lead_w); | |
| 2654 vim_free(smp->sm_oneof_w); | |
| 2655 vim_free(smp->sm_to_w); | |
| 2656 #endif | |
| 2657 } | |
| 344 | 2658 ga_clear(gap); |
| 323 | 2659 |
| 339 | 2660 for (i = 0; i < lp->sl_prefixcnt; ++i) |
| 2661 vim_free(lp->sl_prefprog[i]); | |
| 376 | 2662 lp->sl_prefixcnt = 0; |
| 339 | 2663 vim_free(lp->sl_prefprog); |
| 376 | 2664 lp->sl_prefprog = NULL; |
| 2665 | |
| 714 | 2666 vim_free(lp->sl_info); |
| 2667 lp->sl_info = NULL; | |
| 2668 | |
| 376 | 2669 vim_free(lp->sl_midword); |
| 2670 lp->sl_midword = NULL; | |
| 339 | 2671 |
| 483 | 2672 vim_free(lp->sl_compprog); |
| 1762 | 2673 vim_free(lp->sl_comprules); |
| 483 | 2674 vim_free(lp->sl_compstartflags); |
| 485 | 2675 vim_free(lp->sl_compallflags); |
| 483 | 2676 lp->sl_compprog = NULL; |
| 1762 | 2677 lp->sl_comprules = NULL; |
| 483 | 2678 lp->sl_compstartflags = NULL; |
| 485 | 2679 lp->sl_compallflags = NULL; |
| 483 | 2680 |
| 2681 vim_free(lp->sl_syllable); | |
| 2682 lp->sl_syllable = NULL; | |
| 2683 ga_clear(&lp->sl_syl_items); | |
| 481 | 2684 |
| 809 | 2685 ga_clear_strings(&lp->sl_comppat); |
| 2686 | |
| 625 | 2687 hash_clear_all(&lp->sl_wordcount, WC_KEY_OFF); |
| 2688 hash_init(&lp->sl_wordcount); | |
| 2689 | |
| 2690 #ifdef FEAT_MBYTE | |
| 2691 hash_clear_all(&lp->sl_map_hash, 0); | |
| 2692 #endif | |
| 2693 | |
| 2694 /* Clear info from .sug file. */ | |
| 2695 slang_clear_sug(lp); | |
| 483 | 2696 |
| 2697 lp->sl_compmax = MAXWLEN; | |
| 501 | 2698 lp->sl_compminlen = 0; |
| 483 | 2699 lp->sl_compsylmax = MAXWLEN; |
| 2700 lp->sl_regions[0] = NUL; | |
| 236 | 2701 } |
| 2702 | |
| 2703 /* | |
| 625 | 2704 * Clear the info from the .sug file in "lp". |
| 2705 */ | |
| 2706 static void | |
| 2707 slang_clear_sug(lp) | |
| 2708 slang_T *lp; | |
| 2709 { | |
| 2710 vim_free(lp->sl_sbyts); | |
| 2711 lp->sl_sbyts = NULL; | |
| 2712 vim_free(lp->sl_sidxs); | |
| 2713 lp->sl_sidxs = NULL; | |
| 2714 close_spellbuf(lp->sl_sugbuf); | |
| 2715 lp->sl_sugbuf = NULL; | |
| 2716 lp->sl_sugloaded = FALSE; | |
| 2717 lp->sl_sugtime = 0; | |
| 2718 } | |
| 2719 | |
| 2720 /* | |
| 307 | 2721 * Load one spell file and store the info into a slang_T. |
| 236 | 2722 * Invoked through do_in_runtimepath(). |
| 2723 */ | |
| 2724 static void | |
| 310 | 2725 spell_load_cb(fname, cookie) |
| 236 | 2726 char_u *fname; |
| 501 | 2727 void *cookie; |
| 2728 { | |
| 2729 spelload_T *slp = (spelload_T *)cookie; | |
| 2730 slang_T *slang; | |
| 2731 | |
| 2732 slang = spell_load_file(fname, slp->sl_lang, NULL, FALSE); | |
| 2733 if (slang != NULL) | |
| 2734 { | |
| 2735 /* When a previously loaded file has NOBREAK also use it for the | |
| 2736 * ".add" files. */ | |
| 2737 if (slp->sl_nobreak && slang->sl_add) | |
| 2738 slang->sl_nobreak = TRUE; | |
| 2739 else if (slang->sl_nobreak) | |
| 2740 slp->sl_nobreak = TRUE; | |
| 2741 | |
| 2742 slp->sl_slang = slang; | |
| 2743 } | |
| 310 | 2744 } |
| 2745 | |
| 2746 /* | |
| 2747 * Load one spell file and store the info into a slang_T. | |
| 2748 * | |
| 625 | 2749 * This is invoked in three ways: |
| 310 | 2750 * - From spell_load_cb() to load a spell file for the first time. "lang" is |
| 2751 * the language name, "old_lp" is NULL. Will allocate an slang_T. | |
| 2752 * - To reload a spell file that was changed. "lang" is NULL and "old_lp" | |
| 2753 * points to the existing slang_T. | |
| 625 | 2754 * - Just after writing a .spl file; it's read back to produce the .sug file. |
| 626 | 2755 * "old_lp" is NULL and "lang" is NULL. Will allocate an slang_T. |
| 2756 * | |
| 323 | 2757 * Returns the slang_T the spell file was loaded into. NULL for error. |
| 310 | 2758 */ |
| 323 | 2759 static slang_T * |
| 2760 spell_load_file(fname, lang, old_lp, silent) | |
| 310 | 2761 char_u *fname; |
| 2762 char_u *lang; | |
| 2763 slang_T *old_lp; | |
| 323 | 2764 int silent; /* no error if file doesn't exist */ |
| 310 | 2765 { |
| 236 | 2766 FILE *fd; |
| 483 | 2767 char_u buf[VIMSPELLMAGICL]; |
| 236 | 2768 char_u *p; |
| 2769 int i; | |
| 339 | 2770 int n; |
| 300 | 2771 int len; |
| 236 | 2772 char_u *save_sourcing_name = sourcing_name; |
| 2773 linenr_T save_sourcing_lnum = sourcing_lnum; | |
| 307 | 2774 slang_T *lp = NULL; |
| 344 | 2775 int c = 0; |
| 483 | 2776 int res; |
| 236 | 2777 |
| 310 | 2778 fd = mch_fopen((char *)fname, "r"); |
| 236 | 2779 if (fd == NULL) |
| 2780 { | |
| 323 | 2781 if (!silent) |
| 2782 EMSG2(_(e_notopen), fname); | |
| 2783 else if (p_verbose > 2) | |
| 2784 { | |
| 2785 verbose_enter(); | |
| 2786 smsg((char_u *)e_notopen, fname); | |
| 2787 verbose_leave(); | |
| 2788 } | |
| 255 | 2789 goto endFAIL; |
| 236 | 2790 } |
| 310 | 2791 if (p_verbose > 2) |
| 2792 { | |
| 2793 verbose_enter(); | |
| 2794 smsg((char_u *)_("Reading spell file \"%s\""), fname); | |
| 2795 verbose_leave(); | |
| 2796 } | |
| 2797 | |
| 2798 if (old_lp == NULL) | |
| 2799 { | |
| 2800 lp = slang_alloc(lang); | |
| 2801 if (lp == NULL) | |
| 2802 goto endFAIL; | |
| 2803 | |
| 2804 /* Remember the file name, used to reload the file when it's updated. */ | |
| 2805 lp->sl_fname = vim_strsave(fname); | |
| 2806 if (lp->sl_fname == NULL) | |
| 2807 goto endFAIL; | |
| 2808 | |
| 2660 | 2809 /* Check for .add.spl (_add.spl for VMS). */ |
| 2810 lp->sl_add = strstr((char *)gettail(fname), SPL_FNAME_ADD) != NULL; | |
| 310 | 2811 } |
| 2812 else | |
| 2813 lp = old_lp; | |
| 307 | 2814 |
| 236 | 2815 /* Set sourcing_name, so that error messages mention the file name. */ |
| 2816 sourcing_name = fname; | |
| 2817 sourcing_lnum = 0; | |
| 2818 | |
| 625 | 2819 /* |
| 2820 * <HEADER>: <fileID> | |
| 339 | 2821 */ |
| 236 | 2822 for (i = 0; i < VIMSPELLMAGICL; ++i) |
| 2823 buf[i] = getc(fd); /* <fileID> */ | |
| 2824 if (STRNCMP(buf, VIMSPELLMAGIC, VIMSPELLMAGICL) != 0) | |
| 2825 { | |
| 483 | 2826 EMSG(_("E757: This does not look like a spell file")); |
| 381 | 2827 goto endFAIL; |
| 483 | 2828 } |
| 2829 c = getc(fd); /* <versionnr> */ | |
| 2830 if (c < VIMSPELLVERSION) | |
| 2831 { | |
| 2832 EMSG(_("E771: Old spell file, needs to be updated")); | |
| 381 | 2833 goto endFAIL; |
| 2834 } | |
| 483 | 2835 else if (c > VIMSPELLVERSION) |
| 2836 { | |
| 2837 EMSG(_("E772: Spell file is for newer version of Vim")); | |
| 481 | 2838 goto endFAIL; |
| 483 | 2839 } |
| 2840 | |
| 2841 | |
| 2842 /* | |
| 2843 * <SECTIONS>: <section> ... <sectionend> | |
| 2844 * <section>: <sectionID> <sectionflags> <sectionlen> (section contents) | |
| 2845 */ | |
| 2846 for (;;) | |
| 2847 { | |
| 2848 n = getc(fd); /* <sectionID> or <sectionend> */ | |
| 2849 if (n == SN_END) | |
| 2850 break; | |
| 2851 c = getc(fd); /* <sectionflags> */ | |
| 695 | 2852 len = get4c(fd); /* <sectionlen> */ |
| 483 | 2853 if (len < 0) |
| 2854 goto truncerr; | |
| 2855 | |
| 2856 res = 0; | |
| 2857 switch (n) | |
| 2858 { | |
| 714 | 2859 case SN_INFO: |
| 2860 lp->sl_info = read_string(fd, len); /* <infotext> */ | |
| 2861 if (lp->sl_info == NULL) | |
| 2862 goto endFAIL; | |
| 2863 break; | |
| 2864 | |
| 483 | 2865 case SN_REGION: |
| 2866 res = read_region_section(fd, lp, len); | |
| 2867 break; | |
| 2868 | |
| 2869 case SN_CHARFLAGS: | |
| 2870 res = read_charflags_section(fd); | |
| 2871 break; | |
| 2872 | |
| 2873 case SN_MIDWORD: | |
| 2874 lp->sl_midword = read_string(fd, len); /* <midword> */ | |
| 2875 if (lp->sl_midword == NULL) | |
| 2876 goto endFAIL; | |
| 2877 break; | |
| 2878 | |
| 2879 case SN_PREFCOND: | |
| 2880 res = read_prefcond_section(fd, lp); | |
| 2881 break; | |
| 2882 | |
| 2883 case SN_REP: | |
| 625 | 2884 res = read_rep_section(fd, &lp->sl_rep, lp->sl_rep_first); |
| 2885 break; | |
| 2886 | |
| 2887 case SN_REPSAL: | |
| 2888 res = read_rep_section(fd, &lp->sl_repsal, lp->sl_repsal_first); | |
| 483 | 2889 break; |
| 2890 | |
| 2891 case SN_SAL: | |
| 2892 res = read_sal_section(fd, lp); | |
| 2893 break; | |
| 2894 | |
| 2895 case SN_SOFO: | |
| 2896 res = read_sofo_section(fd, lp); | |
| 2897 break; | |
| 2898 | |
| 2899 case SN_MAP: | |
| 2900 p = read_string(fd, len); /* <mapstr> */ | |
| 2901 if (p == NULL) | |
| 2902 goto endFAIL; | |
| 2903 set_map_str(lp, p); | |
| 2904 vim_free(p); | |
| 2905 break; | |
| 2906 | |
| 625 | 2907 case SN_WORDS: |
| 2908 res = read_words_section(fd, lp, len); | |
| 2909 break; | |
| 2910 | |
| 2911 case SN_SUGFILE: | |
|
2229
d45902a5c61c
Fix a few more things for persistent undo.
Bram Moolenaar <bram@vim.org>
parents:
2214
diff
changeset
|
2912 lp->sl_sugtime = get8ctime(fd); /* <timestamp> */ |
| 625 | 2913 break; |
| 2914 | |
| 698 | 2915 case SN_NOSPLITSUGS: |
| 2916 lp->sl_nosplitsugs = TRUE; /* <timestamp> */ | |
| 2917 break; | |
| 2918 | |
| 483 | 2919 case SN_COMPOUND: |
| 2920 res = read_compound(fd, lp, len); | |
| 2921 break; | |
| 2922 | |
| 492 | 2923 case SN_NOBREAK: |
| 2924 lp->sl_nobreak = TRUE; | |
| 2925 break; | |
| 2926 | |
| 483 | 2927 case SN_SYLLABLE: |
| 2928 lp->sl_syllable = read_string(fd, len); /* <syllable> */ | |
| 2929 if (lp->sl_syllable == NULL) | |
| 2930 goto endFAIL; | |
| 2931 if (init_syl_tab(lp) == FAIL) | |
| 2932 goto endFAIL; | |
| 2933 break; | |
| 2934 | |
| 2935 default: | |
| 2936 /* Unsupported section. When it's required give an error | |
| 2937 * message. When it's not required skip the contents. */ | |
| 2938 if (c & SNF_REQUIRED) | |
| 2939 { | |
| 2940 EMSG(_("E770: Unsupported section in spell file")); | |
| 2941 goto endFAIL; | |
| 2942 } | |
| 2943 while (--len >= 0) | |
| 2944 if (getc(fd) < 0) | |
| 2945 goto truncerr; | |
| 2946 break; | |
| 2947 } | |
| 625 | 2948 someerror: |
| 483 | 2949 if (res == SP_FORMERROR) |
| 2950 { | |
| 2951 EMSG(_(e_format)); | |
| 381 | 2952 goto endFAIL; |
| 344 | 2953 } |
| 483 | 2954 if (res == SP_TRUNCERROR) |
| 2955 { | |
| 2956 truncerr: | |
| 2957 EMSG(_(e_spell_trunc)); | |
| 375 | 2958 goto endFAIL; |
| 483 | 2959 } |
| 495 | 2960 if (res == SP_OTHERERROR) |
| 483 | 2961 goto endFAIL; |
| 389 | 2962 } |
| 236 | 2963 |
| 625 | 2964 /* <LWORDTREE> */ |
| 2965 res = spell_read_tree(fd, &lp->sl_fbyts, &lp->sl_fidxs, FALSE, 0); | |
| 2966 if (res != 0) | |
| 2967 goto someerror; | |
| 2968 | |
| 2969 /* <KWORDTREE> */ | |
| 2970 res = spell_read_tree(fd, &lp->sl_kbyts, &lp->sl_kidxs, FALSE, 0); | |
| 2971 if (res != 0) | |
| 2972 goto someerror; | |
| 2973 | |
| 2974 /* <PREFIXTREE> */ | |
| 2975 res = spell_read_tree(fd, &lp->sl_pbyts, &lp->sl_pidxs, TRUE, | |
| 2976 lp->sl_prefixcnt); | |
| 2977 if (res != 0) | |
| 2978 goto someerror; | |
| 243 | 2979 |
| 310 | 2980 /* For a new file link it in the list of spell files. */ |
| 626 | 2981 if (old_lp == NULL && lang != NULL) |
| 310 | 2982 { |
| 2983 lp->sl_next = first_lang; | |
| 2984 first_lang = lp; | |
| 2985 } | |
| 307 | 2986 |
| 255 | 2987 goto endOK; |
| 2988 | |
| 2989 endFAIL: | |
| 310 | 2990 if (lang != NULL) |
| 2991 /* truncating the name signals the error to spell_load_lang() */ | |
| 2992 *lang = NUL; | |
| 2993 if (lp != NULL && old_lp == NULL) | |
| 307 | 2994 slang_free(lp); |
| 500 | 2995 lp = NULL; |
| 255 | 2996 |
| 2997 endOK: | |
| 236 | 2998 if (fd != NULL) |
| 2999 fclose(fd); | |
| 3000 sourcing_name = save_sourcing_name; | |
| 3001 sourcing_lnum = save_sourcing_lnum; | |
| 323 | 3002 |
| 3003 return lp; | |
| 236 | 3004 } |
| 3005 | |
| 376 | 3006 /* |
| 3007 * Read a length field from "fd" in "cnt_bytes" bytes. | |
| 381 | 3008 * Allocate memory, read the string into it and add a NUL at the end. |
| 376 | 3009 * Returns NULL when the count is zero. |
| 483 | 3010 * Sets "*cntp" to SP_*ERROR when there is an error, length of the result |
| 3011 * otherwise. | |
| 376 | 3012 */ |
| 3013 static char_u * | |
| 389 | 3014 read_cnt_string(fd, cnt_bytes, cntp) |
| 376 | 3015 FILE *fd; |
| 3016 int cnt_bytes; | |
| 389 | 3017 int *cntp; |
| 376 | 3018 { |
| 3019 int cnt = 0; | |
| 3020 int i; | |
| 3021 char_u *str; | |
| 3022 | |
| 3023 /* read the length bytes, MSB first */ | |
| 3024 for (i = 0; i < cnt_bytes; ++i) | |
| 3025 cnt = (cnt << 8) + getc(fd); | |
| 3026 if (cnt < 0) | |
| 3027 { | |
| 483 | 3028 *cntp = SP_TRUNCERROR; |
| 376 | 3029 return NULL; |
| 3030 } | |
| 389 | 3031 *cntp = cnt; |
| 3032 if (cnt == 0) | |
| 3033 return NULL; /* nothing to read, return NULL */ | |
| 376 | 3034 |
| 483 | 3035 str = read_string(fd, cnt); |
| 3036 if (str == NULL) | |
| 495 | 3037 *cntp = SP_OTHERERROR; |
| 483 | 3038 return str; |
| 3039 } | |
| 3040 | |
| 3041 /* | |
| 3042 * Read SN_REGION: <regionname> ... | |
| 3043 * Return SP_*ERROR flags. | |
| 3044 */ | |
| 3045 static int | |
| 3046 read_region_section(fd, lp, len) | |
| 3047 FILE *fd; | |
| 3048 slang_T *lp; | |
| 3049 int len; | |
| 3050 { | |
| 3051 int i; | |
| 3052 | |
| 3053 if (len > 16) | |
| 3054 return SP_FORMERROR; | |
| 3055 for (i = 0; i < len; ++i) | |
| 3056 lp->sl_regions[i] = getc(fd); /* <regionname> */ | |
| 3057 lp->sl_regions[len] = NUL; | |
| 3058 return 0; | |
| 3059 } | |
| 3060 | |
| 3061 /* | |
| 3062 * Read SN_CHARFLAGS section: <charflagslen> <charflags> | |
| 3063 * <folcharslen> <folchars> | |
| 3064 * Return SP_*ERROR flags. | |
| 3065 */ | |
| 3066 static int | |
| 3067 read_charflags_section(fd) | |
| 3068 FILE *fd; | |
| 3069 { | |
| 3070 char_u *flags; | |
| 3071 char_u *fol; | |
| 3072 int flagslen, follen; | |
| 3073 | |
| 3074 /* <charflagslen> <charflags> */ | |
| 3075 flags = read_cnt_string(fd, 1, &flagslen); | |
| 3076 if (flagslen < 0) | |
| 3077 return flagslen; | |
| 3078 | |
| 3079 /* <folcharslen> <folchars> */ | |
| 3080 fol = read_cnt_string(fd, 2, &follen); | |
| 3081 if (follen < 0) | |
| 3082 { | |
| 3083 vim_free(flags); | |
| 3084 return follen; | |
| 3085 } | |
| 3086 | |
| 3087 /* Set the word-char flags and fill SPELL_ISUPPER() table. */ | |
| 3088 if (flags != NULL && fol != NULL) | |
| 3089 set_spell_charflags(flags, flagslen, fol); | |
| 3090 | |
| 3091 vim_free(flags); | |
| 3092 vim_free(fol); | |
| 3093 | |
| 3094 /* When <charflagslen> is zero then <fcharlen> must also be zero. */ | |
| 3095 if ((flags == NULL) != (fol == NULL)) | |
| 3096 return SP_FORMERROR; | |
| 3097 return 0; | |
| 3098 } | |
| 3099 | |
| 3100 /* | |
| 3101 * Read SN_PREFCOND section. | |
| 3102 * Return SP_*ERROR flags. | |
| 3103 */ | |
| 3104 static int | |
| 3105 read_prefcond_section(fd, lp) | |
| 3106 FILE *fd; | |
| 3107 slang_T *lp; | |
| 3108 { | |
| 3109 int cnt; | |
| 3110 int i; | |
| 3111 int n; | |
| 3112 char_u *p; | |
| 3113 char_u buf[MAXWLEN + 1]; | |
| 3114 | |
| 3115 /* <prefcondcnt> <prefcond> ... */ | |
| 695 | 3116 cnt = get2c(fd); /* <prefcondcnt> */ |
| 483 | 3117 if (cnt <= 0) |
| 3118 return SP_FORMERROR; | |
| 3119 | |
| 3120 lp->sl_prefprog = (regprog_T **)alloc_clear( | |
| 3121 (unsigned)sizeof(regprog_T *) * cnt); | |
| 3122 if (lp->sl_prefprog == NULL) | |
| 495 | 3123 return SP_OTHERERROR; |
| 483 | 3124 lp->sl_prefixcnt = cnt; |
| 3125 | |
| 376 | 3126 for (i = 0; i < cnt; ++i) |
| 483 | 3127 { |
| 3128 /* <prefcond> : <condlen> <condstr> */ | |
| 3129 n = getc(fd); /* <condlen> */ | |
| 3130 if (n < 0 || n >= MAXWLEN) | |
| 3131 return SP_FORMERROR; | |
| 3132 | |
| 3133 /* When <condlen> is zero we have an empty condition. Otherwise | |
| 3134 * compile the regexp program used to check for the condition. */ | |
| 3135 if (n > 0) | |
| 3136 { | |
| 3137 buf[0] = '^'; /* always match at one position only */ | |
| 3138 p = buf + 1; | |
| 3139 while (n-- > 0) | |
| 3140 *p++ = getc(fd); /* <condstr> */ | |
| 3141 *p = NUL; | |
| 3142 lp->sl_prefprog[i] = vim_regcomp(buf, RE_MAGIC + RE_STRING); | |
| 3143 } | |
| 3144 } | |
| 3145 return 0; | |
| 3146 } | |
| 3147 | |
| 3148 /* | |
| 625 | 3149 * Read REP or REPSAL items section from "fd": <repcount> <rep> ... |
| 483 | 3150 * Return SP_*ERROR flags. |
| 3151 */ | |
| 3152 static int | |
| 625 | 3153 read_rep_section(fd, gap, first) |
| 483 | 3154 FILE *fd; |
| 625 | 3155 garray_T *gap; |
| 3156 short *first; | |
| 483 | 3157 { |
| 3158 int cnt; | |
| 3159 fromto_T *ftp; | |
| 3160 int i; | |
| 3161 | |
| 695 | 3162 cnt = get2c(fd); /* <repcount> */ |
| 483 | 3163 if (cnt < 0) |
| 3164 return SP_TRUNCERROR; | |
| 3165 | |
| 3166 if (ga_grow(gap, cnt) == FAIL) | |
| 495 | 3167 return SP_OTHERERROR; |
| 483 | 3168 |
| 3169 /* <rep> : <repfromlen> <repfrom> <reptolen> <repto> */ | |
| 3170 for (; gap->ga_len < cnt; ++gap->ga_len) | |
| 3171 { | |
| 3172 ftp = &((fromto_T *)gap->ga_data)[gap->ga_len]; | |
| 3173 ftp->ft_from = read_cnt_string(fd, 1, &i); | |
| 3174 if (i < 0) | |
| 3175 return i; | |
| 3176 if (i == 0) | |
| 3177 return SP_FORMERROR; | |
| 3178 ftp->ft_to = read_cnt_string(fd, 1, &i); | |
| 3179 if (i <= 0) | |
| 3180 { | |
| 3181 vim_free(ftp->ft_from); | |
| 3182 if (i < 0) | |
| 3183 return i; | |
| 3184 return SP_FORMERROR; | |
| 3185 } | |
| 3186 } | |
| 3187 | |
| 3188 /* Fill the first-index table. */ | |
| 3189 for (i = 0; i < 256; ++i) | |
| 3190 first[i] = -1; | |
| 3191 for (i = 0; i < gap->ga_len; ++i) | |
| 3192 { | |
| 3193 ftp = &((fromto_T *)gap->ga_data)[i]; | |
| 3194 if (first[*ftp->ft_from] == -1) | |
| 3195 first[*ftp->ft_from] = i; | |
| 3196 } | |
| 3197 return 0; | |
| 3198 } | |
| 3199 | |
| 3200 /* | |
| 3201 * Read SN_SAL section: <salflags> <salcount> <sal> ... | |
| 3202 * Return SP_*ERROR flags. | |
| 3203 */ | |
| 3204 static int | |
| 3205 read_sal_section(fd, slang) | |
| 3206 FILE *fd; | |
| 3207 slang_T *slang; | |
| 3208 { | |
| 3209 int i; | |
| 3210 int cnt; | |
| 3211 garray_T *gap; | |
| 3212 salitem_T *smp; | |
| 3213 int ccnt; | |
| 3214 char_u *p; | |
| 485 | 3215 int c = NUL; |
| 483 | 3216 |
| 3217 slang->sl_sofo = FALSE; | |
| 3218 | |
| 3219 i = getc(fd); /* <salflags> */ | |
| 3220 if (i & SAL_F0LLOWUP) | |
| 3221 slang->sl_followup = TRUE; | |
| 3222 if (i & SAL_COLLAPSE) | |
| 3223 slang->sl_collapse = TRUE; | |
| 3224 if (i & SAL_REM_ACCENTS) | |
| 3225 slang->sl_rem_accents = TRUE; | |
| 3226 | |
| 695 | 3227 cnt = get2c(fd); /* <salcount> */ |
| 483 | 3228 if (cnt < 0) |
| 3229 return SP_TRUNCERROR; | |
| 3230 | |
| 3231 gap = &slang->sl_sal; | |
| 3232 ga_init2(gap, sizeof(salitem_T), 10); | |
| 548 | 3233 if (ga_grow(gap, cnt + 1) == FAIL) |
| 495 | 3234 return SP_OTHERERROR; |
| 483 | 3235 |
| 3236 /* <sal> : <salfromlen> <salfrom> <saltolen> <salto> */ | |
| 3237 for (; gap->ga_len < cnt; ++gap->ga_len) | |
| 3238 { | |
| 3239 smp = &((salitem_T *)gap->ga_data)[gap->ga_len]; | |
| 3240 ccnt = getc(fd); /* <salfromlen> */ | |
| 3241 if (ccnt < 0) | |
| 3242 return SP_TRUNCERROR; | |
| 3243 if ((p = alloc(ccnt + 2)) == NULL) | |
| 495 | 3244 return SP_OTHERERROR; |
| 483 | 3245 smp->sm_lead = p; |
| 3246 | |
| 3247 /* Read up to the first special char into sm_lead. */ | |
| 3248 for (i = 0; i < ccnt; ++i) | |
| 3249 { | |
| 3250 c = getc(fd); /* <salfrom> */ | |
| 3251 if (vim_strchr((char_u *)"0123456789(-<^$", c) != NULL) | |
| 3252 break; | |
| 3253 *p++ = c; | |
| 3254 } | |
| 835 | 3255 smp->sm_leadlen = (int)(p - smp->sm_lead); |
| 483 | 3256 *p++ = NUL; |
| 3257 | |
| 3258 /* Put (abc) chars in sm_oneof, if any. */ | |
| 3259 if (c == '(') | |
| 3260 { | |
| 3261 smp->sm_oneof = p; | |
| 3262 for (++i; i < ccnt; ++i) | |
| 3263 { | |
| 3264 c = getc(fd); /* <salfrom> */ | |
| 3265 if (c == ')') | |
| 3266 break; | |
| 3267 *p++ = c; | |
| 3268 } | |
| 3269 *p++ = NUL; | |
| 3270 if (++i < ccnt) | |
| 3271 c = getc(fd); | |
| 3272 } | |
| 3273 else | |
| 3274 smp->sm_oneof = NULL; | |
| 3275 | |
| 3276 /* Any following chars go in sm_rules. */ | |
| 3277 smp->sm_rules = p; | |
| 3278 if (i < ccnt) | |
| 3279 /* store the char we got while checking for end of sm_lead */ | |
| 3280 *p++ = c; | |
| 3281 for (++i; i < ccnt; ++i) | |
| 3282 *p++ = getc(fd); /* <salfrom> */ | |
| 3283 *p++ = NUL; | |
| 3284 | |
| 3285 /* <saltolen> <salto> */ | |
| 3286 smp->sm_to = read_cnt_string(fd, 1, &ccnt); | |
| 3287 if (ccnt < 0) | |
| 3288 { | |
| 3289 vim_free(smp->sm_lead); | |
| 3290 return ccnt; | |
| 3291 } | |
| 3292 | |
| 3293 #ifdef FEAT_MBYTE | |
| 3294 if (has_mbyte) | |
| 3295 { | |
| 3296 /* convert the multi-byte strings to wide char strings */ | |
| 3297 smp->sm_lead_w = mb_str2wide(smp->sm_lead); | |
| 3298 smp->sm_leadlen = mb_charlen(smp->sm_lead); | |
| 3299 if (smp->sm_oneof == NULL) | |
| 3300 smp->sm_oneof_w = NULL; | |
| 3301 else | |
| 3302 smp->sm_oneof_w = mb_str2wide(smp->sm_oneof); | |
| 3303 if (smp->sm_to == NULL) | |
| 3304 smp->sm_to_w = NULL; | |
| 3305 else | |
| 3306 smp->sm_to_w = mb_str2wide(smp->sm_to); | |
| 3307 if (smp->sm_lead_w == NULL | |
| 3308 || (smp->sm_oneof_w == NULL && smp->sm_oneof != NULL) | |
| 3309 || (smp->sm_to_w == NULL && smp->sm_to != NULL)) | |
| 3310 { | |
| 3311 vim_free(smp->sm_lead); | |
| 3312 vim_free(smp->sm_to); | |
| 3313 vim_free(smp->sm_lead_w); | |
| 3314 vim_free(smp->sm_oneof_w); | |
| 3315 vim_free(smp->sm_to_w); | |
| 495 | 3316 return SP_OTHERERROR; |
| 483 | 3317 } |
| 3318 } | |
| 3319 #endif | |
| 3320 } | |
| 3321 | |
| 548 | 3322 if (gap->ga_len > 0) |
| 3323 { | |
| 3324 /* Add one extra entry to mark the end with an empty sm_lead. Avoids | |
| 3325 * that we need to check the index every time. */ | |
| 3326 smp = &((salitem_T *)gap->ga_data)[gap->ga_len]; | |
| 3327 if ((p = alloc(1)) == NULL) | |
| 3328 return SP_OTHERERROR; | |
| 3329 p[0] = NUL; | |
| 3330 smp->sm_lead = p; | |
| 3331 smp->sm_leadlen = 0; | |
| 3332 smp->sm_oneof = NULL; | |
| 3333 smp->sm_rules = p; | |
| 3334 smp->sm_to = NULL; | |
| 3335 #ifdef FEAT_MBYTE | |
| 3336 if (has_mbyte) | |
| 3337 { | |
| 3338 smp->sm_lead_w = mb_str2wide(smp->sm_lead); | |
| 3339 smp->sm_leadlen = 0; | |
| 3340 smp->sm_oneof_w = NULL; | |
| 3341 smp->sm_to_w = NULL; | |
| 3342 } | |
| 3343 #endif | |
| 3344 ++gap->ga_len; | |
| 3345 } | |
| 3346 | |
| 483 | 3347 /* Fill the first-index table. */ |
| 3348 set_sal_first(slang); | |
| 3349 | |
| 3350 return 0; | |
| 3351 } | |
| 3352 | |
| 3353 /* | |
| 625 | 3354 * Read SN_WORDS: <word> ... |
| 3355 * Return SP_*ERROR flags. | |
| 3356 */ | |
| 3357 static int | |
| 3358 read_words_section(fd, lp, len) | |
| 3359 FILE *fd; | |
| 3360 slang_T *lp; | |
| 3361 int len; | |
| 3362 { | |
| 3363 int done = 0; | |
| 3364 int i; | |
| 1619 | 3365 int c; |
| 625 | 3366 char_u word[MAXWLEN]; |
| 3367 | |
| 3368 while (done < len) | |
| 3369 { | |
| 3370 /* Read one word at a time. */ | |
| 3371 for (i = 0; ; ++i) | |
| 3372 { | |
| 1619 | 3373 c = getc(fd); |
| 3374 if (c == EOF) | |
| 3375 return SP_TRUNCERROR; | |
| 3376 word[i] = c; | |
| 625 | 3377 if (word[i] == NUL) |
| 3378 break; | |
| 3379 if (i == MAXWLEN - 1) | |
| 3380 return SP_FORMERROR; | |
| 3381 } | |
| 3382 | |
| 3383 /* Init the count to 10. */ | |
| 3384 count_common_word(lp, word, -1, 10); | |
| 3385 done += i + 1; | |
| 3386 } | |
| 3387 return 0; | |
| 3388 } | |
| 3389 | |
| 3390 /* | |
| 3391 * Add a word to the hashtable of common words. | |
| 3392 * If it's already there then the counter is increased. | |
| 3393 */ | |
| 3394 static void | |
| 3395 count_common_word(lp, word, len, count) | |
| 3396 slang_T *lp; | |
| 3397 char_u *word; | |
| 3398 int len; /* word length, -1 for upto NUL */ | |
| 3399 int count; /* 1 to count once, 10 to init */ | |
| 3400 { | |
| 3401 hash_T hash; | |
| 3402 hashitem_T *hi; | |
| 3403 wordcount_T *wc; | |
| 3404 char_u buf[MAXWLEN]; | |
| 3405 char_u *p; | |
| 3406 | |
| 3407 if (len == -1) | |
| 3408 p = word; | |
| 3409 else | |
| 3410 { | |
| 3411 vim_strncpy(buf, word, len); | |
| 3412 p = buf; | |
| 3413 } | |
| 3414 | |
| 3415 hash = hash_hash(p); | |
| 3416 hi = hash_lookup(&lp->sl_wordcount, p, hash); | |
| 3417 if (HASHITEM_EMPTY(hi)) | |
| 3418 { | |
| 835 | 3419 wc = (wordcount_T *)alloc((unsigned)(sizeof(wordcount_T) + STRLEN(p))); |
| 625 | 3420 if (wc == NULL) |
| 3421 return; | |
| 3422 STRCPY(wc->wc_word, p); | |
| 3423 wc->wc_count = count; | |
| 3424 hash_add_item(&lp->sl_wordcount, hi, wc->wc_word, hash); | |
| 3425 } | |
| 3426 else | |
| 3427 { | |
| 3428 wc = HI2WC(hi); | |
| 3429 if ((wc->wc_count += count) < (unsigned)count) /* check for overflow */ | |
| 3430 wc->wc_count = MAXWORDCOUNT; | |
| 3431 } | |
| 3432 } | |
| 3433 | |
| 3434 /* | |
| 3435 * Adjust the score of common words. | |
| 3436 */ | |
| 3437 static int | |
| 3438 score_wordcount_adj(slang, score, word, split) | |
| 3439 slang_T *slang; | |
| 3440 int score; | |
| 3441 char_u *word; | |
| 3442 int split; /* word was split, less bonus */ | |
| 3443 { | |
| 3444 hashitem_T *hi; | |
| 3445 wordcount_T *wc; | |
| 3446 int bonus; | |
| 3447 int newscore; | |
| 3448 | |
| 3449 hi = hash_find(&slang->sl_wordcount, word); | |
| 3450 if (!HASHITEM_EMPTY(hi)) | |
| 3451 { | |
| 3452 wc = HI2WC(hi); | |
| 3453 if (wc->wc_count < SCORE_THRES2) | |
| 3454 bonus = SCORE_COMMON1; | |
| 3455 else if (wc->wc_count < SCORE_THRES3) | |
| 3456 bonus = SCORE_COMMON2; | |
| 3457 else | |
| 3458 bonus = SCORE_COMMON3; | |
| 3459 if (split) | |
| 3460 newscore = score - bonus / 2; | |
| 3461 else | |
| 3462 newscore = score - bonus; | |
| 3463 if (newscore < 0) | |
| 3464 return 0; | |
| 3465 return newscore; | |
| 3466 } | |
| 3467 return score; | |
| 3468 } | |
| 3469 | |
| 3470 /* | |
| 483 | 3471 * SN_SOFO: <sofofromlen> <sofofrom> <sofotolen> <sofoto> |
| 3472 * Return SP_*ERROR flags. | |
| 3473 */ | |
| 3474 static int | |
| 3475 read_sofo_section(fd, slang) | |
| 3476 FILE *fd; | |
| 3477 slang_T *slang; | |
| 3478 { | |
| 3479 int cnt; | |
| 3480 char_u *from, *to; | |
| 3481 int res; | |
| 3482 | |
| 3483 slang->sl_sofo = TRUE; | |
| 3484 | |
| 3485 /* <sofofromlen> <sofofrom> */ | |
| 3486 from = read_cnt_string(fd, 2, &cnt); | |
| 3487 if (cnt < 0) | |
| 3488 return cnt; | |
| 3489 | |
| 3490 /* <sofotolen> <sofoto> */ | |
| 3491 to = read_cnt_string(fd, 2, &cnt); | |
| 3492 if (cnt < 0) | |
| 3493 { | |
| 3494 vim_free(from); | |
| 3495 return cnt; | |
| 3496 } | |
| 3497 | |
| 3498 /* Store the info in slang->sl_sal and/or slang->sl_sal_first. */ | |
| 3499 if (from != NULL && to != NULL) | |
| 3500 res = set_sofo(slang, from, to); | |
| 3501 else if (from != NULL || to != NULL) | |
| 3502 res = SP_FORMERROR; /* only one of two strings is an error */ | |
| 3503 else | |
| 3504 res = 0; | |
| 3505 | |
| 3506 vim_free(from); | |
| 3507 vim_free(to); | |
| 3508 return res; | |
| 3509 } | |
| 3510 | |
| 3511 /* | |
| 3512 * Read the compound section from the .spl file: | |
| 809 | 3513 * <compmax> <compminlen> <compsylmax> <compoptions> <compflags> |
| 483 | 3514 * Returns SP_*ERROR flags. |
| 3515 */ | |
| 3516 static int | |
| 3517 read_compound(fd, slang, len) | |
| 3518 FILE *fd; | |
| 3519 slang_T *slang; | |
| 3520 int len; | |
| 3521 { | |
| 3522 int todo = len; | |
| 3523 int c; | |
| 3524 int atstart; | |
| 3525 char_u *pat; | |
| 3526 char_u *pp; | |
| 3527 char_u *cp; | |
| 485 | 3528 char_u *ap; |
| 1762 | 3529 char_u *crp; |
| 809 | 3530 int cnt; |
| 3531 garray_T *gap; | |
| 483 | 3532 |
| 3533 if (todo < 2) | |
| 3534 return SP_FORMERROR; /* need at least two bytes */ | |
| 3535 | |
| 3536 --todo; | |
| 3537 c = getc(fd); /* <compmax> */ | |
| 3538 if (c < 2) | |
| 3539 c = MAXWLEN; | |
| 3540 slang->sl_compmax = c; | |
| 3541 | |
| 3542 --todo; | |
| 3543 c = getc(fd); /* <compminlen> */ | |
| 3544 if (c < 1) | |
| 501 | 3545 c = 0; |
| 483 | 3546 slang->sl_compminlen = c; |
| 3547 | |
| 3548 --todo; | |
| 3549 c = getc(fd); /* <compsylmax> */ | |
| 3550 if (c < 1) | |
| 3551 c = MAXWLEN; | |
| 3552 slang->sl_compsylmax = c; | |
| 3553 | |
| 809 | 3554 c = getc(fd); /* <compoptions> */ |
| 3555 if (c != 0) | |
| 3556 ungetc(c, fd); /* be backwards compatible with Vim 7.0b */ | |
| 3557 else | |
| 3558 { | |
| 3559 --todo; | |
| 3560 c = getc(fd); /* only use the lower byte for now */ | |
| 3561 --todo; | |
| 3562 slang->sl_compoptions = c; | |
| 3563 | |
| 3564 gap = &slang->sl_comppat; | |
| 3565 c = get2c(fd); /* <comppatcount> */ | |
| 3566 todo -= 2; | |
| 3567 ga_init2(gap, sizeof(char_u *), c); | |
| 3568 if (ga_grow(gap, c) == OK) | |
| 3569 while (--c >= 0) | |
| 3570 { | |
| 3571 ((char_u **)(gap->ga_data))[gap->ga_len++] = | |
| 3572 read_cnt_string(fd, 1, &cnt); | |
| 3573 /* <comppatlen> <comppattext> */ | |
| 3574 if (cnt < 0) | |
| 3575 return cnt; | |
| 822 | 3576 todo -= cnt + 1; |
| 3577 } | |
| 3578 } | |
| 3579 if (todo < 0) | |
| 3580 return SP_FORMERROR; | |
| 809 | 3581 |
| 714 | 3582 /* Turn the COMPOUNDRULE items into a regexp pattern: |
| 483 | 3583 * "a[bc]/a*b+" -> "^\(a[bc]\|a*b\+\)$". |
| 495 | 3584 * Inserting backslashes may double the length, "^\(\)$<Nul>" is 7 bytes. |
| 3585 * Conversion to utf-8 may double the size. */ | |
| 3586 c = todo * 2 + 7; | |
| 3587 #ifdef FEAT_MBYTE | |
| 3588 if (enc_utf8) | |
| 3589 c += todo * 2; | |
| 3590 #endif | |
| 3591 pat = alloc((unsigned)c); | |
| 483 | 3592 if (pat == NULL) |
| 495 | 3593 return SP_OTHERERROR; |
| 483 | 3594 |
| 485 | 3595 /* We also need a list of all flags that can appear at the start and one |
| 3596 * for all flags. */ | |
| 483 | 3597 cp = alloc(todo + 1); |
| 3598 if (cp == NULL) | |
| 3599 { | |
| 3600 vim_free(pat); | |
| 495 | 3601 return SP_OTHERERROR; |
| 483 | 3602 } |
| 3603 slang->sl_compstartflags = cp; | |
| 3604 *cp = NUL; | |
| 3605 | |
| 485 | 3606 ap = alloc(todo + 1); |
| 3607 if (ap == NULL) | |
| 3608 { | |
| 3609 vim_free(pat); | |
| 495 | 3610 return SP_OTHERERROR; |
| 485 | 3611 } |
| 3612 slang->sl_compallflags = ap; | |
| 3613 *ap = NUL; | |
| 3614 | |
| 1762 | 3615 /* And a list of all patterns in their original form, for checking whether |
| 3616 * compounding may work in match_compoundrule(). This is freed when we | |
| 3617 * encounter a wildcard, the check doesn't work then. */ | |
| 3618 crp = alloc(todo + 1); | |
| 3619 slang->sl_comprules = crp; | |
| 3620 | |
| 483 | 3621 pp = pat; |
| 3622 *pp++ = '^'; | |
| 3623 *pp++ = '\\'; | |
| 3624 *pp++ = '('; | |
| 3625 | |
| 3626 atstart = 1; | |
| 3627 while (todo-- > 0) | |
| 3628 { | |
| 3629 c = getc(fd); /* <compflags> */ | |
| 1619 | 3630 if (c == EOF) |
| 3631 { | |
| 3632 vim_free(pat); | |
| 3633 return SP_TRUNCERROR; | |
| 3634 } | |
| 485 | 3635 |
| 3636 /* Add all flags to "sl_compallflags". */ | |
| 2691 | 3637 if (vim_strchr((char_u *)"?*+[]/", c) == NULL |
| 495 | 3638 && !byte_in_str(slang->sl_compallflags, c)) |
| 485 | 3639 { |
| 3640 *ap++ = c; | |
| 3641 *ap = NUL; | |
| 3642 } | |
| 3643 | |
| 483 | 3644 if (atstart != 0) |
| 3645 { | |
| 3646 /* At start of item: copy flags to "sl_compstartflags". For a | |
| 3647 * [abc] item set "atstart" to 2 and copy up to the ']'. */ | |
| 3648 if (c == '[') | |
| 3649 atstart = 2; | |
| 3650 else if (c == ']') | |
| 3651 atstart = 0; | |
| 3652 else | |
| 3653 { | |
| 495 | 3654 if (!byte_in_str(slang->sl_compstartflags, c)) |
| 483 | 3655 { |
| 3656 *cp++ = c; | |
| 3657 *cp = NUL; | |
| 3658 } | |
| 3659 if (atstart == 1) | |
| 3660 atstart = 0; | |
| 3661 } | |
| 3662 } | |
| 1762 | 3663 |
| 3664 /* Copy flag to "sl_comprules", unless we run into a wildcard. */ | |
| 3665 if (crp != NULL) | |
| 3666 { | |
| 2691 | 3667 if (c == '?' || c == '+' || c == '*') |
| 1762 | 3668 { |
| 3669 vim_free(slang->sl_comprules); | |
| 3670 slang->sl_comprules = NULL; | |
| 3671 crp = NULL; | |
| 3672 } | |
| 3673 else | |
| 3674 *crp++ = c; | |
| 3675 } | |
| 3676 | |
| 483 | 3677 if (c == '/') /* slash separates two items */ |
| 3678 { | |
| 3679 *pp++ = '\\'; | |
| 3680 *pp++ = '|'; | |
| 3681 atstart = 1; | |
| 3682 } | |
| 3683 else /* normal char, "[abc]" and '*' are copied as-is */ | |
| 3684 { | |
| 2691 | 3685 if (c == '?' || c == '+' || c == '~') |
| 3686 *pp++ = '\\'; /* "a?" becomes "a\?", "a+" becomes "a\+" */ | |
| 495 | 3687 #ifdef FEAT_MBYTE |
| 3688 if (enc_utf8) | |
| 3689 pp += mb_char2bytes(c, pp); | |
| 3690 else | |
| 3691 #endif | |
| 3692 *pp++ = c; | |
| 483 | 3693 } |
| 3694 } | |
| 3695 | |
| 3696 *pp++ = '\\'; | |
| 3697 *pp++ = ')'; | |
| 3698 *pp++ = '$'; | |
| 3699 *pp = NUL; | |
| 3700 | |
| 1762 | 3701 if (crp != NULL) |
| 3702 *crp = NUL; | |
| 3703 | |
| 483 | 3704 slang->sl_compprog = vim_regcomp(pat, RE_MAGIC + RE_STRING + RE_STRICT); |
| 3705 vim_free(pat); | |
| 3706 if (slang->sl_compprog == NULL) | |
| 3707 return SP_FORMERROR; | |
| 3708 | |
| 3709 return 0; | |
| 3710 } | |
| 3711 | |
| 495 | 3712 /* |
| 498 | 3713 * Return TRUE if byte "n" appears in "str". |
| 495 | 3714 * Like strchr() but independent of locale. |
| 3715 */ | |
| 3716 static int | |
| 498 | 3717 byte_in_str(str, n) |
| 495 | 3718 char_u *str; |
| 498 | 3719 int n; |
| 495 | 3720 { |
| 3721 char_u *p; | |
| 3722 | |
| 3723 for (p = str; *p != NUL; ++p) | |
| 498 | 3724 if (*p == n) |
| 495 | 3725 return TRUE; |
| 3726 return FALSE; | |
| 3727 } | |
| 3728 | |
| 483 | 3729 #define SY_MAXLEN 30 |
| 3730 typedef struct syl_item_S | |
| 3731 { | |
| 3732 char_u sy_chars[SY_MAXLEN]; /* the sequence of chars */ | |
| 3733 int sy_len; | |
| 3734 } syl_item_T; | |
| 3735 | |
| 3736 /* | |
| 3737 * Truncate "slang->sl_syllable" at the first slash and put the following items | |
| 3738 * in "slang->sl_syl_items". | |
| 3739 */ | |
| 3740 static int | |
| 3741 init_syl_tab(slang) | |
| 3742 slang_T *slang; | |
| 3743 { | |
| 3744 char_u *p; | |
| 3745 char_u *s; | |
| 3746 int l; | |
| 3747 syl_item_T *syl; | |
| 3748 | |
| 3749 ga_init2(&slang->sl_syl_items, sizeof(syl_item_T), 4); | |
| 3750 p = vim_strchr(slang->sl_syllable, '/'); | |
| 3751 while (p != NULL) | |
| 3752 { | |
| 3753 *p++ = NUL; | |
| 495 | 3754 if (*p == NUL) /* trailing slash */ |
| 483 | 3755 break; |
| 3756 s = p; | |
| 3757 p = vim_strchr(p, '/'); | |
| 3758 if (p == NULL) | |
| 835 | 3759 l = (int)STRLEN(s); |
| 483 | 3760 else |
| 835 | 3761 l = (int)(p - s); |
| 483 | 3762 if (l >= SY_MAXLEN) |
| 3763 return SP_FORMERROR; | |
| 3764 if (ga_grow(&slang->sl_syl_items, 1) == FAIL) | |
| 495 | 3765 return SP_OTHERERROR; |
| 483 | 3766 syl = ((syl_item_T *)slang->sl_syl_items.ga_data) |
| 3767 + slang->sl_syl_items.ga_len++; | |
| 3768 vim_strncpy(syl->sy_chars, s, l); | |
| 3769 syl->sy_len = l; | |
| 3770 } | |
| 3771 return OK; | |
| 3772 } | |
| 3773 | |
| 3774 /* | |
| 3775 * Count the number of syllables in "word". | |
| 3776 * When "word" contains spaces the syllables after the last space are counted. | |
| 3777 * Returns zero if syllables are not defines. | |
| 3778 */ | |
| 3779 static int | |
| 3780 count_syllables(slang, word) | |
| 3781 slang_T *slang; | |
| 3782 char_u *word; | |
| 3783 { | |
| 3784 int cnt = 0; | |
| 3785 int skip = FALSE; | |
| 3786 char_u *p; | |
| 3787 int len; | |
| 3788 int i; | |
| 3789 syl_item_T *syl; | |
| 3790 int c; | |
| 3791 | |
| 3792 if (slang->sl_syllable == NULL) | |
| 3793 return 0; | |
| 3794 | |
| 3795 for (p = word; *p != NUL; p += len) | |
| 3796 { | |
| 3797 /* When running into a space reset counter. */ | |
| 3798 if (*p == ' ') | |
| 3799 { | |
| 3800 len = 1; | |
| 3801 cnt = 0; | |
| 3802 continue; | |
| 3803 } | |
| 3804 | |
| 3805 /* Find longest match of syllable items. */ | |
| 3806 len = 0; | |
| 3807 for (i = 0; i < slang->sl_syl_items.ga_len; ++i) | |
| 3808 { | |
| 3809 syl = ((syl_item_T *)slang->sl_syl_items.ga_data) + i; | |
| 3810 if (syl->sy_len > len | |
| 3811 && STRNCMP(p, syl->sy_chars, syl->sy_len) == 0) | |
| 3812 len = syl->sy_len; | |
| 3813 } | |
| 3814 if (len != 0) /* found a match, count syllable */ | |
| 3815 { | |
| 3816 ++cnt; | |
| 3817 skip = FALSE; | |
| 3818 } | |
| 3819 else | |
| 3820 { | |
| 3821 /* No recognized syllable item, at least a syllable char then? */ | |
| 3822 #ifdef FEAT_MBYTE | |
| 3823 c = mb_ptr2char(p); | |
| 3824 len = (*mb_ptr2len)(p); | |
| 3825 #else | |
| 3826 c = *p; | |
| 3827 len = 1; | |
| 3828 #endif | |
| 3829 if (vim_strchr(slang->sl_syllable, c) == NULL) | |
| 3830 skip = FALSE; /* No, search for next syllable */ | |
| 3831 else if (!skip) | |
| 3832 { | |
| 3833 ++cnt; /* Yes, count it */ | |
| 3834 skip = TRUE; /* don't count following syllable chars */ | |
| 3835 } | |
| 3836 } | |
| 3837 } | |
| 3838 return cnt; | |
| 376 | 3839 } |
| 3840 | |
| 381 | 3841 /* |
| 3842 * Set the SOFOFROM and SOFOTO items in language "lp". | |
| 483 | 3843 * Returns SP_*ERROR flags when there is something wrong. |
| 381 | 3844 */ |
| 3845 static int | |
| 3846 set_sofo(lp, from, to) | |
| 3847 slang_T *lp; | |
| 3848 char_u *from; | |
| 3849 char_u *to; | |
| 3850 { | |
| 3851 int i; | |
| 3852 | |
| 3853 #ifdef FEAT_MBYTE | |
| 3854 garray_T *gap; | |
| 3855 char_u *s; | |
| 3856 char_u *p; | |
| 3857 int c; | |
| 3858 int *inp; | |
| 3859 | |
| 3860 if (has_mbyte) | |
| 3861 { | |
| 3862 /* Use "sl_sal" as an array with 256 pointers to a list of wide | |
| 3863 * characters. The index is the low byte of the character. | |
| 3864 * The list contains from-to pairs with a terminating NUL. | |
| 3865 * sl_sal_first[] is used for latin1 "from" characters. */ | |
| 3866 gap = &lp->sl_sal; | |
| 3867 ga_init2(gap, sizeof(int *), 1); | |
| 3868 if (ga_grow(gap, 256) == FAIL) | |
| 495 | 3869 return SP_OTHERERROR; |
| 381 | 3870 vim_memset(gap->ga_data, 0, sizeof(int *) * 256); |
| 3871 gap->ga_len = 256; | |
| 3872 | |
| 3873 /* First count the number of items for each list. Temporarily use | |
| 3874 * sl_sal_first[] for this. */ | |
| 3875 for (p = from, s = to; *p != NUL && *s != NUL; ) | |
| 3876 { | |
| 474 | 3877 c = mb_cptr2char_adv(&p); |
| 3878 mb_cptr_adv(s); | |
| 381 | 3879 if (c >= 256) |
| 3880 ++lp->sl_sal_first[c & 0xff]; | |
| 3881 } | |
| 3882 if (*p != NUL || *s != NUL) /* lengths differ */ | |
| 483 | 3883 return SP_FORMERROR; |
| 381 | 3884 |
| 3885 /* Allocate the lists. */ | |
| 3886 for (i = 0; i < 256; ++i) | |
| 3887 if (lp->sl_sal_first[i] > 0) | |
| 3888 { | |
| 3889 p = alloc(sizeof(int) * (lp->sl_sal_first[i] * 2 + 1)); | |
| 3890 if (p == NULL) | |
| 495 | 3891 return SP_OTHERERROR; |
| 381 | 3892 ((int **)gap->ga_data)[i] = (int *)p; |
| 3893 *(int *)p = 0; | |
| 3894 } | |
| 3895 | |
| 3896 /* Put the characters up to 255 in sl_sal_first[] the rest in a sl_sal | |
| 3897 * list. */ | |
| 3898 vim_memset(lp->sl_sal_first, 0, sizeof(salfirst_T) * 256); | |
| 3899 for (p = from, s = to; *p != NUL && *s != NUL; ) | |
| 3900 { | |
| 474 | 3901 c = mb_cptr2char_adv(&p); |
| 3902 i = mb_cptr2char_adv(&s); | |
| 381 | 3903 if (c >= 256) |
| 3904 { | |
| 3905 /* Append the from-to chars at the end of the list with | |
| 3906 * the low byte. */ | |
| 3907 inp = ((int **)gap->ga_data)[c & 0xff]; | |
| 3908 while (*inp != 0) | |
| 3909 ++inp; | |
| 3910 *inp++ = c; /* from char */ | |
| 3911 *inp++ = i; /* to char */ | |
| 3912 *inp++ = NUL; /* NUL at the end */ | |
| 3913 } | |
| 3914 else | |
| 3915 /* mapping byte to char is done in sl_sal_first[] */ | |
| 3916 lp->sl_sal_first[c] = i; | |
| 3917 } | |
| 3918 } | |
| 3919 else | |
| 3920 #endif | |
| 3921 { | |
| 3922 /* mapping bytes to bytes is done in sl_sal_first[] */ | |
| 3923 if (STRLEN(from) != STRLEN(to)) | |
| 483 | 3924 return SP_FORMERROR; |
| 381 | 3925 |
| 3926 for (i = 0; to[i] != NUL; ++i) | |
| 3927 lp->sl_sal_first[from[i]] = to[i]; | |
| 3928 lp->sl_sal.ga_len = 1; /* indicates we have soundfolding */ | |
| 3929 } | |
| 3930 | |
| 483 | 3931 return 0; |
| 381 | 3932 } |
| 3933 | |
| 3934 /* | |
| 3935 * Fill the first-index table for "lp". | |
| 3936 */ | |
| 3937 static void | |
| 3938 set_sal_first(lp) | |
| 3939 slang_T *lp; | |
| 3940 { | |
| 3941 salfirst_T *sfirst; | |
| 3942 int i; | |
| 3943 salitem_T *smp; | |
| 3944 int c; | |
| 3945 garray_T *gap = &lp->sl_sal; | |
| 3946 | |
| 3947 sfirst = lp->sl_sal_first; | |
| 3948 for (i = 0; i < 256; ++i) | |
| 3949 sfirst[i] = -1; | |
| 3950 smp = (salitem_T *)gap->ga_data; | |
| 3951 for (i = 0; i < gap->ga_len; ++i) | |
| 3952 { | |
| 3953 #ifdef FEAT_MBYTE | |
| 3954 if (has_mbyte) | |
| 3955 /* Use the lowest byte of the first character. For latin1 it's | |
| 3956 * the character, for other encodings it should differ for most | |
| 3957 * characters. */ | |
| 3958 c = *smp[i].sm_lead_w & 0xff; | |
| 3959 else | |
| 3960 #endif | |
| 3961 c = *smp[i].sm_lead; | |
| 3962 if (sfirst[c] == -1) | |
| 3963 { | |
| 3964 sfirst[c] = i; | |
| 3965 #ifdef FEAT_MBYTE | |
| 3966 if (has_mbyte) | |
| 3967 { | |
| 3968 int n; | |
| 3969 | |
| 3970 /* Make sure all entries with this byte are following each | |
| 3971 * other. Move the ones that are in the wrong position. Do | |
| 3972 * keep the same ordering! */ | |
| 3973 while (i + 1 < gap->ga_len | |
| 3974 && (*smp[i + 1].sm_lead_w & 0xff) == c) | |
| 3975 /* Skip over entry with same index byte. */ | |
| 3976 ++i; | |
| 3977 | |
| 3978 for (n = 1; i + n < gap->ga_len; ++n) | |
| 3979 if ((*smp[i + n].sm_lead_w & 0xff) == c) | |
| 3980 { | |
| 3981 salitem_T tsal; | |
| 3982 | |
| 3983 /* Move entry with same index byte after the entries | |
| 3984 * we already found. */ | |
| 3985 ++i; | |
| 3986 --n; | |
| 3987 tsal = smp[i + n]; | |
| 3988 mch_memmove(smp + i + 1, smp + i, | |
| 3989 sizeof(salitem_T) * n); | |
| 3990 smp[i] = tsal; | |
| 3991 } | |
| 3992 } | |
| 3993 #endif | |
| 3994 } | |
| 3995 } | |
| 3996 } | |
| 376 | 3997 |
| 372 | 3998 #ifdef FEAT_MBYTE |
| 3999 /* | |
| 4000 * Turn a multi-byte string into a wide character string. | |
| 4001 * Return it in allocated memory (NULL for out-of-memory) | |
| 4002 */ | |
| 4003 static int * | |
| 4004 mb_str2wide(s) | |
| 4005 char_u *s; | |
| 4006 { | |
| 4007 int *res; | |
| 4008 char_u *p; | |
| 4009 int i = 0; | |
| 4010 | |
| 4011 res = (int *)alloc(sizeof(int) * (mb_charlen(s) + 1)); | |
| 4012 if (res != NULL) | |
| 4013 { | |
| 4014 for (p = s; *p != NUL; ) | |
| 4015 res[i++] = mb_ptr2char_adv(&p); | |
| 4016 res[i] = NUL; | |
| 4017 } | |
| 4018 return res; | |
| 4019 } | |
| 4020 #endif | |
| 4021 | |
| 236 | 4022 /* |
| 625 | 4023 * Read a tree from the .spl or .sug file. |
| 4024 * Allocates the memory and stores pointers in "bytsp" and "idxsp". | |
| 4025 * This is skipped when the tree has zero length. | |
| 4026 * Returns zero when OK, SP_ value for an error. | |
| 4027 */ | |
| 4028 static int | |
| 4029 spell_read_tree(fd, bytsp, idxsp, prefixtree, prefixcnt) | |
| 4030 FILE *fd; | |
| 4031 char_u **bytsp; | |
| 4032 idx_T **idxsp; | |
| 4033 int prefixtree; /* TRUE for the prefix tree */ | |
| 4034 int prefixcnt; /* when "prefixtree" is TRUE: prefix count */ | |
| 4035 { | |
| 4036 int len; | |
| 4037 int idx; | |
| 4038 char_u *bp; | |
| 4039 idx_T *ip; | |
| 4040 | |
| 4041 /* The tree size was computed when writing the file, so that we can | |
| 4042 * allocate it as one long block. <nodecount> */ | |
| 695 | 4043 len = get4c(fd); |
| 625 | 4044 if (len < 0) |
| 4045 return SP_TRUNCERROR; | |
| 4046 if (len > 0) | |
| 4047 { | |
| 4048 /* Allocate the byte array. */ | |
| 4049 bp = lalloc((long_u)len, TRUE); | |
| 4050 if (bp == NULL) | |
| 4051 return SP_OTHERERROR; | |
| 4052 *bytsp = bp; | |
| 4053 | |
| 4054 /* Allocate the index array. */ | |
| 4055 ip = (idx_T *)lalloc_clear((long_u)(len * sizeof(int)), TRUE); | |
| 4056 if (ip == NULL) | |
| 4057 return SP_OTHERERROR; | |
| 4058 *idxsp = ip; | |
| 4059 | |
| 4060 /* Recursively read the tree and store it in the array. */ | |
| 4061 idx = read_tree_node(fd, bp, ip, len, 0, prefixtree, prefixcnt); | |
| 4062 if (idx < 0) | |
| 4063 return idx; | |
| 4064 } | |
| 4065 return 0; | |
| 4066 } | |
| 4067 | |
| 4068 /* | |
| 300 | 4069 * Read one row of siblings from the spell file and store it in the byte array |
| 4070 * "byts" and index array "idxs". Recursively read the children. | |
| 4071 * | |
| 625 | 4072 * NOTE: The code here must match put_node()! |
| 4073 * | |
| 4074 * Returns the index (>= 0) following the siblings. | |
| 4075 * Returns SP_TRUNCERROR if the file is shorter than expected. | |
| 4076 * Returns SP_FORMERROR if there is a format error. | |
| 236 | 4077 */ |
| 324 | 4078 static idx_T |
| 625 | 4079 read_tree_node(fd, byts, idxs, maxidx, startidx, prefixtree, maxprefcondnr) |
| 300 | 4080 FILE *fd; |
| 4081 char_u *byts; | |
| 324 | 4082 idx_T *idxs; |
| 300 | 4083 int maxidx; /* size of arrays */ |
| 324 | 4084 idx_T startidx; /* current index in "byts" and "idxs" */ |
| 339 | 4085 int prefixtree; /* TRUE for reading PREFIXTREE */ |
| 4086 int maxprefcondnr; /* maximum for <prefcondnr> */ | |
| 236 | 4087 { |
| 300 | 4088 int len; |
| 4089 int i; | |
| 4090 int n; | |
| 324 | 4091 idx_T idx = startidx; |
| 300 | 4092 int c; |
| 366 | 4093 int c2; |
| 300 | 4094 #define SHARED_MASK 0x8000000 |
| 236 | 4095 |
| 300 | 4096 len = getc(fd); /* <siblingcount> */ |
| 4097 if (len <= 0) | |
| 625 | 4098 return SP_TRUNCERROR; |
| 300 | 4099 |
| 4100 if (startidx + len >= maxidx) | |
| 625 | 4101 return SP_FORMERROR; |
| 300 | 4102 byts[idx++] = len; |
| 4103 | |
| 4104 /* Read the byte values, flag/region bytes and shared indexes. */ | |
| 4105 for (i = 1; i <= len; ++i) | |
| 236 | 4106 { |
| 300 | 4107 c = getc(fd); /* <byte> */ |
| 4108 if (c < 0) | |
| 625 | 4109 return SP_TRUNCERROR; |
| 300 | 4110 if (c <= BY_SPECIAL) |
| 4111 { | |
| 366 | 4112 if (c == BY_NOFLAGS && !prefixtree) |
| 300 | 4113 { |
| 4114 /* No flags, all regions. */ | |
| 4115 idxs[idx] = 0; | |
| 4116 c = 0; | |
| 4117 } | |
| 390 | 4118 else if (c != BY_INDEX) |
| 300 | 4119 { |
| 339 | 4120 if (prefixtree) |
| 4121 { | |
| 455 | 4122 /* Read the optional pflags byte, the prefix ID and the |
| 4123 * condition nr. In idxs[] store the prefix ID in the low | |
| 4124 * byte, the condition index shifted up 8 bits, the flags | |
| 4125 * shifted up 24 bits. */ | |
| 4126 if (c == BY_FLAGS) | |
| 4127 c = getc(fd) << 24; /* <pflags> */ | |
| 4128 else | |
| 4129 c = 0; | |
| 4130 | |
| 481 | 4131 c |= getc(fd); /* <affixID> */ |
| 455 | 4132 |
| 695 | 4133 n = get2c(fd); /* <prefcondnr> */ |
| 339 | 4134 if (n >= maxprefcondnr) |
| 625 | 4135 return SP_FORMERROR; |
| 455 | 4136 c |= (n << 8); |
| 339 | 4137 } |
| 390 | 4138 else /* c must be BY_FLAGS or BY_FLAGS2 */ |
| 339 | 4139 { |
| 4140 /* Read flags and optional region and prefix ID. In | |
| 390 | 4141 * idxs[] the flags go in the low two bytes, region above |
| 4142 * that and prefix ID above the region. */ | |
| 4143 c2 = c; | |
| 339 | 4144 c = getc(fd); /* <flags> */ |
| 390 | 4145 if (c2 == BY_FLAGS2) |
| 4146 c = (getc(fd) << 8) + c; /* <flags2> */ | |
| 339 | 4147 if (c & WF_REGION) |
| 390 | 4148 c = (getc(fd) << 16) + c; /* <region> */ |
| 481 | 4149 if (c & WF_AFX) |
| 4150 c = (getc(fd) << 24) + c; /* <affixID> */ | |
| 339 | 4151 } |
| 4152 | |
| 300 | 4153 idxs[idx] = c; |
| 4154 c = 0; | |
| 4155 } | |
| 4156 else /* c == BY_INDEX */ | |
| 4157 { | |
| 4158 /* <nodeidx> */ | |
| 695 | 4159 n = get3c(fd); |
| 300 | 4160 if (n < 0 || n >= maxidx) |
| 625 | 4161 return SP_FORMERROR; |
| 300 | 4162 idxs[idx] = n + SHARED_MASK; |
| 4163 c = getc(fd); /* <xbyte> */ | |
| 4164 } | |
| 4165 } | |
| 4166 byts[idx++] = c; | |
| 236 | 4167 } |
| 4168 | |
| 300 | 4169 /* Recursively read the children for non-shared siblings. |
| 4170 * Skip the end-of-word ones (zero byte value) and the shared ones (and | |
| 4171 * remove SHARED_MASK) */ | |
| 4172 for (i = 1; i <= len; ++i) | |
| 4173 if (byts[startidx + i] != 0) | |
| 4174 { | |
| 4175 if (idxs[startidx + i] & SHARED_MASK) | |
| 4176 idxs[startidx + i] &= ~SHARED_MASK; | |
| 4177 else | |
| 4178 { | |
| 4179 idxs[startidx + i] = idx; | |
| 625 | 4180 idx = read_tree_node(fd, byts, idxs, maxidx, idx, |
| 339 | 4181 prefixtree, maxprefcondnr); |
| 300 | 4182 if (idx < 0) |
| 4183 break; | |
| 4184 } | |
| 4185 } | |
| 236 | 4186 |
| 300 | 4187 return idx; |
| 236 | 4188 } |
| 4189 | |
| 4190 /* | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4191 * Parse 'spelllang' and set w_s->b_langp accordingly. |
| 351 | 4192 * Returns NULL if it's OK, an error message otherwise. |
| 236 | 4193 */ |
| 4194 char_u * | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4195 did_set_spelllang(wp) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4196 win_T *wp; |
| 236 | 4197 { |
| 4198 garray_T ga; | |
| 351 | 4199 char_u *splp; |
| 236 | 4200 char_u *region; |
| 416 | 4201 char_u region_cp[3]; |
| 355 | 4202 int filename; |
| 236 | 4203 int region_mask; |
| 503 | 4204 slang_T *slang; |
| 236 | 4205 int c; |
| 351 | 4206 char_u lang[MAXWLEN + 1]; |
| 323 | 4207 char_u spf_name[MAXPATHL]; |
| 351 | 4208 int len; |
| 4209 char_u *p; | |
| 381 | 4210 int round; |
| 385 | 4211 char_u *spf; |
| 389 | 4212 char_u *use_region = NULL; |
| 4213 int dont_use_region = FALSE; | |
| 501 | 4214 int nobreak = FALSE; |
| 503 | 4215 int i, j; |
| 4216 langp_T *lp, *lp2; | |
| 1185 | 4217 static int recursive = FALSE; |
| 4218 char_u *ret_msg = NULL; | |
| 4219 char_u *spl_copy; | |
| 4220 | |
| 4221 /* We don't want to do this recursively. May happen when a language is | |
| 4222 * not available and the SpellFileMissing autocommand opens a new buffer | |
| 4223 * in which 'spell' is set. */ | |
| 4224 if (recursive) | |
| 4225 return NULL; | |
| 4226 recursive = TRUE; | |
| 236 | 4227 |
| 4228 ga_init2(&ga, sizeof(langp_T), 2); | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4229 clear_midword(wp); |
| 236 | 4230 |
| 1185 | 4231 /* Make a copy of 'spellang', the SpellFileMissing autocommands may change |
| 4232 * it under our fingers. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4233 spl_copy = vim_strsave(wp->w_s->b_p_spl); |
| 1185 | 4234 if (spl_copy == NULL) |
| 4235 goto theend; | |
| 4236 | |
| 351 | 4237 /* loop over comma separated language names. */ |
| 1185 | 4238 for (splp = spl_copy; *splp != NUL; ) |
| 351 | 4239 { |
| 4240 /* Get one language name. */ | |
| 4241 copy_option_part(&splp, lang, MAXWLEN, ","); | |
| 240 | 4242 region = NULL; |
| 835 | 4243 len = (int)STRLEN(lang); |
| 355 | 4244 |
| 4245 /* If the name ends in ".spl" use it as the name of the spell file. | |
| 4246 * If there is a region name let "region" point to it and remove it | |
| 4247 * from the name. */ | |
| 4248 if (len > 4 && fnamecmp(lang + len - 4, ".spl") == 0) | |
| 236 | 4249 { |
| 355 | 4250 filename = TRUE; |
| 4251 | |
| 416 | 4252 /* Locate a region and remove it from the file name. */ |
| 4253 p = vim_strchr(gettail(lang), '_'); | |
| 4254 if (p != NULL && ASCII_ISALPHA(p[1]) && ASCII_ISALPHA(p[2]) | |
| 4255 && !ASCII_ISALPHA(p[3])) | |
| 4256 { | |
| 4257 vim_strncpy(region_cp, p + 1, 2); | |
| 4258 mch_memmove(p, p + 3, len - (p - lang) - 2); | |
| 4259 len -= 3; | |
| 4260 region = region_cp; | |
| 4261 } | |
| 4262 else | |
| 4263 dont_use_region = TRUE; | |
| 4264 | |
| 355 | 4265 /* Check if we loaded this language before. */ |
| 503 | 4266 for (slang = first_lang; slang != NULL; slang = slang->sl_next) |
| 4267 if (fullpathcmp(lang, slang->sl_fname, FALSE) == FPC_SAME) | |
| 355 | 4268 break; |
| 236 | 4269 } |
| 355 | 4270 else |
| 4271 { | |
| 4272 filename = FALSE; | |
| 4273 if (len > 3 && lang[len - 3] == '_') | |
| 4274 { | |
| 4275 region = lang + len - 2; | |
| 4276 len -= 3; | |
| 4277 lang[len] = NUL; | |
| 389 | 4278 } |
| 4279 else | |
| 4280 dont_use_region = TRUE; | |
| 355 | 4281 |
| 4282 /* Check if we loaded this language before. */ | |
| 503 | 4283 for (slang = first_lang; slang != NULL; slang = slang->sl_next) |
| 4284 if (STRICMP(lang, slang->sl_name) == 0) | |
| 355 | 4285 break; |
| 4286 } | |
| 236 | 4287 |
| 416 | 4288 if (region != NULL) |
| 4289 { | |
| 4290 /* If the region differs from what was used before then don't | |
| 4291 * use it for 'spellfile'. */ | |
| 4292 if (use_region != NULL && STRCMP(region, use_region) != 0) | |
| 4293 dont_use_region = TRUE; | |
| 4294 use_region = region; | |
| 4295 } | |
| 4296 | |
| 351 | 4297 /* If not found try loading the language now. */ |
| 503 | 4298 if (slang == NULL) |
| 355 | 4299 { |
| 4300 if (filename) | |
| 4301 (void)spell_load_file(lang, lang, NULL, FALSE); | |
| 4302 else | |
| 1185 | 4303 { |
| 355 | 4304 spell_load_lang(lang); |
| 1185 | 4305 #ifdef FEAT_AUTOCMD |
| 4306 /* SpellFileMissing autocommands may do anything, including | |
| 4307 * destroying the buffer we are using... */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4308 if (!buf_valid(wp->w_buffer)) |
| 1185 | 4309 { |
| 4310 ret_msg = (char_u *)"E797: SpellFileMissing autocommand deleted buffer"; | |
| 4311 goto theend; | |
| 4312 } | |
| 4313 #endif | |
| 4314 } | |
| 355 | 4315 } |
| 236 | 4316 |
| 307 | 4317 /* |
| 351 | 4318 * Loop over the languages, there can be several files for "lang". |
| 307 | 4319 */ |
| 503 | 4320 for (slang = first_lang; slang != NULL; slang = slang->sl_next) |
| 4321 if (filename ? fullpathcmp(lang, slang->sl_fname, FALSE) == FPC_SAME | |
| 4322 : STRICMP(lang, slang->sl_name) == 0) | |
| 236 | 4323 { |
| 316 | 4324 region_mask = REGION_ALL; |
| 355 | 4325 if (!filename && region != NULL) |
| 236 | 4326 { |
| 307 | 4327 /* find region in sl_regions */ |
| 503 | 4328 c = find_region(slang->sl_regions, region); |
| 307 | 4329 if (c == REGION_ALL) |
| 4330 { | |
| 503 | 4331 if (slang->sl_add) |
| 389 | 4332 { |
| 503 | 4333 if (*slang->sl_regions != NUL) |
| 389 | 4334 /* This addition file is for other regions. */ |
| 4335 region_mask = 0; | |
| 4336 } | |
| 4337 else | |
| 4338 /* This is probably an error. Give a warning and | |
| 4339 * accept the words anyway. */ | |
| 351 | 4340 smsg((char_u *) |
| 4341 _("Warning: region %s not supported"), | |
| 4342 region); | |
| 307 | 4343 } |
| 4344 else | |
| 4345 region_mask = 1 << c; | |
| 236 | 4346 } |
| 307 | 4347 |
| 389 | 4348 if (region_mask != 0) |
| 307 | 4349 { |
| 389 | 4350 if (ga_grow(&ga, 1) == FAIL) |
| 4351 { | |
| 4352 ga_clear(&ga); | |
| 1185 | 4353 ret_msg = e_outofmem; |
| 4354 goto theend; | |
| 389 | 4355 } |
| 503 | 4356 LANGP_ENTRY(ga, ga.ga_len)->lp_slang = slang; |
| 389 | 4357 LANGP_ENTRY(ga, ga.ga_len)->lp_region = region_mask; |
| 4358 ++ga.ga_len; | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4359 use_midword(slang, wp); |
| 503 | 4360 if (slang->sl_nobreak) |
| 501 | 4361 nobreak = TRUE; |
| 307 | 4362 } |
| 385 | 4363 } |
| 4364 } | |
| 4365 | |
| 4366 /* round 0: load int_wordlist, if possible. | |
| 4367 * round 1: load first name in 'spellfile'. | |
| 4368 * round 2: load second name in 'spellfile. | |
| 4369 * etc. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4370 spf = curwin->w_s->b_p_spf; |
| 385 | 4371 for (round = 0; round == 0 || *spf != NUL; ++round) |
| 4372 { | |
| 4373 if (round == 0) | |
| 4374 { | |
| 4375 /* Internal wordlist, if there is one. */ | |
| 4376 if (int_wordlist == NULL) | |
| 381 | 4377 continue; |
| 385 | 4378 int_wordlist_spl(spf_name); |
| 381 | 4379 } |
| 4380 else | |
| 4381 { | |
| 385 | 4382 /* One entry in 'spellfile'. */ |
| 4383 copy_option_part(&spf, spf_name, MAXPATHL - 5, ","); | |
| 4384 STRCAT(spf_name, ".spl"); | |
| 4385 | |
| 4386 /* If it was already found above then skip it. */ | |
| 4387 for (c = 0; c < ga.ga_len; ++c) | |
| 500 | 4388 { |
| 4389 p = LANGP_ENTRY(ga, c)->lp_slang->sl_fname; | |
| 4390 if (p != NULL && fullpathcmp(spf_name, p, FALSE) == FPC_SAME) | |
| 385 | 4391 break; |
| 500 | 4392 } |
| 385 | 4393 if (c < ga.ga_len) |
| 381 | 4394 continue; |
| 4395 } | |
| 4396 | |
| 351 | 4397 /* Check if it was loaded already. */ |
| 503 | 4398 for (slang = first_lang; slang != NULL; slang = slang->sl_next) |
| 4399 if (fullpathcmp(spf_name, slang->sl_fname, FALSE) == FPC_SAME) | |
| 323 | 4400 break; |
| 503 | 4401 if (slang == NULL) |
| 323 | 4402 { |
| 351 | 4403 /* Not loaded, try loading it now. The language name includes the |
| 385 | 4404 * region name, the region is ignored otherwise. for int_wordlist |
| 4405 * use an arbitrary name. */ | |
| 4406 if (round == 0) | |
| 4407 STRCPY(lang, "internal wordlist"); | |
| 4408 else | |
| 4409 { | |
| 4410 vim_strncpy(lang, gettail(spf_name), MAXWLEN); | |
| 381 | 4411 p = vim_strchr(lang, '.'); |
| 4412 if (p != NULL) | |
| 4413 *p = NUL; /* truncate at ".encoding.add" */ | |
| 4414 } | |
| 503 | 4415 slang = spell_load_file(spf_name, lang, NULL, TRUE); |
| 501 | 4416 |
| 4417 /* If one of the languages has NOBREAK we assume the addition | |
| 4418 * files also have this. */ | |
| 503 | 4419 if (slang != NULL && nobreak) |
| 4420 slang->sl_nobreak = TRUE; | |
| 4421 } | |
| 4422 if (slang != NULL && ga_grow(&ga, 1) == OK) | |
| 323 | 4423 { |
| 389 | 4424 region_mask = REGION_ALL; |
| 4425 if (use_region != NULL && !dont_use_region) | |
| 4426 { | |
| 4427 /* find region in sl_regions */ | |
| 503 | 4428 c = find_region(slang->sl_regions, use_region); |
| 389 | 4429 if (c != REGION_ALL) |
| 4430 region_mask = 1 << c; | |
| 503 | 4431 else if (*slang->sl_regions != NUL) |
| 389 | 4432 /* This spell file is for other regions. */ |
| 4433 region_mask = 0; | |
| 4434 } | |
| 4435 | |
| 4436 if (region_mask != 0) | |
| 4437 { | |
| 503 | 4438 LANGP_ENTRY(ga, ga.ga_len)->lp_slang = slang; |
| 4439 LANGP_ENTRY(ga, ga.ga_len)->lp_sallang = NULL; | |
| 4440 LANGP_ENTRY(ga, ga.ga_len)->lp_replang = NULL; | |
| 389 | 4441 LANGP_ENTRY(ga, ga.ga_len)->lp_region = region_mask; |
| 4442 ++ga.ga_len; | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4443 use_midword(slang, wp); |
| 389 | 4444 } |
| 323 | 4445 } |
| 4446 } | |
| 4447 | |
| 236 | 4448 /* Everything is fine, store the new b_langp value. */ |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4449 ga_clear(&wp->w_s->b_langp); |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4450 wp->w_s->b_langp = ga; |
| 236 | 4451 |
| 503 | 4452 /* For each language figure out what language to use for sound folding and |
| 4453 * REP items. If the language doesn't support it itself use another one | |
| 4454 * with the same name. E.g. for "en-math" use "en". */ | |
| 4455 for (i = 0; i < ga.ga_len; ++i) | |
| 4456 { | |
| 4457 lp = LANGP_ENTRY(ga, i); | |
| 4458 | |
| 4459 /* sound folding */ | |
| 4460 if (lp->lp_slang->sl_sal.ga_len > 0) | |
| 4461 /* language does sound folding itself */ | |
| 4462 lp->lp_sallang = lp->lp_slang; | |
| 4463 else | |
| 4464 /* find first similar language that does sound folding */ | |
| 4465 for (j = 0; j < ga.ga_len; ++j) | |
| 4466 { | |
| 4467 lp2 = LANGP_ENTRY(ga, j); | |
| 4468 if (lp2->lp_slang->sl_sal.ga_len > 0 | |
| 4469 && STRNCMP(lp->lp_slang->sl_name, | |
| 4470 lp2->lp_slang->sl_name, 2) == 0) | |
| 4471 { | |
| 4472 lp->lp_sallang = lp2->lp_slang; | |
| 4473 break; | |
| 4474 } | |
| 4475 } | |
| 4476 | |
| 4477 /* REP items */ | |
| 4478 if (lp->lp_slang->sl_rep.ga_len > 0) | |
| 4479 /* language has REP items itself */ | |
| 4480 lp->lp_replang = lp->lp_slang; | |
| 4481 else | |
| 625 | 4482 /* find first similar language that has REP items */ |
| 503 | 4483 for (j = 0; j < ga.ga_len; ++j) |
| 4484 { | |
| 4485 lp2 = LANGP_ENTRY(ga, j); | |
| 4486 if (lp2->lp_slang->sl_rep.ga_len > 0 | |
| 4487 && STRNCMP(lp->lp_slang->sl_name, | |
| 4488 lp2->lp_slang->sl_name, 2) == 0) | |
| 4489 { | |
| 4490 lp->lp_replang = lp2->lp_slang; | |
| 4491 break; | |
| 4492 } | |
| 4493 } | |
| 4494 } | |
| 4495 | |
| 1185 | 4496 theend: |
| 4497 vim_free(spl_copy); | |
| 4498 recursive = FALSE; | |
| 4499 return ret_msg; | |
| 236 | 4500 } |
| 4501 | |
| 4502 /* | |
| 376 | 4503 * Clear the midword characters for buffer "buf". |
| 4504 */ | |
| 4505 static void | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4506 clear_midword(wp) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4507 win_T *wp; |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4508 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4509 vim_memset(wp->w_s->b_spell_ismw, 0, 256); |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4510 #ifdef FEAT_MBYTE |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4511 vim_free(wp->w_s->b_spell_ismw_mb); |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4512 wp->w_s->b_spell_ismw_mb = NULL; |
| 376 | 4513 #endif |
| 4514 } | |
| 4515 | |
| 4516 /* | |
| 4517 * Use the "sl_midword" field of language "lp" for buffer "buf". | |
| 4518 * They add up to any currently used midword characters. | |
| 4519 */ | |
| 4520 static void | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4521 use_midword(lp, wp) |
| 376 | 4522 slang_T *lp; |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4523 win_T *wp; |
| 376 | 4524 { |
| 4525 char_u *p; | |
| 4526 | |
| 389 | 4527 if (lp->sl_midword == NULL) /* there aren't any */ |
| 4528 return; | |
| 4529 | |
| 376 | 4530 for (p = lp->sl_midword; *p != NUL; ) |
| 4531 #ifdef FEAT_MBYTE | |
| 4532 if (has_mbyte) | |
| 4533 { | |
| 4534 int c, l, n; | |
| 4535 char_u *bp; | |
| 4536 | |
| 4537 c = mb_ptr2char(p); | |
| 474 | 4538 l = (*mb_ptr2len)(p); |
| 4539 if (c < 256 && l <= 2) | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4540 wp->w_s->b_spell_ismw[c] = TRUE; |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4541 else if (wp->w_s->b_spell_ismw_mb == NULL) |
| 376 | 4542 /* First multi-byte char in "b_spell_ismw_mb". */ |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4543 wp->w_s->b_spell_ismw_mb = vim_strnsave(p, l); |
| 376 | 4544 else |
| 4545 { | |
| 4546 /* Append multi-byte chars to "b_spell_ismw_mb". */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4547 n = (int)STRLEN(wp->w_s->b_spell_ismw_mb); |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4548 bp = vim_strnsave(wp->w_s->b_spell_ismw_mb, n + l); |
| 376 | 4549 if (bp != NULL) |
| 4550 { | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4551 vim_free(wp->w_s->b_spell_ismw_mb); |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4552 wp->w_s->b_spell_ismw_mb = bp; |
| 376 | 4553 vim_strncpy(bp + n, p, l); |
| 4554 } | |
| 4555 } | |
| 4556 p += l; | |
| 4557 } | |
| 4558 else | |
| 4559 #endif | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4560 wp->w_s->b_spell_ismw[*p++] = TRUE; |
| 376 | 4561 } |
| 4562 | |
| 4563 /* | |
| 236 | 4564 * Find the region "region[2]" in "rp" (points to "sl_regions"). |
| 4565 * Each region is simply stored as the two characters of it's name. | |
| 381 | 4566 * Returns the index if found (first is 0), REGION_ALL if not found. |
| 236 | 4567 */ |
| 4568 static int | |
| 4569 find_region(rp, region) | |
| 4570 char_u *rp; | |
| 4571 char_u *region; | |
| 4572 { | |
| 4573 int i; | |
| 4574 | |
| 4575 for (i = 0; ; i += 2) | |
| 4576 { | |
| 4577 if (rp[i] == NUL) | |
| 4578 return REGION_ALL; | |
| 4579 if (rp[i] == region[0] && rp[i + 1] == region[1]) | |
| 4580 break; | |
| 4581 } | |
| 4582 return i / 2; | |
| 4583 } | |
| 4584 | |
| 4585 /* | |
| 323 | 4586 * Return case type of word: |
| 236 | 4587 * w word 0 |
| 300 | 4588 * Word WF_ONECAP |
| 4589 * W WORD WF_ALLCAP | |
| 4590 * WoRd wOrd WF_KEEPCAP | |
| 236 | 4591 */ |
| 4592 static int | |
| 4593 captype(word, end) | |
| 4594 char_u *word; | |
| 323 | 4595 char_u *end; /* When NULL use up to NUL byte. */ |
| 236 | 4596 { |
| 4597 char_u *p; | |
| 4598 int c; | |
| 4599 int firstcap; | |
| 4600 int allcap; | |
| 4601 int past_second = FALSE; /* past second word char */ | |
| 4602 | |
| 4603 /* find first letter */ | |
| 376 | 4604 for (p = word; !spell_iswordp_nmw(p); mb_ptr_adv(p)) |
| 323 | 4605 if (end == NULL ? *p == NUL : p >= end) |
| 236 | 4606 return 0; /* only non-word characters, illegal word */ |
| 4607 #ifdef FEAT_MBYTE | |
| 310 | 4608 if (has_mbyte) |
| 4609 c = mb_ptr2char_adv(&p); | |
| 4610 else | |
| 236 | 4611 #endif |
| 310 | 4612 c = *p++; |
| 324 | 4613 firstcap = allcap = SPELL_ISUPPER(c); |
| 236 | 4614 |
| 4615 /* | |
| 4616 * Need to check all letters to find a word with mixed upper/lower. | |
| 4617 * But a word with an upper char only at start is a ONECAP. | |
| 4618 */ | |
| 323 | 4619 for ( ; end == NULL ? *p != NUL : p < end; mb_ptr_adv(p)) |
| 376 | 4620 if (spell_iswordp_nmw(p)) |
| 236 | 4621 { |
| 455 | 4622 c = PTR2CHAR(p); |
| 324 | 4623 if (!SPELL_ISUPPER(c)) |
| 236 | 4624 { |
| 4625 /* UUl -> KEEPCAP */ | |
| 4626 if (past_second && allcap) | |
| 300 | 4627 return WF_KEEPCAP; |
| 236 | 4628 allcap = FALSE; |
| 4629 } | |
| 4630 else if (!allcap) | |
| 4631 /* UlU -> KEEPCAP */ | |
| 300 | 4632 return WF_KEEPCAP; |
| 236 | 4633 past_second = TRUE; |
| 4634 } | |
| 4635 | |
| 4636 if (allcap) | |
| 300 | 4637 return WF_ALLCAP; |
| 236 | 4638 if (firstcap) |
| 300 | 4639 return WF_ONECAP; |
| 236 | 4640 return 0; |
| 4641 } | |
| 4642 | |
| 474 | 4643 /* |
| 4644 * Like captype() but for a KEEPCAP word add ONECAP if the word starts with a | |
| 4645 * capital. So that make_case_word() can turn WOrd into Word. | |
| 4646 * Add ALLCAP for "WOrD". | |
| 4647 */ | |
| 4648 static int | |
| 4649 badword_captype(word, end) | |
| 4650 char_u *word; | |
| 4651 char_u *end; | |
| 4652 { | |
| 4653 int flags = captype(word, end); | |
| 475 | 4654 int c; |
| 474 | 4655 int l, u; |
| 4656 int first; | |
| 4657 char_u *p; | |
| 4658 | |
| 4659 if (flags & WF_KEEPCAP) | |
| 4660 { | |
| 4661 /* Count the number of UPPER and lower case letters. */ | |
| 4662 l = u = 0; | |
| 4663 first = FALSE; | |
| 4664 for (p = word; p < end; mb_ptr_adv(p)) | |
| 4665 { | |
| 475 | 4666 c = PTR2CHAR(p); |
| 4667 if (SPELL_ISUPPER(c)) | |
| 474 | 4668 { |
| 4669 ++u; | |
| 4670 if (p == word) | |
| 4671 first = TRUE; | |
| 4672 } | |
| 4673 else | |
| 4674 ++l; | |
| 4675 } | |
| 4676 | |
| 4677 /* If there are more UPPER than lower case letters suggest an | |
| 4678 * ALLCAP word. Otherwise, if the first letter is UPPER then | |
| 4679 * suggest ONECAP. Exception: "ALl" most likely should be "All", | |
| 4680 * require three upper case letters. */ | |
| 4681 if (u > l && u > 2) | |
| 4682 flags |= WF_ALLCAP; | |
| 4683 else if (first) | |
| 4684 flags |= WF_ONECAP; | |
| 633 | 4685 |
| 4686 if (u >= 2 && l >= 2) /* maCARONI maCAroni */ | |
| 4687 flags |= WF_MIXCAP; | |
| 474 | 4688 } |
| 4689 return flags; | |
| 4690 } | |
| 4691 | |
| 355 | 4692 # if defined(FEAT_MBYTE) || defined(EXITFREE) || defined(PROTO) |
| 4693 /* | |
| 4694 * Free all languages. | |
| 4695 */ | |
| 4696 void | |
| 4697 spell_free_all() | |
| 4698 { | |
| 503 | 4699 slang_T *slang; |
| 355 | 4700 buf_T *buf; |
| 385 | 4701 char_u fname[MAXPATHL]; |
| 355 | 4702 |
| 2580 | 4703 /* Go through all buffers and handle 'spelllang'. <VN> */ |
| 355 | 4704 for (buf = firstbuf; buf != NULL; buf = buf->b_next) |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4705 ga_clear(&buf->b_s.b_langp); |
| 355 | 4706 |
| 4707 while (first_lang != NULL) | |
| 4708 { | |
| 503 | 4709 slang = first_lang; |
| 4710 first_lang = slang->sl_next; | |
| 4711 slang_free(slang); | |
| 355 | 4712 } |
| 366 | 4713 |
| 385 | 4714 if (int_wordlist != NULL) |
| 4715 { | |
| 4716 /* Delete the internal wordlist and its .spl file */ | |
| 4717 mch_remove(int_wordlist); | |
| 4718 int_wordlist_spl(fname); | |
| 4719 mch_remove(fname); | |
| 4720 vim_free(int_wordlist); | |
| 4721 int_wordlist = NULL; | |
| 381 | 4722 } |
| 4723 | |
| 366 | 4724 init_spell_chartab(); |
| 626 | 4725 |
| 4726 vim_free(repl_to); | |
| 4727 repl_to = NULL; | |
| 4728 vim_free(repl_from); | |
| 4729 repl_from = NULL; | |
| 355 | 4730 } |
| 4731 # endif | |
| 4732 | |
| 236 | 4733 # if defined(FEAT_MBYTE) || defined(PROTO) |
| 4734 /* | |
| 4735 * Clear all spelling tables and reload them. | |
| 307 | 4736 * Used after 'encoding' is set and when ":mkspell" was used. |
| 236 | 4737 */ |
| 4738 void | |
| 4739 spell_reload() | |
| 4740 { | |
| 316 | 4741 win_T *wp; |
| 236 | 4742 |
| 358 | 4743 /* Initialize the table for spell_iswordp(). */ |
| 236 | 4744 init_spell_chartab(); |
| 4745 | |
| 4746 /* Unload all allocated memory. */ | |
| 355 | 4747 spell_free_all(); |
| 236 | 4748 |
| 4749 /* Go through all buffers and handle 'spelllang'. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4750 for (wp = firstwin; wp != NULL; wp = wp->w_next) |
| 236 | 4751 { |
| 316 | 4752 /* Only load the wordlists when 'spelllang' is set and there is a |
| 4753 * window for this buffer in which 'spell' is set. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4754 if (*wp->w_s->b_p_spl != NUL) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4755 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4756 if (wp->w_p_spell) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4757 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4758 (void)did_set_spelllang(wp); |
| 316 | 4759 # ifdef FEAT_WINDOWS |
| 4760 break; | |
| 4761 # endif | |
| 4762 } | |
| 4763 } | |
| 236 | 4764 } |
| 4765 } | |
| 4766 # endif | |
| 4767 | |
| 310 | 4768 /* |
| 4769 * Reload the spell file "fname" if it's loaded. | |
| 4770 */ | |
| 4771 static void | |
| 323 | 4772 spell_reload_one(fname, added_word) |
| 310 | 4773 char_u *fname; |
| 323 | 4774 int added_word; /* invoked through "zg" */ |
| 310 | 4775 { |
| 503 | 4776 slang_T *slang; |
| 323 | 4777 int didit = FALSE; |
| 310 | 4778 |
| 503 | 4779 for (slang = first_lang; slang != NULL; slang = slang->sl_next) |
| 4780 { | |
| 4781 if (fullpathcmp(fname, slang->sl_fname, FALSE) == FPC_SAME) | |
| 4782 { | |
| 4783 slang_clear(slang); | |
| 4784 if (spell_load_file(fname, NULL, slang, FALSE) == NULL) | |
| 500 | 4785 /* reloading failed, clear the language */ |
| 503 | 4786 slang_clear(slang); |
| 737 | 4787 redraw_all_later(SOME_VALID); |
| 323 | 4788 didit = TRUE; |
| 310 | 4789 } |
| 500 | 4790 } |
| 323 | 4791 |
| 4792 /* When "zg" was used and the file wasn't loaded yet, should redo | |
| 626 | 4793 * 'spelllang' to load it now. */ |
| 323 | 4794 if (added_word && !didit) |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
4795 did_set_spelllang(curwin); |
| 310 | 4796 } |
| 4797 | |
| 4798 | |
| 236 | 4799 /* |
| 4800 * Functions for ":mkspell". | |
| 4801 */ | |
| 4802 | |
| 300 | 4803 #define MAXLINELEN 500 /* Maximum length in bytes of a line in a .aff |
| 236 | 4804 and .dic file. */ |
| 4805 /* | |
| 4806 * Main structure to store the contents of a ".aff" file. | |
| 4807 */ | |
| 4808 typedef struct afffile_S | |
| 4809 { | |
| 4810 char_u *af_enc; /* "SET", normalized, alloc'ed string or NULL */ | |
| 498 | 4811 int af_flagtype; /* AFT_CHAR, AFT_LONG, AFT_NUM or AFT_CAPLONG */ |
| 621 | 4812 unsigned af_rare; /* RARE ID for rare word */ |
| 4813 unsigned af_keepcase; /* KEEPCASE ID for keep-case word */ | |
| 495 | 4814 unsigned af_bad; /* BAD ID for banned word */ |
| 4815 unsigned af_needaffix; /* NEEDAFFIX ID */ | |
| 820 | 4816 unsigned af_circumfix; /* CIRCUMFIX ID */ |
| 500 | 4817 unsigned af_needcomp; /* NEEDCOMPOUND ID */ |
| 809 | 4818 unsigned af_comproot; /* COMPOUNDROOT ID */ |
| 4819 unsigned af_compforbid; /* COMPOUNDFORBIDFLAG ID */ | |
| 4820 unsigned af_comppermit; /* COMPOUNDPERMITFLAG ID */ | |
| 698 | 4821 unsigned af_nosuggest; /* NOSUGGEST ID */ |
| 809 | 4822 int af_pfxpostpone; /* postpone prefixes without chop string and |
| 4823 without flags */ | |
| 236 | 4824 hashtab_T af_pref; /* hashtable for prefixes, affheader_T */ |
| 4825 hashtab_T af_suff; /* hashtable for suffixes, affheader_T */ | |
| 495 | 4826 hashtab_T af_comp; /* hashtable for compound flags, compitem_T */ |
| 236 | 4827 } afffile_T; |
| 4828 | |
| 495 | 4829 #define AFT_CHAR 0 /* flags are one character */ |
| 498 | 4830 #define AFT_LONG 1 /* flags are two characters */ |
| 4831 #define AFT_CAPLONG 2 /* flags are one or two characters */ | |
| 4832 #define AFT_NUM 3 /* flags are numbers, comma separated */ | |
| 495 | 4833 |
| 236 | 4834 typedef struct affentry_S affentry_T; |
| 4835 /* Affix entry from ".aff" file. Used for prefixes and suffixes. */ | |
| 4836 struct affentry_S | |
| 4837 { | |
| 4838 affentry_T *ae_next; /* next affix with same name/number */ | |
| 4839 char_u *ae_chop; /* text to chop off basic word (can be NULL) */ | |
| 4840 char_u *ae_add; /* text to add to basic word (can be NULL) */ | |
| 809 | 4841 char_u *ae_flags; /* flags on the affix (can be NULL) */ |
| 236 | 4842 char_u *ae_cond; /* condition (NULL for ".") */ |
| 4843 regprog_T *ae_prog; /* regexp program for ae_cond or NULL */ | |
| 822 | 4844 char ae_compforbid; /* COMPOUNDFORBIDFLAG found */ |
| 4845 char ae_comppermit; /* COMPOUNDPERMITFLAG found */ | |
| 300 | 4846 }; |
| 4847 | |
| 495 | 4848 #ifdef FEAT_MBYTE |
| 4849 # define AH_KEY_LEN 17 /* 2 x 8 bytes + NUL */ | |
| 4850 #else | |
| 498 | 4851 # define AH_KEY_LEN 7 /* 6 digits + NUL */ |
| 495 | 4852 #endif |
| 455 | 4853 |
| 300 | 4854 /* Affix header from ".aff" file. Used for af_pref and af_suff. */ |
| 4855 typedef struct affheader_S | |
| 4856 { | |
| 495 | 4857 char_u ah_key[AH_KEY_LEN]; /* key for hashtab == name of affix */ |
| 4858 unsigned ah_flag; /* affix name as number, uses "af_flagtype" */ | |
| 4859 int ah_newID; /* prefix ID after renumbering; 0 if not used */ | |
| 300 | 4860 int ah_combine; /* suffix may combine with prefix */ |
| 498 | 4861 int ah_follows; /* another affix block should be following */ |
| 300 | 4862 affentry_T *ah_first; /* first affix entry */ |
| 4863 } affheader_T; | |
| 4864 | |
| 4865 #define HI2AH(hi) ((affheader_T *)(hi)->hi_key) | |
| 4866 | |
| 495 | 4867 /* Flag used in compound items. */ |
| 4868 typedef struct compitem_S | |
| 4869 { | |
| 4870 char_u ci_key[AH_KEY_LEN]; /* key for hashtab == name of compound */ | |
| 4871 unsigned ci_flag; /* affix name as number, uses "af_flagtype" */ | |
| 4872 int ci_newID; /* affix ID after renumbering. */ | |
| 4873 } compitem_T; | |
| 4874 | |
| 4875 #define HI2CI(hi) ((compitem_T *)(hi)->hi_key) | |
| 4876 | |
| 300 | 4877 /* |
| 4878 * Structure that is used to store the items in the word tree. This avoids | |
| 474 | 4879 * the need to keep track of each allocated thing, everything is freed all at |
| 4880 * once after ":mkspell" is done. | |
| 1808 | 4881 * Note: "sb_next" must be just before "sb_data" to make sure the alignment of |
| 4882 * "sb_data" is correct for systems where pointers must be aligned on | |
| 4883 * pointer-size boundaries and sizeof(pointer) > sizeof(int) (e.g., Sparc). | |
| 300 | 4884 */ |
| 4885 #define SBLOCKSIZE 16000 /* size of sb_data */ | |
| 4886 typedef struct sblock_S sblock_T; | |
| 4887 struct sblock_S | |
| 4888 { | |
| 1808 | 4889 int sb_used; /* nr of bytes already in use */ |
| 300 | 4890 sblock_T *sb_next; /* next block in list */ |
| 4891 char_u sb_data[1]; /* data, actually longer */ | |
| 236 | 4892 }; |
| 4893 | |
| 4894 /* | |
| 300 | 4895 * A node in the tree. |
| 236 | 4896 */ |
| 300 | 4897 typedef struct wordnode_S wordnode_T; |
| 4898 struct wordnode_S | |
| 236 | 4899 { |
| 346 | 4900 union /* shared to save space */ |
| 4901 { | |
| 469 | 4902 char_u hashkey[6]; /* the hash key, only used while compressing */ |
| 346 | 4903 int index; /* index in written nodes (valid after first |
| 4904 round) */ | |
| 4905 } wn_u1; | |
| 4906 union /* shared to save space */ | |
| 4907 { | |
| 4908 wordnode_T *next; /* next node with same hash key */ | |
| 4909 wordnode_T *wnode; /* parent node that will write this node */ | |
| 4910 } wn_u2; | |
| 300 | 4911 wordnode_T *wn_child; /* child (next byte in word) */ |
| 4912 wordnode_T *wn_sibling; /* next sibling (alternate byte in word, | |
| 4913 always sorted) */ | |
| 474 | 4914 int wn_refs; /* Nr. of references to this node. Only |
| 4915 relevant for first node in a list of | |
| 4916 siblings, in following siblings it is | |
| 4917 always one. */ | |
| 300 | 4918 char_u wn_byte; /* Byte for this node. NUL for word end */ |
| 625 | 4919 |
| 4920 /* Info for when "wn_byte" is NUL. | |
| 4921 * In PREFIXTREE "wn_region" is used for the prefcondnr. | |
| 4922 * In the soundfolded word tree "wn_flags" has the MSW of the wordnr and | |
| 4923 * "wn_region" the LSW of the wordnr. */ | |
| 4924 char_u wn_affixID; /* supported/required prefix ID or 0 */ | |
| 4925 short_u wn_flags; /* WF_ flags */ | |
| 4926 short wn_region; /* region mask */ | |
| 4927 | |
| 469 | 4928 #ifdef SPELL_PRINTTREE |
| 4929 int wn_nr; /* sequence nr for printing */ | |
| 4930 #endif | |
| 236 | 4931 }; |
| 4932 | |
| 390 | 4933 #define WN_MASK 0xffff /* mask relevant bits of "wn_flags" */ |
| 4934 | |
| 300 | 4935 #define HI2WN(hi) (wordnode_T *)((hi)->hi_key) |
| 236 | 4936 |
| 300 | 4937 /* |
| 4938 * Info used while reading the spell files. | |
| 4939 */ | |
| 4940 typedef struct spellinfo_S | |
| 249 | 4941 { |
| 300 | 4942 wordnode_T *si_foldroot; /* tree with case-folded words */ |
| 334 | 4943 long si_foldwcount; /* nr of words in si_foldroot */ |
| 474 | 4944 |
| 300 | 4945 wordnode_T *si_keeproot; /* tree with keep-case words */ |
| 334 | 4946 long si_keepwcount; /* nr of words in si_keeproot */ |
| 474 | 4947 |
| 339 | 4948 wordnode_T *si_prefroot; /* tree with postponed prefixes */ |
| 474 | 4949 |
| 625 | 4950 long si_sugtree; /* creating the soundfolding trie */ |
| 4951 | |
| 300 | 4952 sblock_T *si_blocks; /* memory blocks used */ |
| 482 | 4953 long si_blocks_cnt; /* memory blocks allocated */ |
| 2691 | 4954 int si_did_emsg; /* TRUE when ran out of memory */ |
| 4955 | |
| 482 | 4956 long si_compress_cnt; /* words to add before lowering |
| 4957 compression limit */ | |
| 474 | 4958 wordnode_T *si_first_free; /* List of nodes that have been freed during |
| 4959 compression, linked by "wn_child" field. */ | |
| 482 | 4960 long si_free_count; /* number of nodes in si_first_free */ |
| 474 | 4961 #ifdef SPELL_PRINTTREE |
| 4962 int si_wordnode_nr; /* sequence nr for nodes */ | |
| 4963 #endif | |
| 625 | 4964 buf_T *si_spellbuf; /* buffer used to store soundfold word table */ |
| 474 | 4965 |
| 300 | 4966 int si_ascii; /* handling only ASCII words */ |
| 310 | 4967 int si_add; /* addition file */ |
| 351 | 4968 int si_clear_chartab; /* when TRUE clear char tables */ |
| 300 | 4969 int si_region; /* region mask */ |
| 4970 vimconv_T si_conv; /* for conversion to 'encoding' */ | |
| 302 | 4971 int si_memtot; /* runtime memory used */ |
| 310 | 4972 int si_verbose; /* verbose messages */ |
| 474 | 4973 int si_msg_count; /* number of words added since last message */ |
| 714 | 4974 char_u *si_info; /* info text chars or NULL */ |
| 316 | 4975 int si_region_count; /* number of regions supported (1 when there |
| 4976 are no regions) */ | |
| 2593 | 4977 char_u si_region_name[17]; /* region names; used only if |
| 483 | 4978 * si_region_count > 1) */ |
| 323 | 4979 |
| 4980 garray_T si_rep; /* list of fromto_T entries from REP lines */ | |
| 625 | 4981 garray_T si_repsal; /* list of fromto_T entries from REPSAL lines */ |
| 323 | 4982 garray_T si_sal; /* list of fromto_T entries from SAL lines */ |
| 375 | 4983 char_u *si_sofofr; /* SOFOFROM text */ |
| 4984 char_u *si_sofoto; /* SOFOTO text */ | |
| 625 | 4985 int si_nosugfile; /* NOSUGFILE item found */ |
| 698 | 4986 int si_nosplitsugs; /* NOSPLITSUGS item found */ |
| 323 | 4987 int si_followup; /* soundsalike: ? */ |
| 4988 int si_collapse; /* soundsalike: ? */ | |
| 625 | 4989 hashtab_T si_commonwords; /* hashtable for common words */ |
| 4990 time_t si_sugtime; /* timestamp for .sug file */ | |
| 323 | 4991 int si_rem_accents; /* soundsalike: remove accents */ |
| 4992 garray_T si_map; /* MAP info concatenated */ | |
| 495 | 4993 char_u *si_midword; /* MIDWORD chars or NULL */ |
| 483 | 4994 int si_compmax; /* max nr of words for compounding */ |
| 481 | 4995 int si_compminlen; /* minimal length for compounding */ |
| 483 | 4996 int si_compsylmax; /* max nr of syllables for compounding */ |
| 809 | 4997 int si_compoptions; /* COMP_ flags */ |
| 4998 garray_T si_comppat; /* CHECKCOMPOUNDPATTERN items, each stored as | |
| 4999 a string */ | |
| 481 | 5000 char_u *si_compflags; /* flags used for compounding */ |
| 492 | 5001 char_u si_nobreak; /* NOBREAK */ |
| 483 | 5002 char_u *si_syllable; /* syllable string */ |
| 339 | 5003 garray_T si_prefcond; /* table with conditions for postponed |
| 5004 * prefixes, each stored as a string */ | |
| 495 | 5005 int si_newprefID; /* current value for ah_newID */ |
| 500 | 5006 int si_newcompID; /* current value for compound ID */ |
| 300 | 5007 } spellinfo_T; |
| 249 | 5008 |
| 474 | 5009 static afffile_T *spell_read_aff __ARGS((spellinfo_T *spin, char_u *fname)); |
| 1762 | 5010 static int is_aff_rule __ARGS((char_u **items, int itemcnt, char *rulename, int mincount)); |
| 822 | 5011 static void aff_process_flags __ARGS((afffile_T *affile, affentry_T *entry)); |
| 714 | 5012 static int spell_info_item __ARGS((char_u *s)); |
| 495 | 5013 static unsigned affitem2flag __ARGS((int flagtype, char_u *item, char_u *fname, int lnum)); |
| 5014 static unsigned get_affitem __ARGS((int flagtype, char_u **pp)); | |
| 5015 static void process_compflags __ARGS((spellinfo_T *spin, afffile_T *aff, char_u *compflags)); | |
| 500 | 5016 static void check_renumber __ARGS((spellinfo_T *spin)); |
| 495 | 5017 static int flag_in_afflist __ARGS((int flagtype, char_u *afflist, unsigned flag)); |
| 5018 static void aff_check_number __ARGS((int spinval, int affval, char *name)); | |
| 5019 static void aff_check_string __ARGS((char_u *spinval, char_u *affval, char *name)); | |
| 339 | 5020 static int str_equal __ARGS((char_u *s1, char_u *s2)); |
| 323 | 5021 static void add_fromto __ARGS((spellinfo_T *spin, garray_T *gap, char_u *from, char_u *to)); |
| 5022 static int sal_to_bool __ARGS((char_u *s)); | |
| 240 | 5023 static int has_non_ascii __ARGS((char_u *s)); |
| 300 | 5024 static void spell_free_aff __ARGS((afffile_T *aff)); |
| 474 | 5025 static int spell_read_dic __ARGS((spellinfo_T *spin, char_u *fname, afffile_T *affile)); |
| 820 | 5026 static int get_affix_flags __ARGS((afffile_T *affile, char_u *afflist)); |
| 483 | 5027 static int get_pfxlist __ARGS((afffile_T *affile, char_u *afflist, char_u *store_afflist)); |
| 495 | 5028 static void get_compflags __ARGS((afffile_T *affile, char_u *afflist, char_u *store_afflist)); |
| 820 | 5029 static int store_aff_word __ARGS((spellinfo_T *spin, char_u *word, char_u *afflist, afffile_T *affile, hashtab_T *ht, hashtab_T *xht, int condit, int flags, char_u *pfxlist, int pfxlen)); |
| 474 | 5030 static int spell_read_wordfile __ARGS((spellinfo_T *spin, char_u *fname)); |
| 5031 static void *getroom __ARGS((spellinfo_T *spin, size_t len, int align)); | |
| 5032 static char_u *getroom_save __ARGS((spellinfo_T *spin, char_u *s)); | |
| 300 | 5033 static void free_blocks __ARGS((sblock_T *bl)); |
| 474 | 5034 static wordnode_T *wordtree_alloc __ARGS((spellinfo_T *spin)); |
| 483 | 5035 static int store_word __ARGS((spellinfo_T *spin, char_u *word, int flags, int region, char_u *pfxlist, int need_affix)); |
| 481 | 5036 static int tree_add_word __ARGS((spellinfo_T *spin, char_u *word, wordnode_T *tree, int flags, int region, int affixID)); |
| 474 | 5037 static wordnode_T *get_wordnode __ARGS((spellinfo_T *spin)); |
| 625 | 5038 static int deref_wordnode __ARGS((spellinfo_T *spin, wordnode_T *node)); |
| 474 | 5039 static void free_wordnode __ARGS((spellinfo_T *spin, wordnode_T *n)); |
| 5040 static void wordtree_compress __ARGS((spellinfo_T *spin, wordnode_T *root)); | |
| 5041 static int node_compress __ARGS((spellinfo_T *spin, wordnode_T *node, hashtab_T *ht, int *tot)); | |
| 300 | 5042 static int node_equal __ARGS((wordnode_T *n1, wordnode_T *n2)); |
| 500 | 5043 static int write_vim_spell __ARGS((spellinfo_T *spin, char_u *fname)); |
| 346 | 5044 static void clear_node __ARGS((wordnode_T *node)); |
| 1757 | 5045 static int put_node __ARGS((FILE *fd, wordnode_T *node, int idx, int regionmask, int prefixtree)); |
| 625 | 5046 static void spell_make_sugfile __ARGS((spellinfo_T *spin, char_u *wfname)); |
| 5047 static int sug_filltree __ARGS((spellinfo_T *spin, slang_T *slang)); | |
| 5048 static int sug_maketable __ARGS((spellinfo_T *spin)); | |
| 5049 static int sug_filltable __ARGS((spellinfo_T *spin, wordnode_T *node, int startwordnr, garray_T *gap)); | |
| 5050 static int offset2bytes __ARGS((int nr, char_u *buf)); | |
| 5051 static int bytes2offset __ARGS((char_u **pp)); | |
| 5052 static void sug_write __ARGS((spellinfo_T *spin, char_u *fname)); | |
| 323 | 5053 static void mkspell __ARGS((int fcount, char_u **fnames, int ascii, int overwrite, int added_word)); |
| 625 | 5054 static void spell_message __ARGS((spellinfo_T *spin, char_u *str)); |
| 310 | 5055 static void init_spellfile __ARGS((void)); |
| 236 | 5056 |
| 455 | 5057 /* In the postponed prefixes tree wn_flags is used to store the WFP_ flags, |
| 5058 * but it must be negative to indicate the prefix tree to tree_add_word(). | |
| 5059 * Use a negative number with the lower 8 bits zero. */ | |
| 5060 #define PFX_FLAGS -256 | |
| 390 | 5061 |
| 820 | 5062 /* flags for "condit" argument of store_aff_word() */ |
| 5063 #define CONDIT_COMB 1 /* affix must combine */ | |
| 5064 #define CONDIT_CFIX 2 /* affix must have CIRCUMFIX flag */ | |
| 5065 #define CONDIT_SUF 4 /* add a suffix for matching flags */ | |
| 5066 #define CONDIT_AFF 8 /* word already has an affix */ | |
| 5067 | |
| 483 | 5068 /* |
| 5069 * Tunable parameters for when the tree is compressed. See 'mkspellmem'. | |
| 5070 */ | |
| 5071 static long compress_start = 30000; /* memory / SBLOCKSIZE */ | |
| 5072 static long compress_inc = 100; /* memory / SBLOCKSIZE */ | |
| 5073 static long compress_added = 500000; /* word count */ | |
| 5074 | |
| 469 | 5075 #ifdef SPELL_PRINTTREE |
| 5076 /* | |
| 5077 * For debugging the tree code: print the current tree in a (more or less) | |
| 5078 * readable format, so that we can see what happens when adding a word and/or | |
| 5079 * compressing the tree. | |
| 5080 * Based on code from Olaf Seibert. | |
| 5081 */ | |
| 5082 #define PRINTLINESIZE 1000 | |
| 5083 #define PRINTWIDTH 6 | |
| 5084 | |
| 5085 #define PRINTSOME(l, depth, fmt, a1, a2) vim_snprintf(l + depth * PRINTWIDTH, \ | |
| 5086 PRINTLINESIZE - PRINTWIDTH * depth, fmt, a1, a2) | |
| 5087 | |
| 5088 static char line1[PRINTLINESIZE]; | |
| 5089 static char line2[PRINTLINESIZE]; | |
| 5090 static char line3[PRINTLINESIZE]; | |
| 5091 | |
| 5092 static void | |
| 5093 spell_clear_flags(wordnode_T *node) | |
| 5094 { | |
| 5095 wordnode_T *np; | |
| 5096 | |
| 5097 for (np = node; np != NULL; np = np->wn_sibling) | |
| 5098 { | |
| 5099 np->wn_u1.index = FALSE; | |
| 5100 spell_clear_flags(np->wn_child); | |
| 5101 } | |
| 5102 } | |
| 5103 | |
| 5104 static void | |
| 5105 spell_print_node(wordnode_T *node, int depth) | |
| 5106 { | |
| 5107 if (node->wn_u1.index) | |
| 5108 { | |
| 5109 /* Done this node before, print the reference. */ | |
| 5110 PRINTSOME(line1, depth, "(%d)", node->wn_nr, 0); | |
| 5111 PRINTSOME(line2, depth, " ", 0, 0); | |
| 5112 PRINTSOME(line3, depth, " ", 0, 0); | |
| 5113 msg(line1); | |
| 5114 msg(line2); | |
| 5115 msg(line3); | |
| 5116 } | |
| 5117 else | |
| 5118 { | |
| 5119 node->wn_u1.index = TRUE; | |
| 5120 | |
| 5121 if (node->wn_byte != NUL) | |
| 5122 { | |
| 5123 if (node->wn_child != NULL) | |
| 5124 PRINTSOME(line1, depth, " %c -> ", node->wn_byte, 0); | |
| 5125 else | |
| 5126 /* Cannot happen? */ | |
| 5127 PRINTSOME(line1, depth, " %c ???", node->wn_byte, 0); | |
| 5128 } | |
| 5129 else | |
| 5130 PRINTSOME(line1, depth, " $ ", 0, 0); | |
| 5131 | |
| 5132 PRINTSOME(line2, depth, "%d/%d ", node->wn_nr, node->wn_refs); | |
| 5133 | |
| 5134 if (node->wn_sibling != NULL) | |
| 5135 PRINTSOME(line3, depth, " | ", 0, 0); | |
| 5136 else | |
| 5137 PRINTSOME(line3, depth, " ", 0, 0); | |
| 5138 | |
| 5139 if (node->wn_byte == NUL) | |
| 5140 { | |
| 5141 msg(line1); | |
| 5142 msg(line2); | |
| 5143 msg(line3); | |
| 5144 } | |
| 5145 | |
| 5146 /* do the children */ | |
| 5147 if (node->wn_byte != NUL && node->wn_child != NULL) | |
| 5148 spell_print_node(node->wn_child, depth + 1); | |
| 5149 | |
| 5150 /* do the siblings */ | |
| 5151 if (node->wn_sibling != NULL) | |
| 5152 { | |
| 5153 /* get rid of all parent details except | */ | |
| 5154 STRCPY(line1, line3); | |
| 5155 STRCPY(line2, line3); | |
| 5156 spell_print_node(node->wn_sibling, depth); | |
| 5157 } | |
| 5158 } | |
| 5159 } | |
| 5160 | |
| 5161 static void | |
| 5162 spell_print_tree(wordnode_T *root) | |
| 5163 { | |
| 5164 if (root != NULL) | |
| 5165 { | |
| 5166 /* Clear the "wn_u1.index" fields, used to remember what has been | |
| 5167 * done. */ | |
| 5168 spell_clear_flags(root); | |
| 5169 | |
| 5170 /* Recursively print the tree. */ | |
| 5171 spell_print_node(root, 0); | |
| 5172 } | |
| 5173 } | |
| 5174 #endif /* SPELL_PRINTTREE */ | |
| 5175 | |
| 236 | 5176 /* |
| 323 | 5177 * Read the affix file "fname". |
| 316 | 5178 * Returns an afffile_T, NULL for complete failure. |
| 236 | 5179 */ |
| 5180 static afffile_T * | |
| 474 | 5181 spell_read_aff(spin, fname) |
| 5182 spellinfo_T *spin; | |
| 236 | 5183 char_u *fname; |
| 5184 { | |
| 5185 FILE *fd; | |
| 5186 afffile_T *aff; | |
| 5187 char_u rline[MAXLINELEN]; | |
| 5188 char_u *line; | |
| 5189 char_u *pc = NULL; | |
| 625 | 5190 #define MAXITEMCNT 30 |
| 334 | 5191 char_u *(items[MAXITEMCNT]); |
| 236 | 5192 int itemcnt; |
| 5193 char_u *p; | |
| 5194 int lnum = 0; | |
| 5195 affheader_T *cur_aff = NULL; | |
| 495 | 5196 int did_postpone_prefix = FALSE; |
| 236 | 5197 int aff_todo = 0; |
| 5198 hashtab_T *tp; | |
| 255 | 5199 char_u *low = NULL; |
| 5200 char_u *fol = NULL; | |
| 5201 char_u *upp = NULL; | |
| 323 | 5202 int do_rep; |
| 625 | 5203 int do_repsal; |
| 323 | 5204 int do_sal; |
| 944 | 5205 int do_mapline; |
| 323 | 5206 int found_map = FALSE; |
| 324 | 5207 hashitem_T *hi; |
| 455 | 5208 int l; |
| 495 | 5209 int compminlen = 0; /* COMPOUNDMIN value */ |
| 5210 int compsylmax = 0; /* COMPOUNDSYLMAX value */ | |
| 809 | 5211 int compoptions = 0; /* COMP_ flags */ |
| 5212 int compmax = 0; /* COMPOUNDWORDMAX value */ | |
| 714 | 5213 char_u *compflags = NULL; /* COMPOUNDFLAG and COMPOUNDRULE |
| 495 | 5214 concatenated */ |
| 5215 char_u *midword = NULL; /* MIDWORD value */ | |
| 5216 char_u *syllable = NULL; /* SYLLABLE value */ | |
| 5217 char_u *sofofrom = NULL; /* SOFOFROM value */ | |
| 5218 char_u *sofoto = NULL; /* SOFOTO value */ | |
| 236 | 5219 |
| 300 | 5220 /* |
| 5221 * Open the file. | |
| 5222 */ | |
| 310 | 5223 fd = mch_fopen((char *)fname, "r"); |
| 236 | 5224 if (fd == NULL) |
| 5225 { | |
| 5226 EMSG2(_(e_notopen), fname); | |
| 5227 return NULL; | |
| 5228 } | |
| 5229 | |
| 625 | 5230 vim_snprintf((char *)IObuff, IOSIZE, _("Reading affix file %s ..."), fname); |
| 5231 spell_message(spin, IObuff); | |
| 236 | 5232 |
| 323 | 5233 /* Only do REP lines when not done in another .aff file already. */ |
| 5234 do_rep = spin->si_rep.ga_len == 0; | |
| 5235 | |
| 625 | 5236 /* Only do REPSAL lines when not done in another .aff file already. */ |
| 5237 do_repsal = spin->si_repsal.ga_len == 0; | |
| 5238 | |
| 323 | 5239 /* Only do SAL lines when not done in another .aff file already. */ |
| 5240 do_sal = spin->si_sal.ga_len == 0; | |
| 5241 | |
| 5242 /* Only do MAP lines when not done in another .aff file already. */ | |
| 944 | 5243 do_mapline = spin->si_map.ga_len == 0; |
| 323 | 5244 |
| 300 | 5245 /* |
| 5246 * Allocate and init the afffile_T structure. | |
| 5247 */ | |
| 474 | 5248 aff = (afffile_T *)getroom(spin, sizeof(afffile_T), TRUE); |
| 236 | 5249 if (aff == NULL) |
| 840 | 5250 { |
| 5251 fclose(fd); | |
| 236 | 5252 return NULL; |
| 840 | 5253 } |
| 236 | 5254 hash_init(&aff->af_pref); |
| 5255 hash_init(&aff->af_suff); | |
| 495 | 5256 hash_init(&aff->af_comp); |
| 236 | 5257 |
| 5258 /* | |
| 5259 * Read all the lines in the file one by one. | |
| 5260 */ | |
| 255 | 5261 while (!vim_fgets(rline, MAXLINELEN, fd) && !got_int) |
| 236 | 5262 { |
| 255 | 5263 line_breakcheck(); |
| 236 | 5264 ++lnum; |
| 5265 | |
| 5266 /* Skip comment lines. */ | |
| 5267 if (*rline == '#') | |
| 5268 continue; | |
| 5269 | |
| 5270 /* Convert from "SET" to 'encoding' when needed. */ | |
| 5271 vim_free(pc); | |
| 310 | 5272 #ifdef FEAT_MBYTE |
| 300 | 5273 if (spin->si_conv.vc_type != CONV_NONE) |
| 236 | 5274 { |
| 300 | 5275 pc = string_convert(&spin->si_conv, rline, NULL); |
| 255 | 5276 if (pc == NULL) |
| 5277 { | |
| 5278 smsg((char_u *)_("Conversion failure for word in %s line %d: %s"), | |
| 5279 fname, lnum, rline); | |
| 5280 continue; | |
| 5281 } | |
| 236 | 5282 line = pc; |
| 5283 } | |
| 5284 else | |
| 310 | 5285 #endif |
| 236 | 5286 { |
| 5287 pc = NULL; | |
| 5288 line = rline; | |
| 5289 } | |
| 5290 | |
| 5291 /* Split the line up in white separated items. Put a NUL after each | |
| 5292 * item. */ | |
| 5293 itemcnt = 0; | |
| 5294 for (p = line; ; ) | |
| 5295 { | |
| 5296 while (*p != NUL && *p <= ' ') /* skip white space and CR/NL */ | |
| 5297 ++p; | |
| 5298 if (*p == NUL) | |
| 5299 break; | |
| 334 | 5300 if (itemcnt == MAXITEMCNT) /* too many items */ |
| 300 | 5301 break; |
| 236 | 5302 items[itemcnt++] = p; |
| 714 | 5303 /* A few items have arbitrary text argument, don't split them. */ |
| 5304 if (itemcnt == 2 && spell_info_item(items[0])) | |
| 5305 while (*p >= ' ' || *p == TAB) /* skip until CR/NL */ | |
| 5306 ++p; | |
| 5307 else | |
| 5308 while (*p > ' ') /* skip until white space or CR/NL */ | |
| 5309 ++p; | |
| 236 | 5310 if (*p == NUL) |
| 5311 break; | |
| 5312 *p++ = NUL; | |
| 5313 } | |
| 5314 | |
| 5315 /* Handle non-empty lines. */ | |
| 5316 if (itemcnt > 0) | |
| 5317 { | |
| 1762 | 5318 if (is_aff_rule(items, itemcnt, "SET", 2) && aff->af_enc == NULL) |
| 236 | 5319 { |
| 310 | 5320 #ifdef FEAT_MBYTE |
| 300 | 5321 /* Setup for conversion from "ENC" to 'encoding'. */ |
| 5322 aff->af_enc = enc_canonize(items[1]); | |
| 5323 if (aff->af_enc != NULL && !spin->si_ascii | |
| 5324 && convert_setup(&spin->si_conv, aff->af_enc, | |
| 5325 p_enc) == FAIL) | |
| 5326 smsg((char_u *)_("Conversion in %s not supported: from %s to %s"), | |
| 5327 fname, aff->af_enc, p_enc); | |
| 375 | 5328 spin->si_conv.vc_fail = TRUE; |
| 310 | 5329 #else |
| 5330 smsg((char_u *)_("Conversion in %s not supported"), fname); | |
| 5331 #endif | |
| 236 | 5332 } |
| 1762 | 5333 else if (is_aff_rule(items, itemcnt, "FLAG", 2) |
| 495 | 5334 && aff->af_flagtype == AFT_CHAR) |
| 5335 { | |
| 5336 if (STRCMP(items[1], "long") == 0) | |
| 498 | 5337 aff->af_flagtype = AFT_LONG; |
| 495 | 5338 else if (STRCMP(items[1], "num") == 0) |
| 498 | 5339 aff->af_flagtype = AFT_NUM; |
| 5340 else if (STRCMP(items[1], "caplong") == 0) | |
| 5341 aff->af_flagtype = AFT_CAPLONG; | |
| 495 | 5342 else |
| 5343 smsg((char_u *)_("Invalid value for FLAG in %s line %d: %s"), | |
| 5344 fname, lnum, items[1]); | |
| 621 | 5345 if (aff->af_rare != 0 |
| 5346 || aff->af_keepcase != 0 | |
| 5347 || aff->af_bad != 0 | |
| 500 | 5348 || aff->af_needaffix != 0 |
| 820 | 5349 || aff->af_circumfix != 0 |
| 500 | 5350 || aff->af_needcomp != 0 |
| 809 | 5351 || aff->af_comproot != 0 |
| 698 | 5352 || aff->af_nosuggest != 0 |
| 500 | 5353 || compflags != NULL |
| 495 | 5354 || aff->af_suff.ht_used > 0 |
| 5355 || aff->af_pref.ht_used > 0) | |
| 5356 smsg((char_u *)_("FLAG after using flags in %s line %d: %s"), | |
| 5357 fname, lnum, items[1]); | |
| 5358 } | |
| 714 | 5359 else if (spell_info_item(items[0])) |
| 5360 { | |
| 5361 p = (char_u *)getroom(spin, | |
| 5362 (spin->si_info == NULL ? 0 : STRLEN(spin->si_info)) | |
| 5363 + STRLEN(items[0]) | |
| 5364 + STRLEN(items[1]) + 3, FALSE); | |
| 5365 if (p != NULL) | |
| 5366 { | |
| 5367 if (spin->si_info != NULL) | |
| 5368 { | |
| 5369 STRCPY(p, spin->si_info); | |
| 5370 STRCAT(p, "\n"); | |
| 5371 } | |
| 5372 STRCAT(p, items[0]); | |
| 5373 STRCAT(p, " "); | |
| 5374 STRCAT(p, items[1]); | |
| 5375 spin->si_info = p; | |
| 5376 } | |
| 5377 } | |
| 1762 | 5378 else if (is_aff_rule(items, itemcnt, "MIDWORD", 2) |
| 495 | 5379 && midword == NULL) |
| 5380 { | |
| 5381 midword = getroom_save(spin, items[1]); | |
| 366 | 5382 } |
| 1762 | 5383 else if (is_aff_rule(items, itemcnt, "TRY", 2)) |
| 300 | 5384 { |
| 323 | 5385 /* ignored, we look in the tree for what chars may appear */ |
| 300 | 5386 } |
| 621 | 5387 /* TODO: remove "RAR" later */ |
| 1762 | 5388 else if ((is_aff_rule(items, itemcnt, "RAR", 2) |
| 5389 || is_aff_rule(items, itemcnt, "RARE", 2)) | |
| 5390 && aff->af_rare == 0) | |
| 621 | 5391 { |
| 5392 aff->af_rare = affitem2flag(aff->af_flagtype, items[1], | |
| 495 | 5393 fname, lnum); |
| 307 | 5394 } |
| 621 | 5395 /* TODO: remove "KEP" later */ |
| 1762 | 5396 else if ((is_aff_rule(items, itemcnt, "KEP", 2) |
| 5397 || is_aff_rule(items, itemcnt, "KEEPCASE", 2)) | |
| 621 | 5398 && aff->af_keepcase == 0) |
| 5399 { | |
| 5400 aff->af_keepcase = affitem2flag(aff->af_flagtype, items[1], | |
| 495 | 5401 fname, lnum); |
| 307 | 5402 } |
| 1762 | 5403 else if ((is_aff_rule(items, itemcnt, "BAD", 2) |
| 5404 || is_aff_rule(items, itemcnt, "FORBIDDENWORD", 2)) | |
| 5405 && aff->af_bad == 0) | |
| 346 | 5406 { |
| 495 | 5407 aff->af_bad = affitem2flag(aff->af_flagtype, items[1], |
| 5408 fname, lnum); | |
| 346 | 5409 } |
| 1762 | 5410 else if (is_aff_rule(items, itemcnt, "NEEDAFFIX", 2) |
| 483 | 5411 && aff->af_needaffix == 0) |
| 5412 { | |
| 495 | 5413 aff->af_needaffix = affitem2flag(aff->af_flagtype, items[1], |
| 5414 fname, lnum); | |
| 483 | 5415 } |
| 1762 | 5416 else if (is_aff_rule(items, itemcnt, "CIRCUMFIX", 2) |
| 820 | 5417 && aff->af_circumfix == 0) |
| 5418 { | |
| 5419 aff->af_circumfix = affitem2flag(aff->af_flagtype, items[1], | |
| 5420 fname, lnum); | |
| 5421 } | |
| 1762 | 5422 else if (is_aff_rule(items, itemcnt, "NOSUGGEST", 2) |
| 698 | 5423 && aff->af_nosuggest == 0) |
| 5424 { | |
| 5425 aff->af_nosuggest = affitem2flag(aff->af_flagtype, items[1], | |
| 5426 fname, lnum); | |
| 5427 } | |
| 1762 | 5428 else if ((is_aff_rule(items, itemcnt, "NEEDCOMPOUND", 2) |
| 5429 || is_aff_rule(items, itemcnt, "ONLYINCOMPOUND", 2)) | |
| 500 | 5430 && aff->af_needcomp == 0) |
| 5431 { | |
| 5432 aff->af_needcomp = affitem2flag(aff->af_flagtype, items[1], | |
| 5433 fname, lnum); | |
| 5434 } | |
| 1762 | 5435 else if (is_aff_rule(items, itemcnt, "COMPOUNDROOT", 2) |
| 809 | 5436 && aff->af_comproot == 0) |
| 5437 { | |
| 5438 aff->af_comproot = affitem2flag(aff->af_flagtype, items[1], | |
| 5439 fname, lnum); | |
| 5440 } | |
| 1762 | 5441 else if (is_aff_rule(items, itemcnt, "COMPOUNDFORBIDFLAG", 2) |
| 5442 && aff->af_compforbid == 0) | |
| 809 | 5443 { |
| 5444 aff->af_compforbid = affitem2flag(aff->af_flagtype, items[1], | |
| 5445 fname, lnum); | |
| 822 | 5446 if (aff->af_pref.ht_used > 0) |
| 5447 smsg((char_u *)_("Defining COMPOUNDFORBIDFLAG after PFX item may give wrong results in %s line %d"), | |
| 5448 fname, lnum); | |
| 809 | 5449 } |
| 1762 | 5450 else if (is_aff_rule(items, itemcnt, "COMPOUNDPERMITFLAG", 2) |
| 5451 && aff->af_comppermit == 0) | |
| 809 | 5452 { |
| 5453 aff->af_comppermit = affitem2flag(aff->af_flagtype, items[1], | |
| 5454 fname, lnum); | |
| 822 | 5455 if (aff->af_pref.ht_used > 0) |
| 5456 smsg((char_u *)_("Defining COMPOUNDPERMITFLAG after PFX item may give wrong results in %s line %d"), | |
| 5457 fname, lnum); | |
| 809 | 5458 } |
| 1762 | 5459 else if (is_aff_rule(items, itemcnt, "COMPOUNDFLAG", 2) |
| 495 | 5460 && compflags == NULL) |
| 5461 { | |
| 714 | 5462 /* Turn flag "c" into COMPOUNDRULE compatible string "c+", |
| 495 | 5463 * "Na" into "Na+", "1234" into "1234+". */ |
| 5464 p = getroom(spin, STRLEN(items[1]) + 2, FALSE); | |
| 483 | 5465 if (p != NULL) |
| 5466 { | |
| 495 | 5467 STRCPY(p, items[1]); |
| 5468 STRCAT(p, "+"); | |
| 5469 compflags = p; | |
| 5470 } | |
| 481 | 5471 } |
| 1762 | 5472 else if (is_aff_rule(items, itemcnt, "COMPOUNDRULES", 2)) |
| 5473 { | |
| 5474 /* We don't use the count, but do check that it's a number and | |
| 5475 * not COMPOUNDRULE mistyped. */ | |
| 5476 if (atoi((char *)items[1]) == 0) | |
| 5477 smsg((char_u *)_("Wrong COMPOUNDRULES value in %s line %d: %s"), | |
| 5478 fname, lnum, items[1]); | |
| 5479 } | |
| 5480 else if (is_aff_rule(items, itemcnt, "COMPOUNDRULE", 2)) | |
| 483 | 5481 { |
| 2691 | 5482 /* Don't use the first rule if it is a number. */ |
| 5483 if (compflags != NULL || *skipdigits(items[1]) != NUL) | |
| 5484 { | |
| 5485 /* Concatenate this string to previously defined ones, | |
| 5486 * using a slash to separate them. */ | |
| 5487 l = (int)STRLEN(items[1]) + 1; | |
| 495 | 5488 if (compflags != NULL) |
| 2691 | 5489 l += (int)STRLEN(compflags) + 1; |
| 5490 p = getroom(spin, l, FALSE); | |
| 5491 if (p != NULL) | |
| 483 | 5492 { |
| 2691 | 5493 if (compflags != NULL) |
| 5494 { | |
| 5495 STRCPY(p, compflags); | |
| 5496 STRCAT(p, "/"); | |
| 5497 } | |
| 5498 STRCAT(p, items[1]); | |
| 5499 compflags = p; | |
| 483 | 5500 } |
| 5501 } | |
| 5502 } | |
| 1762 | 5503 else if (is_aff_rule(items, itemcnt, "COMPOUNDWORDMAX", 2) |
| 495 | 5504 && compmax == 0) |
| 5505 { | |
| 5506 compmax = atoi((char *)items[1]); | |
| 5507 if (compmax == 0) | |
| 809 | 5508 smsg((char_u *)_("Wrong COMPOUNDWORDMAX value in %s line %d: %s"), |
| 483 | 5509 fname, lnum, items[1]); |
| 481 | 5510 } |
| 1762 | 5511 else if (is_aff_rule(items, itemcnt, "COMPOUNDMIN", 2) |
| 495 | 5512 && compminlen == 0) |
| 5513 { | |
| 5514 compminlen = atoi((char *)items[1]); | |
| 5515 if (compminlen == 0) | |
| 481 | 5516 smsg((char_u *)_("Wrong COMPOUNDMIN value in %s line %d: %s"), |
| 5517 fname, lnum, items[1]); | |
| 5518 } | |
| 1762 | 5519 else if (is_aff_rule(items, itemcnt, "COMPOUNDSYLMAX", 2) |
| 495 | 5520 && compsylmax == 0) |
| 5521 { | |
| 5522 compsylmax = atoi((char *)items[1]); | |
| 5523 if (compsylmax == 0) | |
| 483 | 5524 smsg((char_u *)_("Wrong COMPOUNDSYLMAX value in %s line %d: %s"), |
| 5525 fname, lnum, items[1]); | |
| 5526 } | |
| 1762 | 5527 else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDDUP", 1)) |
| 809 | 5528 { |
| 5529 compoptions |= COMP_CHECKDUP; | |
| 5530 } | |
| 1762 | 5531 else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDREP", 1)) |
| 809 | 5532 { |
| 5533 compoptions |= COMP_CHECKREP; | |
| 5534 } | |
| 1762 | 5535 else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDCASE", 1)) |
| 809 | 5536 { |
| 5537 compoptions |= COMP_CHECKCASE; | |
| 5538 } | |
| 1762 | 5539 else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDTRIPLE", 1)) |
| 809 | 5540 { |
| 5541 compoptions |= COMP_CHECKTRIPLE; | |
| 5542 } | |
| 1762 | 5543 else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDPATTERN", 2)) |
| 809 | 5544 { |
| 5545 if (atoi((char *)items[1]) == 0) | |
| 5546 smsg((char_u *)_("Wrong CHECKCOMPOUNDPATTERN value in %s line %d: %s"), | |
| 5547 fname, lnum, items[1]); | |
| 5548 } | |
| 1762 | 5549 else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDPATTERN", 3)) |
| 809 | 5550 { |
| 5551 garray_T *gap = &spin->si_comppat; | |
| 5552 int i; | |
| 5553 | |
| 5554 /* Only add the couple if it isn't already there. */ | |
| 5555 for (i = 0; i < gap->ga_len - 1; i += 2) | |
| 5556 if (STRCMP(((char_u **)(gap->ga_data))[i], items[1]) == 0 | |
| 5557 && STRCMP(((char_u **)(gap->ga_data))[i + 1], | |
| 5558 items[2]) == 0) | |
| 5559 break; | |
| 5560 if (i >= gap->ga_len && ga_grow(gap, 2) == OK) | |
| 5561 { | |
| 5562 ((char_u **)(gap->ga_data))[gap->ga_len++] | |
| 5563 = getroom_save(spin, items[1]); | |
| 5564 ((char_u **)(gap->ga_data))[gap->ga_len++] | |
| 5565 = getroom_save(spin, items[2]); | |
| 5566 } | |
| 5567 } | |
| 1762 | 5568 else if (is_aff_rule(items, itemcnt, "SYLLABLE", 2) |
| 495 | 5569 && syllable == NULL) |
| 5570 { | |
| 5571 syllable = getroom_save(spin, items[1]); | |
| 483 | 5572 } |
| 1762 | 5573 else if (is_aff_rule(items, itemcnt, "NOBREAK", 1)) |
| 492 | 5574 { |
| 5575 spin->si_nobreak = TRUE; | |
| 5576 } | |
| 1762 | 5577 else if (is_aff_rule(items, itemcnt, "NOSPLITSUGS", 1)) |
| 698 | 5578 { |
| 5579 spin->si_nosplitsugs = TRUE; | |
| 5580 } | |
| 1762 | 5581 else if (is_aff_rule(items, itemcnt, "NOSUGFILE", 1)) |
| 625 | 5582 { |
| 5583 spin->si_nosugfile = TRUE; | |
| 5584 } | |
| 1762 | 5585 else if (is_aff_rule(items, itemcnt, "PFXPOSTPONE", 1)) |
| 339 | 5586 { |
| 5587 aff->af_pfxpostpone = TRUE; | |
| 5588 } | |
| 236 | 5589 else if ((STRCMP(items[0], "PFX") == 0 |
| 5590 || STRCMP(items[0], "SFX") == 0) | |
| 5591 && aff_todo == 0 | |
| 334 | 5592 && itemcnt >= 4) |
| 236 | 5593 { |
| 498 | 5594 int lasti = 4; |
| 5595 char_u key[AH_KEY_LEN]; | |
| 5596 | |
| 5597 if (*items[0] == 'P') | |
| 5598 tp = &aff->af_pref; | |
| 5599 else | |
| 5600 tp = &aff->af_suff; | |
| 5601 | |
| 5602 /* Myspell allows the same affix name to be used multiple | |
| 5603 * times. The affix files that do this have an undocumented | |
| 5604 * "S" flag on all but the last block, thus we check for that | |
| 5605 * and store it in ah_follows. */ | |
| 5606 vim_strncpy(key, items[1], AH_KEY_LEN - 1); | |
| 5607 hi = hash_find(tp, key); | |
| 5608 if (!HASHITEM_EMPTY(hi)) | |
| 5609 { | |
| 5610 cur_aff = HI2AH(hi); | |
| 5611 if (cur_aff->ah_combine != (*items[2] == 'Y')) | |
| 5612 smsg((char_u *)_("Different combining flag in continued affix block in %s line %d: %s"), | |
| 5613 fname, lnum, items[1]); | |
| 5614 if (!cur_aff->ah_follows) | |
| 5615 smsg((char_u *)_("Duplicate affix in %s line %d: %s"), | |
| 5616 fname, lnum, items[1]); | |
| 5617 } | |
| 5618 else | |
| 5619 { | |
| 5620 /* New affix letter. */ | |
| 5621 cur_aff = (affheader_T *)getroom(spin, | |
| 5622 sizeof(affheader_T), TRUE); | |
| 5623 if (cur_aff == NULL) | |
| 5624 break; | |
| 5625 cur_aff->ah_flag = affitem2flag(aff->af_flagtype, items[1], | |
| 5626 fname, lnum); | |
| 5627 if (cur_aff->ah_flag == 0 || STRLEN(items[1]) >= AH_KEY_LEN) | |
| 5628 break; | |
| 5629 if (cur_aff->ah_flag == aff->af_bad | |
| 621 | 5630 || cur_aff->ah_flag == aff->af_rare |
| 5631 || cur_aff->ah_flag == aff->af_keepcase | |
| 500 | 5632 || cur_aff->ah_flag == aff->af_needaffix |
| 820 | 5633 || cur_aff->ah_flag == aff->af_circumfix |
| 698 | 5634 || cur_aff->ah_flag == aff->af_nosuggest |
| 809 | 5635 || cur_aff->ah_flag == aff->af_needcomp |
| 5636 || cur_aff->ah_flag == aff->af_comproot) | |
| 698 | 5637 smsg((char_u *)_("Affix also used for BAD/RARE/KEEPCASE/NEEDAFFIX/NEEDCOMPOUND/NOSUGGEST in %s line %d: %s"), |
| 498 | 5638 fname, lnum, items[1]); |
| 5639 STRCPY(cur_aff->ah_key, items[1]); | |
| 5640 hash_add(tp, cur_aff->ah_key); | |
| 5641 | |
| 5642 cur_aff->ah_combine = (*items[2] == 'Y'); | |
| 5643 } | |
| 5644 | |
| 5645 /* Check for the "S" flag, which apparently means that another | |
| 5646 * block with the same affix name is following. */ | |
| 5647 if (itemcnt > lasti && STRCMP(items[lasti], "S") == 0) | |
| 5648 { | |
| 5649 ++lasti; | |
| 5650 cur_aff->ah_follows = TRUE; | |
| 5651 } | |
| 5652 else | |
| 5653 cur_aff->ah_follows = FALSE; | |
| 5654 | |
| 334 | 5655 /* Myspell allows extra text after the item, but that might |
| 5656 * mean mistakes go unnoticed. Require a comment-starter. */ | |
| 498 | 5657 if (itemcnt > lasti && *items[lasti] != '#') |
| 809 | 5658 smsg((char_u *)_(e_afftrailing), fname, lnum, items[lasti]); |
| 334 | 5659 |
| 498 | 5660 if (STRCMP(items[2], "Y") != 0 && STRCMP(items[2], "N") != 0) |
| 236 | 5661 smsg((char_u *)_("Expected Y or N in %s line %d: %s"), |
| 5662 fname, lnum, items[2]); | |
| 339 | 5663 |
| 498 | 5664 if (*items[0] == 'P' && aff->af_pfxpostpone) |
| 5665 { | |
| 5666 if (cur_aff->ah_newID == 0) | |
| 495 | 5667 { |
| 5668 /* Use a new number in the .spl file later, to be able | |
| 5669 * to handle multiple .aff files. */ | |
| 500 | 5670 check_renumber(spin); |
| 495 | 5671 cur_aff->ah_newID = ++spin->si_newprefID; |
| 5672 | |
| 5673 /* We only really use ah_newID if the prefix is | |
| 5674 * postponed. We know that only after handling all | |
| 5675 * the items. */ | |
| 5676 did_postpone_prefix = FALSE; | |
| 5677 } | |
| 498 | 5678 else |
| 5679 /* Did use the ID in a previous block. */ | |
| 5680 did_postpone_prefix = TRUE; | |
| 5681 } | |
| 5682 | |
| 300 | 5683 aff_todo = atoi((char *)items[3]); |
| 236 | 5684 } |
| 5685 else if ((STRCMP(items[0], "PFX") == 0 | |
| 5686 || STRCMP(items[0], "SFX") == 0) | |
| 5687 && aff_todo > 0 | |
| 5688 && STRCMP(cur_aff->ah_key, items[1]) == 0 | |
| 334 | 5689 && itemcnt >= 5) |
| 236 | 5690 { |
| 5691 affentry_T *aff_entry; | |
| 455 | 5692 int upper = FALSE; |
| 366 | 5693 int lasti = 5; |
| 5694 | |
| 334 | 5695 /* Myspell allows extra text after the item, but that might |
| 809 | 5696 * mean mistakes go unnoticed. Require a comment-starter. |
| 5697 * Hunspell uses a "-" item. */ | |
| 5698 if (itemcnt > lasti && *items[lasti] != '#' | |
| 5699 && (STRCMP(items[lasti], "-") != 0 | |
| 5700 || itemcnt != lasti + 1)) | |
| 482 | 5701 smsg((char_u *)_(e_afftrailing), fname, lnum, items[lasti]); |
| 334 | 5702 |
| 236 | 5703 /* New item for an affix letter. */ |
| 5704 --aff_todo; | |
| 474 | 5705 aff_entry = (affentry_T *)getroom(spin, |
| 447 | 5706 sizeof(affentry_T), TRUE); |
| 236 | 5707 if (aff_entry == NULL) |
| 5708 break; | |
| 240 | 5709 |
| 236 | 5710 if (STRCMP(items[2], "0") != 0) |
| 474 | 5711 aff_entry->ae_chop = getroom_save(spin, items[2]); |
| 236 | 5712 if (STRCMP(items[3], "0") != 0) |
| 809 | 5713 { |
| 474 | 5714 aff_entry->ae_add = getroom_save(spin, items[3]); |
| 236 | 5715 |
| 822 | 5716 /* Recognize flags on the affix: abcd/XYZ */ |
| 809 | 5717 aff_entry->ae_flags = vim_strchr(aff_entry->ae_add, '/'); |
| 5718 if (aff_entry->ae_flags != NULL) | |
| 822 | 5719 { |
| 809 | 5720 *aff_entry->ae_flags++ = NUL; |
| 822 | 5721 aff_process_flags(aff, aff_entry); |
| 5722 } | |
| 809 | 5723 } |
| 5724 | |
| 300 | 5725 /* Don't use an affix entry with non-ASCII characters when |
| 5726 * "spin->si_ascii" is TRUE. */ | |
| 5727 if (!spin->si_ascii || !(has_non_ascii(aff_entry->ae_chop) | |
| 240 | 5728 || has_non_ascii(aff_entry->ae_add))) |
| 5729 { | |
| 5730 aff_entry->ae_next = cur_aff->ah_first; | |
| 5731 cur_aff->ah_first = aff_entry; | |
| 300 | 5732 |
| 5733 if (STRCMP(items[4], ".") != 0) | |
| 5734 { | |
| 5735 char_u buf[MAXLINELEN]; | |
| 5736 | |
| 474 | 5737 aff_entry->ae_cond = getroom_save(spin, items[4]); |
| 300 | 5738 if (*items[0] == 'P') |
| 5739 sprintf((char *)buf, "^%s", items[4]); | |
| 5740 else | |
| 5741 sprintf((char *)buf, "%s$", items[4]); | |
| 5742 aff_entry->ae_prog = vim_regcomp(buf, | |
| 481 | 5743 RE_MAGIC + RE_STRING + RE_STRICT); |
| 5744 if (aff_entry->ae_prog == NULL) | |
| 5745 smsg((char_u *)_("Broken condition in %s line %d: %s"), | |
| 5746 fname, lnum, items[4]); | |
| 300 | 5747 } |
| 339 | 5748 |
| 5749 /* For postponed prefixes we need an entry in si_prefcond | |
| 809 | 5750 * for the condition. Use an existing one if possible. |
| 822 | 5751 * Can't be done for an affix with flags, ignoring |
| 5752 * COMPOUNDFORBIDFLAG and COMPOUNDPERMITFLAG. */ | |
| 809 | 5753 if (*items[0] == 'P' && aff->af_pfxpostpone |
| 5754 && aff_entry->ae_flags == NULL) | |
| 339 | 5755 { |
| 455 | 5756 /* When the chop string is one lower-case letter and |
| 5757 * the add string ends in the upper-case letter we set | |
| 5758 * the "upper" flag, clear "ae_chop" and remove the | |
| 5759 * letters from "ae_add". The condition must either | |
| 5760 * be empty or start with the same letter. */ | |
| 5761 if (aff_entry->ae_chop != NULL | |
| 5762 && aff_entry->ae_add != NULL | |
| 5763 #ifdef FEAT_MBYTE | |
| 474 | 5764 && aff_entry->ae_chop[(*mb_ptr2len)( |
| 455 | 5765 aff_entry->ae_chop)] == NUL |
| 5766 #else | |
| 5767 && aff_entry->ae_chop[1] == NUL | |
| 5768 #endif | |
| 5769 ) | |
| 339 | 5770 { |
| 455 | 5771 int c, c_up; |
| 5772 | |
| 5773 c = PTR2CHAR(aff_entry->ae_chop); | |
| 5774 c_up = SPELL_TOUPPER(c); | |
| 5775 if (c_up != c | |
| 5776 && (aff_entry->ae_cond == NULL | |
| 5777 || PTR2CHAR(aff_entry->ae_cond) == c)) | |
| 5778 { | |
| 5779 p = aff_entry->ae_add | |
| 5780 + STRLEN(aff_entry->ae_add); | |
| 5781 mb_ptr_back(aff_entry->ae_add, p); | |
| 5782 if (PTR2CHAR(p) == c_up) | |
| 5783 { | |
| 5784 upper = TRUE; | |
| 5785 aff_entry->ae_chop = NULL; | |
| 5786 *p = NUL; | |
| 5787 | |
| 5788 /* The condition is matched with the | |
| 5789 * actual word, thus must check for the | |
| 5790 * upper-case letter. */ | |
| 5791 if (aff_entry->ae_cond != NULL) | |
| 5792 { | |
| 5793 char_u buf[MAXLINELEN]; | |
| 5794 #ifdef FEAT_MBYTE | |
| 5795 if (has_mbyte) | |
| 5796 { | |
| 5797 onecap_copy(items[4], buf, TRUE); | |
| 5798 aff_entry->ae_cond = getroom_save( | |
| 474 | 5799 spin, buf); |
| 455 | 5800 } |
| 5801 else | |
| 5802 #endif | |
| 5803 *aff_entry->ae_cond = c_up; | |
| 5804 if (aff_entry->ae_cond != NULL) | |
| 5805 { | |
| 5806 sprintf((char *)buf, "^%s", | |
| 339 | 5807 aff_entry->ae_cond); |
| 455 | 5808 vim_free(aff_entry->ae_prog); |
| 5809 aff_entry->ae_prog = vim_regcomp( | |
| 5810 buf, RE_MAGIC + RE_STRING); | |
| 5811 } | |
| 5812 } | |
| 5813 } | |
| 5814 } | |
| 339 | 5815 } |
| 5816 | |
| 809 | 5817 if (aff_entry->ae_chop == NULL |
| 5818 && aff_entry->ae_flags == NULL) | |
| 390 | 5819 { |
| 455 | 5820 int idx; |
| 5821 char_u **pp; | |
| 5822 int n; | |
| 5823 | |
| 495 | 5824 /* Find a previously used condition. */ |
| 455 | 5825 for (idx = spin->si_prefcond.ga_len - 1; idx >= 0; |
| 5826 --idx) | |
| 5827 { | |
| 5828 p = ((char_u **)spin->si_prefcond.ga_data)[idx]; | |
| 5829 if (str_equal(p, aff_entry->ae_cond)) | |
| 5830 break; | |
| 5831 } | |
| 5832 if (idx < 0 && ga_grow(&spin->si_prefcond, 1) == OK) | |
| 5833 { | |
| 5834 /* Not found, add a new condition. */ | |
| 5835 idx = spin->si_prefcond.ga_len++; | |
| 5836 pp = ((char_u **)spin->si_prefcond.ga_data) | |
| 5837 + idx; | |
| 5838 if (aff_entry->ae_cond == NULL) | |
| 5839 *pp = NULL; | |
| 5840 else | |
| 474 | 5841 *pp = getroom_save(spin, |
| 455 | 5842 aff_entry->ae_cond); |
| 5843 } | |
| 5844 | |
| 5845 /* Add the prefix to the prefix tree. */ | |
| 5846 if (aff_entry->ae_add == NULL) | |
| 5847 p = (char_u *)""; | |
| 390 | 5848 else |
| 455 | 5849 p = aff_entry->ae_add; |
| 809 | 5850 |
| 455 | 5851 /* PFX_FLAGS is a negative number, so that |
| 5852 * tree_add_word() knows this is the prefix tree. */ | |
| 5853 n = PFX_FLAGS; | |
| 5854 if (!cur_aff->ah_combine) | |
| 5855 n |= WFP_NC; | |
| 5856 if (upper) | |
| 5857 n |= WFP_UP; | |
| 822 | 5858 if (aff_entry->ae_comppermit) |
| 5859 n |= WFP_COMPPERMIT; | |
| 5860 if (aff_entry->ae_compforbid) | |
| 5861 n |= WFP_COMPFORBID; | |
| 474 | 5862 tree_add_word(spin, p, spin->si_prefroot, n, |
| 5863 idx, cur_aff->ah_newID); | |
| 495 | 5864 did_postpone_prefix = TRUE; |
| 5865 } | |
| 5866 | |
| 5867 /* Didn't actually use ah_newID, backup si_newprefID. */ | |
| 5868 if (aff_todo == 0 && !did_postpone_prefix) | |
| 5869 { | |
| 5870 --spin->si_newprefID; | |
| 5871 cur_aff->ah_newID = 0; | |
| 390 | 5872 } |
| 339 | 5873 } |
| 240 | 5874 } |
| 236 | 5875 } |
| 1762 | 5876 else if (is_aff_rule(items, itemcnt, "FOL", 2) && fol == NULL) |
| 495 | 5877 { |
| 5878 fol = vim_strsave(items[1]); | |
| 5879 } | |
| 1762 | 5880 else if (is_aff_rule(items, itemcnt, "LOW", 2) && low == NULL) |
| 495 | 5881 { |
| 5882 low = vim_strsave(items[1]); | |
| 5883 } | |
| 1762 | 5884 else if (is_aff_rule(items, itemcnt, "UPP", 2) && upp == NULL) |
| 495 | 5885 { |
| 5886 upp = vim_strsave(items[1]); | |
| 255 | 5887 } |
| 1762 | 5888 else if (is_aff_rule(items, itemcnt, "REP", 2) |
| 5889 || is_aff_rule(items, itemcnt, "REPSAL", 2)) | |
| 625 | 5890 { |
| 5891 /* Ignore REP/REPSAL count */; | |
| 323 | 5892 if (!isdigit(*items[1])) |
| 625 | 5893 smsg((char_u *)_("Expected REP(SAL) count in %s line %d"), |
| 323 | 5894 fname, lnum); |
| 5895 } | |
| 625 | 5896 else if ((STRCMP(items[0], "REP") == 0 |
| 5897 || STRCMP(items[0], "REPSAL") == 0) | |
| 5898 && itemcnt >= 3) | |
| 5899 { | |
| 5900 /* REP/REPSAL item */ | |
| 482 | 5901 /* Myspell ignores extra arguments, we require it starts with |
| 5902 * # to detect mistakes. */ | |
| 5903 if (itemcnt > 3 && items[3][0] != '#') | |
| 5904 smsg((char_u *)_(e_afftrailing), fname, lnum, items[3]); | |
| 625 | 5905 if (items[0][3] == 'S' ? do_repsal : do_rep) |
| 532 | 5906 { |
| 5907 /* Replace underscore with space (can't include a space | |
| 5908 * directly). */ | |
| 5909 for (p = items[1]; *p != NUL; mb_ptr_adv(p)) | |
| 5910 if (*p == '_') | |
| 5911 *p = ' '; | |
| 5912 for (p = items[2]; *p != NUL; mb_ptr_adv(p)) | |
| 5913 if (*p == '_') | |
| 5914 *p = ' '; | |
| 625 | 5915 add_fromto(spin, items[0][3] == 'S' |
| 5916 ? &spin->si_repsal | |
| 5917 : &spin->si_rep, items[1], items[2]); | |
| 532 | 5918 } |
| 323 | 5919 } |
| 1762 | 5920 else if (is_aff_rule(items, itemcnt, "MAP", 2)) |
| 323 | 5921 { |
| 5922 /* MAP item or count */ | |
| 5923 if (!found_map) | |
| 5924 { | |
| 5925 /* First line contains the count. */ | |
| 5926 found_map = TRUE; | |
| 5927 if (!isdigit(*items[1])) | |
| 5928 smsg((char_u *)_("Expected MAP count in %s line %d"), | |
| 5929 fname, lnum); | |
| 5930 } | |
| 944 | 5931 else if (do_mapline) |
| 323 | 5932 { |
| 346 | 5933 int c; |
| 5934 | |
| 5935 /* Check that every character appears only once. */ | |
| 5936 for (p = items[1]; *p != NUL; ) | |
| 5937 { | |
| 5938 #ifdef FEAT_MBYTE | |
| 5939 c = mb_ptr2char_adv(&p); | |
| 5940 #else | |
| 5941 c = *p++; | |
| 5942 #endif | |
| 5943 if ((spin->si_map.ga_len > 0 | |
| 5944 && vim_strchr(spin->si_map.ga_data, c) | |
| 5945 != NULL) | |
| 5946 || vim_strchr(p, c) != NULL) | |
| 5947 smsg((char_u *)_("Duplicate character in MAP in %s line %d"), | |
| 5948 fname, lnum); | |
| 5949 } | |
| 5950 | |
| 323 | 5951 /* We simply concatenate all the MAP strings, separated by |
| 5952 * slashes. */ | |
| 5953 ga_concat(&spin->si_map, items[1]); | |
| 5954 ga_append(&spin->si_map, '/'); | |
| 5955 } | |
| 5956 } | |
| 1762 | 5957 /* Accept "SAL from to" and "SAL from to #comment". */ |
| 5958 else if (is_aff_rule(items, itemcnt, "SAL", 3)) | |
| 323 | 5959 { |
| 5960 if (do_sal) | |
| 5961 { | |
| 5962 /* SAL item (sounds-a-like) | |
| 5963 * Either one of the known keys or a from-to pair. */ | |
| 5964 if (STRCMP(items[1], "followup") == 0) | |
| 5965 spin->si_followup = sal_to_bool(items[2]); | |
| 5966 else if (STRCMP(items[1], "collapse_result") == 0) | |
| 5967 spin->si_collapse = sal_to_bool(items[2]); | |
| 5968 else if (STRCMP(items[1], "remove_accents") == 0) | |
| 5969 spin->si_rem_accents = sal_to_bool(items[2]); | |
| 5970 else | |
| 5971 /* when "to" is "_" it means empty */ | |
| 5972 add_fromto(spin, &spin->si_sal, items[1], | |
| 5973 STRCMP(items[2], "_") == 0 ? (char_u *)"" | |
| 5974 : items[2]); | |
| 5975 } | |
| 236 | 5976 } |
| 1762 | 5977 else if (is_aff_rule(items, itemcnt, "SOFOFROM", 2) |
| 495 | 5978 && sofofrom == NULL) |
| 5979 { | |
| 5980 sofofrom = getroom_save(spin, items[1]); | |
| 375 | 5981 } |
| 1762 | 5982 else if (is_aff_rule(items, itemcnt, "SOFOTO", 2) |
| 495 | 5983 && sofoto == NULL) |
| 5984 { | |
| 5985 sofoto = getroom_save(spin, items[1]); | |
| 375 | 5986 } |
| 625 | 5987 else if (STRCMP(items[0], "COMMON") == 0) |
| 5988 { | |
| 5989 int i; | |
| 5990 | |
| 5991 for (i = 1; i < itemcnt; ++i) | |
| 5992 { | |
| 5993 if (HASHITEM_EMPTY(hash_find(&spin->si_commonwords, | |
| 5994 items[i]))) | |
| 5995 { | |
| 5996 p = vim_strsave(items[i]); | |
| 5997 if (p == NULL) | |
| 5998 break; | |
| 5999 hash_add(&spin->si_commonwords, p); | |
| 6000 } | |
| 6001 } | |
| 6002 } | |
| 300 | 6003 else |
| 481 | 6004 smsg((char_u *)_("Unrecognized or duplicate item in %s line %d: %s"), |
| 236 | 6005 fname, lnum, items[0]); |
| 6006 } | |
| 6007 } | |
| 6008 | |
| 255 | 6009 if (fol != NULL || low != NULL || upp != NULL) |
| 6010 { | |
| 351 | 6011 if (spin->si_clear_chartab) |
| 6012 { | |
| 6013 /* Clear the char type tables, don't want to use any of the | |
| 6014 * currently used spell properties. */ | |
| 6015 init_spell_chartab(); | |
| 6016 spin->si_clear_chartab = FALSE; | |
| 6017 } | |
| 6018 | |
| 316 | 6019 /* |
| 6020 * Don't write a word table for an ASCII file, so that we don't check | |
| 6021 * for conflicts with a word table that matches 'encoding'. | |
| 324 | 6022 * Don't write one for utf-8 either, we use utf_*() and |
| 316 | 6023 * mb_get_class(), the list of chars in the file will be incomplete. |
| 6024 */ | |
| 6025 if (!spin->si_ascii | |
| 6026 #ifdef FEAT_MBYTE | |
| 6027 && !enc_utf8 | |
| 6028 #endif | |
| 6029 ) | |
| 260 | 6030 { |
| 6031 if (fol == NULL || low == NULL || upp == NULL) | |
| 6032 smsg((char_u *)_("Missing FOL/LOW/UPP line in %s"), fname); | |
| 6033 else | |
| 316 | 6034 (void)set_spell_chartab(fol, low, upp); |
| 260 | 6035 } |
| 255 | 6036 |
| 6037 vim_free(fol); | |
| 6038 vim_free(low); | |
| 6039 vim_free(upp); | |
| 6040 } | |
| 6041 | |
| 481 | 6042 /* Use compound specifications of the .aff file for the spell info. */ |
| 495 | 6043 if (compmax != 0) |
| 6044 { | |
| 809 | 6045 aff_check_number(spin->si_compmax, compmax, "COMPOUNDWORDMAX"); |
| 495 | 6046 spin->si_compmax = compmax; |
| 6047 } | |
| 6048 | |
| 6049 if (compminlen != 0) | |
| 6050 { | |
| 6051 aff_check_number(spin->si_compminlen, compminlen, "COMPOUNDMIN"); | |
| 6052 spin->si_compminlen = compminlen; | |
| 6053 } | |
| 6054 | |
| 6055 if (compsylmax != 0) | |
| 6056 { | |
| 6057 if (syllable == NULL) | |
| 6058 smsg((char_u *)_("COMPOUNDSYLMAX used without SYLLABLE")); | |
| 6059 aff_check_number(spin->si_compsylmax, compsylmax, "COMPOUNDSYLMAX"); | |
| 6060 spin->si_compsylmax = compsylmax; | |
| 6061 } | |
| 6062 | |
| 809 | 6063 if (compoptions != 0) |
| 6064 { | |
| 6065 aff_check_number(spin->si_compoptions, compoptions, "COMPOUND options"); | |
| 6066 spin->si_compoptions |= compoptions; | |
| 6067 } | |
| 6068 | |
| 495 | 6069 if (compflags != NULL) |
| 6070 process_compflags(spin, aff, compflags); | |
| 6071 | |
| 6072 /* Check that we didn't use too many renumbered flags. */ | |
| 500 | 6073 if (spin->si_newcompID < spin->si_newprefID) |
| 6074 { | |
| 6075 if (spin->si_newcompID == 127 || spin->si_newcompID == 255) | |
| 495 | 6076 MSG(_("Too many postponed prefixes")); |
| 500 | 6077 else if (spin->si_newprefID == 0 || spin->si_newprefID == 127) |
| 495 | 6078 MSG(_("Too many compound flags")); |
| 481 | 6079 else |
| 1705 | 6080 MSG(_("Too many postponed prefixes and/or compound flags")); |
| 495 | 6081 } |
| 6082 | |
| 6083 if (syllable != NULL) | |
| 6084 { | |
| 6085 aff_check_string(spin->si_syllable, syllable, "SYLLABLE"); | |
| 6086 spin->si_syllable = syllable; | |
| 6087 } | |
| 6088 | |
| 6089 if (sofofrom != NULL || sofoto != NULL) | |
| 6090 { | |
| 6091 if (sofofrom == NULL || sofoto == NULL) | |
| 6092 smsg((char_u *)_("Missing SOFO%s line in %s"), | |
| 6093 sofofrom == NULL ? "FROM" : "TO", fname); | |
| 6094 else if (spin->si_sal.ga_len > 0) | |
| 6095 smsg((char_u *)_("Both SAL and SOFO lines in %s"), fname); | |
| 483 | 6096 else |
| 495 | 6097 { |
| 6098 aff_check_string(spin->si_sofofr, sofofrom, "SOFOFROM"); | |
| 6099 aff_check_string(spin->si_sofoto, sofoto, "SOFOTO"); | |
| 6100 spin->si_sofofr = sofofrom; | |
| 6101 spin->si_sofoto = sofoto; | |
| 6102 } | |
| 6103 } | |
| 6104 | |
| 6105 if (midword != NULL) | |
| 6106 { | |
| 6107 aff_check_string(spin->si_midword, midword, "MIDWORD"); | |
| 6108 spin->si_midword = midword; | |
| 483 | 6109 } |
| 6110 | |
| 236 | 6111 vim_free(pc); |
| 6112 fclose(fd); | |
| 6113 return aff; | |
| 6114 } | |
| 6115 | |
| 6116 /* | |
| 1762 | 6117 * Return TRUE when items[0] equals "rulename", there are "mincount" items or |
| 6118 * a comment is following after item "mincount". | |
| 6119 */ | |
| 6120 static int | |
| 6121 is_aff_rule(items, itemcnt, rulename, mincount) | |
| 6122 char_u **items; | |
| 6123 int itemcnt; | |
| 6124 char *rulename; | |
| 6125 int mincount; | |
| 6126 { | |
| 6127 return (STRCMP(items[0], rulename) == 0 | |
| 6128 && (itemcnt == mincount | |
| 6129 || (itemcnt > mincount && items[mincount][0] == '#'))); | |
| 6130 } | |
| 6131 | |
| 6132 /* | |
| 822 | 6133 * For affix "entry" move COMPOUNDFORBIDFLAG and COMPOUNDPERMITFLAG from |
| 6134 * ae_flags to ae_comppermit and ae_compforbid. | |
| 6135 */ | |
| 6136 static void | |
| 6137 aff_process_flags(affile, entry) | |
| 6138 afffile_T *affile; | |
| 6139 affentry_T *entry; | |
| 6140 { | |
| 6141 char_u *p; | |
| 6142 char_u *prevp; | |
| 825 | 6143 unsigned flag; |
| 822 | 6144 |
| 6145 if (entry->ae_flags != NULL | |
| 6146 && (affile->af_compforbid != 0 || affile->af_comppermit != 0)) | |
| 6147 { | |
| 6148 for (p = entry->ae_flags; *p != NUL; ) | |
| 6149 { | |
| 6150 prevp = p; | |
| 6151 flag = get_affitem(affile->af_flagtype, &p); | |
| 6152 if (flag == affile->af_comppermit || flag == affile->af_compforbid) | |
| 6153 { | |
| 1619 | 6154 STRMOVE(prevp, p); |
| 822 | 6155 p = prevp; |
| 6156 if (flag == affile->af_comppermit) | |
| 6157 entry->ae_comppermit = TRUE; | |
| 6158 else | |
| 6159 entry->ae_compforbid = TRUE; | |
| 6160 } | |
| 6161 if (affile->af_flagtype == AFT_NUM && *p == ',') | |
| 6162 ++p; | |
| 6163 } | |
| 6164 if (*entry->ae_flags == NUL) | |
| 6165 entry->ae_flags = NULL; /* nothing left */ | |
| 6166 } | |
| 6167 } | |
| 6168 | |
| 6169 /* | |
| 714 | 6170 * Return TRUE if "s" is the name of an info item in the affix file. |
| 6171 */ | |
| 6172 static int | |
| 6173 spell_info_item(s) | |
| 6174 char_u *s; | |
| 6175 { | |
| 6176 return STRCMP(s, "NAME") == 0 | |
| 6177 || STRCMP(s, "HOME") == 0 | |
| 6178 || STRCMP(s, "VERSION") == 0 | |
| 6179 || STRCMP(s, "AUTHOR") == 0 | |
| 6180 || STRCMP(s, "EMAIL") == 0 | |
| 6181 || STRCMP(s, "COPYRIGHT") == 0; | |
| 6182 } | |
| 6183 | |
| 6184 /* | |
| 495 | 6185 * Turn an affix flag name into a number, according to the FLAG type. |
| 6186 * returns zero for failure. | |
| 6187 */ | |
| 6188 static unsigned | |
| 6189 affitem2flag(flagtype, item, fname, lnum) | |
| 6190 int flagtype; | |
| 6191 char_u *item; | |
| 6192 char_u *fname; | |
| 6193 int lnum; | |
| 6194 { | |
| 6195 unsigned res; | |
| 6196 char_u *p = item; | |
| 6197 | |
| 6198 res = get_affitem(flagtype, &p); | |
| 6199 if (res == 0) | |
| 6200 { | |
| 498 | 6201 if (flagtype == AFT_NUM) |
| 495 | 6202 smsg((char_u *)_("Flag is not a number in %s line %d: %s"), |
| 6203 fname, lnum, item); | |
| 6204 else | |
| 6205 smsg((char_u *)_("Illegal flag in %s line %d: %s"), | |
| 6206 fname, lnum, item); | |
| 6207 } | |
| 6208 if (*p != NUL) | |
| 6209 { | |
| 6210 smsg((char_u *)_(e_affname), fname, lnum, item); | |
| 6211 return 0; | |
| 6212 } | |
| 6213 | |
| 6214 return res; | |
| 6215 } | |
| 6216 | |
| 6217 /* | |
| 6218 * Get one affix name from "*pp" and advance the pointer. | |
| 6219 * Returns zero for an error, still advances the pointer then. | |
| 6220 */ | |
| 6221 static unsigned | |
| 6222 get_affitem(flagtype, pp) | |
| 6223 int flagtype; | |
| 6224 char_u **pp; | |
| 6225 { | |
| 6226 int res; | |
| 6227 | |
| 498 | 6228 if (flagtype == AFT_NUM) |
| 495 | 6229 { |
| 6230 if (!VIM_ISDIGIT(**pp)) | |
| 6231 { | |
| 498 | 6232 ++*pp; /* always advance, avoid getting stuck */ |
| 495 | 6233 return 0; |
| 6234 } | |
| 6235 res = getdigits(pp); | |
| 6236 } | |
| 6237 else | |
| 6238 { | |
| 6239 #ifdef FEAT_MBYTE | |
| 6240 res = mb_ptr2char_adv(pp); | |
| 6241 #else | |
| 6242 res = *(*pp)++; | |
| 6243 #endif | |
| 498 | 6244 if (flagtype == AFT_LONG || (flagtype == AFT_CAPLONG |
| 495 | 6245 && res >= 'A' && res <= 'Z')) |
| 6246 { | |
| 6247 if (**pp == NUL) | |
| 6248 return 0; | |
| 6249 #ifdef FEAT_MBYTE | |
| 6250 res = mb_ptr2char_adv(pp) + (res << 16); | |
| 6251 #else | |
| 6252 res = *(*pp)++ + (res << 16); | |
| 6253 #endif | |
| 6254 } | |
| 6255 } | |
| 6256 return res; | |
| 6257 } | |
| 6258 | |
| 6259 /* | |
| 6260 * Process the "compflags" string used in an affix file and append it to | |
| 6261 * spin->si_compflags. | |
| 6262 * The processing involves changing the affix names to ID numbers, so that | |
| 6263 * they fit in one byte. | |
| 6264 */ | |
| 6265 static void | |
| 6266 process_compflags(spin, aff, compflags) | |
| 6267 spellinfo_T *spin; | |
| 6268 afffile_T *aff; | |
| 6269 char_u *compflags; | |
| 6270 { | |
| 6271 char_u *p; | |
| 6272 char_u *prevp; | |
| 6273 unsigned flag; | |
| 6274 compitem_T *ci; | |
| 6275 int id; | |
| 6276 int len; | |
| 6277 char_u *tp; | |
| 6278 char_u key[AH_KEY_LEN]; | |
| 6279 hashitem_T *hi; | |
| 6280 | |
| 6281 /* Make room for the old and the new compflags, concatenated with a / in | |
| 6282 * between. Processing it makes it shorter, but we don't know by how | |
| 6283 * much, thus allocate the maximum. */ | |
| 835 | 6284 len = (int)STRLEN(compflags) + 1; |
| 495 | 6285 if (spin->si_compflags != NULL) |
| 835 | 6286 len += (int)STRLEN(spin->si_compflags) + 1; |
| 495 | 6287 p = getroom(spin, len, FALSE); |
| 6288 if (p == NULL) | |
| 6289 return; | |
| 6290 if (spin->si_compflags != NULL) | |
| 6291 { | |
| 6292 STRCPY(p, spin->si_compflags); | |
| 6293 STRCAT(p, "/"); | |
| 6294 } | |
| 6295 spin->si_compflags = p; | |
| 6296 tp = p + STRLEN(p); | |
| 6297 | |
| 6298 for (p = compflags; *p != NUL; ) | |
| 6299 { | |
| 2691 | 6300 if (vim_strchr((char_u *)"/?*+[]", *p) != NULL) |
| 495 | 6301 /* Copy non-flag characters directly. */ |
| 6302 *tp++ = *p++; | |
| 6303 else | |
| 6304 { | |
| 6305 /* First get the flag number, also checks validity. */ | |
| 6306 prevp = p; | |
| 6307 flag = get_affitem(aff->af_flagtype, &p); | |
| 6308 if (flag != 0) | |
| 6309 { | |
| 6310 /* Find the flag in the hashtable. If it was used before, use | |
| 6311 * the existing ID. Otherwise add a new entry. */ | |
| 6312 vim_strncpy(key, prevp, p - prevp); | |
| 6313 hi = hash_find(&aff->af_comp, key); | |
| 6314 if (!HASHITEM_EMPTY(hi)) | |
| 6315 id = HI2CI(hi)->ci_newID; | |
| 6316 else | |
| 6317 { | |
| 6318 ci = (compitem_T *)getroom(spin, sizeof(compitem_T), TRUE); | |
| 6319 if (ci == NULL) | |
| 6320 break; | |
| 6321 STRCPY(ci->ci_key, key); | |
| 6322 ci->ci_flag = flag; | |
| 6323 /* Avoid using a flag ID that has a special meaning in a | |
| 6324 * regexp (also inside []). */ | |
| 6325 do | |
| 6326 { | |
| 500 | 6327 check_renumber(spin); |
| 6328 id = spin->si_newcompID--; | |
| 2691 | 6329 } while (vim_strchr((char_u *)"/?*+[]\\-^", id) != NULL); |
| 495 | 6330 ci->ci_newID = id; |
| 6331 hash_add(&aff->af_comp, ci->ci_key); | |
| 6332 } | |
| 6333 *tp++ = id; | |
| 6334 } | |
| 498 | 6335 if (aff->af_flagtype == AFT_NUM && *p == ',') |
| 495 | 6336 ++p; |
| 6337 } | |
| 6338 } | |
| 6339 | |
| 6340 *tp = NUL; | |
| 6341 } | |
| 6342 | |
| 6343 /* | |
| 500 | 6344 * Check that the new IDs for postponed affixes and compounding don't overrun |
| 6345 * each other. We have almost 255 available, but start at 0-127 to avoid | |
| 6346 * using two bytes for utf-8. When the 0-127 range is used up go to 128-255. | |
| 6347 * When that is used up an error message is given. | |
| 6348 */ | |
| 6349 static void | |
| 6350 check_renumber(spin) | |
| 6351 spellinfo_T *spin; | |
| 6352 { | |
| 6353 if (spin->si_newprefID == spin->si_newcompID && spin->si_newcompID < 128) | |
| 6354 { | |
| 6355 spin->si_newprefID = 127; | |
| 6356 spin->si_newcompID = 255; | |
| 6357 } | |
| 6358 } | |
| 6359 | |
| 6360 /* | |
| 495 | 6361 * Return TRUE if flag "flag" appears in affix list "afflist". |
| 6362 */ | |
| 6363 static int | |
| 6364 flag_in_afflist(flagtype, afflist, flag) | |
| 6365 int flagtype; | |
| 6366 char_u *afflist; | |
| 6367 unsigned flag; | |
| 6368 { | |
| 6369 char_u *p; | |
| 6370 unsigned n; | |
| 6371 | |
| 6372 switch (flagtype) | |
| 6373 { | |
| 6374 case AFT_CHAR: | |
| 6375 return vim_strchr(afflist, flag) != NULL; | |
| 6376 | |
| 498 | 6377 case AFT_CAPLONG: |
| 6378 case AFT_LONG: | |
| 495 | 6379 for (p = afflist; *p != NUL; ) |
| 6380 { | |
| 6381 #ifdef FEAT_MBYTE | |
| 6382 n = mb_ptr2char_adv(&p); | |
| 6383 #else | |
| 6384 n = *p++; | |
| 6385 #endif | |
| 498 | 6386 if ((flagtype == AFT_LONG || (n >= 'A' && n <= 'Z')) |
| 495 | 6387 && *p != NUL) |
| 6388 #ifdef FEAT_MBYTE | |
| 6389 n = mb_ptr2char_adv(&p) + (n << 16); | |
| 6390 #else | |
| 6391 n = *p++ + (n << 16); | |
| 6392 #endif | |
| 6393 if (n == flag) | |
| 6394 return TRUE; | |
| 6395 } | |
| 6396 break; | |
| 6397 | |
| 498 | 6398 case AFT_NUM: |
| 495 | 6399 for (p = afflist; *p != NUL; ) |
| 6400 { | |
| 6401 n = getdigits(&p); | |
| 6402 if (n == flag) | |
| 6403 return TRUE; | |
| 6404 if (*p != NUL) /* skip over comma */ | |
| 6405 ++p; | |
| 6406 } | |
| 6407 break; | |
| 6408 } | |
| 6409 return FALSE; | |
| 6410 } | |
| 6411 | |
| 6412 /* | |
| 6413 * Give a warning when "spinval" and "affval" numbers are set and not the same. | |
| 6414 */ | |
| 6415 static void | |
| 6416 aff_check_number(spinval, affval, name) | |
| 6417 int spinval; | |
| 6418 int affval; | |
| 6419 char *name; | |
| 6420 { | |
| 6421 if (spinval != 0 && spinval != affval) | |
| 6422 smsg((char_u *)_("%s value differs from what is used in another .aff file"), name); | |
| 6423 } | |
| 6424 | |
| 6425 /* | |
| 6426 * Give a warning when "spinval" and "affval" strings are set and not the same. | |
| 6427 */ | |
| 6428 static void | |
| 6429 aff_check_string(spinval, affval, name) | |
| 6430 char_u *spinval; | |
| 6431 char_u *affval; | |
| 6432 char *name; | |
| 6433 { | |
| 6434 if (spinval != NULL && STRCMP(spinval, affval) != 0) | |
| 6435 smsg((char_u *)_("%s value differs from what is used in another .aff file"), name); | |
| 6436 } | |
| 6437 | |
| 6438 /* | |
| 339 | 6439 * Return TRUE if strings "s1" and "s2" are equal. Also consider both being |
| 6440 * NULL as equal. | |
| 6441 */ | |
| 6442 static int | |
| 6443 str_equal(s1, s2) | |
| 6444 char_u *s1; | |
| 6445 char_u *s2; | |
| 6446 { | |
| 6447 if (s1 == NULL || s2 == NULL) | |
| 6448 return s1 == s2; | |
| 6449 return STRCMP(s1, s2) == 0; | |
| 6450 } | |
| 6451 | |
| 6452 /* | |
| 323 | 6453 * Add a from-to item to "gap". Used for REP and SAL items. |
| 6454 * They are stored case-folded. | |
| 6455 */ | |
| 6456 static void | |
| 6457 add_fromto(spin, gap, from, to) | |
| 6458 spellinfo_T *spin; | |
| 6459 garray_T *gap; | |
| 6460 char_u *from; | |
| 6461 char_u *to; | |
| 6462 { | |
| 6463 fromto_T *ftp; | |
| 6464 char_u word[MAXWLEN]; | |
| 6465 | |
| 6466 if (ga_grow(gap, 1) == OK) | |
| 6467 { | |
| 6468 ftp = ((fromto_T *)gap->ga_data) + gap->ga_len; | |
| 835 | 6469 (void)spell_casefold(from, (int)STRLEN(from), word, MAXWLEN); |
| 474 | 6470 ftp->ft_from = getroom_save(spin, word); |
| 835 | 6471 (void)spell_casefold(to, (int)STRLEN(to), word, MAXWLEN); |
| 474 | 6472 ftp->ft_to = getroom_save(spin, word); |
| 323 | 6473 ++gap->ga_len; |
| 6474 } | |
| 6475 } | |
| 6476 | |
| 6477 /* | |
| 6478 * Convert a boolean argument in a SAL line to TRUE or FALSE; | |
| 6479 */ | |
| 6480 static int | |
| 6481 sal_to_bool(s) | |
| 6482 char_u *s; | |
| 6483 { | |
| 6484 return STRCMP(s, "1") == 0 || STRCMP(s, "true") == 0; | |
| 6485 } | |
| 6486 | |
| 6487 /* | |
| 240 | 6488 * Return TRUE if string "s" contains a non-ASCII character (128 or higher). |
| 6489 * When "s" is NULL FALSE is returned. | |
| 6490 */ | |
| 6491 static int | |
| 6492 has_non_ascii(s) | |
| 6493 char_u *s; | |
| 6494 { | |
| 6495 char_u *p; | |
| 6496 | |
| 6497 if (s != NULL) | |
| 6498 for (p = s; *p != NUL; ++p) | |
| 6499 if (*p >= 128) | |
| 6500 return TRUE; | |
| 6501 return FALSE; | |
| 6502 } | |
| 6503 | |
| 6504 /* | |
| 236 | 6505 * Free the structure filled by spell_read_aff(). |
| 6506 */ | |
| 6507 static void | |
| 6508 spell_free_aff(aff) | |
| 6509 afffile_T *aff; | |
| 6510 { | |
| 6511 hashtab_T *ht; | |
| 6512 hashitem_T *hi; | |
| 6513 int todo; | |
| 6514 affheader_T *ah; | |
| 300 | 6515 affentry_T *ae; |
| 236 | 6516 |
| 6517 vim_free(aff->af_enc); | |
| 6518 | |
| 339 | 6519 /* All this trouble to free the "ae_prog" items... */ |
| 236 | 6520 for (ht = &aff->af_pref; ; ht = &aff->af_suff) |
| 6521 { | |
| 835 | 6522 todo = (int)ht->ht_used; |
| 236 | 6523 for (hi = ht->ht_array; todo > 0; ++hi) |
| 6524 { | |
| 6525 if (!HASHITEM_EMPTY(hi)) | |
| 6526 { | |
| 6527 --todo; | |
| 6528 ah = HI2AH(hi); | |
| 300 | 6529 for (ae = ah->ah_first; ae != NULL; ae = ae->ae_next) |
| 6530 vim_free(ae->ae_prog); | |
| 236 | 6531 } |
| 6532 } | |
| 6533 if (ht == &aff->af_suff) | |
| 6534 break; | |
| 6535 } | |
| 300 | 6536 |
| 236 | 6537 hash_clear(&aff->af_pref); |
| 6538 hash_clear(&aff->af_suff); | |
| 495 | 6539 hash_clear(&aff->af_comp); |
| 236 | 6540 } |
| 6541 | |
| 6542 /* | |
| 300 | 6543 * Read dictionary file "fname". |
| 236 | 6544 * Returns OK or FAIL; |
| 6545 */ | |
| 6546 static int | |
| 474 | 6547 spell_read_dic(spin, fname, affile) |
| 6548 spellinfo_T *spin; | |
| 236 | 6549 char_u *fname; |
| 300 | 6550 afffile_T *affile; |
| 236 | 6551 { |
| 300 | 6552 hashtab_T ht; |
| 236 | 6553 char_u line[MAXLINELEN]; |
| 481 | 6554 char_u *p; |
| 300 | 6555 char_u *afflist; |
| 483 | 6556 char_u store_afflist[MAXWLEN]; |
| 6557 int pfxlen; | |
| 6558 int need_affix; | |
| 300 | 6559 char_u *dw; |
| 236 | 6560 char_u *pc; |
| 6561 char_u *w; | |
| 6562 int l; | |
| 6563 hash_T hash; | |
| 6564 hashitem_T *hi; | |
| 6565 FILE *fd; | |
| 6566 int lnum = 1; | |
| 300 | 6567 int non_ascii = 0; |
| 6568 int retval = OK; | |
| 6569 char_u message[MAXLINELEN + MAXWLEN]; | |
| 307 | 6570 int flags; |
| 481 | 6571 int duplicate = 0; |
| 236 | 6572 |
| 300 | 6573 /* |
| 6574 * Open the file. | |
| 6575 */ | |
| 310 | 6576 fd = mch_fopen((char *)fname, "r"); |
| 236 | 6577 if (fd == NULL) |
| 6578 { | |
| 6579 EMSG2(_(e_notopen), fname); | |
| 6580 return FAIL; | |
| 6581 } | |
| 6582 | |
| 300 | 6583 /* The hashtable is only used to detect duplicated words. */ |
| 6584 hash_init(&ht); | |
| 6585 | |
| 625 | 6586 vim_snprintf((char *)IObuff, IOSIZE, |
| 6587 _("Reading dictionary file %s ..."), fname); | |
| 6588 spell_message(spin, IObuff); | |
| 236 | 6589 |
| 474 | 6590 /* start with a message for the first line */ |
| 6591 spin->si_msg_count = 999999; | |
| 6592 | |
| 236 | 6593 /* Read and ignore the first line: word count. */ |
| 6594 (void)vim_fgets(line, MAXLINELEN, fd); | |
| 324 | 6595 if (!vim_isdigit(*skipwhite(line))) |
| 236 | 6596 EMSG2(_("E760: No word count in %s"), fname); |
| 6597 | |
| 6598 /* | |
| 6599 * Read all the lines in the file one by one. | |
| 6600 * The words are converted to 'encoding' here, before being added to | |
| 6601 * the hashtable. | |
| 6602 */ | |
| 255 | 6603 while (!vim_fgets(line, MAXLINELEN, fd) && !got_int) |
| 236 | 6604 { |
| 255 | 6605 line_breakcheck(); |
| 236 | 6606 ++lnum; |
| 455 | 6607 if (line[0] == '#' || line[0] == '/') |
| 351 | 6608 continue; /* comment line */ |
| 236 | 6609 |
| 300 | 6610 /* Remove CR, LF and white space from the end. White space halfway |
| 6611 * the word is kept to allow e.g., "et al.". */ | |
| 835 | 6612 l = (int)STRLEN(line); |
| 236 | 6613 while (l > 0 && line[l - 1] <= ' ') |
| 6614 --l; | |
| 6615 if (l == 0) | |
| 6616 continue; /* empty line */ | |
| 6617 line[l] = NUL; | |
| 6618 | |
| 310 | 6619 #ifdef FEAT_MBYTE |
| 236 | 6620 /* Convert from "SET" to 'encoding' when needed. */ |
| 300 | 6621 if (spin->si_conv.vc_type != CONV_NONE) |
| 236 | 6622 { |
| 300 | 6623 pc = string_convert(&spin->si_conv, line, NULL); |
| 255 | 6624 if (pc == NULL) |
| 6625 { | |
| 6626 smsg((char_u *)_("Conversion failure for word in %s line %d: %s"), | |
| 6627 fname, lnum, line); | |
| 6628 continue; | |
| 6629 } | |
| 236 | 6630 w = pc; |
| 6631 } | |
| 6632 else | |
| 310 | 6633 #endif |
| 236 | 6634 { |
| 6635 pc = NULL; | |
| 6636 w = line; | |
| 6637 } | |
| 6638 | |
| 822 | 6639 /* Truncate the word at the "/", set "afflist" to what follows. |
| 6640 * Replace "\/" by "/" and "\\" by "\". */ | |
| 6641 afflist = NULL; | |
| 6642 for (p = w; *p != NUL; mb_ptr_adv(p)) | |
| 6643 { | |
| 6644 if (*p == '\\' && (p[1] == '\\' || p[1] == '/')) | |
| 1619 | 6645 STRMOVE(p, p + 1); |
| 822 | 6646 else if (*p == '/') |
| 6647 { | |
| 6648 *p = NUL; | |
| 6649 afflist = p + 1; | |
| 6650 break; | |
| 6651 } | |
| 6652 } | |
| 6653 | |
| 6654 /* Skip non-ASCII words when "spin->si_ascii" is TRUE. */ | |
| 6655 if (spin->si_ascii && has_non_ascii(w)) | |
| 6656 { | |
| 6657 ++non_ascii; | |
| 825 | 6658 vim_free(pc); |
| 822 | 6659 continue; |
| 6660 } | |
| 6661 | |
| 474 | 6662 /* This takes time, print a message every 10000 words. */ |
| 6663 if (spin->si_verbose && spin->si_msg_count > 10000) | |
| 6664 { | |
| 6665 spin->si_msg_count = 0; | |
| 339 | 6666 vim_snprintf((char *)message, sizeof(message), |
| 6667 _("line %6d, word %6d - %s"), | |
| 6668 lnum, spin->si_foldwcount + spin->si_keepwcount, w); | |
| 6669 msg_start(); | |
| 6670 msg_puts_long_attr(message, 0); | |
| 6671 msg_clr_eos(); | |
| 6672 msg_didout = FALSE; | |
| 6673 msg_col = 0; | |
| 6674 out_flush(); | |
| 6675 } | |
| 6676 | |
| 300 | 6677 /* Store the word in the hashtable to be able to find duplicates. */ |
| 474 | 6678 dw = (char_u *)getroom_save(spin, w); |
| 236 | 6679 if (dw == NULL) |
| 825 | 6680 { |
| 300 | 6681 retval = FAIL; |
| 825 | 6682 vim_free(pc); |
| 236 | 6683 break; |
| 825 | 6684 } |
| 236 | 6685 |
| 300 | 6686 hash = hash_hash(dw); |
| 6687 hi = hash_lookup(&ht, dw, hash); | |
| 236 | 6688 if (!HASHITEM_EMPTY(hi)) |
| 481 | 6689 { |
| 6690 if (p_verbose > 0) | |
| 6691 smsg((char_u *)_("Duplicate word in %s line %d: %s"), | |
| 375 | 6692 fname, lnum, dw); |
| 481 | 6693 else if (duplicate == 0) |
| 6694 smsg((char_u *)_("First duplicate word in %s line %d: %s"), | |
| 6695 fname, lnum, dw); | |
| 6696 ++duplicate; | |
| 6697 } | |
| 236 | 6698 else |
| 300 | 6699 hash_add_item(&ht, hi, dw, hash); |
| 6700 | |
| 307 | 6701 flags = 0; |
| 483 | 6702 store_afflist[0] = NUL; |
| 6703 pfxlen = 0; | |
| 6704 need_affix = FALSE; | |
| 307 | 6705 if (afflist != NULL) |
| 6706 { | |
| 820 | 6707 /* Extract flags from the affix list. */ |
| 6708 flags |= get_affix_flags(affile, afflist); | |
| 6709 | |
| 495 | 6710 if (affile->af_needaffix != 0 && flag_in_afflist( |
| 6711 affile->af_flagtype, afflist, affile->af_needaffix)) | |
| 483 | 6712 need_affix = TRUE; |
| 339 | 6713 |
| 6714 if (affile->af_pfxpostpone) | |
| 6715 /* Need to store the list of prefix IDs with the word. */ | |
| 483 | 6716 pfxlen = get_pfxlist(affile, afflist, store_afflist); |
| 6717 | |
| 6718 if (spin->si_compflags != NULL) | |
| 6719 /* Need to store the list of compound flags with the word. | |
| 6720 * Concatenate them to the list of prefix IDs. */ | |
| 495 | 6721 get_compflags(affile, afflist, store_afflist + pfxlen); |
| 307 | 6722 } |
| 6723 | |
| 300 | 6724 /* Add the word to the word tree(s). */ |
| 483 | 6725 if (store_word(spin, dw, flags, spin->si_region, |
| 6726 store_afflist, need_affix) == FAIL) | |
| 300 | 6727 retval = FAIL; |
| 236 | 6728 |
| 300 | 6729 if (afflist != NULL) |
| 6730 { | |
| 6731 /* Find all matching suffixes and add the resulting words. | |
| 6732 * Additionally do matching prefixes that combine. */ | |
| 474 | 6733 if (store_aff_word(spin, dw, afflist, affile, |
| 307 | 6734 &affile->af_suff, &affile->af_pref, |
| 820 | 6735 CONDIT_SUF, flags, store_afflist, pfxlen) == FAIL) |
| 300 | 6736 retval = FAIL; |
| 6737 | |
| 6738 /* Find all matching prefixes and add the resulting words. */ | |
| 474 | 6739 if (store_aff_word(spin, dw, afflist, affile, |
| 339 | 6740 &affile->af_pref, NULL, |
| 820 | 6741 CONDIT_SUF, flags, store_afflist, pfxlen) == FAIL) |
| 300 | 6742 retval = FAIL; |
| 6743 } | |
| 825 | 6744 |
| 6745 vim_free(pc); | |
| 236 | 6746 } |
| 6747 | |
| 481 | 6748 if (duplicate > 0) |
| 6749 smsg((char_u *)_("%d duplicate word(s) in %s"), duplicate, fname); | |
| 300 | 6750 if (spin->si_ascii && non_ascii > 0) |
| 481 | 6751 smsg((char_u *)_("Ignored %d word(s) with non-ASCII characters in %s"), |
| 6752 non_ascii, fname); | |
| 300 | 6753 hash_clear(&ht); |
| 6754 | |
| 236 | 6755 fclose(fd); |
| 300 | 6756 return retval; |
| 236 | 6757 } |
| 6758 | |
| 6759 /* | |
| 820 | 6760 * Check for affix flags in "afflist" that are turned into word flags. |
| 6761 * Return WF_ flags. | |
| 6762 */ | |
| 6763 static int | |
| 6764 get_affix_flags(affile, afflist) | |
| 6765 afffile_T *affile; | |
| 6766 char_u *afflist; | |
| 6767 { | |
| 6768 int flags = 0; | |
| 6769 | |
| 6770 if (affile->af_keepcase != 0 && flag_in_afflist( | |
| 6771 affile->af_flagtype, afflist, affile->af_keepcase)) | |
| 6772 flags |= WF_KEEPCAP | WF_FIXCAP; | |
| 6773 if (affile->af_rare != 0 && flag_in_afflist( | |
| 6774 affile->af_flagtype, afflist, affile->af_rare)) | |
| 6775 flags |= WF_RARE; | |
| 6776 if (affile->af_bad != 0 && flag_in_afflist( | |
| 6777 affile->af_flagtype, afflist, affile->af_bad)) | |
| 6778 flags |= WF_BANNED; | |
| 6779 if (affile->af_needcomp != 0 && flag_in_afflist( | |
| 6780 affile->af_flagtype, afflist, affile->af_needcomp)) | |
| 6781 flags |= WF_NEEDCOMP; | |
| 6782 if (affile->af_comproot != 0 && flag_in_afflist( | |
| 6783 affile->af_flagtype, afflist, affile->af_comproot)) | |
| 6784 flags |= WF_COMPROOT; | |
| 6785 if (affile->af_nosuggest != 0 && flag_in_afflist( | |
| 6786 affile->af_flagtype, afflist, affile->af_nosuggest)) | |
| 6787 flags |= WF_NOSUGGEST; | |
| 6788 return flags; | |
| 6789 } | |
| 6790 | |
| 6791 /* | |
| 339 | 6792 * Get the list of prefix IDs from the affix list "afflist". |
| 6793 * Used for PFXPOSTPONE. | |
| 483 | 6794 * Put the resulting flags in "store_afflist[MAXWLEN]" with a terminating NUL |
| 6795 * and return the number of affixes. | |
| 6796 */ | |
| 6797 static int | |
| 6798 get_pfxlist(affile, afflist, store_afflist) | |
| 339 | 6799 afffile_T *affile; |
| 6800 char_u *afflist; | |
| 483 | 6801 char_u *store_afflist; |
| 339 | 6802 { |
| 6803 char_u *p; | |
| 495 | 6804 char_u *prevp; |
| 483 | 6805 int cnt = 0; |
| 495 | 6806 int id; |
| 6807 char_u key[AH_KEY_LEN]; | |
| 339 | 6808 hashitem_T *hi; |
| 6809 | |
| 495 | 6810 for (p = afflist; *p != NUL; ) |
| 6811 { | |
| 6812 prevp = p; | |
| 6813 if (get_affitem(affile->af_flagtype, &p) != 0) | |
| 6814 { | |
| 6815 /* A flag is a postponed prefix flag if it appears in "af_pref" | |
| 6816 * and it's ID is not zero. */ | |
| 6817 vim_strncpy(key, prevp, p - prevp); | |
| 6818 hi = hash_find(&affile->af_pref, key); | |
| 6819 if (!HASHITEM_EMPTY(hi)) | |
| 6820 { | |
| 6821 id = HI2AH(hi)->ah_newID; | |
| 6822 if (id != 0) | |
| 6823 store_afflist[cnt++] = id; | |
| 6824 } | |
| 6825 } | |
| 498 | 6826 if (affile->af_flagtype == AFT_NUM && *p == ',') |
| 495 | 6827 ++p; |
| 483 | 6828 } |
| 6829 | |
| 6830 store_afflist[cnt] = NUL; | |
| 6831 return cnt; | |
| 339 | 6832 } |
| 6833 | |
| 6834 /* | |
| 495 | 6835 * Get the list of compound IDs from the affix list "afflist" that are used |
| 6836 * for compound words. | |
| 483 | 6837 * Puts the flags in "store_afflist[]". |
| 6838 */ | |
| 6839 static void | |
| 495 | 6840 get_compflags(affile, afflist, store_afflist) |
| 6841 afffile_T *affile; | |
| 481 | 6842 char_u *afflist; |
| 483 | 6843 char_u *store_afflist; |
| 481 | 6844 { |
| 6845 char_u *p; | |
| 495 | 6846 char_u *prevp; |
| 483 | 6847 int cnt = 0; |
| 495 | 6848 char_u key[AH_KEY_LEN]; |
| 6849 hashitem_T *hi; | |
| 6850 | |
| 6851 for (p = afflist; *p != NUL; ) | |
| 6852 { | |
| 6853 prevp = p; | |
| 6854 if (get_affitem(affile->af_flagtype, &p) != 0) | |
| 6855 { | |
| 6856 /* A flag is a compound flag if it appears in "af_comp". */ | |
| 6857 vim_strncpy(key, prevp, p - prevp); | |
| 6858 hi = hash_find(&affile->af_comp, key); | |
| 6859 if (!HASHITEM_EMPTY(hi)) | |
| 6860 store_afflist[cnt++] = HI2CI(hi)->ci_newID; | |
| 6861 } | |
| 498 | 6862 if (affile->af_flagtype == AFT_NUM && *p == ',') |
| 495 | 6863 ++p; |
| 6864 } | |
| 483 | 6865 |
| 6866 store_afflist[cnt] = NUL; | |
| 481 | 6867 } |
| 6868 | |
| 6869 /* | |
| 300 | 6870 * Apply affixes to a word and store the resulting words. |
| 6871 * "ht" is the hashtable with affentry_T that need to be applied, either | |
| 6872 * prefixes or suffixes. | |
| 6873 * "xht", when not NULL, is the prefix hashtable, to be used additionally on | |
| 6874 * the resulting words for combining affixes. | |
| 6875 * | |
| 6876 * Returns FAIL when out of memory. | |
| 236 | 6877 */ |
| 300 | 6878 static int |
| 820 | 6879 store_aff_word(spin, word, afflist, affile, ht, xht, condit, flags, |
| 483 | 6880 pfxlist, pfxlen) |
| 474 | 6881 spellinfo_T *spin; /* spell info */ |
| 300 | 6882 char_u *word; /* basic word start */ |
| 6883 char_u *afflist; /* list of names of supported affixes */ | |
| 339 | 6884 afffile_T *affile; |
| 300 | 6885 hashtab_T *ht; |
| 6886 hashtab_T *xht; | |
| 820 | 6887 int condit; /* CONDIT_SUF et al. */ |
| 307 | 6888 int flags; /* flags for the word */ |
| 339 | 6889 char_u *pfxlist; /* list of prefix IDs */ |
| 483 | 6890 int pfxlen; /* nr of flags in "pfxlist" for prefixes, rest |
| 6891 * is compound flags */ | |
| 236 | 6892 { |
| 6893 int todo; | |
| 6894 hashitem_T *hi; | |
| 300 | 6895 affheader_T *ah; |
| 6896 affentry_T *ae; | |
| 6897 regmatch_T regmatch; | |
| 6898 char_u newword[MAXWLEN]; | |
| 6899 int retval = OK; | |
| 820 | 6900 int i, j; |
| 300 | 6901 char_u *p; |
| 366 | 6902 int use_flags; |
| 390 | 6903 char_u *use_pfxlist; |
| 820 | 6904 int use_pfxlen; |
| 6905 int need_affix; | |
| 6906 char_u store_afflist[MAXWLEN]; | |
| 483 | 6907 char_u pfx_pfxlist[MAXWLEN]; |
| 6908 size_t wordlen = STRLEN(word); | |
| 820 | 6909 int use_condit; |
| 236 | 6910 |
| 835 | 6911 todo = (int)ht->ht_used; |
| 300 | 6912 for (hi = ht->ht_array; todo > 0 && retval == OK; ++hi) |
| 236 | 6913 { |
| 6914 if (!HASHITEM_EMPTY(hi)) | |
| 6915 { | |
| 6916 --todo; | |
| 300 | 6917 ah = HI2AH(hi); |
| 236 | 6918 |
| 300 | 6919 /* Check that the affix combines, if required, and that the word |
| 6920 * supports this affix. */ | |
| 820 | 6921 if (((condit & CONDIT_COMB) == 0 || ah->ah_combine) |
| 6922 && flag_in_afflist(affile->af_flagtype, afflist, | |
| 6923 ah->ah_flag)) | |
| 236 | 6924 { |
| 300 | 6925 /* Loop over all affix entries with this name. */ |
| 6926 for (ae = ah->ah_first; ae != NULL; ae = ae->ae_next) | |
| 236 | 6927 { |
| 300 | 6928 /* Check the condition. It's not logical to match case |
| 6929 * here, but it is required for compatibility with | |
| 339 | 6930 * Myspell. |
| 481 | 6931 * Another requirement from Myspell is that the chop |
| 6932 * string is shorter than the word itself. | |
| 339 | 6933 * For prefixes, when "PFXPOSTPONE" was used, only do |
| 820 | 6934 * prefixes with a chop string and/or flags. |
| 6935 * When a previously added affix had CIRCUMFIX this one | |
| 6936 * must have it too, if it had not then this one must not | |
| 6937 * have one either. */ | |
| 300 | 6938 regmatch.regprog = ae->ae_prog; |
| 6939 regmatch.rm_ic = FALSE; | |
| 339 | 6940 if ((xht != NULL || !affile->af_pfxpostpone |
| 809 | 6941 || ae->ae_chop != NULL |
| 6942 || ae->ae_flags != NULL) | |
| 481 | 6943 && (ae->ae_chop == NULL |
| 6944 || STRLEN(ae->ae_chop) < wordlen) | |
| 339 | 6945 && (ae->ae_prog == NULL |
| 820 | 6946 || vim_regexec(®match, word, (colnr_T)0)) |
| 6947 && (((condit & CONDIT_CFIX) == 0) | |
| 6948 == ((condit & CONDIT_AFF) == 0 | |
| 6949 || ae->ae_flags == NULL | |
| 6950 || !flag_in_afflist(affile->af_flagtype, | |
| 6951 ae->ae_flags, affile->af_circumfix)))) | |
| 300 | 6952 { |
| 6953 /* Match. Remove the chop and add the affix. */ | |
| 6954 if (xht == NULL) | |
| 240 | 6955 { |
| 300 | 6956 /* prefix: chop/add at the start of the word */ |
| 6957 if (ae->ae_add == NULL) | |
| 6958 *newword = NUL; | |
| 6959 else | |
| 2768 | 6960 vim_strncpy(newword, ae->ae_add, MAXWLEN - 1); |
| 300 | 6961 p = word; |
| 6962 if (ae->ae_chop != NULL) | |
| 310 | 6963 { |
| 300 | 6964 /* Skip chop string. */ |
| 310 | 6965 #ifdef FEAT_MBYTE |
| 6966 if (has_mbyte) | |
| 324 | 6967 { |
| 310 | 6968 i = mb_charlen(ae->ae_chop); |
| 324 | 6969 for ( ; i > 0; --i) |
| 6970 mb_ptr_adv(p); | |
| 6971 } | |
| 310 | 6972 else |
| 6973 #endif | |
| 324 | 6974 p += STRLEN(ae->ae_chop); |
| 310 | 6975 } |
| 300 | 6976 STRCAT(newword, p); |
| 6977 } | |
| 6978 else | |
| 6979 { | |
| 6980 /* suffix: chop/add at the end of the word */ | |
| 2768 | 6981 vim_strncpy(newword, word, MAXWLEN - 1); |
| 300 | 6982 if (ae->ae_chop != NULL) |
| 6983 { | |
| 6984 /* Remove chop string. */ | |
| 6985 p = newword + STRLEN(newword); | |
| 835 | 6986 i = (int)MB_CHARLEN(ae->ae_chop); |
| 310 | 6987 for ( ; i > 0; --i) |
| 300 | 6988 mb_ptr_back(newword, p); |
| 6989 *p = NUL; | |
| 6990 } | |
| 6991 if (ae->ae_add != NULL) | |
| 6992 STRCAT(newword, ae->ae_add); | |
| 240 | 6993 } |
| 6994 | |
| 820 | 6995 use_flags = flags; |
| 6996 use_pfxlist = pfxlist; | |
| 6997 use_pfxlen = pfxlen; | |
| 6998 need_affix = FALSE; | |
| 6999 use_condit = condit | CONDIT_COMB | CONDIT_AFF; | |
| 7000 if (ae->ae_flags != NULL) | |
| 7001 { | |
| 7002 /* Extract flags from the affix list. */ | |
| 7003 use_flags |= get_affix_flags(affile, ae->ae_flags); | |
| 7004 | |
| 7005 if (affile->af_needaffix != 0 && flag_in_afflist( | |
| 7006 affile->af_flagtype, ae->ae_flags, | |
| 7007 affile->af_needaffix)) | |
| 7008 need_affix = TRUE; | |
| 7009 | |
| 7010 /* When there is a CIRCUMFIX flag the other affix | |
| 7011 * must also have it and we don't add the word | |
| 7012 * with one affix. */ | |
| 7013 if (affile->af_circumfix != 0 && flag_in_afflist( | |
| 7014 affile->af_flagtype, ae->ae_flags, | |
| 7015 affile->af_circumfix)) | |
| 7016 { | |
| 7017 use_condit |= CONDIT_CFIX; | |
| 7018 if ((condit & CONDIT_CFIX) == 0) | |
| 7019 need_affix = TRUE; | |
| 7020 } | |
| 7021 | |
| 7022 if (affile->af_pfxpostpone | |
| 7023 || spin->si_compflags != NULL) | |
| 7024 { | |
| 7025 if (affile->af_pfxpostpone) | |
| 7026 /* Get prefix IDS from the affix list. */ | |
| 7027 use_pfxlen = get_pfxlist(affile, | |
| 7028 ae->ae_flags, store_afflist); | |
| 7029 else | |
| 7030 use_pfxlen = 0; | |
| 7031 use_pfxlist = store_afflist; | |
| 7032 | |
| 7033 /* Combine the prefix IDs. Avoid adding the | |
| 7034 * same ID twice. */ | |
| 7035 for (i = 0; i < pfxlen; ++i) | |
| 7036 { | |
| 7037 for (j = 0; j < use_pfxlen; ++j) | |
| 7038 if (pfxlist[i] == use_pfxlist[j]) | |
| 7039 break; | |
| 7040 if (j == use_pfxlen) | |
| 7041 use_pfxlist[use_pfxlen++] = pfxlist[i]; | |
| 7042 } | |
| 7043 | |
| 7044 if (spin->si_compflags != NULL) | |
| 7045 /* Get compound IDS from the affix list. */ | |
| 7046 get_compflags(affile, ae->ae_flags, | |
| 7047 use_pfxlist + use_pfxlen); | |
| 7048 | |
| 7049 /* Combine the list of compound flags. | |
| 7050 * Concatenate them to the prefix IDs list. | |
| 7051 * Avoid adding the same ID twice. */ | |
| 7052 for (i = pfxlen; pfxlist[i] != NUL; ++i) | |
| 7053 { | |
| 7054 for (j = use_pfxlen; | |
| 7055 use_pfxlist[j] != NUL; ++j) | |
| 7056 if (pfxlist[i] == use_pfxlist[j]) | |
| 7057 break; | |
| 7058 if (use_pfxlist[j] == NUL) | |
| 7059 { | |
| 7060 use_pfxlist[j++] = pfxlist[i]; | |
| 7061 use_pfxlist[j] = NUL; | |
| 7062 } | |
| 7063 } | |
| 7064 } | |
| 7065 } | |
| 483 | 7066 |
| 819 | 7067 /* Obey a "COMPOUNDFORBIDFLAG" of the affix: don't |
| 809 | 7068 * use the compound flags. */ |
| 822 | 7069 if (use_pfxlist != NULL && ae->ae_compforbid) |
| 483 | 7070 { |
| 820 | 7071 vim_strncpy(pfx_pfxlist, use_pfxlist, use_pfxlen); |
| 483 | 7072 use_pfxlist = pfx_pfxlist; |
| 7073 } | |
| 390 | 7074 |
| 7075 /* When there are postponed prefixes... */ | |
| 393 | 7076 if (spin->si_prefroot != NULL |
| 7077 && spin->si_prefroot->wn_sibling != NULL) | |
| 390 | 7078 { |
| 7079 /* ... add a flag to indicate an affix was used. */ | |
| 7080 use_flags |= WF_HAS_AFF; | |
| 7081 | |
| 7082 /* ... don't use a prefix list if combining | |
| 483 | 7083 * affixes is not allowed. But do use the |
| 7084 * compound flags after them. */ | |
| 827 | 7085 if (!ah->ah_combine && use_pfxlist != NULL) |
| 820 | 7086 use_pfxlist += use_pfxlen; |
| 390 | 7087 } |
| 366 | 7088 |
| 819 | 7089 /* When compounding is supported and there is no |
| 7090 * "COMPOUNDPERMITFLAG" then forbid compounding on the | |
| 7091 * side where the affix is applied. */ | |
| 822 | 7092 if (spin->si_compflags != NULL && !ae->ae_comppermit) |
| 819 | 7093 { |
| 7094 if (xht != NULL) | |
| 7095 use_flags |= WF_NOCOMPAFT; | |
| 7096 else | |
| 7097 use_flags |= WF_NOCOMPBEF; | |
| 7098 } | |
| 7099 | |
| 300 | 7100 /* Store the modified word. */ |
| 474 | 7101 if (store_word(spin, newword, use_flags, |
| 820 | 7102 spin->si_region, use_pfxlist, |
| 7103 need_affix) == FAIL) | |
| 300 | 7104 retval = FAIL; |
| 236 | 7105 |
| 820 | 7106 /* When added a prefix or a first suffix and the affix |
| 7107 * has flags may add a(nother) suffix. RECURSIVE! */ | |
| 7108 if ((condit & CONDIT_SUF) && ae->ae_flags != NULL) | |
| 7109 if (store_aff_word(spin, newword, ae->ae_flags, | |
| 7110 affile, &affile->af_suff, xht, | |
| 7111 use_condit & (xht == NULL | |
| 7112 ? ~0 : ~CONDIT_SUF), | |
| 7113 use_flags, use_pfxlist, pfxlen) == FAIL) | |
| 7114 retval = FAIL; | |
| 7115 | |
| 300 | 7116 /* When added a suffix and combining is allowed also |
| 820 | 7117 * try adding a prefix additionally. Both for the |
| 7118 * word flags and for the affix flags. RECURSIVE! */ | |
| 300 | 7119 if (xht != NULL && ah->ah_combine) |
| 820 | 7120 { |
| 7121 if (store_aff_word(spin, newword, | |
| 7122 afflist, affile, | |
| 7123 xht, NULL, use_condit, | |
| 7124 use_flags, use_pfxlist, | |
| 7125 pfxlen) == FAIL | |
| 7126 || (ae->ae_flags != NULL | |
| 7127 && store_aff_word(spin, newword, | |
| 7128 ae->ae_flags, affile, | |
| 7129 xht, NULL, use_condit, | |
| 7130 use_flags, use_pfxlist, | |
| 7131 pfxlen) == FAIL)) | |
| 300 | 7132 retval = FAIL; |
| 820 | 7133 } |
| 236 | 7134 } |
| 7135 } | |
| 7136 } | |
| 7137 } | |
| 7138 } | |
| 7139 | |
| 7140 return retval; | |
| 7141 } | |
| 7142 | |
| 7143 /* | |
| 300 | 7144 * Read a file with a list of words. |
| 236 | 7145 */ |
| 7146 static int | |
| 474 | 7147 spell_read_wordfile(spin, fname) |
| 7148 spellinfo_T *spin; | |
| 300 | 7149 char_u *fname; |
| 236 | 7150 { |
| 300 | 7151 FILE *fd; |
| 7152 long lnum = 0; | |
| 7153 char_u rline[MAXLINELEN]; | |
| 7154 char_u *line; | |
| 7155 char_u *pc = NULL; | |
| 381 | 7156 char_u *p; |
| 300 | 7157 int l; |
| 7158 int retval = OK; | |
| 7159 int did_word = FALSE; | |
| 7160 int non_ascii = 0; | |
| 307 | 7161 int flags; |
| 316 | 7162 int regionmask; |
| 236 | 7163 |
| 300 | 7164 /* |
| 7165 * Open the file. | |
| 7166 */ | |
| 310 | 7167 fd = mch_fopen((char *)fname, "r"); |
| 300 | 7168 if (fd == NULL) |
| 236 | 7169 { |
| 300 | 7170 EMSG2(_(e_notopen), fname); |
| 7171 return FAIL; | |
| 236 | 7172 } |
| 7173 | |
| 625 | 7174 vim_snprintf((char *)IObuff, IOSIZE, _("Reading word file %s ..."), fname); |
| 7175 spell_message(spin, IObuff); | |
| 300 | 7176 |
| 7177 /* | |
| 7178 * Read all the lines in the file one by one. | |
| 7179 */ | |
| 7180 while (!vim_fgets(rline, MAXLINELEN, fd) && !got_int) | |
| 7181 { | |
| 7182 line_breakcheck(); | |
| 7183 ++lnum; | |
| 7184 | |
| 7185 /* Skip comment lines. */ | |
| 7186 if (*rline == '#') | |
| 7187 continue; | |
| 7188 | |
| 7189 /* Remove CR, LF and white space from the end. */ | |
| 835 | 7190 l = (int)STRLEN(rline); |
| 300 | 7191 while (l > 0 && rline[l - 1] <= ' ') |
| 7192 --l; | |
| 7193 if (l == 0) | |
| 7194 continue; /* empty or blank line */ | |
| 7195 rline[l] = NUL; | |
| 7196 | |
| 859 | 7197 /* Convert from "/encoding={encoding}" to 'encoding' when needed. */ |
| 300 | 7198 vim_free(pc); |
| 310 | 7199 #ifdef FEAT_MBYTE |
| 300 | 7200 if (spin->si_conv.vc_type != CONV_NONE) |
| 7201 { | |
| 7202 pc = string_convert(&spin->si_conv, rline, NULL); | |
| 7203 if (pc == NULL) | |
| 7204 { | |
| 7205 smsg((char_u *)_("Conversion failure for word in %s line %d: %s"), | |
| 7206 fname, lnum, rline); | |
| 7207 continue; | |
| 7208 } | |
| 7209 line = pc; | |
| 7210 } | |
| 7211 else | |
| 310 | 7212 #endif |
| 300 | 7213 { |
| 7214 pc = NULL; | |
| 7215 line = rline; | |
| 7216 } | |
| 7217 | |
| 307 | 7218 if (*line == '/') |
| 300 | 7219 { |
| 307 | 7220 ++line; |
| 7221 if (STRNCMP(line, "encoding=", 9) == 0) | |
| 300 | 7222 { |
| 7223 if (spin->si_conv.vc_type != CONV_NONE) | |
| 316 | 7224 smsg((char_u *)_("Duplicate /encoding= line ignored in %s line %d: %s"), |
| 7225 fname, lnum, line - 1); | |
| 300 | 7226 else if (did_word) |
| 316 | 7227 smsg((char_u *)_("/encoding= line after word ignored in %s line %d: %s"), |
| 7228 fname, lnum, line - 1); | |
| 300 | 7229 else |
| 7230 { | |
| 310 | 7231 #ifdef FEAT_MBYTE |
| 7232 char_u *enc; | |
| 7233 | |
| 300 | 7234 /* Setup for conversion to 'encoding'. */ |
| 859 | 7235 line += 9; |
| 316 | 7236 enc = enc_canonize(line); |
| 300 | 7237 if (enc != NULL && !spin->si_ascii |
| 7238 && convert_setup(&spin->si_conv, enc, | |
| 7239 p_enc) == FAIL) | |
| 7240 smsg((char_u *)_("Conversion in %s not supported: from %s to %s"), | |
| 316 | 7241 fname, line, p_enc); |
| 300 | 7242 vim_free(enc); |
| 375 | 7243 spin->si_conv.vc_fail = TRUE; |
| 310 | 7244 #else |
| 7245 smsg((char_u *)_("Conversion in %s not supported"), fname); | |
| 7246 #endif | |
| 300 | 7247 } |
| 307 | 7248 continue; |
| 300 | 7249 } |
| 307 | 7250 |
| 316 | 7251 if (STRNCMP(line, "regions=", 8) == 0) |
| 7252 { | |
| 7253 if (spin->si_region_count > 1) | |
| 7254 smsg((char_u *)_("Duplicate /regions= line ignored in %s line %d: %s"), | |
| 7255 fname, lnum, line); | |
| 7256 else | |
| 7257 { | |
| 7258 line += 8; | |
| 7259 if (STRLEN(line) > 16) | |
| 7260 smsg((char_u *)_("Too many regions in %s line %d: %s"), | |
| 7261 fname, lnum, line); | |
| 7262 else | |
| 7263 { | |
| 835 | 7264 spin->si_region_count = (int)STRLEN(line) / 2; |
| 316 | 7265 STRCPY(spin->si_region_name, line); |
| 389 | 7266 |
| 7267 /* Adjust the mask for a word valid in all regions. */ | |
| 7268 spin->si_region = (1 << spin->si_region_count) - 1; | |
| 316 | 7269 } |
| 7270 } | |
| 7271 continue; | |
| 7272 } | |
| 7273 | |
| 381 | 7274 smsg((char_u *)_("/ line ignored in %s line %d: %s"), |
| 7275 fname, lnum, line - 1); | |
| 7276 continue; | |
| 7277 } | |
| 7278 | |
| 7279 flags = 0; | |
| 7280 regionmask = spin->si_region; | |
| 7281 | |
| 7282 /* Check for flags and region after a slash. */ | |
| 7283 p = vim_strchr(line, '/'); | |
| 7284 if (p != NULL) | |
| 7285 { | |
| 7286 *p++ = NUL; | |
| 7287 while (*p != NUL) | |
| 7288 { | |
| 7289 if (*p == '=') /* keep-case word */ | |
| 389 | 7290 flags |= WF_KEEPCAP | WF_FIXCAP; |
| 381 | 7291 else if (*p == '!') /* Bad, bad, wicked word. */ |
| 7292 flags |= WF_BANNED; | |
| 7293 else if (*p == '?') /* Rare word. */ | |
| 7294 flags |= WF_RARE; | |
| 7295 else if (VIM_ISDIGIT(*p)) /* region number(s) */ | |
| 316 | 7296 { |
| 381 | 7297 if ((flags & WF_REGION) == 0) /* first one */ |
| 7298 regionmask = 0; | |
| 7299 flags |= WF_REGION; | |
| 7300 | |
| 7301 l = *p - '0'; | |
| 316 | 7302 if (l > spin->si_region_count) |
| 7303 { | |
| 7304 smsg((char_u *)_("Invalid region nr in %s line %d: %s"), | |
| 381 | 7305 fname, lnum, p); |
| 316 | 7306 break; |
| 7307 } | |
| 7308 regionmask |= 1 << (l - 1); | |
| 7309 } | |
| 381 | 7310 else |
| 7311 { | |
| 7312 smsg((char_u *)_("Unrecognized flags in %s line %d: %s"), | |
| 7313 fname, lnum, p); | |
| 7314 break; | |
| 7315 } | |
| 7316 ++p; | |
| 307 | 7317 } |
| 300 | 7318 } |
| 7319 | |
| 7320 /* Skip non-ASCII words when "spin->si_ascii" is TRUE. */ | |
| 7321 if (spin->si_ascii && has_non_ascii(line)) | |
| 7322 { | |
| 7323 ++non_ascii; | |
| 7324 continue; | |
| 7325 } | |
| 7326 | |
| 7327 /* Normal word: store it. */ | |
| 483 | 7328 if (store_word(spin, line, flags, regionmask, NULL, FALSE) == FAIL) |
| 300 | 7329 { |
| 7330 retval = FAIL; | |
| 7331 break; | |
| 7332 } | |
| 7333 did_word = TRUE; | |
| 7334 } | |
| 7335 | |
| 7336 vim_free(pc); | |
| 7337 fclose(fd); | |
| 7338 | |
| 625 | 7339 if (spin->si_ascii && non_ascii > 0) |
| 7340 { | |
| 7341 vim_snprintf((char *)IObuff, IOSIZE, | |
| 7342 _("Ignored %d words with non-ASCII characters"), non_ascii); | |
| 7343 spell_message(spin, IObuff); | |
| 7344 } | |
| 7345 | |
| 300 | 7346 return retval; |
| 236 | 7347 } |
| 7348 | |
| 7349 /* | |
| 300 | 7350 * Get part of an sblock_T, "len" bytes long. |
| 474 | 7351 * This avoids calling free() for every little struct we use (and keeping |
| 7352 * track of them). | |
| 300 | 7353 * The memory is cleared to all zeros. |
| 7354 * Returns NULL when out of memory. | |
| 7355 */ | |
| 7356 static void * | |
| 474 | 7357 getroom(spin, len, align) |
| 7358 spellinfo_T *spin; | |
| 447 | 7359 size_t len; /* length needed */ |
| 7360 int align; /* align for pointer */ | |
| 300 | 7361 { |
| 7362 char_u *p; | |
| 474 | 7363 sblock_T *bl = spin->si_blocks; |
| 300 | 7364 |
| 447 | 7365 if (align && bl != NULL) |
| 7366 /* Round size up for alignment. On some systems structures need to be | |
| 7367 * aligned to the size of a pointer (e.g., SPARC). */ | |
| 7368 bl->sb_used = (bl->sb_used + sizeof(char *) - 1) | |
| 7369 & ~(sizeof(char *) - 1); | |
| 7370 | |
| 300 | 7371 if (bl == NULL || bl->sb_used + len > SBLOCKSIZE) |
| 7372 { | |
| 2691 | 7373 if (len >= SBLOCKSIZE) |
| 7374 bl = NULL; | |
| 7375 else | |
| 7376 /* Allocate a block of memory. It is not freed until much later. */ | |
| 7377 bl = (sblock_T *)alloc_clear( | |
| 7378 (unsigned)(sizeof(sblock_T) + SBLOCKSIZE)); | |
| 300 | 7379 if (bl == NULL) |
| 2691 | 7380 { |
| 7381 if (!spin->si_did_emsg) | |
| 7382 { | |
| 7383 EMSG(_("E845: Insufficient memory, word list will be incomplete")); | |
| 7384 spin->si_did_emsg = TRUE; | |
| 7385 } | |
| 300 | 7386 return NULL; |
| 2691 | 7387 } |
| 474 | 7388 bl->sb_next = spin->si_blocks; |
| 7389 spin->si_blocks = bl; | |
| 300 | 7390 bl->sb_used = 0; |
| 482 | 7391 ++spin->si_blocks_cnt; |
| 300 | 7392 } |
| 7393 | |
| 7394 p = bl->sb_data + bl->sb_used; | |
| 835 | 7395 bl->sb_used += (int)len; |
| 300 | 7396 |
| 7397 return p; | |
| 7398 } | |
| 7399 | |
| 7400 /* | |
| 7401 * Make a copy of a string into memory allocated with getroom(). | |
| 2691 | 7402 * Returns NULL when out of memory. |
| 300 | 7403 */ |
| 7404 static char_u * | |
| 474 | 7405 getroom_save(spin, s) |
| 7406 spellinfo_T *spin; | |
| 300 | 7407 char_u *s; |
| 7408 { | |
| 7409 char_u *sc; | |
| 7410 | |
| 474 | 7411 sc = (char_u *)getroom(spin, STRLEN(s) + 1, FALSE); |
| 300 | 7412 if (sc != NULL) |
| 7413 STRCPY(sc, s); | |
| 7414 return sc; | |
| 7415 } | |
| 7416 | |
| 7417 | |
| 7418 /* | |
| 7419 * Free the list of allocated sblock_T. | |
| 236 | 7420 */ |
| 7421 static void | |
| 300 | 7422 free_blocks(bl) |
| 7423 sblock_T *bl; | |
| 236 | 7424 { |
| 300 | 7425 sblock_T *next; |
| 236 | 7426 |
| 300 | 7427 while (bl != NULL) |
| 236 | 7428 { |
| 300 | 7429 next = bl->sb_next; |
| 7430 vim_free(bl); | |
| 7431 bl = next; | |
| 236 | 7432 } |
| 7433 } | |
| 7434 | |
| 7435 /* | |
| 300 | 7436 * Allocate the root of a word tree. |
| 2691 | 7437 * Returns NULL when out of memory. |
| 236 | 7438 */ |
| 300 | 7439 static wordnode_T * |
| 474 | 7440 wordtree_alloc(spin) |
| 7441 spellinfo_T *spin; | |
| 7442 { | |
| 7443 return (wordnode_T *)getroom(spin, sizeof(wordnode_T), TRUE); | |
| 236 | 7444 } |
| 7445 | |
| 7446 /* | |
| 300 | 7447 * Store a word in the tree(s). |
| 390 | 7448 * Always store it in the case-folded tree. For a keep-case word this is |
| 7449 * useful when the word can also be used with all caps (no WF_FIXCAP flag) and | |
| 7450 * used to find suggestions. | |
| 300 | 7451 * For a keep-case word also store it in the keep-case tree. |
| 482 | 7452 * When "pfxlist" is not NULL store the word for each postponed prefix ID and |
| 7453 * compound flag. | |
| 236 | 7454 */ |
| 7455 static int | |
| 483 | 7456 store_word(spin, word, flags, region, pfxlist, need_affix) |
| 474 | 7457 spellinfo_T *spin; |
| 300 | 7458 char_u *word; |
| 307 | 7459 int flags; /* extra flags, WF_BANNED */ |
| 316 | 7460 int region; /* supported region(s) */ |
| 339 | 7461 char_u *pfxlist; /* list of prefix IDs or NULL */ |
| 483 | 7462 int need_affix; /* only store word with affix ID */ |
| 236 | 7463 { |
| 835 | 7464 int len = (int)STRLEN(word); |
| 300 | 7465 int ct = captype(word, word + len); |
| 7466 char_u foldword[MAXWLEN]; | |
| 339 | 7467 int res = OK; |
| 7468 char_u *p; | |
| 236 | 7469 |
| 323 | 7470 (void)spell_casefold(word, len, foldword, MAXWLEN); |
| 339 | 7471 for (p = pfxlist; res == OK; ++p) |
| 7472 { | |
| 483 | 7473 if (!need_affix || (p != NULL && *p != NUL)) |
| 7474 res = tree_add_word(spin, foldword, spin->si_foldroot, ct | flags, | |
| 474 | 7475 region, p == NULL ? 0 : *p); |
| 339 | 7476 if (p == NULL || *p == NUL) |
| 7477 break; | |
| 7478 } | |
| 334 | 7479 ++spin->si_foldwcount; |
| 307 | 7480 |
| 474 | 7481 if (res == OK && (ct == WF_KEEPCAP || (flags & WF_KEEPCAP))) |
| 334 | 7482 { |
| 339 | 7483 for (p = pfxlist; res == OK; ++p) |
| 7484 { | |
| 483 | 7485 if (!need_affix || (p != NULL && *p != NUL)) |
| 7486 res = tree_add_word(spin, word, spin->si_keeproot, flags, | |
| 474 | 7487 region, p == NULL ? 0 : *p); |
| 339 | 7488 if (p == NULL || *p == NUL) |
| 7489 break; | |
| 7490 } | |
| 334 | 7491 ++spin->si_keepwcount; |
| 7492 } | |
| 300 | 7493 return res; |
| 236 | 7494 } |
| 7495 | |
| 7496 /* | |
| 300 | 7497 * Add word "word" to a word tree at "root". |
| 625 | 7498 * When "flags" < 0 we are adding to the prefix tree where "flags" is used for |
| 366 | 7499 * "rare" and "region" is the condition nr. |
| 255 | 7500 * Returns FAIL when out of memory. |
| 236 | 7501 */ |
| 255 | 7502 static int |
| 481 | 7503 tree_add_word(spin, word, root, flags, region, affixID) |
| 474 | 7504 spellinfo_T *spin; |
| 300 | 7505 char_u *word; |
| 7506 wordnode_T *root; | |
| 7507 int flags; | |
| 7508 int region; | |
| 481 | 7509 int affixID; |
| 236 | 7510 { |
| 300 | 7511 wordnode_T *node = root; |
| 7512 wordnode_T *np; | |
| 469 | 7513 wordnode_T *copyp, **copyprev; |
| 300 | 7514 wordnode_T **prev = NULL; |
| 7515 int i; | |
| 255 | 7516 |
| 300 | 7517 /* Add each byte of the word to the tree, including the NUL at the end. */ |
| 7518 for (i = 0; ; ++i) | |
| 255 | 7519 { |
| 469 | 7520 /* When there is more than one reference to this node we need to make |
| 7521 * a copy, so that we can modify it. Copy the whole list of siblings | |
| 7522 * (we don't optimize for a partly shared list of siblings). */ | |
| 7523 if (node != NULL && node->wn_refs > 1) | |
| 7524 { | |
| 7525 --node->wn_refs; | |
| 7526 copyprev = prev; | |
| 7527 for (copyp = node; copyp != NULL; copyp = copyp->wn_sibling) | |
| 7528 { | |
| 7529 /* Allocate a new node and copy the info. */ | |
| 474 | 7530 np = get_wordnode(spin); |
| 469 | 7531 if (np == NULL) |
| 7532 return FAIL; | |
| 7533 np->wn_child = copyp->wn_child; | |
| 7534 if (np->wn_child != NULL) | |
| 7535 ++np->wn_child->wn_refs; /* child gets extra ref */ | |
| 7536 np->wn_byte = copyp->wn_byte; | |
| 7537 if (np->wn_byte == NUL) | |
| 7538 { | |
| 7539 np->wn_flags = copyp->wn_flags; | |
| 7540 np->wn_region = copyp->wn_region; | |
| 481 | 7541 np->wn_affixID = copyp->wn_affixID; |
| 469 | 7542 } |
| 7543 | |
| 7544 /* Link the new node in the list, there will be one ref. */ | |
| 7545 np->wn_refs = 1; | |
| 840 | 7546 if (copyprev != NULL) |
| 7547 *copyprev = np; | |
| 469 | 7548 copyprev = &np->wn_sibling; |
| 7549 | |
| 7550 /* Let "node" point to the head of the copied list. */ | |
| 7551 if (copyp == node) | |
| 7552 node = np; | |
| 7553 } | |
| 7554 } | |
| 7555 | |
| 300 | 7556 /* Look for the sibling that has the same character. They are sorted |
| 7557 * on byte value, thus stop searching when a sibling is found with a | |
| 339 | 7558 * higher byte value. For zero bytes (end of word) the sorting is |
| 481 | 7559 * done on flags and then on affixID. */ |
| 339 | 7560 while (node != NULL |
| 7561 && (node->wn_byte < word[i] | |
| 7562 || (node->wn_byte == NUL | |
| 7563 && (flags < 0 | |
| 625 | 7564 ? node->wn_affixID < (unsigned)affixID |
| 7565 : (node->wn_flags < (unsigned)(flags & WN_MASK) | |
| 390 | 7566 || (node->wn_flags == (flags & WN_MASK) |
| 625 | 7567 && (spin->si_sugtree |
| 7568 ? (node->wn_region & 0xffff) < region | |
| 7569 : node->wn_affixID | |
| 7570 < (unsigned)affixID))))))) | |
| 236 | 7571 { |
| 300 | 7572 prev = &node->wn_sibling; |
| 7573 node = *prev; | |
| 236 | 7574 } |
| 339 | 7575 if (node == NULL |
| 7576 || node->wn_byte != word[i] | |
| 7577 || (word[i] == NUL | |
| 7578 && (flags < 0 | |
| 625 | 7579 || spin->si_sugtree |
| 390 | 7580 || node->wn_flags != (flags & WN_MASK) |
| 481 | 7581 || node->wn_affixID != affixID))) |
| 255 | 7582 { |
| 300 | 7583 /* Allocate a new node. */ |
| 474 | 7584 np = get_wordnode(spin); |
| 300 | 7585 if (np == NULL) |
| 7586 return FAIL; | |
| 7587 np->wn_byte = word[i]; | |
| 469 | 7588 |
| 7589 /* If "node" is NULL this is a new child or the end of the sibling | |
| 7590 * list: ref count is one. Otherwise use ref count of sibling and | |
| 7591 * make ref count of sibling one (matters when inserting in front | |
| 7592 * of the list of siblings). */ | |
| 7593 if (node == NULL) | |
| 7594 np->wn_refs = 1; | |
| 7595 else | |
| 7596 { | |
| 7597 np->wn_refs = node->wn_refs; | |
| 7598 node->wn_refs = 1; | |
| 7599 } | |
|
2299
92e31a1b0c2f
Fix crash in generating spell .sug file.
Bram Moolenaar <bram@vim.org>
parents:
2250
diff
changeset
|
7600 if (prev != NULL) |
|
92e31a1b0c2f
Fix crash in generating spell .sug file.
Bram Moolenaar <bram@vim.org>
parents:
2250
diff
changeset
|
7601 *prev = np; |
| 300 | 7602 np->wn_sibling = node; |
| 7603 node = np; | |
| 255 | 7604 } |
| 300 | 7605 |
| 7606 if (word[i] == NUL) | |
| 7607 { | |
| 7608 node->wn_flags = flags; | |
| 7609 node->wn_region |= region; | |
| 481 | 7610 node->wn_affixID = affixID; |
| 300 | 7611 break; |
| 7612 } | |
| 7613 prev = &node->wn_child; | |
| 7614 node = *prev; | |
| 255 | 7615 } |
| 469 | 7616 #ifdef SPELL_PRINTTREE |
| 7617 smsg("Added \"%s\"", word); | |
| 7618 spell_print_tree(root->wn_sibling); | |
| 7619 #endif | |
| 7620 | |
| 474 | 7621 /* count nr of words added since last message */ |
| 7622 ++spin->si_msg_count; | |
| 7623 | |
| 482 | 7624 if (spin->si_compress_cnt > 1) |
| 7625 { | |
| 7626 if (--spin->si_compress_cnt == 1) | |
| 7627 /* Did enough words to lower the block count limit. */ | |
| 483 | 7628 spin->si_blocks_cnt += compress_inc; |
| 482 | 7629 } |
| 7630 | |
| 469 | 7631 /* |
| 482 | 7632 * When we have allocated lots of memory we need to compress the word tree |
| 7633 * to free up some room. But compression is slow, and we might actually | |
| 7634 * need that room, thus only compress in the following situations: | |
| 7635 * 1. When not compressed before (si_compress_cnt == 0): when using | |
| 483 | 7636 * "compress_start" blocks. |
| 7637 * 2. When compressed before and used "compress_inc" blocks before | |
| 7638 * adding "compress_added" words (si_compress_cnt > 1). | |
| 7639 * 3. When compressed before, added "compress_added" words | |
| 482 | 7640 * (si_compress_cnt == 1) and the number of free nodes drops below the |
| 7641 * maximum word length. | |
| 469 | 7642 */ |
| 482 | 7643 #ifndef SPELL_PRINTTREE |
| 7644 if (spin->si_compress_cnt == 1 | |
| 7645 ? spin->si_free_count < MAXWLEN | |
| 483 | 7646 : spin->si_blocks_cnt >= compress_start) |
| 482 | 7647 #endif |
| 7648 { | |
| 7649 /* Decrement the block counter. The effect is that we compress again | |
| 483 | 7650 * when the freed up room has been used and another "compress_inc" |
| 7651 * blocks have been allocated. Unless "compress_added" words have | |
| 482 | 7652 * been added, then the limit is put back again. */ |
| 483 | 7653 spin->si_blocks_cnt -= compress_inc; |
| 7654 spin->si_compress_cnt = compress_added; | |
| 482 | 7655 |
| 474 | 7656 if (spin->si_verbose) |
| 7657 { | |
| 7658 msg_start(); | |
| 7659 msg_puts((char_u *)_(msg_compressing)); | |
| 7660 msg_clr_eos(); | |
| 7661 msg_didout = FALSE; | |
| 7662 msg_col = 0; | |
| 7663 out_flush(); | |
| 7664 } | |
| 482 | 7665 |
| 7666 /* Compress both trees. Either they both have many nodes, which makes | |
| 7667 * compression useful, or one of them is small, which means | |
| 625 | 7668 * compression goes fast. But when filling the souldfold word tree |
| 7669 * there is no keep-case tree. */ | |
| 482 | 7670 wordtree_compress(spin, spin->si_foldroot); |
| 625 | 7671 if (affixID >= 0) |
| 7672 wordtree_compress(spin, spin->si_keeproot); | |
| 469 | 7673 } |
| 255 | 7674 |
| 7675 return OK; | |
| 236 | 7676 } |
| 7677 | |
| 469 | 7678 /* |
| 483 | 7679 * Check the 'mkspellmem' option. Return FAIL if it's wrong. |
| 7680 * Sets "sps_flags". | |
| 7681 */ | |
| 7682 int | |
| 7683 spell_check_msm() | |
| 7684 { | |
| 7685 char_u *p = p_msm; | |
| 7686 long start = 0; | |
| 944 | 7687 long incr = 0; |
| 483 | 7688 long added = 0; |
| 7689 | |
| 7690 if (!VIM_ISDIGIT(*p)) | |
| 7691 return FAIL; | |
| 7692 /* block count = (value * 1024) / SBLOCKSIZE (but avoid overflow)*/ | |
| 7693 start = (getdigits(&p) * 10) / (SBLOCKSIZE / 102); | |
| 7694 if (*p != ',') | |
| 7695 return FAIL; | |
| 7696 ++p; | |
| 7697 if (!VIM_ISDIGIT(*p)) | |
| 7698 return FAIL; | |
| 944 | 7699 incr = (getdigits(&p) * 102) / (SBLOCKSIZE / 10); |
| 483 | 7700 if (*p != ',') |
| 7701 return FAIL; | |
| 7702 ++p; | |
| 7703 if (!VIM_ISDIGIT(*p)) | |
| 7704 return FAIL; | |
| 7705 added = getdigits(&p) * 1024; | |
| 7706 if (*p != NUL) | |
| 7707 return FAIL; | |
| 7708 | |
| 944 | 7709 if (start == 0 || incr == 0 || added == 0 || incr > start) |
| 483 | 7710 return FAIL; |
| 7711 | |
| 7712 compress_start = start; | |
| 944 | 7713 compress_inc = incr; |
| 483 | 7714 compress_added = added; |
| 7715 return OK; | |
| 7716 } | |
| 7717 | |
| 7718 | |
| 7719 /* | |
| 469 | 7720 * Get a wordnode_T, either from the list of previously freed nodes or |
| 7721 * allocate a new one. | |
| 2691 | 7722 * Returns NULL when out of memory. |
| 469 | 7723 */ |
| 7724 static wordnode_T * | |
| 474 | 7725 get_wordnode(spin) |
| 7726 spellinfo_T *spin; | |
| 469 | 7727 { |
| 7728 wordnode_T *n; | |
| 7729 | |
| 474 | 7730 if (spin->si_first_free == NULL) |
| 7731 n = (wordnode_T *)getroom(spin, sizeof(wordnode_T), TRUE); | |
| 469 | 7732 else |
| 7733 { | |
| 474 | 7734 n = spin->si_first_free; |
| 7735 spin->si_first_free = n->wn_child; | |
| 469 | 7736 vim_memset(n, 0, sizeof(wordnode_T)); |
| 482 | 7737 --spin->si_free_count; |
| 469 | 7738 } |
| 7739 #ifdef SPELL_PRINTTREE | |
| 2691 | 7740 if (n != NULL) |
| 7741 n->wn_nr = ++spin->si_wordnode_nr; | |
| 469 | 7742 #endif |
| 7743 return n; | |
| 7744 } | |
| 7745 | |
| 7746 /* | |
| 474 | 7747 * Decrement the reference count on a node (which is the head of a list of |
| 7748 * siblings). If the reference count becomes zero free the node and its | |
| 7749 * siblings. | |
| 625 | 7750 * Returns the number of nodes actually freed. |
| 7751 */ | |
| 7752 static int | |
| 474 | 7753 deref_wordnode(spin, node) |
| 7754 spellinfo_T *spin; | |
| 7755 wordnode_T *node; | |
| 7756 { | |
| 625 | 7757 wordnode_T *np; |
| 7758 int cnt = 0; | |
| 474 | 7759 |
| 7760 if (--node->wn_refs == 0) | |
| 625 | 7761 { |
| 474 | 7762 for (np = node; np != NULL; np = np->wn_sibling) |
| 7763 { | |
| 7764 if (np->wn_child != NULL) | |
| 625 | 7765 cnt += deref_wordnode(spin, np->wn_child); |
| 474 | 7766 free_wordnode(spin, np); |
| 625 | 7767 ++cnt; |
| 7768 } | |
| 7769 ++cnt; /* length field */ | |
| 7770 } | |
| 7771 return cnt; | |
| 474 | 7772 } |
| 7773 | |
| 7774 /* | |
| 469 | 7775 * Free a wordnode_T for re-use later. |
| 474 | 7776 * Only the "wn_child" field becomes invalid. |
| 469 | 7777 */ |
| 7778 static void | |
| 474 | 7779 free_wordnode(spin, n) |
| 7780 spellinfo_T *spin; | |
| 7781 wordnode_T *n; | |
| 7782 { | |
| 7783 n->wn_child = spin->si_first_free; | |
| 7784 spin->si_first_free = n; | |
| 482 | 7785 ++spin->si_free_count; |
| 469 | 7786 } |
| 7787 | |
| 236 | 7788 /* |
| 300 | 7789 * Compress a tree: find tails that are identical and can be shared. |
| 7790 */ | |
| 7791 static void | |
| 474 | 7792 wordtree_compress(spin, root) |
| 7793 spellinfo_T *spin; | |
| 300 | 7794 wordnode_T *root; |
| 7795 { | |
| 7796 hashtab_T ht; | |
| 7797 int n; | |
| 7798 int tot = 0; | |
| 469 | 7799 int perc; |
| 300 | 7800 |
| 474 | 7801 /* Skip the root itself, it's not actually used. The first sibling is the |
| 7802 * start of the tree. */ | |
| 7803 if (root->wn_sibling != NULL) | |
| 300 | 7804 { |
| 7805 hash_init(&ht); | |
| 474 | 7806 n = node_compress(spin, root->wn_sibling, &ht, &tot); |
| 469 | 7807 |
| 7808 #ifndef SPELL_PRINTTREE | |
| 310 | 7809 if (spin->si_verbose || p_verbose > 2) |
| 469 | 7810 #endif |
| 310 | 7811 { |
| 469 | 7812 if (tot > 1000000) |
| 7813 perc = (tot - n) / (tot / 100); | |
| 482 | 7814 else if (tot == 0) |
| 7815 perc = 0; | |
| 469 | 7816 else |
| 7817 perc = (tot - n) * 100 / tot; | |
| 625 | 7818 vim_snprintf((char *)IObuff, IOSIZE, |
| 7819 _("Compressed %d of %d nodes; %d (%d%%) remaining"), | |
| 7820 n, tot, tot - n, perc); | |
| 7821 spell_message(spin, IObuff); | |
| 310 | 7822 } |
| 469 | 7823 #ifdef SPELL_PRINTTREE |
| 474 | 7824 spell_print_tree(root->wn_sibling); |
| 469 | 7825 #endif |
| 300 | 7826 hash_clear(&ht); |
| 7827 } | |
| 7828 } | |
| 7829 | |
| 7830 /* | |
| 7831 * Compress a node, its siblings and its children, depth first. | |
| 7832 * Returns the number of compressed nodes. | |
| 236 | 7833 */ |
| 255 | 7834 static int |
| 474 | 7835 node_compress(spin, node, ht, tot) |
| 7836 spellinfo_T *spin; | |
| 300 | 7837 wordnode_T *node; |
| 7838 hashtab_T *ht; | |
| 7839 int *tot; /* total count of nodes before compressing, | |
| 7840 incremented while going through the tree */ | |
| 236 | 7841 { |
| 300 | 7842 wordnode_T *np; |
| 7843 wordnode_T *tp; | |
| 7844 wordnode_T *child; | |
| 7845 hash_T hash; | |
| 236 | 7846 hashitem_T *hi; |
| 300 | 7847 int len = 0; |
| 7848 unsigned nr, n; | |
| 7849 int compressed = 0; | |
| 236 | 7850 |
| 300 | 7851 /* |
| 7852 * Go through the list of siblings. Compress each child and then try | |
| 7853 * finding an identical child to replace it. | |
| 7854 * Note that with "child" we mean not just the node that is pointed to, | |
| 625 | 7855 * but the whole list of siblings of which the child node is the first. |
| 300 | 7856 */ |
| 482 | 7857 for (np = node; np != NULL && !got_int; np = np->wn_sibling) |
| 236 | 7858 { |
| 300 | 7859 ++len; |
| 7860 if ((child = np->wn_child) != NULL) | |
| 7861 { | |
| 625 | 7862 /* Compress the child first. This fills hashkey. */ |
| 474 | 7863 compressed += node_compress(spin, child, ht, tot); |
| 300 | 7864 |
| 7865 /* Try to find an identical child. */ | |
| 346 | 7866 hash = hash_hash(child->wn_u1.hashkey); |
| 7867 hi = hash_lookup(ht, child->wn_u1.hashkey, hash); | |
| 300 | 7868 if (!HASHITEM_EMPTY(hi)) |
| 7869 { | |
| 625 | 7870 /* There are children we encountered before with a hash value |
| 7871 * identical to the current child. Now check if there is one | |
| 7872 * that is really identical. */ | |
| 346 | 7873 for (tp = HI2WN(hi); tp != NULL; tp = tp->wn_u2.next) |
| 300 | 7874 if (node_equal(child, tp)) |
| 7875 { | |
| 7876 /* Found one! Now use that child in place of the | |
| 469 | 7877 * current one. This means the current child and all |
| 7878 * its siblings is unlinked from the tree. */ | |
| 7879 ++tp->wn_refs; | |
| 625 | 7880 compressed += deref_wordnode(spin, child); |
| 300 | 7881 np->wn_child = tp; |
| 7882 break; | |
| 7883 } | |
| 7884 if (tp == NULL) | |
| 7885 { | |
| 7886 /* No other child with this hash value equals the child of | |
| 7887 * the node, add it to the linked list after the first | |
| 7888 * item. */ | |
| 7889 tp = HI2WN(hi); | |
| 346 | 7890 child->wn_u2.next = tp->wn_u2.next; |
| 7891 tp->wn_u2.next = child; | |
| 300 | 7892 } |
| 7893 } | |
| 7894 else | |
| 7895 /* No other child has this hash value, add it to the | |
| 7896 * hashtable. */ | |
| 346 | 7897 hash_add_item(ht, hi, child->wn_u1.hashkey, hash); |
| 300 | 7898 } |
| 236 | 7899 } |
| 625 | 7900 *tot += len + 1; /* add one for the node that stores the length */ |
| 300 | 7901 |
| 7902 /* | |
| 7903 * Make a hash key for the node and its siblings, so that we can quickly | |
| 7904 * find a lookalike node. This must be done after compressing the sibling | |
| 7905 * list, otherwise the hash key would become invalid by the compression. | |
| 7906 */ | |
| 346 | 7907 node->wn_u1.hashkey[0] = len; |
| 300 | 7908 nr = 0; |
| 7909 for (np = node; np != NULL; np = np->wn_sibling) | |
| 236 | 7910 { |
| 300 | 7911 if (np->wn_byte == NUL) |
| 481 | 7912 /* end node: use wn_flags, wn_region and wn_affixID */ |
| 7913 n = np->wn_flags + (np->wn_region << 8) + (np->wn_affixID << 16); | |
| 300 | 7914 else |
| 7915 /* byte node: use the byte value and the child pointer */ | |
| 835 | 7916 n = (unsigned)(np->wn_byte + ((long_u)np->wn_child << 8)); |
| 300 | 7917 nr = nr * 101 + n; |
| 236 | 7918 } |
| 300 | 7919 |
| 7920 /* Avoid NUL bytes, it terminates the hash key. */ | |
| 7921 n = nr & 0xff; | |
| 346 | 7922 node->wn_u1.hashkey[1] = n == 0 ? 1 : n; |
| 300 | 7923 n = (nr >> 8) & 0xff; |
| 346 | 7924 node->wn_u1.hashkey[2] = n == 0 ? 1 : n; |
| 300 | 7925 n = (nr >> 16) & 0xff; |
| 346 | 7926 node->wn_u1.hashkey[3] = n == 0 ? 1 : n; |
| 300 | 7927 n = (nr >> 24) & 0xff; |
| 346 | 7928 node->wn_u1.hashkey[4] = n == 0 ? 1 : n; |
| 7929 node->wn_u1.hashkey[5] = NUL; | |
| 300 | 7930 |
| 482 | 7931 /* Check for CTRL-C pressed now and then. */ |
| 7932 fast_breakcheck(); | |
| 7933 | |
| 300 | 7934 return compressed; |
| 7935 } | |
| 7936 | |
| 7937 /* | |
| 7938 * Return TRUE when two nodes have identical siblings and children. | |
| 7939 */ | |
| 7940 static int | |
| 7941 node_equal(n1, n2) | |
| 7942 wordnode_T *n1; | |
| 7943 wordnode_T *n2; | |
| 7944 { | |
| 7945 wordnode_T *p1; | |
| 7946 wordnode_T *p2; | |
| 7947 | |
| 7948 for (p1 = n1, p2 = n2; p1 != NULL && p2 != NULL; | |
| 7949 p1 = p1->wn_sibling, p2 = p2->wn_sibling) | |
| 7950 if (p1->wn_byte != p2->wn_byte | |
| 7951 || (p1->wn_byte == NUL | |
| 7952 ? (p1->wn_flags != p2->wn_flags | |
| 339 | 7953 || p1->wn_region != p2->wn_region |
| 481 | 7954 || p1->wn_affixID != p2->wn_affixID) |
| 300 | 7955 : (p1->wn_child != p2->wn_child))) |
| 7956 break; | |
| 7957 | |
| 7958 return p1 == NULL && p2 == NULL; | |
| 236 | 7959 } |
| 7960 | |
| 323 | 7961 static int |
| 7962 #ifdef __BORLANDC__ | |
| 7963 _RTLENTRYF | |
| 7964 #endif | |
| 7965 rep_compare __ARGS((const void *s1, const void *s2)); | |
| 7966 | |
| 7967 /* | |
| 7968 * Function given to qsort() to sort the REP items on "from" string. | |
| 7969 */ | |
| 7970 static int | |
| 7971 #ifdef __BORLANDC__ | |
| 7972 _RTLENTRYF | |
| 7973 #endif | |
| 7974 rep_compare(s1, s2) | |
| 7975 const void *s1; | |
| 7976 const void *s2; | |
| 7977 { | |
| 7978 fromto_T *p1 = (fromto_T *)s1; | |
| 7979 fromto_T *p2 = (fromto_T *)s2; | |
| 7980 | |
| 7981 return STRCMP(p1->ft_from, p2->ft_from); | |
| 7982 } | |
| 7983 | |
| 236 | 7984 /* |
| 483 | 7985 * Write the Vim .spl file "fname". |
| 500 | 7986 * Return FAIL or OK; |
| 7987 */ | |
| 7988 static int | |
| 474 | 7989 write_vim_spell(spin, fname) |
| 7990 spellinfo_T *spin; | |
| 236 | 7991 char_u *fname; |
| 7992 { | |
| 300 | 7993 FILE *fd; |
| 7994 int regionmask; | |
| 236 | 7995 int round; |
| 300 | 7996 wordnode_T *tree; |
| 7997 int nodecount; | |
| 323 | 7998 int i; |
| 7999 int l; | |
| 8000 garray_T *gap; | |
| 8001 fromto_T *ftp; | |
| 8002 char_u *p; | |
| 8003 int rr; | |
| 500 | 8004 int retval = OK; |
| 1759 | 8005 size_t fwv = 1; /* collect return value of fwrite() to avoid |
| 1753 | 8006 warnings from picky compiler */ |
| 236 | 8007 |
| 310 | 8008 fd = mch_fopen((char *)fname, "w"); |
| 300 | 8009 if (fd == NULL) |
| 236 | 8010 { |
| 8011 EMSG2(_(e_notopen), fname); | |
| 500 | 8012 return FAIL; |
| 236 | 8013 } |
| 8014 | |
| 483 | 8015 /* <HEADER>: <fileID> <versionnr> */ |
| 300 | 8016 /* <fileID> */ |
| 1753 | 8017 fwv &= fwrite(VIMSPELLMAGIC, VIMSPELLMAGICL, (size_t)1, fd); |
| 1759 | 8018 if (fwv != (size_t)1) |
| 8019 /* Catch first write error, don't try writing more. */ | |
| 8020 goto theend; | |
| 8021 | |
| 483 | 8022 putc(VIMSPELLVERSION, fd); /* <versionnr> */ |
| 8023 | |
| 8024 /* | |
| 8025 * <SECTIONS>: <section> ... <sectionend> | |
| 8026 */ | |
| 8027 | |
| 714 | 8028 /* SN_INFO: <infotext> */ |
| 8029 if (spin->si_info != NULL) | |
| 8030 { | |
| 8031 putc(SN_INFO, fd); /* <sectionID> */ | |
| 8032 putc(0, fd); /* <sectionflags> */ | |
| 8033 | |
| 835 | 8034 i = (int)STRLEN(spin->si_info); |
| 714 | 8035 put_bytes(fd, (long_u)i, 4); /* <sectionlen> */ |
| 1753 | 8036 fwv &= fwrite(spin->si_info, (size_t)i, (size_t)1, fd); /* <infotext> */ |
| 714 | 8037 } |
| 8038 | |
| 483 | 8039 /* SN_REGION: <regionname> ... |
| 8040 * Write the region names only if there is more than one. */ | |
| 316 | 8041 if (spin->si_region_count > 1) |
| 236 | 8042 { |
| 483 | 8043 putc(SN_REGION, fd); /* <sectionID> */ |
| 8044 putc(SNF_REQUIRED, fd); /* <sectionflags> */ | |
| 8045 l = spin->si_region_count * 2; | |
| 8046 put_bytes(fd, (long_u)l, 4); /* <sectionlen> */ | |
| 1753 | 8047 fwv &= fwrite(spin->si_region_name, (size_t)l, (size_t)1, fd); |
| 483 | 8048 /* <regionname> ... */ |
| 316 | 8049 regionmask = (1 << spin->si_region_count) - 1; |
| 236 | 8050 } |
| 8051 else | |
| 300 | 8052 regionmask = 0; |
| 483 | 8053 |
| 8054 /* SN_CHARFLAGS: <charflagslen> <charflags> <folcharslen> <folchars> | |
| 8055 * | |
| 8056 * The table with character flags and the table for case folding. | |
| 8057 * This makes sure the same characters are recognized as word characters | |
| 8058 * when generating an when using a spell file. | |
| 260 | 8059 * Skip this for ASCII, the table may conflict with the one used for |
| 323 | 8060 * 'encoding'. |
| 8061 * Also skip this for an .add.spl file, the main spell file must contain | |
| 8062 * the table (avoids that it conflicts). File is shorter too. | |
| 8063 */ | |
| 483 | 8064 if (!spin->si_ascii && !spin->si_add) |
| 8065 { | |
| 8066 char_u folchars[128 * 8]; | |
| 8067 int flags; | |
| 8068 | |
| 485 | 8069 putc(SN_CHARFLAGS, fd); /* <sectionID> */ |
| 483 | 8070 putc(SNF_REQUIRED, fd); /* <sectionflags> */ |
| 8071 | |
| 8072 /* Form the <folchars> string first, we need to know its length. */ | |
| 8073 l = 0; | |
| 8074 for (i = 128; i < 256; ++i) | |
| 8075 { | |
| 8076 #ifdef FEAT_MBYTE | |
| 8077 if (has_mbyte) | |
| 8078 l += mb_char2bytes(spelltab.st_fold[i], folchars + l); | |
| 8079 else | |
| 8080 #endif | |
| 8081 folchars[l++] = spelltab.st_fold[i]; | |
| 8082 } | |
| 8083 put_bytes(fd, (long_u)(1 + 128 + 2 + l), 4); /* <sectionlen> */ | |
| 8084 | |
| 8085 fputc(128, fd); /* <charflagslen> */ | |
| 8086 for (i = 128; i < 256; ++i) | |
| 8087 { | |
| 8088 flags = 0; | |
| 8089 if (spelltab.st_isw[i]) | |
| 8090 flags |= CF_WORD; | |
| 8091 if (spelltab.st_isu[i]) | |
| 8092 flags |= CF_UPPER; | |
| 8093 fputc(flags, fd); /* <charflags> */ | |
| 8094 } | |
| 8095 | |
| 8096 put_bytes(fd, (long_u)l, 2); /* <folcharslen> */ | |
| 1753 | 8097 fwv &= fwrite(folchars, (size_t)l, (size_t)1, fd); /* <folchars> */ |
| 483 | 8098 } |
| 8099 | |
| 8100 /* SN_MIDWORD: <midword> */ | |
| 8101 if (spin->si_midword != NULL) | |
| 8102 { | |
| 8103 putc(SN_MIDWORD, fd); /* <sectionID> */ | |
| 8104 putc(SNF_REQUIRED, fd); /* <sectionflags> */ | |
| 8105 | |
| 835 | 8106 i = (int)STRLEN(spin->si_midword); |
| 483 | 8107 put_bytes(fd, (long_u)i, 4); /* <sectionlen> */ |
| 1753 | 8108 fwv &= fwrite(spin->si_midword, (size_t)i, (size_t)1, fd); |
| 8109 /* <midword> */ | |
| 366 | 8110 } |
| 8111 | |
| 483 | 8112 /* SN_PREFCOND: <prefcondcnt> <prefcond> ... */ |
| 8113 if (spin->si_prefcond.ga_len > 0) | |
| 8114 { | |
| 8115 putc(SN_PREFCOND, fd); /* <sectionID> */ | |
| 8116 putc(SNF_REQUIRED, fd); /* <sectionflags> */ | |
| 8117 | |
| 8118 l = write_spell_prefcond(NULL, &spin->si_prefcond); | |
| 8119 put_bytes(fd, (long_u)l, 4); /* <sectionlen> */ | |
| 8120 | |
| 8121 write_spell_prefcond(fd, &spin->si_prefcond); | |
| 8122 } | |
| 8123 | |
| 8124 /* SN_REP: <repcount> <rep> ... | |
| 625 | 8125 * SN_SAL: <salflags> <salcount> <sal> ... |
| 8126 * SN_REPSAL: <repcount> <rep> ... */ | |
| 323 | 8127 |
| 483 | 8128 /* round 1: SN_REP section |
| 625 | 8129 * round 2: SN_SAL section (unless SN_SOFO is used) |
| 8130 * round 3: SN_REPSAL section */ | |
| 8131 for (round = 1; round <= 3; ++round) | |
| 323 | 8132 { |
| 8133 if (round == 1) | |
| 8134 gap = &spin->si_rep; | |
| 625 | 8135 else if (round == 2) |
| 8136 { | |
| 8137 /* Don't write SN_SAL when using a SN_SOFO section */ | |
| 8138 if (spin->si_sofofr != NULL && spin->si_sofoto != NULL) | |
| 8139 continue; | |
| 8140 gap = &spin->si_sal; | |
| 483 | 8141 } |
| 323 | 8142 else |
| 625 | 8143 gap = &spin->si_repsal; |
| 8144 | |
| 8145 /* Don't write the section if there are no items. */ | |
| 8146 if (gap->ga_len == 0) | |
| 8147 continue; | |
| 8148 | |
| 8149 /* Sort the REP/REPSAL items. */ | |
| 8150 if (round != 2) | |
| 8151 qsort(gap->ga_data, (size_t)gap->ga_len, | |
| 8152 sizeof(fromto_T), rep_compare); | |
| 8153 | |
| 8154 i = round == 1 ? SN_REP : (round == 2 ? SN_SAL : SN_REPSAL); | |
| 8155 putc(i, fd); /* <sectionID> */ | |
| 483 | 8156 |
| 8157 /* This is for making suggestions, section is not required. */ | |
| 8158 putc(0, fd); /* <sectionflags> */ | |
| 8159 | |
| 8160 /* Compute the length of what follows. */ | |
| 8161 l = 2; /* count <repcount> or <salcount> */ | |
| 8162 for (i = 0; i < gap->ga_len; ++i) | |
| 8163 { | |
| 8164 ftp = &((fromto_T *)gap->ga_data)[i]; | |
| 835 | 8165 l += 1 + (int)STRLEN(ftp->ft_from); /* count <*fromlen> and <*from> */ |
| 8166 l += 1 + (int)STRLEN(ftp->ft_to); /* count <*tolen> and <*to> */ | |
| 483 | 8167 } |
| 8168 if (round == 2) | |
| 8169 ++l; /* count <salflags> */ | |
| 8170 put_bytes(fd, (long_u)l, 4); /* <sectionlen> */ | |
| 8171 | |
| 8172 if (round == 2) | |
| 8173 { | |
| 323 | 8174 i = 0; |
| 8175 if (spin->si_followup) | |
| 8176 i |= SAL_F0LLOWUP; | |
| 8177 if (spin->si_collapse) | |
| 8178 i |= SAL_COLLAPSE; | |
| 8179 if (spin->si_rem_accents) | |
| 8180 i |= SAL_REM_ACCENTS; | |
| 8181 putc(i, fd); /* <salflags> */ | |
| 8182 } | |
| 8183 | |
| 8184 put_bytes(fd, (long_u)gap->ga_len, 2); /* <repcount> or <salcount> */ | |
| 8185 for (i = 0; i < gap->ga_len; ++i) | |
| 8186 { | |
| 8187 /* <rep> : <repfromlen> <repfrom> <reptolen> <repto> */ | |
| 8188 /* <sal> : <salfromlen> <salfrom> <saltolen> <salto> */ | |
| 8189 ftp = &((fromto_T *)gap->ga_data)[i]; | |
| 8190 for (rr = 1; rr <= 2; ++rr) | |
| 8191 { | |
| 8192 p = rr == 1 ? ftp->ft_from : ftp->ft_to; | |
| 835 | 8193 l = (int)STRLEN(p); |
| 323 | 8194 putc(l, fd); |
| 1758 | 8195 if (l > 0) |
| 8196 fwv &= fwrite(p, l, (size_t)1, fd); | |
| 323 | 8197 } |
| 8198 } | |
| 483 | 8199 |
| 8200 } | |
| 8201 | |
| 8202 /* SN_SOFO: <sofofromlen> <sofofrom> <sofotolen> <sofoto> | |
| 8203 * This is for making suggestions, section is not required. */ | |
| 375 | 8204 if (spin->si_sofofr != NULL && spin->si_sofoto != NULL) |
| 8205 { | |
| 483 | 8206 putc(SN_SOFO, fd); /* <sectionID> */ |
| 8207 putc(0, fd); /* <sectionflags> */ | |
| 375 | 8208 |
| 835 | 8209 l = (int)STRLEN(spin->si_sofofr); |
| 483 | 8210 put_bytes(fd, (long_u)(l + STRLEN(spin->si_sofoto) + 4), 4); |
| 8211 /* <sectionlen> */ | |
| 8212 | |
| 8213 put_bytes(fd, (long_u)l, 2); /* <sofofromlen> */ | |
| 1753 | 8214 fwv &= fwrite(spin->si_sofofr, l, (size_t)1, fd); /* <sofofrom> */ |
| 375 | 8215 |
| 835 | 8216 l = (int)STRLEN(spin->si_sofoto); |
| 483 | 8217 put_bytes(fd, (long_u)l, 2); /* <sofotolen> */ |
| 1753 | 8218 fwv &= fwrite(spin->si_sofoto, l, (size_t)1, fd); /* <sofoto> */ |
| 483 | 8219 } |
| 8220 | |
| 625 | 8221 /* SN_WORDS: <word> ... |
| 8222 * This is for making suggestions, section is not required. */ | |
| 8223 if (spin->si_commonwords.ht_used > 0) | |
| 8224 { | |
| 8225 putc(SN_WORDS, fd); /* <sectionID> */ | |
| 8226 putc(0, fd); /* <sectionflags> */ | |
| 8227 | |
| 8228 /* round 1: count the bytes | |
| 8229 * round 2: write the bytes */ | |
| 8230 for (round = 1; round <= 2; ++round) | |
| 8231 { | |
| 8232 int todo; | |
| 8233 int len = 0; | |
| 8234 hashitem_T *hi; | |
| 8235 | |
| 835 | 8236 todo = (int)spin->si_commonwords.ht_used; |
| 625 | 8237 for (hi = spin->si_commonwords.ht_array; todo > 0; ++hi) |
| 8238 if (!HASHITEM_EMPTY(hi)) | |
| 8239 { | |
| 835 | 8240 l = (int)STRLEN(hi->hi_key) + 1; |
| 625 | 8241 len += l; |
| 8242 if (round == 2) /* <word> */ | |
| 1753 | 8243 fwv &= fwrite(hi->hi_key, (size_t)l, (size_t)1, fd); |
| 625 | 8244 --todo; |
| 8245 } | |
| 8246 if (round == 1) | |
| 8247 put_bytes(fd, (long_u)len, 4); /* <sectionlen> */ | |
| 8248 } | |
| 8249 } | |
| 8250 | |
| 483 | 8251 /* SN_MAP: <mapstr> |
| 8252 * This is for making suggestions, section is not required. */ | |
| 8253 if (spin->si_map.ga_len > 0) | |
| 8254 { | |
| 8255 putc(SN_MAP, fd); /* <sectionID> */ | |
| 8256 putc(0, fd); /* <sectionflags> */ | |
| 8257 l = spin->si_map.ga_len; | |
| 8258 put_bytes(fd, (long_u)l, 4); /* <sectionlen> */ | |
| 1753 | 8259 fwv &= fwrite(spin->si_map.ga_data, (size_t)l, (size_t)1, fd); |
| 483 | 8260 /* <mapstr> */ |
| 8261 } | |
| 8262 | |
| 625 | 8263 /* SN_SUGFILE: <timestamp> |
| 8264 * This is used to notify that a .sug file may be available and at the | |
| 8265 * same time allows for checking that a .sug file that is found matches | |
| 8266 * with this .spl file. That's because the word numbers must be exactly | |
| 8267 * right. */ | |
| 8268 if (!spin->si_nosugfile | |
| 8269 && (spin->si_sal.ga_len > 0 | |
| 8270 || (spin->si_sofofr != NULL && spin->si_sofoto != NULL))) | |
| 8271 { | |
| 8272 putc(SN_SUGFILE, fd); /* <sectionID> */ | |
| 8273 putc(0, fd); /* <sectionflags> */ | |
| 8274 put_bytes(fd, (long_u)8, 4); /* <sectionlen> */ | |
| 8275 | |
| 8276 /* Set si_sugtime and write it to the file. */ | |
| 8277 spin->si_sugtime = time(NULL); | |
|
2229
d45902a5c61c
Fix a few more things for persistent undo.
Bram Moolenaar <bram@vim.org>
parents:
2214
diff
changeset
|
8278 put_time(fd, spin->si_sugtime); /* <timestamp> */ |
| 625 | 8279 } |
| 8280 | |
| 698 | 8281 /* SN_NOSPLITSUGS: nothing |
| 8282 * This is used to notify that no suggestions with word splits are to be | |
| 8283 * made. */ | |
| 8284 if (spin->si_nosplitsugs) | |
| 8285 { | |
| 8286 putc(SN_NOSPLITSUGS, fd); /* <sectionID> */ | |
| 8287 putc(0, fd); /* <sectionflags> */ | |
| 8288 put_bytes(fd, (long_u)0, 4); /* <sectionlen> */ | |
| 8289 } | |
| 8290 | |
| 483 | 8291 /* SN_COMPOUND: compound info. |
| 8292 * We don't mark it required, when not supported all compound words will | |
| 8293 * be bad words. */ | |
| 8294 if (spin->si_compflags != NULL) | |
| 8295 { | |
| 8296 putc(SN_COMPOUND, fd); /* <sectionID> */ | |
| 8297 putc(0, fd); /* <sectionflags> */ | |
| 8298 | |
| 835 | 8299 l = (int)STRLEN(spin->si_compflags); |
| 809 | 8300 for (i = 0; i < spin->si_comppat.ga_len; ++i) |
| 835 | 8301 l += (int)STRLEN(((char_u **)(spin->si_comppat.ga_data))[i]) + 1; |
| 809 | 8302 put_bytes(fd, (long_u)(l + 7), 4); /* <sectionlen> */ |
| 8303 | |
| 483 | 8304 putc(spin->si_compmax, fd); /* <compmax> */ |
| 8305 putc(spin->si_compminlen, fd); /* <compminlen> */ | |
| 8306 putc(spin->si_compsylmax, fd); /* <compsylmax> */ | |
| 809 | 8307 putc(0, fd); /* for Vim 7.0b compatibility */ |
| 8308 putc(spin->si_compoptions, fd); /* <compoptions> */ | |
| 8309 put_bytes(fd, (long_u)spin->si_comppat.ga_len, 2); | |
| 8310 /* <comppatcount> */ | |
| 8311 for (i = 0; i < spin->si_comppat.ga_len; ++i) | |
| 8312 { | |
| 8313 p = ((char_u **)(spin->si_comppat.ga_data))[i]; | |
| 835 | 8314 putc((int)STRLEN(p), fd); /* <comppatlen> */ |
| 1753 | 8315 fwv &= fwrite(p, (size_t)STRLEN(p), (size_t)1, fd); |
| 8316 /* <comppattext> */ | |
| 809 | 8317 } |
| 483 | 8318 /* <compflags> */ |
| 1753 | 8319 fwv &= fwrite(spin->si_compflags, (size_t)STRLEN(spin->si_compflags), |
| 809 | 8320 (size_t)1, fd); |
| 483 | 8321 } |
| 8322 | |
| 492 | 8323 /* SN_NOBREAK: NOBREAK flag */ |
| 8324 if (spin->si_nobreak) | |
| 8325 { | |
| 8326 putc(SN_NOBREAK, fd); /* <sectionID> */ | |
| 8327 putc(0, fd); /* <sectionflags> */ | |
| 8328 | |
| 1188 | 8329 /* It's empty, the presence of the section flags the feature. */ |
| 492 | 8330 put_bytes(fd, (long_u)0, 4); /* <sectionlen> */ |
| 8331 } | |
| 8332 | |
| 483 | 8333 /* SN_SYLLABLE: syllable info. |
| 8334 * We don't mark it required, when not supported syllables will not be | |
| 8335 * counted. */ | |
| 8336 if (spin->si_syllable != NULL) | |
| 8337 { | |
| 8338 putc(SN_SYLLABLE, fd); /* <sectionID> */ | |
| 8339 putc(0, fd); /* <sectionflags> */ | |
| 8340 | |
| 835 | 8341 l = (int)STRLEN(spin->si_syllable); |
| 483 | 8342 put_bytes(fd, (long_u)l, 4); /* <sectionlen> */ |
| 1753 | 8343 fwv &= fwrite(spin->si_syllable, (size_t)l, (size_t)1, fd); |
| 8344 /* <syllable> */ | |
| 483 | 8345 } |
| 8346 | |
| 8347 /* end of <SECTIONS> */ | |
| 8348 putc(SN_END, fd); /* <sectionend> */ | |
| 8349 | |
| 302 | 8350 |
| 236 | 8351 /* |
| 339 | 8352 * <LWORDTREE> <KWORDTREE> <PREFIXTREE> |
| 236 | 8353 */ |
| 323 | 8354 spin->si_memtot = 0; |
| 339 | 8355 for (round = 1; round <= 3; ++round) |
| 236 | 8356 { |
| 339 | 8357 if (round == 1) |
| 469 | 8358 tree = spin->si_foldroot->wn_sibling; |
| 339 | 8359 else if (round == 2) |
| 469 | 8360 tree = spin->si_keeproot->wn_sibling; |
| 339 | 8361 else |
| 469 | 8362 tree = spin->si_prefroot->wn_sibling; |
| 236 | 8363 |
| 346 | 8364 /* Clear the index and wnode fields in the tree. */ |
| 8365 clear_node(tree); | |
| 8366 | |
| 300 | 8367 /* Count the number of nodes. Needed to be able to allocate the |
| 346 | 8368 * memory when reading the nodes. Also fills in index for shared |
| 300 | 8369 * nodes. */ |
| 346 | 8370 nodecount = put_node(NULL, tree, 0, regionmask, round == 3); |
| 236 | 8371 |
| 300 | 8372 /* number of nodes in 4 bytes */ |
| 8373 put_bytes(fd, (long_u)nodecount, 4); /* <nodecount> */ | |
| 302 | 8374 spin->si_memtot += nodecount + nodecount * sizeof(int); |
| 236 | 8375 |
| 300 | 8376 /* Write the nodes. */ |
| 346 | 8377 (void)put_node(fd, tree, 0, regionmask, round == 3); |
| 236 | 8378 } |
| 8379 | |
| 1753 | 8380 /* Write another byte to check for errors (file system full). */ |
| 500 | 8381 if (putc(0, fd) == EOF) |
| 8382 retval = FAIL; | |
| 1759 | 8383 theend: |
| 500 | 8384 if (fclose(fd) == EOF) |
| 8385 retval = FAIL; | |
| 8386 | |
| 1759 | 8387 if (fwv != (size_t)1) |
| 1753 | 8388 retval = FAIL; |
| 8389 if (retval == FAIL) | |
| 8390 EMSG(_(e_write)); | |
| 8391 | |
| 500 | 8392 return retval; |
| 236 | 8393 } |
| 8394 | |
| 8395 /* | |
| 346 | 8396 * Clear the index and wnode fields of "node", it siblings and its |
| 8397 * children. This is needed because they are a union with other items to save | |
| 8398 * space. | |
| 8399 */ | |
| 8400 static void | |
| 8401 clear_node(node) | |
| 8402 wordnode_T *node; | |
| 8403 { | |
| 8404 wordnode_T *np; | |
| 8405 | |
| 8406 if (node != NULL) | |
| 8407 for (np = node; np != NULL; np = np->wn_sibling) | |
| 8408 { | |
| 8409 np->wn_u1.index = 0; | |
| 8410 np->wn_u2.wnode = NULL; | |
| 8411 | |
| 8412 if (np->wn_byte != NUL) | |
| 8413 clear_node(np->wn_child); | |
| 8414 } | |
| 8415 } | |
| 8416 | |
| 8417 | |
| 8418 /* | |
| 300 | 8419 * Dump a word tree at node "node". |
| 8420 * | |
| 8421 * This first writes the list of possible bytes (siblings). Then for each | |
| 8422 * byte recursively write the children. | |
| 8423 * | |
| 625 | 8424 * NOTE: The code here must match the code in read_tree_node(), since |
| 8425 * assumptions are made about the indexes (so that we don't have to write them | |
| 8426 * in the file). | |
| 236 | 8427 * |
| 300 | 8428 * Returns the number of nodes used. |
| 236 | 8429 */ |
| 300 | 8430 static int |
| 944 | 8431 put_node(fd, node, idx, regionmask, prefixtree) |
| 339 | 8432 FILE *fd; /* NULL when only counting */ |
| 300 | 8433 wordnode_T *node; |
| 944 | 8434 int idx; |
| 300 | 8435 int regionmask; |
| 339 | 8436 int prefixtree; /* TRUE for PREFIXTREE */ |
| 236 | 8437 { |
| 944 | 8438 int newindex = idx; |
| 300 | 8439 int siblingcount = 0; |
| 8440 wordnode_T *np; | |
| 236 | 8441 int flags; |
| 300 | 8442 |
| 8443 /* If "node" is zero the tree is empty. */ | |
| 8444 if (node == NULL) | |
| 8445 return 0; | |
| 8446 | |
| 8447 /* Store the index where this node is written. */ | |
| 944 | 8448 node->wn_u1.index = idx; |
| 236 | 8449 |
| 300 | 8450 /* Count the number of siblings. */ |
| 8451 for (np = node; np != NULL; np = np->wn_sibling) | |
| 8452 ++siblingcount; | |
| 236 | 8453 |
| 300 | 8454 /* Write the sibling count. */ |
| 8455 if (fd != NULL) | |
| 8456 putc(siblingcount, fd); /* <siblingcount> */ | |
| 236 | 8457 |
| 300 | 8458 /* Write each sibling byte and optionally extra info. */ |
| 8459 for (np = node; np != NULL; np = np->wn_sibling) | |
| 236 | 8460 { |
| 300 | 8461 if (np->wn_byte == 0) |
| 8462 { | |
| 8463 if (fd != NULL) | |
| 8464 { | |
| 339 | 8465 /* For a NUL byte (end of word) write the flags etc. */ |
| 8466 if (prefixtree) | |
| 300 | 8467 { |
| 481 | 8468 /* In PREFIXTREE write the required affixID and the |
| 390 | 8469 * associated condition nr (stored in wn_region). The |
| 8470 * byte value is misused to store the "rare" and "not | |
| 8471 * combining" flags */ | |
| 455 | 8472 if (np->wn_flags == (short_u)PFX_FLAGS) |
| 8473 putc(BY_NOFLAGS, fd); /* <byte> */ | |
| 8474 else | |
| 8475 { | |
| 390 | 8476 putc(BY_FLAGS, fd); /* <byte> */ |
| 455 | 8477 putc(np->wn_flags, fd); /* <pflags> */ |
| 8478 } | |
| 481 | 8479 putc(np->wn_affixID, fd); /* <affixID> */ |
| 339 | 8480 put_bytes(fd, (long_u)np->wn_region, 2); /* <prefcondnr> */ |
| 300 | 8481 } |
| 8482 else | |
| 8483 { | |
| 339 | 8484 /* For word trees we write the flag/region items. */ |
| 8485 flags = np->wn_flags; | |
| 8486 if (regionmask != 0 && np->wn_region != regionmask) | |
| 8487 flags |= WF_REGION; | |
| 481 | 8488 if (np->wn_affixID != 0) |
| 8489 flags |= WF_AFX; | |
| 339 | 8490 if (flags == 0) |
| 8491 { | |
| 8492 /* word without flags or region */ | |
| 8493 putc(BY_NOFLAGS, fd); /* <byte> */ | |
| 8494 } | |
| 8495 else | |
| 8496 { | |
| 390 | 8497 if (np->wn_flags >= 0x100) |
| 8498 { | |
| 8499 putc(BY_FLAGS2, fd); /* <byte> */ | |
| 8500 putc(flags, fd); /* <flags> */ | |
| 8501 putc((unsigned)flags >> 8, fd); /* <flags2> */ | |
| 8502 } | |
| 8503 else | |
| 8504 { | |
| 8505 putc(BY_FLAGS, fd); /* <byte> */ | |
| 8506 putc(flags, fd); /* <flags> */ | |
| 8507 } | |
| 339 | 8508 if (flags & WF_REGION) |
| 8509 putc(np->wn_region, fd); /* <region> */ | |
| 481 | 8510 if (flags & WF_AFX) |
| 8511 putc(np->wn_affixID, fd); /* <affixID> */ | |
| 339 | 8512 } |
| 300 | 8513 } |
| 8514 } | |
| 8515 } | |
| 8516 else | |
| 8517 { | |
| 346 | 8518 if (np->wn_child->wn_u1.index != 0 |
| 8519 && np->wn_child->wn_u2.wnode != node) | |
| 300 | 8520 { |
| 8521 /* The child is written elsewhere, write the reference. */ | |
| 8522 if (fd != NULL) | |
| 8523 { | |
| 8524 putc(BY_INDEX, fd); /* <byte> */ | |
| 8525 /* <nodeidx> */ | |
| 346 | 8526 put_bytes(fd, (long_u)np->wn_child->wn_u1.index, 3); |
| 300 | 8527 } |
| 8528 } | |
| 346 | 8529 else if (np->wn_child->wn_u2.wnode == NULL) |
| 300 | 8530 /* We will write the child below and give it an index. */ |
| 346 | 8531 np->wn_child->wn_u2.wnode = node; |
| 236 | 8532 |
| 300 | 8533 if (fd != NULL) |
| 8534 if (putc(np->wn_byte, fd) == EOF) /* <byte> or <xbyte> */ | |
| 8535 { | |
| 8536 EMSG(_(e_write)); | |
| 8537 return 0; | |
| 8538 } | |
| 8539 } | |
| 236 | 8540 } |
| 8541 | |
| 300 | 8542 /* Space used in the array when reading: one for each sibling and one for |
| 8543 * the count. */ | |
| 8544 newindex += siblingcount + 1; | |
| 249 | 8545 |
| 300 | 8546 /* Recursively dump the children of each sibling. */ |
| 8547 for (np = node; np != NULL; np = np->wn_sibling) | |
| 346 | 8548 if (np->wn_byte != 0 && np->wn_child->wn_u2.wnode == node) |
| 8549 newindex = put_node(fd, np->wn_child, newindex, regionmask, | |
| 339 | 8550 prefixtree); |
| 249 | 8551 |
| 300 | 8552 return newindex; |
| 236 | 8553 } |
| 8554 | |
| 8555 | |
| 8556 /* | |
| 310 | 8557 * ":mkspell [-ascii] outfile infile ..." |
| 8558 * ":mkspell [-ascii] addfile" | |
| 236 | 8559 */ |
| 8560 void | |
| 8561 ex_mkspell(eap) | |
| 8562 exarg_T *eap; | |
| 8563 { | |
| 8564 int fcount; | |
| 8565 char_u **fnames; | |
| 310 | 8566 char_u *arg = eap->arg; |
| 8567 int ascii = FALSE; | |
| 8568 | |
| 8569 if (STRNCMP(arg, "-ascii", 6) == 0) | |
| 8570 { | |
| 8571 ascii = TRUE; | |
| 8572 arg = skipwhite(arg + 6); | |
| 8573 } | |
| 8574 | |
| 8575 /* Expand all the remaining arguments (e.g., $VIMRUNTIME). */ | |
| 8576 if (get_arglist_exp(arg, &fcount, &fnames) == OK) | |
| 8577 { | |
| 323 | 8578 mkspell(fcount, fnames, ascii, eap->forceit, FALSE); |
| 310 | 8579 FreeWild(fcount, fnames); |
| 8580 } | |
| 8581 } | |
| 8582 | |
| 8583 /* | |
| 625 | 8584 * Create the .sug file. |
| 8585 * Uses the soundfold info in "spin". | |
| 8586 * Writes the file with the name "wfname", with ".spl" changed to ".sug". | |
| 8587 */ | |
| 8588 static void | |
| 8589 spell_make_sugfile(spin, wfname) | |
| 8590 spellinfo_T *spin; | |
| 8591 char_u *wfname; | |
| 8592 { | |
| 2770 | 8593 char_u *fname = NULL; |
| 625 | 8594 int len; |
| 8595 slang_T *slang; | |
| 8596 int free_slang = FALSE; | |
| 8597 | |
| 8598 /* | |
| 8599 * Read back the .spl file that was written. This fills the required | |
| 8600 * info for soundfolding. This also uses less memory than the | |
| 8601 * pointer-linked version of the trie. And it avoids having two versions | |
| 8602 * of the code for the soundfolding stuff. | |
| 8603 * It might have been done already by spell_reload_one(). | |
| 8604 */ | |
| 8605 for (slang = first_lang; slang != NULL; slang = slang->sl_next) | |
| 8606 if (fullpathcmp(wfname, slang->sl_fname, FALSE) == FPC_SAME) | |
| 8607 break; | |
| 8608 if (slang == NULL) | |
| 8609 { | |
| 8610 spell_message(spin, (char_u *)_("Reading back spell file...")); | |
| 8611 slang = spell_load_file(wfname, NULL, NULL, FALSE); | |
| 8612 if (slang == NULL) | |
| 8613 return; | |
| 8614 free_slang = TRUE; | |
| 8615 } | |
| 8616 | |
| 8617 /* | |
| 8618 * Clear the info in "spin" that is used. | |
| 8619 */ | |
| 8620 spin->si_blocks = NULL; | |
| 8621 spin->si_blocks_cnt = 0; | |
| 8622 spin->si_compress_cnt = 0; /* will stay at 0 all the time*/ | |
| 8623 spin->si_free_count = 0; | |
| 8624 spin->si_first_free = NULL; | |
| 8625 spin->si_foldwcount = 0; | |
| 8626 | |
| 8627 /* | |
| 8628 * Go through the trie of good words, soundfold each word and add it to | |
| 8629 * the soundfold trie. | |
| 8630 */ | |
| 8631 spell_message(spin, (char_u *)_("Performing soundfolding...")); | |
| 8632 if (sug_filltree(spin, slang) == FAIL) | |
| 8633 goto theend; | |
| 8634 | |
| 8635 /* | |
| 8636 * Create the table which links each soundfold word with a list of the | |
| 8637 * good words it may come from. Creates buffer "spin->si_spellbuf". | |
| 8638 * This also removes the wordnr from the NUL byte entries to make | |
| 8639 * compression possible. | |
| 8640 */ | |
| 8641 if (sug_maketable(spin) == FAIL) | |
| 8642 goto theend; | |
| 8643 | |
| 8644 smsg((char_u *)_("Number of words after soundfolding: %ld"), | |
| 8645 (long)spin->si_spellbuf->b_ml.ml_line_count); | |
| 8646 | |
| 8647 /* | |
| 8648 * Compress the soundfold trie. | |
| 8649 */ | |
| 8650 spell_message(spin, (char_u *)_(msg_compressing)); | |
| 8651 wordtree_compress(spin, spin->si_foldroot); | |
| 8652 | |
| 8653 /* | |
| 8654 * Write the .sug file. | |
| 8655 * Make the file name by changing ".spl" to ".sug". | |
| 8656 */ | |
| 2770 | 8657 fname = alloc(MAXPATHL); |
| 8658 if (fname == NULL) | |
| 8659 goto theend; | |
| 2768 | 8660 vim_strncpy(fname, wfname, MAXPATHL - 1); |
| 835 | 8661 len = (int)STRLEN(fname); |
| 625 | 8662 fname[len - 2] = 'u'; |
| 8663 fname[len - 1] = 'g'; | |
| 8664 sug_write(spin, fname); | |
| 8665 | |
| 8666 theend: | |
| 2770 | 8667 vim_free(fname); |
| 625 | 8668 if (free_slang) |
| 8669 slang_free(slang); | |
| 8670 free_blocks(spin->si_blocks); | |
| 8671 close_spellbuf(spin->si_spellbuf); | |
| 8672 } | |
| 8673 | |
| 8674 /* | |
| 8675 * Build the soundfold trie for language "slang". | |
| 8676 */ | |
| 8677 static int | |
| 8678 sug_filltree(spin, slang) | |
| 8679 spellinfo_T *spin; | |
| 8680 slang_T *slang; | |
| 8681 { | |
| 8682 char_u *byts; | |
| 8683 idx_T *idxs; | |
| 8684 int depth; | |
| 8685 idx_T arridx[MAXWLEN]; | |
| 8686 int curi[MAXWLEN]; | |
| 8687 char_u tword[MAXWLEN]; | |
| 8688 char_u tsalword[MAXWLEN]; | |
| 8689 int c; | |
| 8690 idx_T n; | |
| 8691 unsigned words_done = 0; | |
| 8692 int wordcount[MAXWLEN]; | |
| 8693 | |
| 8694 /* We use si_foldroot for the souldfolded trie. */ | |
| 8695 spin->si_foldroot = wordtree_alloc(spin); | |
| 8696 if (spin->si_foldroot == NULL) | |
| 8697 return FAIL; | |
| 8698 | |
| 8699 /* let tree_add_word() know we're adding to the soundfolded tree */ | |
| 8700 spin->si_sugtree = TRUE; | |
| 8701 | |
| 8702 /* | |
| 8703 * Go through the whole case-folded tree, soundfold each word and put it | |
| 8704 * in the trie. | |
| 8705 */ | |
| 8706 byts = slang->sl_fbyts; | |
| 8707 idxs = slang->sl_fidxs; | |
| 8708 | |
| 8709 arridx[0] = 0; | |
| 8710 curi[0] = 1; | |
| 8711 wordcount[0] = 0; | |
| 8712 | |
| 8713 depth = 0; | |
| 8714 while (depth >= 0 && !got_int) | |
| 8715 { | |
| 8716 if (curi[depth] > byts[arridx[depth]]) | |
| 8717 { | |
| 8718 /* Done all bytes at this node, go up one level. */ | |
| 8719 idxs[arridx[depth]] = wordcount[depth]; | |
| 8720 if (depth > 0) | |
| 8721 wordcount[depth - 1] += wordcount[depth]; | |
| 8722 | |
| 8723 --depth; | |
| 8724 line_breakcheck(); | |
| 8725 } | |
| 8726 else | |
| 8727 { | |
| 8728 | |
| 8729 /* Do one more byte at this node. */ | |
| 8730 n = arridx[depth] + curi[depth]; | |
| 8731 ++curi[depth]; | |
| 8732 | |
| 8733 c = byts[n]; | |
| 8734 if (c == 0) | |
| 8735 { | |
| 8736 /* Sound-fold the word. */ | |
| 8737 tword[depth] = NUL; | |
| 8738 spell_soundfold(slang, tword, TRUE, tsalword); | |
| 8739 | |
| 8740 /* We use the "flags" field for the MSB of the wordnr, | |
| 8741 * "region" for the LSB of the wordnr. */ | |
| 8742 if (tree_add_word(spin, tsalword, spin->si_foldroot, | |
| 8743 words_done >> 16, words_done & 0xffff, | |
| 8744 0) == FAIL) | |
| 8745 return FAIL; | |
| 8746 | |
| 8747 ++words_done; | |
| 8748 ++wordcount[depth]; | |
| 8749 | |
| 8750 /* Reset the block count each time to avoid compression | |
| 8751 * kicking in. */ | |
| 8752 spin->si_blocks_cnt = 0; | |
| 8753 | |
| 8754 /* Skip over any other NUL bytes (same word with different | |
| 8755 * flags). */ | |
| 8756 while (byts[n + 1] == 0) | |
| 8757 { | |
| 8758 ++n; | |
| 8759 ++curi[depth]; | |
| 8760 } | |
| 8761 } | |
| 8762 else | |
| 8763 { | |
| 8764 /* Normal char, go one level deeper. */ | |
| 8765 tword[depth++] = c; | |
| 8766 arridx[depth] = idxs[n]; | |
| 8767 curi[depth] = 1; | |
| 8768 wordcount[depth] = 0; | |
| 8769 } | |
| 8770 } | |
| 8771 } | |
| 8772 | |
| 8773 smsg((char_u *)_("Total number of words: %d"), words_done); | |
| 8774 | |
| 8775 return OK; | |
| 8776 } | |
| 8777 | |
| 8778 /* | |
| 8779 * Make the table that links each word in the soundfold trie to the words it | |
| 8780 * can be produced from. | |
| 8781 * This is not unlike lines in a file, thus use a memfile to be able to access | |
| 8782 * the table efficiently. | |
| 8783 * Returns FAIL when out of memory. | |
| 8784 */ | |
| 8785 static int | |
| 8786 sug_maketable(spin) | |
| 8787 spellinfo_T *spin; | |
| 8788 { | |
| 8789 garray_T ga; | |
| 8790 int res = OK; | |
| 8791 | |
| 8792 /* Allocate a buffer, open a memline for it and create the swap file | |
| 8793 * (uses a temp file, not a .swp file). */ | |
| 8794 spin->si_spellbuf = open_spellbuf(); | |
| 8795 if (spin->si_spellbuf == NULL) | |
| 8796 return FAIL; | |
| 8797 | |
| 8798 /* Use a buffer to store the line info, avoids allocating many small | |
| 8799 * pieces of memory. */ | |
| 8800 ga_init2(&ga, 1, 100); | |
| 8801 | |
| 8802 /* recursively go through the tree */ | |
| 8803 if (sug_filltable(spin, spin->si_foldroot->wn_sibling, 0, &ga) == -1) | |
| 8804 res = FAIL; | |
| 8805 | |
| 8806 ga_clear(&ga); | |
| 8807 return res; | |
| 8808 } | |
| 8809 | |
| 8810 /* | |
| 8811 * Fill the table for one node and its children. | |
| 8812 * Returns the wordnr at the start of the node. | |
| 8813 * Returns -1 when out of memory. | |
| 8814 */ | |
| 8815 static int | |
| 8816 sug_filltable(spin, node, startwordnr, gap) | |
| 8817 spellinfo_T *spin; | |
| 8818 wordnode_T *node; | |
| 8819 int startwordnr; | |
| 8820 garray_T *gap; /* place to store line of numbers */ | |
| 8821 { | |
| 8822 wordnode_T *p, *np; | |
| 8823 int wordnr = startwordnr; | |
| 8824 int nr; | |
| 8825 int prev_nr; | |
| 8826 | |
| 8827 for (p = node; p != NULL; p = p->wn_sibling) | |
| 8828 { | |
| 8829 if (p->wn_byte == NUL) | |
| 8830 { | |
| 8831 gap->ga_len = 0; | |
| 8832 prev_nr = 0; | |
| 8833 for (np = p; np != NULL && np->wn_byte == NUL; np = np->wn_sibling) | |
| 8834 { | |
| 8835 if (ga_grow(gap, 10) == FAIL) | |
| 8836 return -1; | |
| 8837 | |
| 8838 nr = (np->wn_flags << 16) + (np->wn_region & 0xffff); | |
| 8839 /* Compute the offset from the previous nr and store the | |
| 8840 * offset in a way that it takes a minimum number of bytes. | |
| 8841 * It's a bit like utf-8, but without the need to mark | |
| 8842 * following bytes. */ | |
| 8843 nr -= prev_nr; | |
| 8844 prev_nr += nr; | |
| 8845 gap->ga_len += offset2bytes(nr, | |
| 8846 (char_u *)gap->ga_data + gap->ga_len); | |
| 8847 } | |
| 8848 | |
| 8849 /* add the NUL byte */ | |
| 8850 ((char_u *)gap->ga_data)[gap->ga_len++] = NUL; | |
| 8851 | |
| 8852 if (ml_append_buf(spin->si_spellbuf, (linenr_T)wordnr, | |
| 8853 gap->ga_data, gap->ga_len, TRUE) == FAIL) | |
| 8854 return -1; | |
| 8855 ++wordnr; | |
| 8856 | |
| 8857 /* Remove extra NUL entries, we no longer need them. We don't | |
| 8858 * bother freeing the nodes, the won't be reused anyway. */ | |
| 8859 while (p->wn_sibling != NULL && p->wn_sibling->wn_byte == NUL) | |
| 8860 p->wn_sibling = p->wn_sibling->wn_sibling; | |
| 8861 | |
| 8862 /* Clear the flags on the remaining NUL node, so that compression | |
| 8863 * works a lot better. */ | |
| 8864 p->wn_flags = 0; | |
| 8865 p->wn_region = 0; | |
| 8866 } | |
| 8867 else | |
| 8868 { | |
| 8869 wordnr = sug_filltable(spin, p->wn_child, wordnr, gap); | |
| 8870 if (wordnr == -1) | |
| 8871 return -1; | |
| 8872 } | |
| 8873 } | |
| 8874 return wordnr; | |
| 8875 } | |
| 8876 | |
| 8877 /* | |
| 8878 * Convert an offset into a minimal number of bytes. | |
| 8879 * Similar to utf_char2byters, but use 8 bits in followup bytes and avoid NUL | |
| 8880 * bytes. | |
| 8881 */ | |
| 8882 static int | |
| 8883 offset2bytes(nr, buf) | |
| 8884 int nr; | |
| 8885 char_u *buf; | |
| 8886 { | |
| 8887 int rem; | |
| 8888 int b1, b2, b3, b4; | |
| 8889 | |
| 8890 /* Split the number in parts of base 255. We need to avoid NUL bytes. */ | |
| 8891 b1 = nr % 255 + 1; | |
| 8892 rem = nr / 255; | |
| 8893 b2 = rem % 255 + 1; | |
| 8894 rem = rem / 255; | |
| 8895 b3 = rem % 255 + 1; | |
| 8896 b4 = rem / 255 + 1; | |
| 8897 | |
| 8898 if (b4 > 1 || b3 > 0x1f) /* 4 bytes */ | |
| 8899 { | |
| 8900 buf[0] = 0xe0 + b4; | |
| 8901 buf[1] = b3; | |
| 8902 buf[2] = b2; | |
| 8903 buf[3] = b1; | |
| 8904 return 4; | |
| 8905 } | |
| 8906 if (b3 > 1 || b2 > 0x3f ) /* 3 bytes */ | |
| 8907 { | |
| 8908 buf[0] = 0xc0 + b3; | |
| 8909 buf[1] = b2; | |
| 8910 buf[2] = b1; | |
| 8911 return 3; | |
| 8912 } | |
| 8913 if (b2 > 1 || b1 > 0x7f ) /* 2 bytes */ | |
| 8914 { | |
| 8915 buf[0] = 0x80 + b2; | |
| 8916 buf[1] = b1; | |
| 8917 return 2; | |
| 8918 } | |
| 8919 /* 1 byte */ | |
| 8920 buf[0] = b1; | |
| 8921 return 1; | |
| 8922 } | |
| 8923 | |
| 8924 /* | |
| 8925 * Opposite of offset2bytes(). | |
| 8926 * "pp" points to the bytes and is advanced over it. | |
| 8927 * Returns the offset. | |
| 8928 */ | |
| 8929 static int | |
| 8930 bytes2offset(pp) | |
| 8931 char_u **pp; | |
| 8932 { | |
| 8933 char_u *p = *pp; | |
| 8934 int nr; | |
| 8935 int c; | |
| 8936 | |
| 8937 c = *p++; | |
| 8938 if ((c & 0x80) == 0x00) /* 1 byte */ | |
| 8939 { | |
| 8940 nr = c - 1; | |
| 8941 } | |
| 8942 else if ((c & 0xc0) == 0x80) /* 2 bytes */ | |
| 8943 { | |
| 8944 nr = (c & 0x3f) - 1; | |
| 8945 nr = nr * 255 + (*p++ - 1); | |
| 8946 } | |
| 8947 else if ((c & 0xe0) == 0xc0) /* 3 bytes */ | |
| 8948 { | |
| 8949 nr = (c & 0x1f) - 1; | |
| 8950 nr = nr * 255 + (*p++ - 1); | |
| 8951 nr = nr * 255 + (*p++ - 1); | |
| 8952 } | |
| 8953 else /* 4 bytes */ | |
| 8954 { | |
| 8955 nr = (c & 0x0f) - 1; | |
| 8956 nr = nr * 255 + (*p++ - 1); | |
| 8957 nr = nr * 255 + (*p++ - 1); | |
| 8958 nr = nr * 255 + (*p++ - 1); | |
| 8959 } | |
| 8960 | |
| 8961 *pp = p; | |
| 8962 return nr; | |
| 8963 } | |
| 8964 | |
| 8965 /* | |
| 8966 * Write the .sug file in "fname". | |
| 8967 */ | |
| 8968 static void | |
| 8969 sug_write(spin, fname) | |
| 8970 spellinfo_T *spin; | |
| 8971 char_u *fname; | |
| 8972 { | |
| 8973 FILE *fd; | |
| 8974 wordnode_T *tree; | |
| 8975 int nodecount; | |
| 8976 int wcount; | |
| 8977 char_u *line; | |
| 8978 linenr_T lnum; | |
| 8979 int len; | |
| 8980 | |
| 8981 /* Create the file. Note that an existing file is silently overwritten! */ | |
| 8982 fd = mch_fopen((char *)fname, "w"); | |
| 8983 if (fd == NULL) | |
| 8984 { | |
| 8985 EMSG2(_(e_notopen), fname); | |
| 8986 return; | |
| 8987 } | |
| 8988 | |
| 8989 vim_snprintf((char *)IObuff, IOSIZE, | |
| 8990 _("Writing suggestion file %s ..."), fname); | |
| 8991 spell_message(spin, IObuff); | |
| 8992 | |
| 8993 /* | |
| 8994 * <SUGHEADER>: <fileID> <versionnr> <timestamp> | |
| 8995 */ | |
| 8996 if (fwrite(VIMSUGMAGIC, VIMSUGMAGICL, (size_t)1, fd) != 1) /* <fileID> */ | |
| 8997 { | |
| 8998 EMSG(_(e_write)); | |
| 8999 goto theend; | |
| 9000 } | |
| 9001 putc(VIMSUGVERSION, fd); /* <versionnr> */ | |
| 9002 | |
| 9003 /* Write si_sugtime to the file. */ | |
|
2229
d45902a5c61c
Fix a few more things for persistent undo.
Bram Moolenaar <bram@vim.org>
parents:
2214
diff
changeset
|
9004 put_time(fd, spin->si_sugtime); /* <timestamp> */ |
| 625 | 9005 |
| 9006 /* | |
| 9007 * <SUGWORDTREE> | |
| 9008 */ | |
| 9009 spin->si_memtot = 0; | |
| 9010 tree = spin->si_foldroot->wn_sibling; | |
| 9011 | |
| 9012 /* Clear the index and wnode fields in the tree. */ | |
| 9013 clear_node(tree); | |
| 9014 | |
| 9015 /* Count the number of nodes. Needed to be able to allocate the | |
| 9016 * memory when reading the nodes. Also fills in index for shared | |
| 9017 * nodes. */ | |
| 9018 nodecount = put_node(NULL, tree, 0, 0, FALSE); | |
| 9019 | |
| 9020 /* number of nodes in 4 bytes */ | |
| 9021 put_bytes(fd, (long_u)nodecount, 4); /* <nodecount> */ | |
| 9022 spin->si_memtot += nodecount + nodecount * sizeof(int); | |
| 9023 | |
| 9024 /* Write the nodes. */ | |
| 9025 (void)put_node(fd, tree, 0, 0, FALSE); | |
| 9026 | |
| 9027 /* | |
| 9028 * <SUGTABLE>: <sugwcount> <sugline> ... | |
| 9029 */ | |
| 9030 wcount = spin->si_spellbuf->b_ml.ml_line_count; | |
| 9031 put_bytes(fd, (long_u)wcount, 4); /* <sugwcount> */ | |
| 9032 | |
| 9033 for (lnum = 1; lnum <= (linenr_T)wcount; ++lnum) | |
| 9034 { | |
| 9035 /* <sugline>: <sugnr> ... NUL */ | |
| 9036 line = ml_get_buf(spin->si_spellbuf, lnum, FALSE); | |
| 835 | 9037 len = (int)STRLEN(line) + 1; |
| 625 | 9038 if (fwrite(line, (size_t)len, (size_t)1, fd) == 0) |
| 9039 { | |
| 9040 EMSG(_(e_write)); | |
| 9041 goto theend; | |
| 9042 } | |
| 9043 spin->si_memtot += len; | |
| 9044 } | |
| 9045 | |
| 9046 /* Write another byte to check for errors. */ | |
| 9047 if (putc(0, fd) == EOF) | |
| 9048 EMSG(_(e_write)); | |
| 9049 | |
| 9050 vim_snprintf((char *)IObuff, IOSIZE, | |
| 9051 _("Estimated runtime memory use: %d bytes"), spin->si_memtot); | |
| 9052 spell_message(spin, IObuff); | |
| 9053 | |
| 9054 theend: | |
| 9055 /* close the file */ | |
| 9056 fclose(fd); | |
| 9057 } | |
| 9058 | |
| 9059 /* | |
| 9060 * Open a spell buffer. This is a nameless buffer that is not in the buffer | |
| 9061 * list and only contains text lines. Can use a swapfile to reduce memory | |
| 9062 * use. | |
| 9063 * Most other fields are invalid! Esp. watch out for string options being | |
| 9064 * NULL and there is no undo info. | |
| 9065 * Returns NULL when out of memory. | |
| 9066 */ | |
| 9067 static buf_T * | |
| 9068 open_spellbuf() | |
| 9069 { | |
| 9070 buf_T *buf; | |
| 9071 | |
| 9072 buf = (buf_T *)alloc_clear(sizeof(buf_T)); | |
| 9073 if (buf != NULL) | |
| 9074 { | |
| 9075 buf->b_spell = TRUE; | |
| 9076 buf->b_p_swf = TRUE; /* may create a swap file */ | |
| 9077 ml_open(buf); | |
| 9078 ml_open_file(buf); /* create swap file now */ | |
| 9079 } | |
| 9080 return buf; | |
| 9081 } | |
| 9082 | |
| 9083 /* | |
| 9084 * Close the buffer used for spell info. | |
| 9085 */ | |
| 9086 static void | |
| 9087 close_spellbuf(buf) | |
| 9088 buf_T *buf; | |
| 9089 { | |
| 9090 if (buf != NULL) | |
| 9091 { | |
| 9092 ml_close(buf, TRUE); | |
| 9093 vim_free(buf); | |
| 9094 } | |
| 9095 } | |
| 9096 | |
| 9097 | |
| 9098 /* | |
| 310 | 9099 * Create a Vim spell file from one or more word lists. |
| 9100 * "fnames[0]" is the output file name. | |
| 9101 * "fnames[fcount - 1]" is the last input file name. | |
| 9102 * Exception: when "fnames[0]" ends in ".add" it's used as the input file name | |
| 9103 * and ".spl" is appended to make the output file name. | |
| 9104 */ | |
| 9105 static void | |
| 323 | 9106 mkspell(fcount, fnames, ascii, overwrite, added_word) |
| 310 | 9107 int fcount; |
| 9108 char_u **fnames; | |
| 9109 int ascii; /* -ascii argument given */ | |
| 9110 int overwrite; /* overwrite existing output file */ | |
| 323 | 9111 int added_word; /* invoked through "zg" */ |
| 310 | 9112 { |
| 2770 | 9113 char_u *fname = NULL; |
| 9114 char_u *wfname; | |
| 310 | 9115 char_u **innames; |
| 9116 int incount; | |
| 236 | 9117 afffile_T *(afile[8]); |
| 9118 int i; | |
| 9119 int len; | |
| 9120 struct stat st; | |
| 255 | 9121 int error = FALSE; |
| 300 | 9122 spellinfo_T spin; |
| 9123 | |
| 9124 vim_memset(&spin, 0, sizeof(spin)); | |
| 323 | 9125 spin.si_verbose = !added_word; |
| 310 | 9126 spin.si_ascii = ascii; |
| 323 | 9127 spin.si_followup = TRUE; |
| 9128 spin.si_rem_accents = TRUE; | |
| 9129 ga_init2(&spin.si_rep, (int)sizeof(fromto_T), 20); | |
| 625 | 9130 ga_init2(&spin.si_repsal, (int)sizeof(fromto_T), 20); |
| 323 | 9131 ga_init2(&spin.si_sal, (int)sizeof(fromto_T), 20); |
| 9132 ga_init2(&spin.si_map, (int)sizeof(char_u), 100); | |
| 809 | 9133 ga_init2(&spin.si_comppat, (int)sizeof(char_u *), 20); |
| 339 | 9134 ga_init2(&spin.si_prefcond, (int)sizeof(char_u *), 50); |
| 625 | 9135 hash_init(&spin.si_commonwords); |
| 500 | 9136 spin.si_newcompID = 127; /* start compound ID at first maximum */ |
| 310 | 9137 |
| 9138 /* default: fnames[0] is output file, following are input files */ | |
| 9139 innames = &fnames[1]; | |
| 9140 incount = fcount - 1; | |
| 9141 | |
| 2770 | 9142 wfname = alloc(MAXPATHL); |
| 9143 if (wfname == NULL) | |
| 9144 return; | |
| 9145 | |
| 310 | 9146 if (fcount >= 1) |
| 240 | 9147 { |
| 835 | 9148 len = (int)STRLEN(fnames[0]); |
| 310 | 9149 if (fcount == 1 && len > 4 && STRCMP(fnames[0] + len - 4, ".add") == 0) |
| 9150 { | |
| 9151 /* For ":mkspell path/en.latin1.add" output file is | |
| 9152 * "path/en.latin1.add.spl". */ | |
| 9153 innames = &fnames[0]; | |
| 9154 incount = 1; | |
| 2770 | 9155 vim_snprintf((char *)wfname, MAXPATHL, "%s.spl", fnames[0]); |
| 310 | 9156 } |
| 366 | 9157 else if (fcount == 1) |
| 9158 { | |
| 9159 /* For ":mkspell path/vim" output file is "path/vim.latin1.spl". */ | |
| 9160 innames = &fnames[0]; | |
| 9161 incount = 1; | |
| 2770 | 9162 vim_snprintf((char *)wfname, MAXPATHL, SPL_FNAME_TMPL, |
| 2660 | 9163 fnames[0], spin.si_ascii ? (char_u *)"ascii" : spell_enc()); |
| 366 | 9164 } |
| 310 | 9165 else if (len > 4 && STRCMP(fnames[0] + len - 4, ".spl") == 0) |
| 9166 { | |
| 9167 /* Name ends in ".spl", use as the file name. */ | |
| 2770 | 9168 vim_strncpy(wfname, fnames[0], MAXPATHL - 1); |
| 310 | 9169 } |
| 9170 else | |
| 9171 /* Name should be language, make the file name from it. */ | |
| 2770 | 9172 vim_snprintf((char *)wfname, MAXPATHL, SPL_FNAME_TMPL, |
| 2660 | 9173 fnames[0], spin.si_ascii ? (char_u *)"ascii" : spell_enc()); |
| 310 | 9174 |
| 9175 /* Check for .ascii.spl. */ | |
| 2660 | 9176 if (strstr((char *)gettail(wfname), SPL_FNAME_ASCII) != NULL) |
| 310 | 9177 spin.si_ascii = TRUE; |
| 9178 | |
| 9179 /* Check for .add.spl. */ | |
| 2660 | 9180 if (strstr((char *)gettail(wfname), SPL_FNAME_ADD) != NULL) |
| 310 | 9181 spin.si_add = TRUE; |
| 240 | 9182 } |
| 9183 | |
| 310 | 9184 if (incount <= 0) |
| 236 | 9185 EMSG(_(e_invarg)); /* need at least output and input names */ |
| 351 | 9186 else if (vim_strchr(gettail(wfname), '_') != NULL) |
| 9187 EMSG(_("E751: Output file name must not have region name")); | |
| 310 | 9188 else if (incount > 8) |
| 236 | 9189 EMSG(_("E754: Only up to 8 regions supported")); |
| 9190 else | |
| 9191 { | |
| 9192 /* Check for overwriting before doing things that may take a lot of | |
| 9193 * time. */ | |
| 310 | 9194 if (!overwrite && mch_stat((char *)wfname, &st) >= 0) |
| 236 | 9195 { |
| 9196 EMSG(_(e_exists)); | |
| 2770 | 9197 goto theend; |
| 236 | 9198 } |
| 310 | 9199 if (mch_isdir(wfname)) |
| 236 | 9200 { |
| 310 | 9201 EMSG2(_(e_isadir2), wfname); |
| 2770 | 9202 goto theend; |
| 9203 } | |
| 9204 | |
| 9205 fname = alloc(MAXPATHL); | |
| 9206 if (fname == NULL) | |
| 9207 goto theend; | |
| 236 | 9208 |
| 9209 /* | |
| 9210 * Init the aff and dic pointers. | |
| 9211 * Get the region names if there are more than 2 arguments. | |
| 9212 */ | |
| 310 | 9213 for (i = 0; i < incount; ++i) |
| 236 | 9214 { |
| 310 | 9215 afile[i] = NULL; |
| 300 | 9216 |
| 316 | 9217 if (incount > 1) |
| 236 | 9218 { |
| 835 | 9219 len = (int)STRLEN(innames[i]); |
| 310 | 9220 if (STRLEN(gettail(innames[i])) < 5 |
| 9221 || innames[i][len - 3] != '_') | |
| 236 | 9222 { |
| 310 | 9223 EMSG2(_("E755: Invalid region in %s"), innames[i]); |
| 2770 | 9224 goto theend; |
| 236 | 9225 } |
| 316 | 9226 spin.si_region_name[i * 2] = TOLOWER_ASC(innames[i][len - 2]); |
| 9227 spin.si_region_name[i * 2 + 1] = | |
| 9228 TOLOWER_ASC(innames[i][len - 1]); | |
| 236 | 9229 } |
| 9230 } | |
| 316 | 9231 spin.si_region_count = incount; |
| 236 | 9232 |
| 474 | 9233 spin.si_foldroot = wordtree_alloc(&spin); |
| 9234 spin.si_keeproot = wordtree_alloc(&spin); | |
| 9235 spin.si_prefroot = wordtree_alloc(&spin); | |
| 339 | 9236 if (spin.si_foldroot == NULL |
| 9237 || spin.si_keeproot == NULL | |
| 9238 || spin.si_prefroot == NULL) | |
| 300 | 9239 { |
| 469 | 9240 free_blocks(spin.si_blocks); |
| 2770 | 9241 goto theend; |
| 300 | 9242 } |
| 9243 | |
| 351 | 9244 /* When not producing a .add.spl file clear the character table when |
| 9245 * we encounter one in the .aff file. This means we dump the current | |
| 9246 * one in the .spl file if the .aff file doesn't define one. That's | |
| 9247 * better than guessing the contents, the table will match a | |
| 9248 * previously loaded spell file. */ | |
| 9249 if (!spin.si_add) | |
| 9250 spin.si_clear_chartab = TRUE; | |
| 9251 | |
| 236 | 9252 /* |
| 9253 * Read all the .aff and .dic files. | |
| 9254 * Text is converted to 'encoding'. | |
| 300 | 9255 * Words are stored in the case-folded and keep-case trees. |
| 236 | 9256 */ |
| 310 | 9257 for (i = 0; i < incount && !error; ++i) |
| 236 | 9258 { |
| 300 | 9259 spin.si_conv.vc_type = CONV_NONE; |
| 310 | 9260 spin.si_region = 1 << i; |
| 9261 | |
| 2770 | 9262 vim_snprintf((char *)fname, MAXPATHL, "%s.aff", innames[i]); |
| 300 | 9263 if (mch_stat((char *)fname, &st) >= 0) |
| 9264 { | |
| 9265 /* Read the .aff file. Will init "spin->si_conv" based on the | |
| 9266 * "SET" line. */ | |
| 474 | 9267 afile[i] = spell_read_aff(&spin, fname); |
| 310 | 9268 if (afile[i] == NULL) |
| 300 | 9269 error = TRUE; |
| 9270 else | |
| 9271 { | |
| 9272 /* Read the .dic file and store the words in the trees. */ | |
| 2770 | 9273 vim_snprintf((char *)fname, MAXPATHL, "%s.dic", |
| 310 | 9274 innames[i]); |
| 474 | 9275 if (spell_read_dic(&spin, fname, afile[i]) == FAIL) |
| 300 | 9276 error = TRUE; |
| 9277 } | |
| 9278 } | |
| 9279 else | |
| 9280 { | |
| 9281 /* No .aff file, try reading the file as a word list. Store | |
| 9282 * the words in the trees. */ | |
| 474 | 9283 if (spell_read_wordfile(&spin, innames[i]) == FAIL) |
| 300 | 9284 error = TRUE; |
| 9285 } | |
| 236 | 9286 |
| 310 | 9287 #ifdef FEAT_MBYTE |
| 236 | 9288 /* Free any conversion stuff. */ |
| 300 | 9289 convert_setup(&spin.si_conv, NULL, NULL); |
| 310 | 9290 #endif |
| 236 | 9291 } |
| 9292 | |
| 492 | 9293 if (spin.si_compflags != NULL && spin.si_nobreak) |
| 9294 MSG(_("Warning: both compounding and NOBREAK specified")); | |
| 9295 | |
| 625 | 9296 if (!error && !got_int) |
| 236 | 9297 { |
| 9298 /* | |
| 300 | 9299 * Combine tails in the tree. |
| 236 | 9300 */ |
| 625 | 9301 spell_message(&spin, (char_u *)_(msg_compressing)); |
| 474 | 9302 wordtree_compress(&spin, spin.si_foldroot); |
| 9303 wordtree_compress(&spin, spin.si_keeproot); | |
| 9304 wordtree_compress(&spin, spin.si_prefroot); | |
| 236 | 9305 } |
| 9306 | |
| 625 | 9307 if (!error && !got_int) |
| 300 | 9308 { |
| 9309 /* | |
| 9310 * Write the info in the spell file. | |
| 9311 */ | |
| 625 | 9312 vim_snprintf((char *)IObuff, IOSIZE, |
| 9313 _("Writing spell file %s ..."), wfname); | |
| 9314 spell_message(&spin, IObuff); | |
| 310 | 9315 |
| 500 | 9316 error = write_vim_spell(&spin, wfname) == FAIL; |
| 474 | 9317 |
| 625 | 9318 spell_message(&spin, (char_u *)_("Done!")); |
| 9319 vim_snprintf((char *)IObuff, IOSIZE, | |
| 9320 _("Estimated runtime memory use: %d bytes"), spin.si_memtot); | |
| 9321 spell_message(&spin, IObuff); | |
| 9322 | |
| 9323 /* | |
| 9324 * If the file is loaded need to reload it. | |
| 9325 */ | |
| 500 | 9326 if (!error) |
| 9327 spell_reload_one(wfname, added_word); | |
| 300 | 9328 } |
| 9329 | |
| 9330 /* Free the allocated memory. */ | |
| 323 | 9331 ga_clear(&spin.si_rep); |
| 625 | 9332 ga_clear(&spin.si_repsal); |
| 323 | 9333 ga_clear(&spin.si_sal); |
| 9334 ga_clear(&spin.si_map); | |
| 809 | 9335 ga_clear(&spin.si_comppat); |
| 339 | 9336 ga_clear(&spin.si_prefcond); |
| 625 | 9337 hash_clear_all(&spin.si_commonwords, 0); |
| 300 | 9338 |
| 9339 /* Free the .aff file structures. */ | |
| 310 | 9340 for (i = 0; i < incount; ++i) |
| 9341 if (afile[i] != NULL) | |
| 9342 spell_free_aff(afile[i]); | |
| 339 | 9343 |
| 9344 /* Free all the bits and pieces at once. */ | |
| 9345 free_blocks(spin.si_blocks); | |
| 625 | 9346 |
| 9347 /* | |
| 9348 * If there is soundfolding info and no NOSUGFILE item create the | |
| 9349 * .sug file with the soundfolded word trie. | |
| 9350 */ | |
| 9351 if (spin.si_sugtime != 0 && !error && !got_int) | |
| 9352 spell_make_sugfile(&spin, wfname); | |
| 9353 | |
| 9354 } | |
| 2770 | 9355 |
| 9356 theend: | |
| 9357 vim_free(fname); | |
| 9358 vim_free(wfname); | |
| 625 | 9359 } |
| 9360 | |
| 9361 /* | |
| 9362 * Display a message for spell file processing when 'verbose' is set or using | |
| 9363 * ":mkspell". "str" can be IObuff. | |
| 9364 */ | |
| 9365 static void | |
| 9366 spell_message(spin, str) | |
| 9367 spellinfo_T *spin; | |
| 9368 char_u *str; | |
| 9369 { | |
| 9370 if (spin->si_verbose || p_verbose > 2) | |
| 9371 { | |
| 9372 if (!spin->si_verbose) | |
| 9373 verbose_enter(); | |
| 9374 MSG(str); | |
| 9375 out_flush(); | |
| 9376 if (!spin->si_verbose) | |
| 9377 verbose_leave(); | |
| 9378 } | |
| 9379 } | |
| 310 | 9380 |
| 9381 /* | |
| 385 | 9382 * ":[count]spellgood {word}" |
| 9383 * ":[count]spellwrong {word}" | |
| 710 | 9384 * ":[count]spellundo {word}" |
| 310 | 9385 */ |
| 9386 void | |
| 9387 ex_spell(eap) | |
| 9388 exarg_T *eap; | |
| 9389 { | |
| 835 | 9390 spell_add_word(eap->arg, (int)STRLEN(eap->arg), eap->cmdidx == CMD_spellwrong, |
| 710 | 9391 eap->forceit ? 0 : (int)eap->line2, |
| 9392 eap->cmdidx == CMD_spellundo); | |
| 236 | 9393 } |
| 9394 | |
| 310 | 9395 /* |
| 9396 * Add "word[len]" to 'spellfile' as a good or bad word. | |
| 9397 */ | |
| 9398 void | |
| 944 | 9399 spell_add_word(word, len, bad, idx, undo) |
| 310 | 9400 char_u *word; |
| 9401 int len; | |
| 9402 int bad; | |
| 944 | 9403 int idx; /* "zG" and "zW": zero, otherwise index in |
| 385 | 9404 'spellfile' */ |
| 710 | 9405 int undo; /* TRUE for "zug", "zuG", "zuw" and "zuW" */ |
| 310 | 9406 { |
| 977 | 9407 FILE *fd = NULL; |
| 385 | 9408 buf_T *buf = NULL; |
| 351 | 9409 int new_spf = FALSE; |
| 381 | 9410 char_u *fname; |
| 2770 | 9411 char_u *fnamebuf = NULL; |
| 385 | 9412 char_u line[MAXWLEN * 2]; |
| 9413 long fpos, fpos_next = 0; | |
| 9414 int i; | |
| 9415 char_u *spf; | |
| 9416 | |
| 944 | 9417 if (idx == 0) /* use internal wordlist */ |
| 385 | 9418 { |
| 9419 if (int_wordlist == NULL) | |
| 9420 { | |
| 9421 int_wordlist = vim_tempname('s'); | |
| 9422 if (int_wordlist == NULL) | |
| 381 | 9423 return; |
| 9424 } | |
| 385 | 9425 fname = int_wordlist; |
| 381 | 9426 } |
| 310 | 9427 else |
| 9428 { | |
| 381 | 9429 /* If 'spellfile' isn't set figure out a good default value. */ |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9430 if (*curwin->w_s->b_p_spf == NUL) |
| 381 | 9431 { |
| 9432 init_spellfile(); | |
| 9433 new_spf = TRUE; | |
| 9434 } | |
| 9435 | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9436 if (*curwin->w_s->b_p_spf == NUL) |
| 381 | 9437 { |
| 523 | 9438 EMSG2(_(e_notset), "spellfile"); |
| 381 | 9439 return; |
| 9440 } | |
| 2770 | 9441 fnamebuf = alloc(MAXPATHL); |
| 9442 if (fnamebuf == NULL) | |
| 9443 return; | |
| 381 | 9444 |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9445 for (spf = curwin->w_s->b_p_spf, i = 1; *spf != NUL; ++i) |
| 385 | 9446 { |
| 9447 copy_option_part(&spf, fnamebuf, MAXPATHL, ","); | |
| 944 | 9448 if (i == idx) |
| 385 | 9449 break; |
| 9450 if (*spf == NUL) | |
| 9451 { | |
| 944 | 9452 EMSGN(_("E765: 'spellfile' does not have %ld entries"), idx); |
| 2770 | 9453 vim_free(fnamebuf); |
| 385 | 9454 return; |
| 9455 } | |
| 9456 } | |
| 9457 | |
| 310 | 9458 /* Check that the user isn't editing the .add file somewhere. */ |
| 385 | 9459 buf = buflist_findname_exp(fnamebuf); |
| 310 | 9460 if (buf != NULL && buf->b_ml.ml_mfp == NULL) |
| 9461 buf = NULL; | |
| 9462 if (buf != NULL && bufIsChanged(buf)) | |
| 381 | 9463 { |
| 310 | 9464 EMSG(_(e_bufloaded)); |
| 2770 | 9465 vim_free(fnamebuf); |
| 381 | 9466 return; |
| 9467 } | |
| 9468 | |
| 385 | 9469 fname = fnamebuf; |
| 9470 } | |
| 9471 | |
| 710 | 9472 if (bad || undo) |
| 9473 { | |
| 9474 /* When the word appears as good word we need to remove that one, | |
| 385 | 9475 * since its flags sort before the one with WF_BANNED. */ |
| 9476 fd = mch_fopen((char *)fname, "r"); | |
| 9477 if (fd != NULL) | |
| 9478 { | |
| 9479 while (!vim_fgets(line, MAXWLEN * 2, fd)) | |
| 9480 { | |
| 9481 fpos = fpos_next; | |
| 9482 fpos_next = ftell(fd); | |
| 9483 if (STRNCMP(word, line, len) == 0 | |
| 9484 && (line[len] == '/' || line[len] < ' ')) | |
| 9485 { | |
| 9486 /* Found duplicate word. Remove it by writing a '#' at | |
| 9487 * the start of the line. Mixing reading and writing | |
| 9488 * doesn't work for all systems, close the file first. */ | |
| 9489 fclose(fd); | |
| 9490 fd = mch_fopen((char *)fname, "r+"); | |
| 9491 if (fd == NULL) | |
| 9492 break; | |
| 9493 if (fseek(fd, fpos, SEEK_SET) == 0) | |
| 710 | 9494 { |
| 385 | 9495 fputc('#', fd); |
| 710 | 9496 if (undo) |
| 974 | 9497 { |
| 9498 home_replace(NULL, fname, NameBuff, MAXPATHL, TRUE); | |
| 844 | 9499 smsg((char_u *)_("Word removed from %s"), NameBuff); |
| 974 | 9500 } |
| 710 | 9501 } |
| 385 | 9502 fseek(fd, fpos_next, SEEK_SET); |
| 9503 } | |
| 9504 } | |
|
2445
04dae202d316
Fixes for coverity warnings.
Bram Moolenaar <bram@vim.org>
parents:
2299
diff
changeset
|
9505 if (fd != NULL) |
|
04dae202d316
Fixes for coverity warnings.
Bram Moolenaar <bram@vim.org>
parents:
2299
diff
changeset
|
9506 fclose(fd); |
| 385 | 9507 } |
| 381 | 9508 } |
| 975 | 9509 |
| 9510 if (!undo) | |
| 710 | 9511 { |
| 9512 fd = mch_fopen((char *)fname, "a"); | |
| 9513 if (fd == NULL && new_spf) | |
| 9514 { | |
| 975 | 9515 char_u *p; |
| 9516 | |
| 710 | 9517 /* We just initialized the 'spellfile' option and can't open the |
| 9518 * file. We may need to create the "spell" directory first. We | |
| 9519 * already checked the runtime directory is writable in | |
| 9520 * init_spellfile(). */ | |
| 975 | 9521 if (!dir_of_file_exists(fname) && (p = gettail_sep(fname)) != fname) |
| 9522 { | |
| 9523 int c = *p; | |
| 9524 | |
| 710 | 9525 /* The directory doesn't exist. Try creating it and opening |
| 9526 * the file again. */ | |
| 975 | 9527 *p = NUL; |
| 9528 vim_mkdir(fname, 0755); | |
| 9529 *p = c; | |
| 710 | 9530 fd = mch_fopen((char *)fname, "a"); |
| 9531 } | |
| 9532 } | |
| 9533 | |
| 9534 if (fd == NULL) | |
| 9535 EMSG2(_(e_notopen), fname); | |
| 310 | 9536 else |
| 710 | 9537 { |
| 9538 if (bad) | |
| 9539 fprintf(fd, "%.*s/!\n", len, word); | |
| 9540 else | |
| 9541 fprintf(fd, "%.*s\n", len, word); | |
| 9542 fclose(fd); | |
| 9543 | |
| 9544 home_replace(NULL, fname, NameBuff, MAXPATHL, TRUE); | |
| 9545 smsg((char_u *)_("Word added to %s"), NameBuff); | |
| 9546 } | |
| 9547 } | |
| 9548 | |
| 9549 if (fd != NULL) | |
| 9550 { | |
| 381 | 9551 /* Update the .add.spl file. */ |
| 9552 mkspell(1, &fname, FALSE, TRUE, TRUE); | |
| 9553 | |
| 9554 /* If the .add file is edited somewhere, reload it. */ | |
| 9555 if (buf != NULL) | |
| 627 | 9556 buf_reload(buf, buf->b_orig_mode); |
| 381 | 9557 |
| 737 | 9558 redraw_all_later(SOME_VALID); |
| 310 | 9559 } |
| 2770 | 9560 vim_free(fnamebuf); |
| 310 | 9561 } |
| 9562 | |
| 9563 /* | |
| 9564 * Initialize 'spellfile' for the current buffer. | |
| 9565 */ | |
| 9566 static void | |
| 9567 init_spellfile() | |
| 9568 { | |
| 2770 | 9569 char_u *buf; |
| 310 | 9570 int l; |
| 500 | 9571 char_u *fname; |
| 310 | 9572 char_u *rtp; |
| 351 | 9573 char_u *lend; |
| 501 | 9574 int aspath = FALSE; |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9575 char_u *lstart = curbuf->b_s.b_p_spl; |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9576 |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9577 if (*curwin->w_s->b_p_spl != NUL && curwin->w_s->b_langp.ga_len > 0) |
| 310 | 9578 { |
| 2770 | 9579 buf = alloc(MAXPATHL); |
| 9580 if (buf == NULL) | |
| 9581 return; | |
| 9582 | |
| 501 | 9583 /* Find the end of the language name. Exclude the region. If there |
| 9584 * is a path separator remember the start of the tail. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9585 for (lend = curwin->w_s->b_p_spl; *lend != NUL |
| 351 | 9586 && vim_strchr((char_u *)",._", *lend) == NULL; ++lend) |
| 501 | 9587 if (vim_ispathsep(*lend)) |
| 9588 { | |
| 9589 aspath = TRUE; | |
| 9590 lstart = lend + 1; | |
| 9591 } | |
| 351 | 9592 |
| 9593 /* Loop over all entries in 'runtimepath'. Use the first one where we | |
| 9594 * are allowed to write. */ | |
| 310 | 9595 rtp = p_rtp; |
| 9596 while (*rtp != NUL) | |
| 9597 { | |
| 501 | 9598 if (aspath) |
| 9599 /* Use directory of an entry with path, e.g., for | |
| 9600 * "/dir/lg.utf-8.spl" use "/dir". */ | |
| 2593 | 9601 vim_strncpy(buf, curbuf->b_s.b_p_spl, |
| 9602 lstart - curbuf->b_s.b_p_spl - 1); | |
| 501 | 9603 else |
| 9604 /* Copy the path from 'runtimepath' to buf[]. */ | |
| 9605 copy_option_part(&rtp, buf, MAXPATHL, ","); | |
| 310 | 9606 if (filewritable(buf) == 2) |
| 9607 { | |
| 316 | 9608 /* Use the first language name from 'spelllang' and the |
| 9609 * encoding used in the first loaded .spl file. */ | |
| 501 | 9610 if (aspath) |
| 2593 | 9611 vim_strncpy(buf, curbuf->b_s.b_p_spl, |
| 9612 lend - curbuf->b_s.b_p_spl); | |
| 501 | 9613 else |
| 9614 { | |
| 819 | 9615 /* Create the "spell" directory if it doesn't exist yet. */ |
| 835 | 9616 l = (int)STRLEN(buf); |
| 819 | 9617 vim_snprintf((char *)buf + l, MAXPATHL - l, "/spell"); |
| 2593 | 9618 if (filewritable(buf) != 2) |
| 819 | 9619 vim_mkdir(buf, 0755); |
| 9620 | |
| 835 | 9621 l = (int)STRLEN(buf); |
| 501 | 9622 vim_snprintf((char *)buf + l, MAXPATHL - l, |
| 819 | 9623 "/%.*s", (int)(lend - lstart), lstart); |
| 501 | 9624 } |
| 835 | 9625 l = (int)STRLEN(buf); |
| 2770 | 9626 fname = LANGP_ENTRY(curwin->w_s->b_langp, 0) |
| 9627 ->lp_slang->sl_fname; | |
| 501 | 9628 vim_snprintf((char *)buf + l, MAXPATHL - l, ".%s.add", |
| 9629 fname != NULL | |
| 9630 && strstr((char *)gettail(fname), ".ascii.") != NULL | |
| 9631 ? (char_u *)"ascii" : spell_enc()); | |
| 310 | 9632 set_option_value((char_u *)"spellfile", 0L, buf, OPT_LOCAL); |
| 9633 break; | |
| 9634 } | |
| 501 | 9635 aspath = FALSE; |
| 310 | 9636 } |
| 2770 | 9637 |
| 9638 vim_free(buf); | |
| 310 | 9639 } |
| 9640 } | |
| 236 | 9641 |
| 300 | 9642 |
| 307 | 9643 /* |
| 9644 * Init the chartab used for spelling for ASCII. | |
| 9645 * EBCDIC is not supported! | |
| 9646 */ | |
| 9647 static void | |
| 9648 clear_spell_chartab(sp) | |
| 9649 spelltab_T *sp; | |
| 9650 { | |
| 324 | 9651 int i; |
| 307 | 9652 |
| 9653 /* Init everything to FALSE. */ | |
| 9654 vim_memset(sp->st_isw, FALSE, sizeof(sp->st_isw)); | |
| 9655 vim_memset(sp->st_isu, FALSE, sizeof(sp->st_isu)); | |
| 9656 for (i = 0; i < 256; ++i) | |
| 324 | 9657 { |
| 307 | 9658 sp->st_fold[i] = i; |
| 324 | 9659 sp->st_upper[i] = i; |
| 9660 } | |
| 307 | 9661 |
| 9662 /* We include digits. A word shouldn't start with a digit, but handling | |
| 9663 * that is done separately. */ | |
| 9664 for (i = '0'; i <= '9'; ++i) | |
| 9665 sp->st_isw[i] = TRUE; | |
| 9666 for (i = 'A'; i <= 'Z'; ++i) | |
| 9667 { | |
| 9668 sp->st_isw[i] = TRUE; | |
| 9669 sp->st_isu[i] = TRUE; | |
| 9670 sp->st_fold[i] = i + 0x20; | |
| 9671 } | |
| 9672 for (i = 'a'; i <= 'z'; ++i) | |
| 324 | 9673 { |
| 307 | 9674 sp->st_isw[i] = TRUE; |
| 324 | 9675 sp->st_upper[i] = i - 0x20; |
| 9676 } | |
| 307 | 9677 } |
| 9678 | |
| 9679 /* | |
| 9680 * Init the chartab used for spelling. Only depends on 'encoding'. | |
| 9681 * Called once while starting up and when 'encoding' changes. | |
| 9682 * The default is to use isalpha(), but the spell file should define the word | |
| 9683 * characters to make it possible that 'encoding' differs from the current | |
| 390 | 9684 * locale. For utf-8 we don't use isalpha() but our own functions. |
| 307 | 9685 */ |
| 9686 void | |
| 9687 init_spell_chartab() | |
| 9688 { | |
| 9689 int i; | |
| 9690 | |
| 9691 did_set_spelltab = FALSE; | |
| 9692 clear_spell_chartab(&spelltab); | |
| 9693 #ifdef FEAT_MBYTE | |
| 9694 if (enc_dbcs) | |
| 9695 { | |
| 9696 /* DBCS: assume double-wide characters are word characters. */ | |
| 9697 for (i = 128; i <= 255; ++i) | |
| 9698 if (MB_BYTE2LEN(i) == 2) | |
| 9699 spelltab.st_isw[i] = TRUE; | |
| 9700 } | |
| 324 | 9701 else if (enc_utf8) |
| 9702 { | |
| 9703 for (i = 128; i < 256; ++i) | |
| 9704 { | |
|
2140
12aba62fa7c6
updated for version 7.2.422
Bram Moolenaar <bram@zimbu.org>
parents:
2046
diff
changeset
|
9705 int f = utf_fold(i); |
|
12aba62fa7c6
updated for version 7.2.422
Bram Moolenaar <bram@zimbu.org>
parents:
2046
diff
changeset
|
9706 int u = utf_toupper(i); |
|
12aba62fa7c6
updated for version 7.2.422
Bram Moolenaar <bram@zimbu.org>
parents:
2046
diff
changeset
|
9707 |
| 324 | 9708 spelltab.st_isu[i] = utf_isupper(i); |
| 9709 spelltab.st_isw[i] = spelltab.st_isu[i] || utf_islower(i); | |
|
2140
12aba62fa7c6
updated for version 7.2.422
Bram Moolenaar <bram@zimbu.org>
parents:
2046
diff
changeset
|
9710 /* The folded/upper-cased value is different between latin1 and |
|
12aba62fa7c6
updated for version 7.2.422
Bram Moolenaar <bram@zimbu.org>
parents:
2046
diff
changeset
|
9711 * utf8 for 0xb5, causing E763 for no good reason. Use the latin1 |
|
12aba62fa7c6
updated for version 7.2.422
Bram Moolenaar <bram@zimbu.org>
parents:
2046
diff
changeset
|
9712 * value for utf-8 to avoid this. */ |
|
12aba62fa7c6
updated for version 7.2.422
Bram Moolenaar <bram@zimbu.org>
parents:
2046
diff
changeset
|
9713 spelltab.st_fold[i] = (f < 256) ? f : i; |
|
12aba62fa7c6
updated for version 7.2.422
Bram Moolenaar <bram@zimbu.org>
parents:
2046
diff
changeset
|
9714 spelltab.st_upper[i] = (u < 256) ? u : i; |
| 324 | 9715 } |
| 9716 } | |
| 307 | 9717 else |
| 9718 #endif | |
| 9719 { | |
| 324 | 9720 /* Rough guess: use locale-dependent library functions. */ |
| 307 | 9721 for (i = 128; i < 256; ++i) |
| 9722 { | |
| 9723 if (MB_ISUPPER(i)) | |
| 9724 { | |
| 324 | 9725 spelltab.st_isw[i] = TRUE; |
| 307 | 9726 spelltab.st_isu[i] = TRUE; |
| 9727 spelltab.st_fold[i] = MB_TOLOWER(i); | |
| 9728 } | |
| 324 | 9729 else if (MB_ISLOWER(i)) |
| 9730 { | |
| 9731 spelltab.st_isw[i] = TRUE; | |
| 9732 spelltab.st_upper[i] = MB_TOUPPER(i); | |
| 9733 } | |
| 307 | 9734 } |
| 9735 } | |
| 9736 } | |
| 9737 | |
| 9738 /* | |
| 9739 * Set the spell character tables from strings in the affix file. | |
| 9740 */ | |
| 9741 static int | |
| 9742 set_spell_chartab(fol, low, upp) | |
| 9743 char_u *fol; | |
| 9744 char_u *low; | |
| 9745 char_u *upp; | |
| 9746 { | |
| 9747 /* We build the new tables here first, so that we can compare with the | |
| 9748 * previous one. */ | |
| 9749 spelltab_T new_st; | |
| 9750 char_u *pf = fol, *pl = low, *pu = upp; | |
| 9751 int f, l, u; | |
| 9752 | |
| 9753 clear_spell_chartab(&new_st); | |
| 9754 | |
| 9755 while (*pf != NUL) | |
| 9756 { | |
| 9757 if (*pl == NUL || *pu == NUL) | |
| 9758 { | |
| 9759 EMSG(_(e_affform)); | |
| 9760 return FAIL; | |
| 9761 } | |
| 9762 #ifdef FEAT_MBYTE | |
| 9763 f = mb_ptr2char_adv(&pf); | |
| 9764 l = mb_ptr2char_adv(&pl); | |
| 9765 u = mb_ptr2char_adv(&pu); | |
| 9766 #else | |
| 9767 f = *pf++; | |
| 9768 l = *pl++; | |
| 9769 u = *pu++; | |
| 9770 #endif | |
| 9771 /* Every character that appears is a word character. */ | |
| 9772 if (f < 256) | |
| 9773 new_st.st_isw[f] = TRUE; | |
| 9774 if (l < 256) | |
| 9775 new_st.st_isw[l] = TRUE; | |
| 9776 if (u < 256) | |
| 9777 new_st.st_isw[u] = TRUE; | |
| 9778 | |
| 9779 /* if "LOW" and "FOL" are not the same the "LOW" char needs | |
| 9780 * case-folding */ | |
| 9781 if (l < 256 && l != f) | |
| 9782 { | |
| 9783 if (f >= 256) | |
| 9784 { | |
| 9785 EMSG(_(e_affrange)); | |
| 9786 return FAIL; | |
| 9787 } | |
| 9788 new_st.st_fold[l] = f; | |
| 9789 } | |
| 9790 | |
| 9791 /* if "UPP" and "FOL" are not the same the "UPP" char needs | |
| 324 | 9792 * case-folding, it's upper case and the "UPP" is the upper case of |
| 9793 * "FOL" . */ | |
| 307 | 9794 if (u < 256 && u != f) |
| 9795 { | |
| 9796 if (f >= 256) | |
| 9797 { | |
| 9798 EMSG(_(e_affrange)); | |
| 9799 return FAIL; | |
| 9800 } | |
| 9801 new_st.st_fold[u] = f; | |
| 9802 new_st.st_isu[u] = TRUE; | |
| 324 | 9803 new_st.st_upper[f] = u; |
| 307 | 9804 } |
| 9805 } | |
| 9806 | |
| 9807 if (*pl != NUL || *pu != NUL) | |
| 9808 { | |
| 9809 EMSG(_(e_affform)); | |
| 9810 return FAIL; | |
| 9811 } | |
| 9812 | |
| 9813 return set_spell_finish(&new_st); | |
| 9814 } | |
| 9815 | |
| 9816 /* | |
| 9817 * Set the spell character tables from strings in the .spl file. | |
| 9818 */ | |
| 483 | 9819 static void |
| 389 | 9820 set_spell_charflags(flags, cnt, fol) |
| 307 | 9821 char_u *flags; |
| 389 | 9822 int cnt; /* length of "flags" */ |
| 9823 char_u *fol; | |
| 307 | 9824 { |
| 9825 /* We build the new tables here first, so that we can compare with the | |
| 9826 * previous one. */ | |
| 9827 spelltab_T new_st; | |
| 9828 int i; | |
| 389 | 9829 char_u *p = fol; |
| 324 | 9830 int c; |
| 307 | 9831 |
| 9832 clear_spell_chartab(&new_st); | |
| 9833 | |
| 389 | 9834 for (i = 0; i < 128; ++i) |
| 9835 { | |
| 9836 if (i < cnt) | |
| 9837 { | |
| 9838 new_st.st_isw[i + 128] = (flags[i] & CF_WORD) != 0; | |
| 9839 new_st.st_isu[i + 128] = (flags[i] & CF_UPPER) != 0; | |
| 9840 } | |
| 9841 | |
| 9842 if (*p != NUL) | |
| 9843 { | |
| 307 | 9844 #ifdef FEAT_MBYTE |
| 389 | 9845 c = mb_ptr2char_adv(&p); |
| 307 | 9846 #else |
| 389 | 9847 c = *p++; |
| 307 | 9848 #endif |
| 389 | 9849 new_st.st_fold[i + 128] = c; |
| 9850 if (i + 128 != c && new_st.st_isu[i + 128] && c < 256) | |
| 9851 new_st.st_upper[c] = i + 128; | |
| 9852 } | |
| 307 | 9853 } |
| 9854 | |
| 483 | 9855 (void)set_spell_finish(&new_st); |
| 307 | 9856 } |
| 9857 | |
| 9858 static int | |
| 9859 set_spell_finish(new_st) | |
| 9860 spelltab_T *new_st; | |
| 9861 { | |
| 9862 int i; | |
| 9863 | |
| 9864 if (did_set_spelltab) | |
| 9865 { | |
| 9866 /* check that it's the same table */ | |
| 9867 for (i = 0; i < 256; ++i) | |
| 9868 { | |
| 9869 if (spelltab.st_isw[i] != new_st->st_isw[i] | |
| 9870 || spelltab.st_isu[i] != new_st->st_isu[i] | |
| 324 | 9871 || spelltab.st_fold[i] != new_st->st_fold[i] |
| 9872 || spelltab.st_upper[i] != new_st->st_upper[i]) | |
| 307 | 9873 { |
| 9874 EMSG(_("E763: Word characters differ between spell files")); | |
| 9875 return FAIL; | |
| 9876 } | |
| 9877 } | |
| 9878 } | |
| 9879 else | |
| 9880 { | |
| 9881 /* copy the new spelltab into the one being used */ | |
| 9882 spelltab = *new_st; | |
| 9883 did_set_spelltab = TRUE; | |
| 9884 } | |
| 9885 | |
| 9886 return OK; | |
| 9887 } | |
| 9888 | |
| 9889 /* | |
| 358 | 9890 * Return TRUE if "p" points to a word character. |
| 366 | 9891 * As a special case we see "midword" characters as word character when it is |
| 358 | 9892 * followed by a word character. This finds they'there but not 'they there'. |
| 366 | 9893 * Thus this only works properly when past the first character of the word. |
| 358 | 9894 */ |
| 9895 static int | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9896 spell_iswordp(p, wp) |
| 358 | 9897 char_u *p; |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9898 win_T *wp; /* buffer used */ |
| 358 | 9899 { |
| 366 | 9900 #ifdef FEAT_MBYTE |
| 358 | 9901 char_u *s; |
| 366 | 9902 int l; |
| 9903 int c; | |
| 9904 | |
| 9905 if (has_mbyte) | |
| 9906 { | |
| 9907 l = MB_BYTE2LEN(*p); | |
| 358 | 9908 s = p; |
| 366 | 9909 if (l == 1) |
| 9910 { | |
| 9911 /* be quick for ASCII */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9912 if (wp->w_s->b_spell_ismw[*p]) |
| 366 | 9913 s = p + 1; /* skip a mid-word character */ |
| 9914 } | |
| 9915 else | |
| 9916 { | |
| 9917 c = mb_ptr2char(p); | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9918 if (c < 256 ? wp->w_s->b_spell_ismw[c] |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9919 : (wp->w_s->b_spell_ismw_mb != NULL |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9920 && vim_strchr(wp->w_s->b_spell_ismw_mb, c) != NULL)) |
| 366 | 9921 s = p + l; |
| 9922 } | |
| 9923 | |
| 390 | 9924 c = mb_ptr2char(s); |
| 9925 if (c > 255) | |
| 1580 | 9926 return spell_mb_isword_class(mb_get_class(s)); |
| 390 | 9927 return spelltab.st_isw[c]; |
| 366 | 9928 } |
| 358 | 9929 #endif |
| 366 | 9930 |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9931 return spelltab.st_isw[wp->w_s->b_spell_ismw[*p] ? p[1] : p[0]]; |
| 376 | 9932 } |
| 9933 | |
| 9934 /* | |
| 9935 * Return TRUE if "p" points to a word character. | |
| 9936 * Unlike spell_iswordp() this doesn't check for "midword" characters. | |
| 9937 */ | |
| 9938 static int | |
| 9939 spell_iswordp_nmw(p) | |
| 9940 char_u *p; | |
| 9941 { | |
| 9942 #ifdef FEAT_MBYTE | |
| 390 | 9943 int c; |
| 9944 | |
| 9945 if (has_mbyte) | |
| 9946 { | |
| 9947 c = mb_ptr2char(p); | |
| 9948 if (c > 255) | |
| 1580 | 9949 return spell_mb_isword_class(mb_get_class(p)); |
| 390 | 9950 return spelltab.st_isw[c]; |
| 9951 } | |
| 376 | 9952 #endif |
| 9953 return spelltab.st_isw[*p]; | |
| 358 | 9954 } |
| 9955 | |
| 372 | 9956 #ifdef FEAT_MBYTE |
| 9957 /* | |
| 1580 | 9958 * Return TRUE if word class indicates a word character. |
| 9959 * Only for characters above 255. | |
| 9960 * Unicode subscript and superscript are not considered word characters. | |
| 9961 */ | |
| 9962 static int | |
| 9963 spell_mb_isword_class(cl) | |
| 9964 int cl; | |
| 9965 { | |
| 9966 return cl >= 2 && cl != 0x2070 && cl != 0x2080; | |
| 9967 } | |
| 9968 | |
| 9969 /* | |
| 372 | 9970 * Return TRUE if "p" points to a word character. |
| 9971 * Wide version of spell_iswordp(). | |
| 9972 */ | |
| 9973 static int | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9974 spell_iswordp_w(p, wp) |
| 372 | 9975 int *p; |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9976 win_T *wp; |
| 372 | 9977 { |
| 9978 int *s; | |
| 9979 | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9980 if (*p < 256 ? wp->w_s->b_spell_ismw[*p] |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9981 : (wp->w_s->b_spell_ismw_mb != NULL |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
9982 && vim_strchr(wp->w_s->b_spell_ismw_mb, *p) != NULL)) |
| 372 | 9983 s = p + 1; |
| 9984 else | |
| 9985 s = p; | |
| 9986 | |
| 390 | 9987 if (*s > 255) |
| 372 | 9988 { |
| 9989 if (enc_utf8) | |
| 1580 | 9990 return spell_mb_isword_class(utf_class(*s)); |
| 372 | 9991 if (enc_dbcs) |
| 9992 return dbcs_class((unsigned)*s >> 8, *s & 0xff) >= 2; | |
| 9993 return 0; | |
| 9994 } | |
| 9995 return spelltab.st_isw[*s]; | |
| 9996 } | |
| 9997 #endif | |
| 9998 | |
| 358 | 9999 /* |
| 339 | 10000 * Write the table with prefix conditions to the .spl file. |
| 483 | 10001 * When "fd" is NULL only count the length of what is written. |
| 10002 */ | |
| 10003 static int | |
| 339 | 10004 write_spell_prefcond(fd, gap) |
| 10005 FILE *fd; | |
| 10006 garray_T *gap; | |
| 10007 { | |
| 10008 int i; | |
| 10009 char_u *p; | |
| 10010 int len; | |
| 483 | 10011 int totlen; |
| 1759 | 10012 size_t x = 1; /* collect return value of fwrite() */ |
| 483 | 10013 |
| 10014 if (fd != NULL) | |
| 10015 put_bytes(fd, (long_u)gap->ga_len, 2); /* <prefcondcnt> */ | |
| 10016 | |
| 10017 totlen = 2 + gap->ga_len; /* length of <prefcondcnt> and <condlen> bytes */ | |
| 339 | 10018 |
| 10019 for (i = 0; i < gap->ga_len; ++i) | |
| 10020 { | |
| 10021 /* <prefcond> : <condlen> <condstr> */ | |
| 10022 p = ((char_u **)gap->ga_data)[i]; | |
| 483 | 10023 if (p != NULL) |
| 339 | 10024 { |
| 835 | 10025 len = (int)STRLEN(p); |
| 483 | 10026 if (fd != NULL) |
| 10027 { | |
| 10028 fputc(len, fd); | |
| 1753 | 10029 x &= fwrite(p, (size_t)len, (size_t)1, fd); |
| 483 | 10030 } |
| 10031 totlen += len; | |
| 10032 } | |
| 10033 else if (fd != NULL) | |
| 10034 fputc(0, fd); | |
| 10035 } | |
| 10036 | |
| 10037 return totlen; | |
| 307 | 10038 } |
| 10039 | |
| 10040 /* | |
| 324 | 10041 * Case-fold "str[len]" into "buf[buflen]". The result is NUL terminated. |
| 10042 * Uses the character definitions from the .spl file. | |
| 307 | 10043 * When using a multi-byte 'encoding' the length may change! |
| 10044 * Returns FAIL when something wrong. | |
| 10045 */ | |
| 10046 static int | |
| 324 | 10047 spell_casefold(str, len, buf, buflen) |
| 10048 char_u *str; | |
| 307 | 10049 int len; |
| 10050 char_u *buf; | |
| 10051 int buflen; | |
| 10052 { | |
| 10053 int i; | |
| 10054 | |
| 10055 if (len >= buflen) | |
| 10056 { | |
| 10057 buf[0] = NUL; | |
| 10058 return FAIL; /* result will not fit */ | |
| 10059 } | |
| 10060 | |
| 10061 #ifdef FEAT_MBYTE | |
| 10062 if (has_mbyte) | |
| 10063 { | |
| 324 | 10064 int outi = 0; |
| 10065 char_u *p; | |
| 307 | 10066 int c; |
| 10067 | |
| 10068 /* Fold one character at a time. */ | |
| 324 | 10069 for (p = str; p < str + len; ) |
| 307 | 10070 { |
| 10071 if (outi + MB_MAXBYTES > buflen) | |
| 10072 { | |
| 10073 buf[outi] = NUL; | |
| 10074 return FAIL; | |
| 10075 } | |
| 474 | 10076 c = mb_cptr2char_adv(&p); |
| 324 | 10077 outi += mb_char2bytes(SPELL_TOFOLD(c), buf + outi); |
| 307 | 10078 } |
| 10079 buf[outi] = NUL; | |
| 10080 } | |
| 10081 else | |
| 10082 #endif | |
| 10083 { | |
| 10084 /* Be quick for non-multibyte encodings. */ | |
| 10085 for (i = 0; i < len; ++i) | |
| 324 | 10086 buf[i] = spelltab.st_fold[str[i]]; |
| 307 | 10087 buf[i] = NUL; |
| 10088 } | |
| 10089 | |
| 10090 return OK; | |
| 10091 } | |
| 10092 | |
| 625 | 10093 /* values for sps_flags */ |
| 372 | 10094 #define SPS_BEST 1 |
| 10095 #define SPS_FAST 2 | |
| 10096 #define SPS_DOUBLE 4 | |
| 10097 | |
| 625 | 10098 static int sps_flags = SPS_BEST; /* flags from 'spellsuggest' */ |
| 10099 static int sps_limit = 9999; /* max nr of suggestions given */ | |
| 372 | 10100 |
| 10101 /* | |
| 10102 * Check the 'spellsuggest' option. Return FAIL if it's wrong. | |
| 483 | 10103 * Sets "sps_flags" and "sps_limit". |
| 372 | 10104 */ |
| 10105 int | |
| 10106 spell_check_sps() | |
| 10107 { | |
| 10108 char_u *p; | |
| 483 | 10109 char_u *s; |
| 372 | 10110 char_u buf[MAXPATHL]; |
| 10111 int f; | |
| 10112 | |
| 10113 sps_flags = 0; | |
| 483 | 10114 sps_limit = 9999; |
| 372 | 10115 |
| 10116 for (p = p_sps; *p != NUL; ) | |
| 10117 { | |
| 10118 copy_option_part(&p, buf, MAXPATHL, ","); | |
| 10119 | |
| 10120 f = 0; | |
| 483 | 10121 if (VIM_ISDIGIT(*buf)) |
| 10122 { | |
| 10123 s = buf; | |
| 10124 sps_limit = getdigits(&s); | |
| 10125 if (*s != NUL && !VIM_ISDIGIT(*s)) | |
| 10126 f = -1; | |
| 10127 } | |
| 10128 else if (STRCMP(buf, "best") == 0) | |
| 372 | 10129 f = SPS_BEST; |
| 10130 else if (STRCMP(buf, "fast") == 0) | |
| 10131 f = SPS_FAST; | |
| 10132 else if (STRCMP(buf, "double") == 0) | |
| 10133 f = SPS_DOUBLE; | |
| 10134 else if (STRNCMP(buf, "expr:", 5) != 0 | |
| 10135 && STRNCMP(buf, "file:", 5) != 0) | |
| 10136 f = -1; | |
| 10137 | |
| 10138 if (f == -1 || (sps_flags != 0 && f != 0)) | |
| 10139 { | |
| 10140 sps_flags = SPS_BEST; | |
| 483 | 10141 sps_limit = 9999; |
| 372 | 10142 return FAIL; |
| 10143 } | |
| 10144 if (f != 0) | |
| 10145 sps_flags = f; | |
| 10146 } | |
| 10147 | |
| 10148 if (sps_flags == 0) | |
| 10149 sps_flags = SPS_BEST; | |
| 10150 | |
| 10151 return OK; | |
| 10152 } | |
| 10153 | |
| 323 | 10154 /* |
| 10155 * "z?": Find badly spelled word under or after the cursor. | |
| 10156 * Give suggestions for the properly spelled word. | |
| 638 | 10157 * In Visual mode use the highlighted word as the bad word. |
| 485 | 10158 * When "count" is non-zero use that suggestion. |
| 323 | 10159 */ |
| 10160 void | |
| 485 | 10161 spell_suggest(count) |
| 10162 int count; | |
| 323 | 10163 { |
| 10164 char_u *line; | |
| 10165 pos_T prev_cursor = curwin->w_cursor; | |
| 10166 char_u wcopy[MAXWLEN + 2]; | |
| 10167 char_u *p; | |
| 10168 int i; | |
| 10169 int c; | |
| 10170 suginfo_T sug; | |
| 10171 suggest_T *stp; | |
| 372 | 10172 int mouse_used; |
| 388 | 10173 int need_cap; |
| 483 | 10174 int limit; |
| 485 | 10175 int selected = count; |
| 638 | 10176 int badlen = 0; |
| 1938 | 10177 int msg_scroll_save = msg_scroll; |
| 638 | 10178 |
| 10179 if (no_spell_checking(curwin)) | |
| 10180 return; | |
| 10181 | |
| 10182 #ifdef FEAT_VISUAL | |
| 10183 if (VIsual_active) | |
| 10184 { | |
| 10185 /* Use the Visually selected text as the bad word. But reject | |
| 10186 * a multi-line selection. */ | |
| 10187 if (curwin->w_cursor.lnum != VIsual.lnum) | |
| 10188 { | |
| 10189 vim_beep(); | |
| 10190 return; | |
| 10191 } | |
| 10192 badlen = (int)curwin->w_cursor.col - (int)VIsual.col; | |
| 10193 if (badlen < 0) | |
| 10194 badlen = -badlen; | |
| 10195 else | |
| 10196 curwin->w_cursor.col = VIsual.col; | |
| 10197 ++badlen; | |
| 10198 end_visual_mode(); | |
| 10199 } | |
| 10200 else | |
| 10201 #endif | |
| 10202 /* Find the start of the badly spelled word. */ | |
| 10203 if (spell_move_to(curwin, FORWARD, TRUE, TRUE, NULL) == 0 | |
| 346 | 10204 || curwin->w_cursor.col > prev_cursor.col) |
| 10205 { | |
| 10206 /* No bad word or it starts after the cursor: use the word under the | |
| 10207 * cursor. */ | |
| 10208 curwin->w_cursor = prev_cursor; | |
| 10209 line = ml_get_curline(); | |
| 10210 p = line + curwin->w_cursor.col; | |
| 10211 /* Backup to before start of word. */ | |
| 390 | 10212 while (p > line && spell_iswordp_nmw(p)) |
| 346 | 10213 mb_ptr_back(line, p); |
| 10214 /* Forward to start of word. */ | |
| 390 | 10215 while (*p != NUL && !spell_iswordp_nmw(p)) |
| 346 | 10216 mb_ptr_adv(p); |
| 10217 | |
| 390 | 10218 if (!spell_iswordp_nmw(p)) /* No word found. */ |
| 346 | 10219 { |
| 10220 beep_flush(); | |
| 10221 return; | |
| 10222 } | |
| 835 | 10223 curwin->w_cursor.col = (colnr_T)(p - line); |
| 323 | 10224 } |
| 10225 | |
| 344 | 10226 /* Get the word and its length. */ |
| 10227 | |
| 388 | 10228 /* Figure out if the word should be capitalised. */ |
| 475 | 10229 need_cap = check_need_cap(curwin->w_cursor.lnum, curwin->w_cursor.col); |
| 10230 | |
|
2046
d433f43849ba
updated for version 7.2.332
Bram Moolenaar <bram@zimbu.org>
parents:
1938
diff
changeset
|
10231 /* Make a copy of current line since autocommands may free the line. */ |
|
d433f43849ba
updated for version 7.2.332
Bram Moolenaar <bram@zimbu.org>
parents:
1938
diff
changeset
|
10232 line = vim_strsave(ml_get_curline()); |
|
d433f43849ba
updated for version 7.2.332
Bram Moolenaar <bram@zimbu.org>
parents:
1938
diff
changeset
|
10233 if (line == NULL) |
|
d433f43849ba
updated for version 7.2.332
Bram Moolenaar <bram@zimbu.org>
parents:
1938
diff
changeset
|
10234 goto skip; |
| 388 | 10235 |
| 483 | 10236 /* Get the list of suggestions. Limit to 'lines' - 2 or the number in |
| 10237 * 'spellsuggest', whatever is smaller. */ | |
| 10238 if (sps_limit > (int)Rows - 2) | |
| 10239 limit = (int)Rows - 2; | |
| 10240 else | |
| 10241 limit = sps_limit; | |
| 638 | 10242 spell_find_suggest(line + curwin->w_cursor.col, badlen, &sug, limit, |
| 625 | 10243 TRUE, need_cap, TRUE); |
| 323 | 10244 |
| 10245 if (sug.su_ga.ga_len == 0) | |
| 10246 MSG(_("Sorry, no suggestions")); | |
| 485 | 10247 else if (count > 0) |
| 10248 { | |
| 10249 if (count > sug.su_ga.ga_len) | |
| 10250 smsg((char_u *)_("Sorry, only %ld suggestions"), | |
| 10251 (long)sug.su_ga.ga_len); | |
| 10252 } | |
| 323 | 10253 else |
| 10254 { | |
| 372 | 10255 vim_free(repl_from); |
| 10256 repl_from = NULL; | |
| 10257 vim_free(repl_to); | |
| 10258 repl_to = NULL; | |
| 10259 | |
| 474 | 10260 #ifdef FEAT_RIGHTLEFT |
| 10261 /* When 'rightleft' is set the list is drawn right-left. */ | |
| 10262 cmdmsg_rl = curwin->w_p_rl; | |
| 10263 if (cmdmsg_rl) | |
| 10264 msg_col = Columns - 1; | |
| 10265 #endif | |
| 10266 | |
| 323 | 10267 /* List the suggestions. */ |
| 10268 msg_start(); | |
| 914 | 10269 msg_row = Rows - 1; /* for when 'cmdheight' > 1 */ |
| 372 | 10270 lines_left = Rows; /* avoid more prompt */ |
| 323 | 10271 vim_snprintf((char *)IObuff, IOSIZE, _("Change \"%.*s\" to:"), |
| 10272 sug.su_badlen, sug.su_badptr); | |
| 474 | 10273 #ifdef FEAT_RIGHTLEFT |
| 10274 if (cmdmsg_rl && STRNCMP(IObuff, "Change", 6) == 0) | |
| 10275 { | |
| 10276 /* And now the rabbit from the high hat: Avoid showing the | |
| 10277 * untranslated message rightleft. */ | |
| 10278 vim_snprintf((char *)IObuff, IOSIZE, ":ot \"%.*s\" egnahC", | |
| 10279 sug.su_badlen, sug.su_badptr); | |
| 10280 } | |
| 10281 #endif | |
| 323 | 10282 msg_puts(IObuff); |
| 10283 msg_clr_eos(); | |
| 10284 msg_putchar('\n'); | |
| 346 | 10285 |
| 323 | 10286 msg_scroll = TRUE; |
| 10287 for (i = 0; i < sug.su_ga.ga_len; ++i) | |
| 10288 { | |
| 344 | 10289 stp = &SUG(sug.su_ga, i); |
| 323 | 10290 |
| 10291 /* The suggested word may replace only part of the bad word, add | |
| 10292 * the not replaced part. */ | |
| 2768 | 10293 vim_strncpy(wcopy, stp->st_word, MAXWLEN); |
| 323 | 10294 if (sug.su_badlen > stp->st_orglen) |
| 625 | 10295 vim_strncpy(wcopy + stp->st_wordlen, |
| 323 | 10296 sug.su_badptr + stp->st_orglen, |
| 10297 sug.su_badlen - stp->st_orglen); | |
| 474 | 10298 vim_snprintf((char *)IObuff, IOSIZE, "%2d", i + 1); |
| 10299 #ifdef FEAT_RIGHTLEFT | |
| 10300 if (cmdmsg_rl) | |
| 10301 rl_mirror(IObuff); | |
| 10302 #endif | |
| 10303 msg_puts(IObuff); | |
| 10304 | |
| 10305 vim_snprintf((char *)IObuff, IOSIZE, " \"%s\"", wcopy); | |
| 346 | 10306 msg_puts(IObuff); |
| 10307 | |
| 10308 /* The word may replace more than "su_badlen". */ | |
| 10309 if (sug.su_badlen < stp->st_orglen) | |
| 10310 { | |
| 10311 vim_snprintf((char *)IObuff, IOSIZE, _(" < \"%.*s\""), | |
| 10312 stp->st_orglen, sug.su_badptr); | |
| 10313 msg_puts(IObuff); | |
| 10314 } | |
| 10315 | |
| 324 | 10316 if (p_verbose > 0) |
| 344 | 10317 { |
| 346 | 10318 /* Add the score. */ |
| 351 | 10319 if (sps_flags & (SPS_DOUBLE | SPS_BEST)) |
| 474 | 10320 vim_snprintf((char *)IObuff, IOSIZE, " (%s%d - %d)", |
| 344 | 10321 stp->st_salscore ? "s " : "", |
| 10322 stp->st_score, stp->st_altscore); | |
| 10323 else | |
| 474 | 10324 vim_snprintf((char *)IObuff, IOSIZE, " (%d)", |
| 346 | 10325 stp->st_score); |
| 474 | 10326 #ifdef FEAT_RIGHTLEFT |
| 10327 if (cmdmsg_rl) | |
| 10328 /* Mirror the numbers, but keep the leading space. */ | |
| 10329 rl_mirror(IObuff + 1); | |
| 10330 #endif | |
| 346 | 10331 msg_advance(30); |
| 10332 msg_puts(IObuff); | |
| 344 | 10333 } |
| 323 | 10334 msg_putchar('\n'); |
| 10335 } | |
| 10336 | |
| 474 | 10337 #ifdef FEAT_RIGHTLEFT |
| 10338 cmdmsg_rl = FALSE; | |
| 10339 msg_col = 0; | |
| 10340 #endif | |
| 323 | 10341 /* Ask for choice. */ |
| 485 | 10342 selected = prompt_for_number(&mouse_used); |
| 372 | 10343 if (mouse_used) |
| 485 | 10344 selected -= lines_left; |
| 1938 | 10345 lines_left = Rows; /* avoid more prompt */ |
| 10346 /* don't delay for 'smd' in normal_cmd() */ | |
| 10347 msg_scroll = msg_scroll_save; | |
| 485 | 10348 } |
| 10349 | |
| 10350 if (selected > 0 && selected <= sug.su_ga.ga_len && u_save_cursor() == OK) | |
| 10351 { | |
| 10352 /* Save the from and to text for :spellrepall. */ | |
| 10353 stp = &SUG(sug.su_ga, selected - 1); | |
| 548 | 10354 if (sug.su_badlen > stp->st_orglen) |
| 10355 { | |
| 10356 /* Replacing less than "su_badlen", append the remainder to | |
| 10357 * repl_to. */ | |
| 10358 repl_from = vim_strnsave(sug.su_badptr, sug.su_badlen); | |
| 10359 vim_snprintf((char *)IObuff, IOSIZE, "%s%.*s", stp->st_word, | |
| 10360 sug.su_badlen - stp->st_orglen, | |
| 10361 sug.su_badptr + stp->st_orglen); | |
| 10362 repl_to = vim_strsave(IObuff); | |
| 10363 } | |
| 10364 else | |
| 10365 { | |
| 10366 /* Replacing su_badlen or more, use the whole word. */ | |
| 10367 repl_from = vim_strnsave(sug.su_badptr, stp->st_orglen); | |
| 10368 repl_to = vim_strsave(stp->st_word); | |
| 10369 } | |
| 485 | 10370 |
| 10371 /* Replace the word. */ | |
| 1938 | 10372 p = alloc((unsigned)STRLEN(line) - stp->st_orglen |
| 10373 + stp->st_wordlen + 1); | |
| 485 | 10374 if (p != NULL) |
| 10375 { | |
| 835 | 10376 c = (int)(sug.su_badptr - line); |
| 485 | 10377 mch_memmove(p, line, c); |
| 10378 STRCPY(p + c, stp->st_word); | |
| 10379 STRCAT(p, sug.su_badptr + stp->st_orglen); | |
| 10380 ml_replace(curwin->w_cursor.lnum, p, FALSE); | |
| 10381 curwin->w_cursor.col = c; | |
| 10382 | |
| 10383 /* For redo we use a change-word command. */ | |
| 10384 ResetRedobuff(); | |
| 10385 AppendToRedobuff((char_u *)"ciw"); | |
| 620 | 10386 AppendToRedobuffLit(p + c, |
| 625 | 10387 stp->st_wordlen + sug.su_badlen - stp->st_orglen); |
| 485 | 10388 AppendCharToRedobuff(ESC); |
| 819 | 10389 |
| 10390 /* After this "p" may be invalid. */ | |
| 10391 changed_bytes(curwin->w_cursor.lnum, c); | |
| 485 | 10392 } |
| 10393 } | |
| 10394 else | |
| 10395 curwin->w_cursor = prev_cursor; | |
| 323 | 10396 |
| 344 | 10397 spell_find_cleanup(&sug); |
|
2046
d433f43849ba
updated for version 7.2.332
Bram Moolenaar <bram@zimbu.org>
parents:
1938
diff
changeset
|
10398 skip: |
|
d433f43849ba
updated for version 7.2.332
Bram Moolenaar <bram@zimbu.org>
parents:
1938
diff
changeset
|
10399 vim_free(line); |
| 344 | 10400 } |
| 10401 | |
| 10402 /* | |
| 475 | 10403 * Check if the word at line "lnum" column "col" is required to start with a |
| 10404 * capital. This uses 'spellcapcheck' of the current buffer. | |
| 10405 */ | |
| 10406 static int | |
| 10407 check_need_cap(lnum, col) | |
| 10408 linenr_T lnum; | |
| 10409 colnr_T col; | |
| 10410 { | |
| 10411 int need_cap = FALSE; | |
| 10412 char_u *line; | |
| 10413 char_u *line_copy = NULL; | |
| 10414 char_u *p; | |
| 10415 colnr_T endcol; | |
| 10416 regmatch_T regmatch; | |
| 10417 | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
10418 if (curwin->w_s->b_cap_prog == NULL) |
| 475 | 10419 return FALSE; |
| 10420 | |
| 10421 line = ml_get_curline(); | |
| 10422 endcol = 0; | |
| 10423 if ((int)(skipwhite(line) - line) >= (int)col) | |
| 10424 { | |
| 10425 /* At start of line, check if previous line is empty or sentence | |
| 10426 * ends there. */ | |
| 10427 if (lnum == 1) | |
| 10428 need_cap = TRUE; | |
| 10429 else | |
| 10430 { | |
| 10431 line = ml_get(lnum - 1); | |
| 10432 if (*skipwhite(line) == NUL) | |
| 10433 need_cap = TRUE; | |
| 10434 else | |
| 10435 { | |
| 10436 /* Append a space in place of the line break. */ | |
| 10437 line_copy = concat_str(line, (char_u *)" "); | |
| 10438 line = line_copy; | |
| 835 | 10439 endcol = (colnr_T)STRLEN(line); |
| 475 | 10440 } |
| 10441 } | |
| 10442 } | |
| 10443 else | |
| 10444 endcol = col; | |
| 10445 | |
| 10446 if (endcol > 0) | |
| 10447 { | |
| 10448 /* Check if sentence ends before the bad word. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
10449 regmatch.regprog = curwin->w_s->b_cap_prog; |
| 475 | 10450 regmatch.rm_ic = FALSE; |
| 10451 p = line + endcol; | |
| 10452 for (;;) | |
| 10453 { | |
| 10454 mb_ptr_back(line, p); | |
| 10455 if (p == line || spell_iswordp_nmw(p)) | |
| 10456 break; | |
| 10457 if (vim_regexec(®match, p, 0) | |
| 10458 && regmatch.endp[0] == line + endcol) | |
| 10459 { | |
| 10460 need_cap = TRUE; | |
| 10461 break; | |
| 10462 } | |
| 10463 } | |
| 10464 } | |
| 10465 | |
| 10466 vim_free(line_copy); | |
| 10467 | |
| 10468 return need_cap; | |
| 10469 } | |
| 10470 | |
| 10471 | |
| 10472 /* | |
| 372 | 10473 * ":spellrepall" |
| 10474 */ | |
| 10475 void | |
| 10476 ex_spellrepall(eap) | |
| 1883 | 10477 exarg_T *eap UNUSED; |
| 372 | 10478 { |
| 10479 pos_T pos = curwin->w_cursor; | |
| 10480 char_u *frompat; | |
| 10481 int addlen; | |
| 10482 char_u *line; | |
| 10483 char_u *p; | |
| 10484 int save_ws = p_ws; | |
| 483 | 10485 linenr_T prev_lnum = 0; |
| 372 | 10486 |
| 10487 if (repl_from == NULL || repl_to == NULL) | |
| 10488 { | |
| 10489 EMSG(_("E752: No previous spell replacement")); | |
| 10490 return; | |
| 10491 } | |
| 835 | 10492 addlen = (int)(STRLEN(repl_to) - STRLEN(repl_from)); |
| 10493 | |
| 10494 frompat = alloc((unsigned)STRLEN(repl_from) + 7); | |
| 372 | 10495 if (frompat == NULL) |
| 10496 return; | |
| 10497 sprintf((char *)frompat, "\\V\\<%s\\>", repl_from); | |
| 10498 p_ws = FALSE; | |
| 10499 | |
| 483 | 10500 sub_nsubs = 0; |
| 10501 sub_nlines = 0; | |
| 372 | 10502 curwin->w_cursor.lnum = 0; |
| 10503 while (!got_int) | |
| 10504 { | |
| 1521 | 10505 if (do_search(NULL, '/', frompat, 1L, SEARCH_KEEP, NULL) == 0 |
| 372 | 10506 || u_save_cursor() == FAIL) |
| 10507 break; | |
| 10508 | |
| 10509 /* Only replace when the right word isn't there yet. This happens | |
| 10510 * when changing "etc" to "etc.". */ | |
| 10511 line = ml_get_curline(); | |
| 10512 if (addlen <= 0 || STRNCMP(line + curwin->w_cursor.col, | |
| 10513 repl_to, STRLEN(repl_to)) != 0) | |
| 10514 { | |
| 835 | 10515 p = alloc((unsigned)STRLEN(line) + addlen + 1); |
| 372 | 10516 if (p == NULL) |
| 10517 break; | |
| 10518 mch_memmove(p, line, curwin->w_cursor.col); | |
| 10519 STRCPY(p + curwin->w_cursor.col, repl_to); | |
| 10520 STRCAT(p, line + curwin->w_cursor.col + STRLEN(repl_from)); | |
| 10521 ml_replace(curwin->w_cursor.lnum, p, FALSE); | |
| 10522 changed_bytes(curwin->w_cursor.lnum, curwin->w_cursor.col); | |
| 483 | 10523 |
| 10524 if (curwin->w_cursor.lnum != prev_lnum) | |
| 10525 { | |
| 10526 ++sub_nlines; | |
| 10527 prev_lnum = curwin->w_cursor.lnum; | |
| 10528 } | |
| 10529 ++sub_nsubs; | |
| 372 | 10530 } |
| 835 | 10531 curwin->w_cursor.col += (colnr_T)STRLEN(repl_to); |
| 372 | 10532 } |
| 10533 | |
| 10534 p_ws = save_ws; | |
| 10535 curwin->w_cursor = pos; | |
| 10536 vim_free(frompat); | |
| 10537 | |
| 483 | 10538 if (sub_nsubs == 0) |
| 372 | 10539 EMSG2(_("E753: Not found: %s"), repl_from); |
| 483 | 10540 else |
| 10541 do_sub_msg(FALSE); | |
| 372 | 10542 } |
| 10543 | |
| 10544 /* | |
| 344 | 10545 * Find spell suggestions for "word". Return them in the growarray "*gap" as |
| 10546 * a list of allocated strings. | |
| 10547 */ | |
| 10548 void | |
| 625 | 10549 spell_suggest_list(gap, word, maxcount, need_cap, interactive) |
| 344 | 10550 garray_T *gap; |
| 10551 char_u *word; | |
| 10552 int maxcount; /* maximum nr of suggestions */ | |
| 475 | 10553 int need_cap; /* 'spellcapcheck' matched */ |
| 625 | 10554 int interactive; |
| 344 | 10555 { |
| 10556 suginfo_T sug; | |
| 10557 int i; | |
| 10558 suggest_T *stp; | |
| 10559 char_u *wcopy; | |
| 10560 | |
| 638 | 10561 spell_find_suggest(word, 0, &sug, maxcount, FALSE, need_cap, interactive); |
| 344 | 10562 |
| 10563 /* Make room in "gap". */ | |
| 10564 ga_init2(gap, sizeof(char_u *), sug.su_ga.ga_len + 1); | |
| 626 | 10565 if (ga_grow(gap, sug.su_ga.ga_len) == OK) |
| 10566 { | |
| 10567 for (i = 0; i < sug.su_ga.ga_len; ++i) | |
| 10568 { | |
| 10569 stp = &SUG(sug.su_ga, i); | |
| 10570 | |
| 10571 /* The suggested word may replace only part of "word", add the not | |
| 10572 * replaced part. */ | |
| 10573 wcopy = alloc(stp->st_wordlen | |
| 835 | 10574 + (unsigned)STRLEN(sug.su_badptr + stp->st_orglen) + 1); |
| 626 | 10575 if (wcopy == NULL) |
| 10576 break; | |
| 10577 STRCPY(wcopy, stp->st_word); | |
| 10578 STRCPY(wcopy + stp->st_wordlen, sug.su_badptr + stp->st_orglen); | |
| 10579 ((char_u **)gap->ga_data)[gap->ga_len++] = wcopy; | |
| 10580 } | |
| 344 | 10581 } |
| 10582 | |
| 10583 spell_find_cleanup(&sug); | |
| 10584 } | |
| 10585 | |
| 10586 /* | |
| 10587 * Find spell suggestions for the word at the start of "badptr". | |
| 10588 * Return the suggestions in "su->su_ga". | |
| 10589 * The maximum number of suggestions is "maxcount". | |
| 10590 * Note: does use info for the current window. | |
| 10591 * This is based on the mechanisms of Aspell, but completely reimplemented. | |
| 10592 */ | |
| 10593 static void | |
| 638 | 10594 spell_find_suggest(badptr, badlen, su, maxcount, banbadword, need_cap, interactive) |
| 344 | 10595 char_u *badptr; |
| 638 | 10596 int badlen; /* length of bad word or 0 if unknown */ |
| 344 | 10597 suginfo_T *su; |
| 10598 int maxcount; | |
| 358 | 10599 int banbadword; /* don't include badword in suggestions */ |
| 388 | 10600 int need_cap; /* word should start with capital */ |
| 625 | 10601 int interactive; |
| 344 | 10602 { |
| 534 | 10603 hlf_T attr = HLF_COUNT; |
| 372 | 10604 char_u buf[MAXPATHL]; |
| 10605 char_u *p; | |
| 10606 int do_combine = FALSE; | |
| 10607 char_u *sps_copy; | |
| 10608 #ifdef FEAT_EVAL | |
| 10609 static int expr_busy = FALSE; | |
| 10610 #endif | |
| 385 | 10611 int c; |
| 503 | 10612 int i; |
| 10613 langp_T *lp; | |
| 344 | 10614 |
| 10615 /* | |
| 10616 * Set the info in "*su". | |
| 10617 */ | |
| 10618 vim_memset(su, 0, sizeof(suginfo_T)); | |
| 10619 ga_init2(&su->su_ga, (int)sizeof(suggest_T), 10); | |
| 10620 ga_init2(&su->su_sga, (int)sizeof(suggest_T), 10); | |
| 355 | 10621 if (*badptr == NUL) |
| 10622 return; | |
| 344 | 10623 hash_init(&su->su_banned); |
| 10624 | |
| 10625 su->su_badptr = badptr; | |
| 638 | 10626 if (badlen != 0) |
| 10627 su->su_badlen = badlen; | |
| 10628 else | |
| 10629 su->su_badlen = spell_check(curwin, su->su_badptr, &attr, NULL, FALSE); | |
| 344 | 10630 su->su_maxcount = maxcount; |
| 372 | 10631 su->su_maxscore = SCORE_MAXINIT; |
| 344 | 10632 |
| 10633 if (su->su_badlen >= MAXWLEN) | |
| 10634 su->su_badlen = MAXWLEN - 1; /* just in case */ | |
| 10635 vim_strncpy(su->su_badword, su->su_badptr, su->su_badlen); | |
| 10636 (void)spell_casefold(su->su_badptr, su->su_badlen, | |
| 10637 su->su_fbadword, MAXWLEN); | |
| 346 | 10638 /* get caps flags for bad word */ |
| 474 | 10639 su->su_badflags = badword_captype(su->su_badptr, |
| 10640 su->su_badptr + su->su_badlen); | |
| 388 | 10641 if (need_cap) |
| 10642 su->su_badflags |= WF_ONECAP; | |
| 344 | 10643 |
| 503 | 10644 /* Find the default language for sound folding. We simply use the first |
| 10645 * one in 'spelllang' that supports sound folding. That's good for when | |
| 10646 * using multiple files for one language, it's not that bad when mixing | |
| 10647 * languages (e.g., "pl,en"). */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
10648 for (i = 0; i < curbuf->b_s.b_langp.ga_len; ++i) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
10649 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
10650 lp = LANGP_ENTRY(curbuf->b_s.b_langp, i); |
| 503 | 10651 if (lp->lp_sallang != NULL) |
| 10652 { | |
| 10653 su->su_sallang = lp->lp_sallang; | |
| 10654 break; | |
| 10655 } | |
| 10656 } | |
| 10657 | |
| 534 | 10658 /* Soundfold the bad word with the default sound folding, so that we don't |
| 10659 * have to do this many times. */ | |
| 10660 if (su->su_sallang != NULL) | |
| 10661 spell_soundfold(su->su_sallang, su->su_fbadword, TRUE, | |
| 10662 su->su_sal_badword); | |
| 10663 | |
| 385 | 10664 /* If the word is not capitalised and spell_check() doesn't consider the |
| 10665 * word to be bad then it might need to be capitalised. Add a suggestion | |
| 10666 * for that. */ | |
| 455 | 10667 c = PTR2CHAR(su->su_badptr); |
| 534 | 10668 if (!SPELL_ISUPPER(c) && attr == HLF_COUNT) |
| 385 | 10669 { |
| 10670 make_case_word(su->su_badword, buf, WF_ONECAP); | |
| 10671 add_suggestion(su, &su->su_ga, buf, su->su_badlen, SCORE_ICASE, | |
| 625 | 10672 0, TRUE, su->su_sallang, FALSE); |
| 385 | 10673 } |
| 10674 | |
| 344 | 10675 /* Ban the bad word itself. It may appear in another region. */ |
| 358 | 10676 if (banbadword) |
| 10677 add_banned(su, su->su_badword); | |
| 344 | 10678 |
| 372 | 10679 /* Make a copy of 'spellsuggest', because the expression may change it. */ |
| 10680 sps_copy = vim_strsave(p_sps); | |
| 10681 if (sps_copy == NULL) | |
| 10682 return; | |
| 10683 | |
| 10684 /* Loop over the items in 'spellsuggest'. */ | |
| 10685 for (p = sps_copy; *p != NUL; ) | |
| 10686 { | |
| 10687 copy_option_part(&p, buf, MAXPATHL, ","); | |
| 10688 | |
| 10689 if (STRNCMP(buf, "expr:", 5) == 0) | |
| 10690 { | |
| 10691 #ifdef FEAT_EVAL | |
| 375 | 10692 /* Evaluate an expression. Skip this when called recursively, |
| 10693 * when using spellsuggest() in the expression. */ | |
| 372 | 10694 if (!expr_busy) |
| 10695 { | |
| 10696 expr_busy = TRUE; | |
| 10697 spell_suggest_expr(su, buf + 5); | |
| 10698 expr_busy = FALSE; | |
| 10699 } | |
| 10700 #endif | |
| 10701 } | |
| 10702 else if (STRNCMP(buf, "file:", 5) == 0) | |
| 10703 /* Use list of suggestions in a file. */ | |
| 10704 spell_suggest_file(su, buf + 5); | |
| 10705 else | |
| 10706 { | |
| 10707 /* Use internal method. */ | |
| 625 | 10708 spell_suggest_intern(su, interactive); |
| 372 | 10709 if (sps_flags & SPS_DOUBLE) |
| 10710 do_combine = TRUE; | |
| 10711 } | |
| 10712 } | |
| 10713 | |
| 10714 vim_free(sps_copy); | |
| 10715 | |
| 10716 if (do_combine) | |
| 10717 /* Combine the two list of suggestions. This must be done last, | |
| 10718 * because sorting changes the order again. */ | |
| 10719 score_combine(su); | |
| 10720 } | |
| 10721 | |
| 10722 #ifdef FEAT_EVAL | |
| 10723 /* | |
| 10724 * Find suggestions by evaluating expression "expr". | |
| 10725 */ | |
| 10726 static void | |
| 10727 spell_suggest_expr(su, expr) | |
| 10728 suginfo_T *su; | |
| 10729 char_u *expr; | |
| 10730 { | |
| 10731 list_T *list; | |
| 10732 listitem_T *li; | |
| 10733 int score; | |
| 10734 char_u *p; | |
| 10735 | |
| 10736 /* The work is split up in a few parts to avoid having to export | |
| 10737 * suginfo_T. | |
| 10738 * First evaluate the expression and get the resulting list. */ | |
| 10739 list = eval_spell_expr(su->su_badword, expr); | |
| 10740 if (list != NULL) | |
| 10741 { | |
| 10742 /* Loop over the items in the list. */ | |
| 10743 for (li = list->lv_first; li != NULL; li = li->li_next) | |
| 10744 if (li->li_tv.v_type == VAR_LIST) | |
| 10745 { | |
| 10746 /* Get the word and the score from the items. */ | |
| 10747 score = get_spellword(li->li_tv.vval.v_list, &p); | |
| 625 | 10748 if (score >= 0 && score <= su->su_maxscore) |
| 10749 add_suggestion(su, &su->su_ga, p, su->su_badlen, | |
| 10750 score, 0, TRUE, su->su_sallang, FALSE); | |
| 372 | 10751 } |
| 10752 list_unref(list); | |
| 10753 } | |
| 10754 | |
| 625 | 10755 /* Remove bogus suggestions, sort and truncate at "maxcount". */ |
| 10756 check_suggestions(su, &su->su_ga); | |
| 372 | 10757 (void)cleanup_suggestions(&su->su_ga, su->su_maxscore, su->su_maxcount); |
| 10758 } | |
| 10759 #endif | |
| 10760 | |
| 10761 /* | |
| 474 | 10762 * Find suggestions in file "fname". Used for "file:" in 'spellsuggest'. |
| 372 | 10763 */ |
| 10764 static void | |
| 10765 spell_suggest_file(su, fname) | |
| 10766 suginfo_T *su; | |
| 10767 char_u *fname; | |
| 10768 { | |
| 10769 FILE *fd; | |
| 10770 char_u line[MAXWLEN * 2]; | |
| 10771 char_u *p; | |
| 10772 int len; | |
| 10773 char_u cword[MAXWLEN]; | |
| 10774 | |
| 10775 /* Open the file. */ | |
| 10776 fd = mch_fopen((char *)fname, "r"); | |
| 10777 if (fd == NULL) | |
| 10778 { | |
| 10779 EMSG2(_(e_notopen), fname); | |
| 10780 return; | |
| 10781 } | |
| 10782 | |
| 10783 /* Read it line by line. */ | |
| 10784 while (!vim_fgets(line, MAXWLEN * 2, fd) && !got_int) | |
| 10785 { | |
| 10786 line_breakcheck(); | |
| 10787 | |
| 10788 p = vim_strchr(line, '/'); | |
| 10789 if (p == NULL) | |
| 10790 continue; /* No Tab found, just skip the line. */ | |
| 10791 *p++ = NUL; | |
| 10792 if (STRICMP(su->su_badword, line) == 0) | |
| 10793 { | |
| 10794 /* Match! Isolate the good word, until CR or NL. */ | |
| 10795 for (len = 0; p[len] >= ' '; ++len) | |
| 10796 ; | |
| 10797 p[len] = NUL; | |
| 10798 | |
| 10799 /* If the suggestion doesn't have specific case duplicate the case | |
| 10800 * of the bad word. */ | |
| 10801 if (captype(p, NULL) == 0) | |
| 10802 { | |
| 10803 make_case_word(p, cword, su->su_badflags); | |
| 10804 p = cword; | |
| 10805 } | |
| 10806 | |
| 10807 add_suggestion(su, &su->su_ga, p, su->su_badlen, | |
| 625 | 10808 SCORE_FILE, 0, TRUE, su->su_sallang, FALSE); |
| 372 | 10809 } |
| 10810 } | |
| 10811 | |
| 10812 fclose(fd); | |
| 10813 | |
| 625 | 10814 /* Remove bogus suggestions, sort and truncate at "maxcount". */ |
| 10815 check_suggestions(su, &su->su_ga); | |
| 372 | 10816 (void)cleanup_suggestions(&su->su_ga, su->su_maxscore, su->su_maxcount); |
| 10817 } | |
| 10818 | |
| 10819 /* | |
| 10820 * Find suggestions for the internal method indicated by "sps_flags". | |
| 10821 */ | |
| 10822 static void | |
| 625 | 10823 spell_suggest_intern(su, interactive) |
| 372 | 10824 suginfo_T *su; |
| 625 | 10825 int interactive; |
| 10826 { | |
| 10827 /* | |
| 10828 * Load the .sug file(s) that are available and not done yet. | |
| 10829 */ | |
| 10830 suggest_load_files(); | |
| 10831 | |
| 344 | 10832 /* |
| 346 | 10833 * 1. Try special cases, such as repeating a word: "the the" -> "the". |
| 344 | 10834 * |
| 10835 * Set a maximum score to limit the combination of operations that is | |
| 10836 * tried. | |
| 10837 */ | |
| 346 | 10838 suggest_try_special(su); |
| 10839 | |
| 10840 /* | |
| 10841 * 2. Try inserting/deleting/swapping/changing a letter, use REP entries | |
| 10842 * from the .aff file and inserting a space (split the word). | |
| 10843 */ | |
| 10844 suggest_try_change(su); | |
| 344 | 10845 |
| 10846 /* For the resulting top-scorers compute the sound-a-like score. */ | |
| 10847 if (sps_flags & SPS_DOUBLE) | |
| 10848 score_comp_sal(su); | |
| 10849 | |
| 10850 /* | |
| 346 | 10851 * 3. Try finding sound-a-like words. |
| 344 | 10852 */ |
| 625 | 10853 if ((sps_flags & SPS_FAST) == 0) |
| 10854 { | |
| 10855 if (sps_flags & SPS_BEST) | |
| 10856 /* Adjust the word score for the suggestions found so far for how | |
| 10857 * they sounds like. */ | |
| 10858 rescore_suggestions(su); | |
| 10859 | |
| 10860 /* | |
|
2046
d433f43849ba
updated for version 7.2.332
Bram Moolenaar <bram@zimbu.org>
parents:
1938
diff
changeset
|
10861 * While going through the soundfold tree "su_maxscore" is the score |
| 625 | 10862 * for the soundfold word, limits the changes that are being tried, |
| 10863 * and "su_sfmaxscore" the rescored score, which is set by | |
| 10864 * cleanup_suggestions(). | |
| 10865 * First find words with a small edit distance, because this is much | |
| 10866 * faster and often already finds the top-N suggestions. If we didn't | |
| 10867 * find many suggestions try again with a higher edit distance. | |
| 10868 * "sl_sounddone" is used to avoid doing the same word twice. | |
| 10869 */ | |
| 10870 suggest_try_soundalike_prep(); | |
| 10871 su->su_maxscore = SCORE_SFMAX1; | |
| 10872 su->su_sfmaxscore = SCORE_MAXINIT * 3; | |
| 346 | 10873 suggest_try_soundalike(su); |
| 625 | 10874 if (su->su_ga.ga_len < SUG_CLEAN_COUNT(su)) |
| 10875 { | |
| 10876 /* We didn't find enough matches, try again, allowing more | |
| 10877 * changes to the soundfold word. */ | |
| 10878 su->su_maxscore = SCORE_SFMAX2; | |
| 10879 suggest_try_soundalike(su); | |
| 10880 if (su->su_ga.ga_len < SUG_CLEAN_COUNT(su)) | |
| 10881 { | |
| 10882 /* Still didn't find enough matches, try again, allowing even | |
| 10883 * more changes to the soundfold word. */ | |
| 10884 su->su_maxscore = SCORE_SFMAX3; | |
| 10885 suggest_try_soundalike(su); | |
| 10886 } | |
| 10887 } | |
| 10888 su->su_maxscore = su->su_sfmaxscore; | |
| 10889 suggest_try_soundalike_finish(); | |
| 10890 } | |
| 10891 | |
| 10892 /* When CTRL-C was hit while searching do show the results. Only clear | |
| 10893 * got_int when using a command, not for spellsuggest(). */ | |
| 344 | 10894 ui_breakcheck(); |
| 625 | 10895 if (interactive && got_int) |
| 344 | 10896 { |
| 10897 (void)vgetc(); | |
| 10898 got_int = FALSE; | |
| 10899 } | |
| 10900 | |
| 372 | 10901 if ((sps_flags & SPS_DOUBLE) == 0 && su->su_ga.ga_len != 0) |
| 344 | 10902 { |
| 10903 if (sps_flags & SPS_BEST) | |
| 10904 /* Adjust the word score for how it sounds like. */ | |
| 10905 rescore_suggestions(su); | |
| 10906 | |
| 625 | 10907 /* Remove bogus suggestions, sort and truncate at "maxcount". */ |
| 10908 check_suggestions(su, &su->su_ga); | |
| 372 | 10909 (void)cleanup_suggestions(&su->su_ga, su->su_maxscore, su->su_maxcount); |
| 344 | 10910 } |
| 10911 } | |
| 10912 | |
| 10913 /* | |
| 625 | 10914 * Load the .sug files for languages that have one and weren't loaded yet. |
| 10915 */ | |
| 10916 static void | |
| 10917 suggest_load_files() | |
| 10918 { | |
| 10919 langp_T *lp; | |
| 10920 int lpi; | |
| 10921 slang_T *slang; | |
| 10922 char_u *dotp; | |
| 10923 FILE *fd; | |
| 10924 char_u buf[MAXWLEN]; | |
| 10925 int i; | |
| 10926 time_t timestamp; | |
| 10927 int wcount; | |
| 10928 int wordnr; | |
| 10929 garray_T ga; | |
| 10930 int c; | |
| 10931 | |
| 10932 /* Do this for all languages that support sound folding. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
10933 for (lpi = 0; lpi < curwin->w_s->b_langp.ga_len; ++lpi) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
10934 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
10935 lp = LANGP_ENTRY(curwin->w_s->b_langp, lpi); |
| 625 | 10936 slang = lp->lp_slang; |
| 10937 if (slang->sl_sugtime != 0 && !slang->sl_sugloaded) | |
| 10938 { | |
| 10939 /* Change ".spl" to ".sug" and open the file. When the file isn't | |
| 10940 * found silently skip it. Do set "sl_sugloaded" so that we | |
| 10941 * don't try again and again. */ | |
| 10942 slang->sl_sugloaded = TRUE; | |
| 10943 | |
| 10944 dotp = vim_strrchr(slang->sl_fname, '.'); | |
| 10945 if (dotp == NULL || fnamecmp(dotp, ".spl") != 0) | |
| 10946 continue; | |
| 10947 STRCPY(dotp, ".sug"); | |
| 822 | 10948 fd = mch_fopen((char *)slang->sl_fname, "r"); |
| 625 | 10949 if (fd == NULL) |
| 10950 goto nextone; | |
| 10951 | |
| 10952 /* | |
| 10953 * <SUGHEADER>: <fileID> <versionnr> <timestamp> | |
| 10954 */ | |
| 10955 for (i = 0; i < VIMSUGMAGICL; ++i) | |
| 10956 buf[i] = getc(fd); /* <fileID> */ | |
| 10957 if (STRNCMP(buf, VIMSUGMAGIC, VIMSUGMAGICL) != 0) | |
| 10958 { | |
| 626 | 10959 EMSG2(_("E778: This does not look like a .sug file: %s"), |
| 625 | 10960 slang->sl_fname); |
| 10961 goto nextone; | |
| 10962 } | |
| 10963 c = getc(fd); /* <versionnr> */ | |
| 10964 if (c < VIMSUGVERSION) | |
| 10965 { | |
| 626 | 10966 EMSG2(_("E779: Old .sug file, needs to be updated: %s"), |
| 625 | 10967 slang->sl_fname); |
| 10968 goto nextone; | |
| 10969 } | |
| 10970 else if (c > VIMSUGVERSION) | |
| 10971 { | |
| 626 | 10972 EMSG2(_("E780: .sug file is for newer version of Vim: %s"), |
| 625 | 10973 slang->sl_fname); |
| 10974 goto nextone; | |
| 10975 } | |
| 10976 | |
| 10977 /* Check the timestamp, it must be exactly the same as the one in | |
| 10978 * the .spl file. Otherwise the word numbers won't match. */ | |
|
2229
d45902a5c61c
Fix a few more things for persistent undo.
Bram Moolenaar <bram@vim.org>
parents:
2214
diff
changeset
|
10979 timestamp = get8ctime(fd); /* <timestamp> */ |
| 625 | 10980 if (timestamp != slang->sl_sugtime) |
| 10981 { | |
| 626 | 10982 EMSG2(_("E781: .sug file doesn't match .spl file: %s"), |
| 625 | 10983 slang->sl_fname); |
| 10984 goto nextone; | |
| 10985 } | |
| 10986 | |
| 10987 /* | |
| 10988 * <SUGWORDTREE>: <wordtree> | |
| 10989 * Read the trie with the soundfolded words. | |
| 10990 */ | |
| 10991 if (spell_read_tree(fd, &slang->sl_sbyts, &slang->sl_sidxs, | |
| 10992 FALSE, 0) != 0) | |
| 10993 { | |
| 10994 someerror: | |
| 626 | 10995 EMSG2(_("E782: error while reading .sug file: %s"), |
| 625 | 10996 slang->sl_fname); |
| 10997 slang_clear_sug(slang); | |
| 10998 goto nextone; | |
| 10999 } | |
| 11000 | |
| 11001 /* | |
| 11002 * <SUGTABLE>: <sugwcount> <sugline> ... | |
| 11003 * | |
| 11004 * Read the table with word numbers. We use a file buffer for | |
| 11005 * this, because it's so much like a file with lines. Makes it | |
| 11006 * possible to swap the info and save on memory use. | |
| 11007 */ | |
| 11008 slang->sl_sugbuf = open_spellbuf(); | |
| 11009 if (slang->sl_sugbuf == NULL) | |
| 11010 goto someerror; | |
| 11011 /* <sugwcount> */ | |
| 695 | 11012 wcount = get4c(fd); |
| 625 | 11013 if (wcount < 0) |
| 11014 goto someerror; | |
| 11015 | |
| 11016 /* Read all the wordnr lists into the buffer, one NUL terminated | |
| 11017 * list per line. */ | |
| 11018 ga_init2(&ga, 1, 100); | |
| 11019 for (wordnr = 0; wordnr < wcount; ++wordnr) | |
| 11020 { | |
| 11021 ga.ga_len = 0; | |
| 11022 for (;;) | |
| 11023 { | |
| 11024 c = getc(fd); /* <sugline> */ | |
| 11025 if (c < 0 || ga_grow(&ga, 1) == FAIL) | |
| 11026 goto someerror; | |
| 11027 ((char_u *)ga.ga_data)[ga.ga_len++] = c; | |
| 11028 if (c == NUL) | |
| 11029 break; | |
| 11030 } | |
| 11031 if (ml_append_buf(slang->sl_sugbuf, (linenr_T)wordnr, | |
| 11032 ga.ga_data, ga.ga_len, TRUE) == FAIL) | |
| 11033 goto someerror; | |
| 11034 } | |
| 11035 ga_clear(&ga); | |
| 11036 | |
| 11037 /* | |
| 11038 * Need to put word counts in the word tries, so that we can find | |
| 11039 * a word by its number. | |
| 11040 */ | |
| 11041 tree_count_words(slang->sl_fbyts, slang->sl_fidxs); | |
| 11042 tree_count_words(slang->sl_sbyts, slang->sl_sidxs); | |
| 11043 | |
| 11044 nextone: | |
| 11045 if (fd != NULL) | |
| 11046 fclose(fd); | |
| 11047 STRCPY(dotp, ".spl"); | |
| 11048 } | |
| 11049 } | |
| 11050 } | |
| 11051 | |
| 11052 | |
| 11053 /* | |
| 11054 * Fill in the wordcount fields for a trie. | |
| 11055 * Returns the total number of words. | |
| 11056 */ | |
| 11057 static void | |
| 11058 tree_count_words(byts, idxs) | |
| 11059 char_u *byts; | |
| 11060 idx_T *idxs; | |
| 11061 { | |
| 11062 int depth; | |
| 11063 idx_T arridx[MAXWLEN]; | |
| 11064 int curi[MAXWLEN]; | |
| 11065 int c; | |
| 11066 idx_T n; | |
| 11067 int wordcount[MAXWLEN]; | |
| 11068 | |
| 11069 arridx[0] = 0; | |
| 11070 curi[0] = 1; | |
| 11071 wordcount[0] = 0; | |
| 11072 depth = 0; | |
| 11073 while (depth >= 0 && !got_int) | |
| 11074 { | |
| 11075 if (curi[depth] > byts[arridx[depth]]) | |
| 11076 { | |
| 11077 /* Done all bytes at this node, go up one level. */ | |
| 11078 idxs[arridx[depth]] = wordcount[depth]; | |
| 11079 if (depth > 0) | |
| 11080 wordcount[depth - 1] += wordcount[depth]; | |
| 11081 | |
| 11082 --depth; | |
| 11083 fast_breakcheck(); | |
| 11084 } | |
| 11085 else | |
| 11086 { | |
| 11087 /* Do one more byte at this node. */ | |
| 11088 n = arridx[depth] + curi[depth]; | |
| 11089 ++curi[depth]; | |
| 11090 | |
| 11091 c = byts[n]; | |
| 11092 if (c == 0) | |
| 11093 { | |
| 11094 /* End of word, count it. */ | |
| 11095 ++wordcount[depth]; | |
| 11096 | |
| 11097 /* Skip over any other NUL bytes (same word with different | |
| 11098 * flags). */ | |
| 11099 while (byts[n + 1] == 0) | |
| 11100 { | |
| 11101 ++n; | |
| 11102 ++curi[depth]; | |
| 11103 } | |
| 11104 } | |
| 11105 else | |
| 11106 { | |
| 11107 /* Normal char, go one level deeper to count the words. */ | |
| 11108 ++depth; | |
| 11109 arridx[depth] = idxs[n]; | |
| 11110 curi[depth] = 1; | |
| 11111 wordcount[depth] = 0; | |
| 11112 } | |
| 11113 } | |
| 11114 } | |
| 11115 } | |
| 11116 | |
| 11117 /* | |
| 344 | 11118 * Free the info put in "*su" by spell_find_suggest(). |
| 11119 */ | |
| 11120 static void | |
| 11121 spell_find_cleanup(su) | |
| 11122 suginfo_T *su; | |
| 11123 { | |
| 11124 int i; | |
| 11125 | |
| 11126 /* Free the suggestions. */ | |
| 11127 for (i = 0; i < su->su_ga.ga_len; ++i) | |
| 11128 vim_free(SUG(su->su_ga, i).st_word); | |
| 11129 ga_clear(&su->su_ga); | |
| 11130 for (i = 0; i < su->su_sga.ga_len; ++i) | |
| 11131 vim_free(SUG(su->su_sga, i).st_word); | |
| 11132 ga_clear(&su->su_sga); | |
| 323 | 11133 |
| 11134 /* Free the banned words. */ | |
| 625 | 11135 hash_clear_all(&su->su_banned, 0); |
| 323 | 11136 } |
| 11137 | |
| 11138 /* | |
| 324 | 11139 * Make a copy of "word", with the first letter upper or lower cased, to |
| 11140 * "wcopy[MAXWLEN]". "word" must not be empty. | |
| 11141 * The result is NUL terminated. | |
| 323 | 11142 */ |
| 11143 static void | |
| 324 | 11144 onecap_copy(word, wcopy, upper) |
| 323 | 11145 char_u *word; |
| 11146 char_u *wcopy; | |
| 11147 int upper; /* TRUE: first letter made upper case */ | |
| 11148 { | |
| 11149 char_u *p; | |
| 11150 int c; | |
| 11151 int l; | |
| 11152 | |
| 11153 p = word; | |
| 11154 #ifdef FEAT_MBYTE | |
| 11155 if (has_mbyte) | |
| 474 | 11156 c = mb_cptr2char_adv(&p); |
| 323 | 11157 else |
| 11158 #endif | |
| 11159 c = *p++; | |
| 11160 if (upper) | |
| 324 | 11161 c = SPELL_TOUPPER(c); |
| 323 | 11162 else |
| 324 | 11163 c = SPELL_TOFOLD(c); |
| 323 | 11164 #ifdef FEAT_MBYTE |
| 11165 if (has_mbyte) | |
| 11166 l = mb_char2bytes(c, wcopy); | |
| 11167 else | |
| 11168 #endif | |
| 11169 { | |
| 11170 l = 1; | |
| 11171 wcopy[0] = c; | |
| 11172 } | |
| 376 | 11173 vim_strncpy(wcopy + l, p, MAXWLEN - l - 1); |
| 323 | 11174 } |
| 11175 | |
| 11176 /* | |
| 324 | 11177 * Make a copy of "word" with all the letters upper cased into |
| 11178 * "wcopy[MAXWLEN]". The result is NUL terminated. | |
| 323 | 11179 */ |
| 11180 static void | |
| 11181 allcap_copy(word, wcopy) | |
| 11182 char_u *word; | |
| 11183 char_u *wcopy; | |
| 11184 { | |
| 11185 char_u *s; | |
| 11186 char_u *d; | |
| 11187 int c; | |
| 11188 | |
| 11189 d = wcopy; | |
| 11190 for (s = word; *s != NUL; ) | |
| 11191 { | |
| 11192 #ifdef FEAT_MBYTE | |
| 11193 if (has_mbyte) | |
| 474 | 11194 c = mb_cptr2char_adv(&s); |
| 323 | 11195 else |
| 11196 #endif | |
| 11197 c = *s++; | |
| 492 | 11198 |
| 11199 #ifdef FEAT_MBYTE | |
| 11200 /* We only change ß to SS when we are certain latin1 is used. It | |
| 11201 * would cause weird errors in other 8-bit encodings. */ | |
| 11202 if (enc_latin1like && c == 0xdf) | |
| 11203 { | |
| 11204 c = 'S'; | |
| 11205 if (d - wcopy >= MAXWLEN - 1) | |
| 11206 break; | |
| 11207 *d++ = c; | |
| 11208 } | |
| 11209 else | |
| 11210 #endif | |
| 11211 c = SPELL_TOUPPER(c); | |
| 323 | 11212 |
| 11213 #ifdef FEAT_MBYTE | |
| 11214 if (has_mbyte) | |
| 11215 { | |
| 11216 if (d - wcopy >= MAXWLEN - MB_MAXBYTES) | |
| 11217 break; | |
| 11218 d += mb_char2bytes(c, d); | |
| 11219 } | |
| 11220 else | |
| 11221 #endif | |
| 11222 { | |
| 11223 if (d - wcopy >= MAXWLEN - 1) | |
| 11224 break; | |
| 11225 *d++ = c; | |
| 11226 } | |
| 11227 } | |
| 11228 *d = NUL; | |
| 11229 } | |
| 11230 | |
| 11231 /* | |
| 346 | 11232 * Try finding suggestions by recognizing specific situations. |
| 11233 */ | |
| 11234 static void | |
| 11235 suggest_try_special(su) | |
| 11236 suginfo_T *su; | |
| 11237 { | |
| 11238 char_u *p; | |
| 376 | 11239 size_t len; |
| 346 | 11240 int c; |
| 11241 char_u word[MAXWLEN]; | |
| 11242 | |
| 11243 /* | |
| 11244 * Recognize a word that is repeated: "the the". | |
| 11245 */ | |
| 11246 p = skiptowhite(su->su_fbadword); | |
| 11247 len = p - su->su_fbadword; | |
| 11248 p = skipwhite(p); | |
| 11249 if (STRLEN(p) == len && STRNCMP(su->su_fbadword, p, len) == 0) | |
| 11250 { | |
| 11251 /* Include badflags: if the badword is onecap or allcap | |
| 11252 * use that for the goodword too: "The the" -> "The". */ | |
| 11253 c = su->su_fbadword[len]; | |
| 11254 su->su_fbadword[len] = NUL; | |
| 11255 make_case_word(su->su_fbadword, word, su->su_badflags); | |
| 11256 su->su_fbadword[len] = c; | |
| 534 | 11257 |
| 11258 /* Give a soundalike score of 0, compute the score as if deleting one | |
| 11259 * character. */ | |
| 11260 add_suggestion(su, &su->su_ga, word, su->su_badlen, | |
| 625 | 11261 RESCORE(SCORE_REP, 0), 0, TRUE, su->su_sallang, FALSE); |
| 346 | 11262 } |
| 11263 } | |
| 11264 | |
| 11265 /* | |
| 323 | 11266 * Try finding suggestions by adding/removing/swapping letters. |
| 11267 */ | |
| 11268 static void | |
| 346 | 11269 suggest_try_change(su) |
| 323 | 11270 suginfo_T *su; |
| 11271 { | |
| 11272 char_u fword[MAXWLEN]; /* copy of the bad word, case-folded */ | |
| 625 | 11273 int n; |
| 11274 char_u *p; | |
| 11275 int lpi; | |
| 323 | 11276 langp_T *lp; |
| 11277 | |
| 11278 /* We make a copy of the case-folded bad word, so that we can modify it | |
| 346 | 11279 * to find matches (esp. REP items). Append some more text, changing |
| 11280 * chars after the bad word may help. */ | |
| 323 | 11281 STRCPY(fword, su->su_fbadword); |
| 835 | 11282 n = (int)STRLEN(fword); |
| 346 | 11283 p = su->su_badptr + su->su_badlen; |
| 835 | 11284 (void)spell_casefold(p, (int)STRLEN(p), fword + n, MAXWLEN - n); |
| 323 | 11285 |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
11286 for (lpi = 0; lpi < curwin->w_s->b_langp.ga_len; ++lpi) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
11287 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
11288 lp = LANGP_ENTRY(curwin->w_s->b_langp, lpi); |
| 482 | 11289 |
| 500 | 11290 /* If reloading a spell file fails it's still in the list but |
| 11291 * everything has been cleared. */ | |
| 625 | 11292 if (lp->lp_slang->sl_fbyts == NULL) |
| 500 | 11293 continue; |
| 11294 | |
| 625 | 11295 /* Try it for this language. Will add possible suggestions. */ |
| 11296 suggest_trie_walk(su, lp, fword, FALSE); | |
| 11297 } | |
| 11298 } | |
| 11299 | |
| 11300 /* Check the maximum score, if we go over it we won't try this change. */ | |
| 11301 #define TRY_DEEPER(su, stack, depth, add) \ | |
| 11302 (stack[depth].ts_score + (add) < su->su_maxscore) | |
| 11303 | |
| 11304 /* | |
| 11305 * Try finding suggestions by adding/removing/swapping letters. | |
| 11306 * | |
| 11307 * This uses a state machine. At each node in the tree we try various | |
| 11308 * operations. When trying if an operation works "depth" is increased and the | |
| 11309 * stack[] is used to store info. This allows combinations, thus insert one | |
| 11310 * character, replace one and delete another. The number of changes is | |
| 11311 * limited by su->su_maxscore. | |
| 11312 * | |
| 11313 * After implementing this I noticed an article by Kemal Oflazer that | |
| 11314 * describes something similar: "Error-tolerant Finite State Recognition with | |
| 11315 * Applications to Morphological Analysis and Spelling Correction" (1996). | |
| 11316 * The implementation in the article is simplified and requires a stack of | |
| 11317 * unknown depth. The implementation here only needs a stack depth equal to | |
| 11318 * the length of the word. | |
| 11319 * | |
| 11320 * This is also used for the sound-folded word, "soundfold" is TRUE then. | |
| 11321 * The mechanism is the same, but we find a match with a sound-folded word | |
| 11322 * that comes from one or more original words. Each of these words may be | |
| 11323 * added, this is done by add_sound_suggest(). | |
| 11324 * Don't use: | |
| 11325 * the prefix tree or the keep-case tree | |
| 11326 * "su->su_badlen" | |
| 11327 * anything to do with upper and lower case | |
| 11328 * anything to do with word or non-word characters ("spell_iswordp()") | |
| 11329 * banned words | |
| 11330 * word flags (rare, region, compounding) | |
| 11331 * word splitting for now | |
| 11332 * "similar_chars()" | |
| 11333 * use "slang->sl_repsal" instead of "lp->lp_replang->sl_rep" | |
| 11334 */ | |
| 11335 static void | |
| 11336 suggest_trie_walk(su, lp, fword, soundfold) | |
| 11337 suginfo_T *su; | |
| 11338 langp_T *lp; | |
| 11339 char_u *fword; | |
| 11340 int soundfold; | |
| 11341 { | |
| 11342 char_u tword[MAXWLEN]; /* good word collected so far */ | |
| 11343 trystate_T stack[MAXWLEN]; | |
| 11344 char_u preword[MAXWLEN * 3]; /* word found with proper case; | |
|
2046
d433f43849ba
updated for version 7.2.332
Bram Moolenaar <bram@zimbu.org>
parents:
1938
diff
changeset
|
11345 * concatenation of prefix compound |
| 625 | 11346 * words and split word. NUL terminated |
| 11347 * when going deeper but not when coming | |
| 11348 * back. */ | |
| 11349 char_u compflags[MAXWLEN]; /* compound flags, one for each word */ | |
| 11350 trystate_T *sp; | |
| 11351 int newscore; | |
| 11352 int score; | |
| 11353 char_u *byts, *fbyts, *pbyts; | |
| 11354 idx_T *idxs, *fidxs, *pidxs; | |
| 11355 int depth; | |
| 11356 int c, c2, c3; | |
| 11357 int n = 0; | |
| 11358 int flags; | |
| 11359 garray_T *gap; | |
| 11360 idx_T arridx; | |
| 11361 int len; | |
| 11362 char_u *p; | |
| 11363 fromto_T *ftp; | |
| 11364 int fl = 0, tl; | |
| 11365 int repextra = 0; /* extra bytes in fword[] from REP item */ | |
| 11366 slang_T *slang = lp->lp_slang; | |
| 11367 int fword_ends; | |
| 11368 int goodword_ends; | |
| 11369 #ifdef DEBUG_TRIEWALK | |
| 11370 /* Stores the name of the change made at each level. */ | |
| 11371 char_u changename[MAXWLEN][80]; | |
| 11372 #endif | |
| 11373 int breakcheckcount = 1000; | |
| 11374 int compound_ok; | |
| 11375 | |
| 11376 /* | |
| 11377 * Go through the whole case-fold tree, try changes at each node. | |
| 11378 * "tword[]" contains the word collected from nodes in the tree. | |
| 11379 * "fword[]" the word we are trying to match with (initially the bad | |
| 11380 * word). | |
| 11381 */ | |
| 11382 depth = 0; | |
| 11383 sp = &stack[0]; | |
| 11384 vim_memset(sp, 0, sizeof(trystate_T)); | |
| 11385 sp->ts_curi = 1; | |
| 11386 | |
| 11387 if (soundfold) | |
| 11388 { | |
| 11389 /* Going through the soundfold tree. */ | |
| 11390 byts = fbyts = slang->sl_sbyts; | |
| 11391 idxs = fidxs = slang->sl_sidxs; | |
| 11392 pbyts = NULL; | |
| 11393 pidxs = NULL; | |
| 11394 sp->ts_prefixdepth = PFD_NOPREFIX; | |
| 11395 sp->ts_state = STATE_START; | |
| 11396 } | |
| 11397 else | |
| 11398 { | |
| 330 | 11399 /* |
| 375 | 11400 * When there are postponed prefixes we need to use these first. At |
| 11401 * the end of the prefix we continue in the case-fold tree. | |
| 11402 */ | |
| 482 | 11403 fbyts = slang->sl_fbyts; |
| 11404 fidxs = slang->sl_fidxs; | |
| 11405 pbyts = slang->sl_pbyts; | |
| 11406 pidxs = slang->sl_pidxs; | |
| 375 | 11407 if (pbyts != NULL) |
| 11408 { | |
| 11409 byts = pbyts; | |
| 11410 idxs = pidxs; | |
| 482 | 11411 sp->ts_prefixdepth = PFD_PREFIXTREE; |
| 375 | 11412 sp->ts_state = STATE_NOPREFIX; /* try without prefix first */ |
| 11413 } | |
| 11414 else | |
| 11415 { | |
| 11416 byts = fbyts; | |
| 11417 idxs = fidxs; | |
| 482 | 11418 sp->ts_prefixdepth = PFD_NOPREFIX; |
| 485 | 11419 sp->ts_state = STATE_START; |
| 375 | 11420 } |
| 625 | 11421 } |
| 11422 | |
| 11423 /* | |
| 11424 * Loop to find all suggestions. At each round we either: | |
| 11425 * - For the current state try one operation, advance "ts_curi", | |
| 11426 * increase "depth". | |
| 11427 * - When a state is done go to the next, set "ts_state". | |
| 11428 * - When all states are tried decrease "depth". | |
| 11429 */ | |
| 11430 while (depth >= 0 && !got_int) | |
| 11431 { | |
| 11432 sp = &stack[depth]; | |
| 11433 switch (sp->ts_state) | |
| 11434 { | |
| 11435 case STATE_START: | |
| 11436 case STATE_NOPREFIX: | |
| 11437 /* | |
| 11438 * Start of node: Deal with NUL bytes, which means | |
| 11439 * tword[] may end here. | |
| 11440 */ | |
| 11441 arridx = sp->ts_arridx; /* current node in the tree */ | |
| 11442 len = byts[arridx]; /* bytes in this node */ | |
| 11443 arridx += sp->ts_curi; /* index of current byte */ | |
| 11444 | |
| 11445 if (sp->ts_prefixdepth == PFD_PREFIXTREE) | |
| 11446 { | |
| 11447 /* Skip over the NUL bytes, we use them later. */ | |
| 11448 for (n = 0; n < len && byts[arridx + n] == 0; ++n) | |
| 11449 ; | |
| 11450 sp->ts_curi += n; | |
| 11451 | |
| 11452 /* Always past NUL bytes now. */ | |
| 11453 n = (int)sp->ts_state; | |
| 11454 sp->ts_state = STATE_ENDNUL; | |
| 11455 sp->ts_save_badflags = su->su_badflags; | |
| 11456 | |
| 11457 /* At end of a prefix or at start of prefixtree: check for | |
| 11458 * following word. */ | |
| 11459 if (byts[arridx] == 0 || n == (int)STATE_NOPREFIX) | |
| 11460 { | |
| 11461 /* Set su->su_badflags to the caps type at this position. | |
| 11462 * Use the caps type until here for the prefix itself. */ | |
| 11463 #ifdef FEAT_MBYTE | |
| 11464 if (has_mbyte) | |
| 11465 n = nofold_len(fword, sp->ts_fidx, su->su_badptr); | |
| 11466 else | |
| 11467 #endif | |
| 11468 n = sp->ts_fidx; | |
| 11469 flags = badword_captype(su->su_badptr, su->su_badptr + n); | |
| 11470 su->su_badflags = badword_captype(su->su_badptr + n, | |
| 455 | 11471 su->su_badptr + su->su_badlen); |
| 625 | 11472 #ifdef DEBUG_TRIEWALK |
| 11473 sprintf(changename[depth], "prefix"); | |
| 11474 #endif | |
| 11475 go_deeper(stack, depth, 0); | |
| 11476 ++depth; | |
| 11477 sp = &stack[depth]; | |
| 11478 sp->ts_prefixdepth = depth - 1; | |
| 11479 byts = fbyts; | |
| 11480 idxs = fidxs; | |
| 11481 sp->ts_arridx = 0; | |
| 11482 | |
| 11483 /* Move the prefix to preword[] with the right case | |
| 11484 * and make find_keepcap_word() works. */ | |
| 11485 tword[sp->ts_twordlen] = NUL; | |
| 11486 make_case_word(tword + sp->ts_splitoff, | |
| 11487 preword + sp->ts_prewordlen, flags); | |
| 835 | 11488 sp->ts_prewordlen = (char_u)STRLEN(preword); |
| 625 | 11489 sp->ts_splitoff = sp->ts_twordlen; |
| 11490 } | |
| 11491 break; | |
| 11492 } | |
| 11493 | |
| 11494 if (sp->ts_curi > len || byts[arridx] != 0) | |
| 11495 { | |
| 11496 /* Past bytes in node and/or past NUL bytes. */ | |
| 11497 sp->ts_state = STATE_ENDNUL; | |
| 11498 sp->ts_save_badflags = su->su_badflags; | |
| 11499 break; | |
| 11500 } | |
| 11501 | |
| 11502 /* | |
| 11503 * End of word in tree. | |
| 11504 */ | |
| 11505 ++sp->ts_curi; /* eat one NUL byte */ | |
| 11506 | |
| 11507 flags = (int)idxs[arridx]; | |
| 698 | 11508 |
| 11509 /* Skip words with the NOSUGGEST flag. */ | |
| 11510 if (flags & WF_NOSUGGEST) | |
| 11511 break; | |
| 11512 | |
| 625 | 11513 fword_ends = (fword[sp->ts_fidx] == NUL |
| 11514 || (soundfold | |
| 11515 ? vim_iswhite(fword[sp->ts_fidx]) | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
11516 : !spell_iswordp(fword + sp->ts_fidx, curwin))); |
| 625 | 11517 tword[sp->ts_twordlen] = NUL; |
| 11518 | |
| 11519 if (sp->ts_prefixdepth <= PFD_NOTSPECIAL | |
| 485 | 11520 && (sp->ts_flags & TSF_PREFIXOK) == 0) |
| 625 | 11521 { |
| 11522 /* There was a prefix before the word. Check that the prefix | |
| 11523 * can be used with this word. */ | |
| 11524 /* Count the length of the NULs in the prefix. If there are | |
| 11525 * none this must be the first try without a prefix. */ | |
| 11526 n = stack[sp->ts_prefixdepth].ts_arridx; | |
| 11527 len = pbyts[n++]; | |
| 11528 for (c = 0; c < len && pbyts[n + c] == 0; ++c) | |
| 11529 ; | |
| 11530 if (c > 0) | |
| 11531 { | |
| 11532 c = valid_word_prefix(c, n, flags, | |
| 483 | 11533 tword + sp->ts_splitoff, slang, FALSE); |
| 625 | 11534 if (c == 0) |
| 11535 break; | |
| 11536 | |
| 11537 /* Use the WF_RARE flag for a rare prefix. */ | |
| 11538 if (c & WF_RAREPFX) | |
| 11539 flags |= WF_RARE; | |
| 11540 | |
| 11541 /* Tricky: when checking for both prefix and compounding | |
| 11542 * we run into the prefix flag first. | |
| 11543 * Remember that it's OK, so that we accept the prefix | |
| 11544 * when arriving at a compound flag. */ | |
| 11545 sp->ts_flags |= TSF_PREFIXOK; | |
| 11546 } | |
| 11547 } | |
| 11548 | |
| 11549 /* Check NEEDCOMPOUND: can't use word without compounding. Do try | |
| 11550 * appending another compound word below. */ | |
| 11551 if (sp->ts_complen == sp->ts_compsplit && fword_ends | |
| 500 | 11552 && (flags & WF_NEEDCOMP)) |
| 625 | 11553 goodword_ends = FALSE; |
| 11554 else | |
| 11555 goodword_ends = TRUE; | |
| 11556 | |
| 11557 p = NULL; | |
| 11558 compound_ok = TRUE; | |
| 11559 if (sp->ts_complen > sp->ts_compsplit) | |
| 11560 { | |
| 11561 if (slang->sl_nobreak) | |
| 11562 { | |
| 11563 /* There was a word before this word. When there was no | |
| 11564 * change in this word (it was correct) add the first word | |
| 11565 * as a suggestion. If this word was corrected too, we | |
| 11566 * need to check if a correct word follows. */ | |
| 11567 if (sp->ts_fidx - sp->ts_splitfidx | |
| 11568 == sp->ts_twordlen - sp->ts_splitoff | |
| 11569 && STRNCMP(fword + sp->ts_splitfidx, | |
| 11570 tword + sp->ts_splitoff, | |
| 11571 sp->ts_fidx - sp->ts_splitfidx) == 0) | |
| 492 | 11572 { |
| 625 | 11573 preword[sp->ts_prewordlen] = NUL; |
| 11574 newscore = score_wordcount_adj(slang, sp->ts_score, | |
| 11575 preword + sp->ts_prewordlen, | |
| 11576 sp->ts_prewordlen > 0); | |
| 11577 /* Add the suggestion if the score isn't too bad. */ | |
| 11578 if (newscore <= su->su_maxscore) | |
| 492 | 11579 add_suggestion(su, &su->su_ga, preword, |
| 503 | 11580 sp->ts_splitfidx - repextra, |
| 625 | 11581 newscore, 0, FALSE, |
| 11582 lp->lp_sallang, FALSE); | |
| 11583 break; | |
| 492 | 11584 } |
| 490 | 11585 } |
| 11586 else | |
| 625 | 11587 { |
| 11588 /* There was a compound word before this word. If this | |
| 11589 * word does not support compounding then give up | |
| 11590 * (splitting is tried for the word without compound | |
| 11591 * flag). */ | |
| 11592 if (((unsigned)flags >> 24) == 0 | |
| 11593 || sp->ts_twordlen - sp->ts_splitoff | |
| 11594 < slang->sl_compminlen) | |
| 11595 break; | |
| 11596 #ifdef FEAT_MBYTE | |
| 11597 /* For multi-byte chars check character length against | |
| 11598 * COMPOUNDMIN. */ | |
| 11599 if (has_mbyte | |
| 11600 && slang->sl_compminlen > 0 | |
| 11601 && mb_charlen(tword + sp->ts_splitoff) | |
| 11602 < slang->sl_compminlen) | |
| 11603 break; | |
| 11604 #endif | |
| 11605 | |
| 11606 compflags[sp->ts_complen] = ((unsigned)flags >> 24); | |
| 11607 compflags[sp->ts_complen + 1] = NUL; | |
| 11608 vim_strncpy(preword + sp->ts_prewordlen, | |
| 11609 tword + sp->ts_splitoff, | |
| 11610 sp->ts_twordlen - sp->ts_splitoff); | |
| 1762 | 11611 |
| 11612 /* Verify CHECKCOMPOUNDPATTERN rules. */ | |
| 11613 if (match_checkcompoundpattern(preword, sp->ts_prewordlen, | |
| 11614 &slang->sl_comppat)) | |
| 11615 compound_ok = FALSE; | |
| 11616 | |
| 11617 if (compound_ok) | |
| 11618 { | |
| 11619 p = preword; | |
| 11620 while (*skiptowhite(p) != NUL) | |
| 11621 p = skipwhite(skiptowhite(p)); | |
| 11622 if (fword_ends && !can_compound(slang, p, | |
| 625 | 11623 compflags + sp->ts_compsplit)) |
| 1762 | 11624 /* Compound is not allowed. But it may still be |
| 11625 * possible if we add another (short) word. */ | |
| 11626 compound_ok = FALSE; | |
| 11627 } | |
| 625 | 11628 |
| 11629 /* Get pointer to last char of previous word. */ | |
| 11630 p = preword + sp->ts_prewordlen; | |
| 11631 mb_ptr_back(preword, p); | |
| 11632 } | |
| 11633 } | |
| 11634 | |
| 11635 /* | |
| 11636 * Form the word with proper case in preword. | |
| 11637 * If there is a word from a previous split, append. | |
| 11638 * For the soundfold tree don't change the case, simply append. | |
| 11639 */ | |
| 11640 if (soundfold) | |
| 11641 STRCPY(preword + sp->ts_prewordlen, tword + sp->ts_splitoff); | |
| 11642 else if (flags & WF_KEEPCAP) | |
| 11643 /* Must find the word in the keep-case tree. */ | |
| 11644 find_keepcap_word(slang, tword + sp->ts_splitoff, | |
| 483 | 11645 preword + sp->ts_prewordlen); |
| 625 | 11646 else |
| 11647 { | |
| 11648 /* Include badflags: If the badword is onecap or allcap | |
| 11649 * use that for the goodword too. But if the badword is | |
| 11650 * allcap and it's only one char long use onecap. */ | |
| 11651 c = su->su_badflags; | |
| 11652 if ((c & WF_ALLCAP) | |
| 11653 #ifdef FEAT_MBYTE | |
| 11654 && su->su_badlen == (*mb_ptr2len)(su->su_badptr) | |
| 346 | 11655 #else |
| 625 | 11656 && su->su_badlen == 1 |
| 11657 #endif | |
| 11658 ) | |
| 11659 c = WF_ONECAP; | |
| 11660 c |= flags; | |
| 11661 | |
| 11662 /* When appending a compound word after a word character don't | |
| 11663 * use Onecap. */ | |
| 11664 if (p != NULL && spell_iswordp_nmw(p)) | |
| 11665 c &= ~WF_ONECAP; | |
| 11666 make_case_word(tword + sp->ts_splitoff, | |
| 490 | 11667 preword + sp->ts_prewordlen, c); |
| 625 | 11668 } |
| 11669 | |
| 11670 if (!soundfold) | |
| 11671 { | |
| 323 | 11672 /* Don't use a banned word. It may appear again as a good |
| 11673 * word, thus remember it. */ | |
| 11674 if (flags & WF_BANNED) | |
| 11675 { | |
| 483 | 11676 add_banned(su, preword + sp->ts_prewordlen); |
| 323 | 11677 break; |
| 11678 } | |
| 534 | 11679 if ((sp->ts_complen == sp->ts_compsplit |
| 625 | 11680 && WAS_BANNED(su, preword + sp->ts_prewordlen)) |
| 11681 || WAS_BANNED(su, preword)) | |
| 534 | 11682 { |
| 11683 if (slang->sl_compprog == NULL) | |
| 11684 break; | |
| 11685 /* the word so far was banned but we may try compounding */ | |
| 11686 goodword_ends = FALSE; | |
| 11687 } | |
| 625 | 11688 } |
| 11689 | |
| 11690 newscore = 0; | |
| 11691 if (!soundfold) /* soundfold words don't have flags */ | |
| 11692 { | |
| 323 | 11693 if ((flags & WF_REGION) |
| 390 | 11694 && (((unsigned)flags >> 16) & lp->lp_region) == 0) |
| 323 | 11695 newscore += SCORE_REGION; |
| 11696 if (flags & WF_RARE) | |
| 11697 newscore += SCORE_RARE; | |
| 11698 | |
| 346 | 11699 if (!spell_valid_case(su->su_badflags, |
| 483 | 11700 captype(preword + sp->ts_prewordlen, NULL))) |
| 323 | 11701 newscore += SCORE_ICASE; |
| 625 | 11702 } |
| 11703 | |
| 11704 /* TODO: how about splitting in the soundfold tree? */ | |
| 11705 if (fword_ends | |
| 11706 && goodword_ends | |
| 11707 && sp->ts_fidx >= sp->ts_fidxtry | |
| 11708 && compound_ok) | |
| 11709 { | |
| 11710 /* The badword also ends: add suggestions. */ | |
| 11711 #ifdef DEBUG_TRIEWALK | |
| 11712 if (soundfold && STRCMP(preword, "smwrd") == 0) | |
| 11713 { | |
| 11714 int j; | |
| 11715 | |
| 11716 /* print the stack of changes that brought us here */ | |
| 11717 smsg("------ %s -------", fword); | |
| 11718 for (j = 0; j < depth; ++j) | |
| 11719 smsg("%s", changename[j]); | |
| 11720 } | |
| 11721 #endif | |
| 11722 if (soundfold) | |
| 11723 { | |
| 11724 /* For soundfolded words we need to find the original | |
| 1188 | 11725 * words, the edit distance and then add them. */ |
| 625 | 11726 add_sound_suggest(su, preword, sp->ts_score, lp); |
| 11727 } | |
|
2457
6c1d10a85398
Avoid illegal memory access in spell suggestion. (Dominique Pelle)
Bram Moolenaar <bram@vim.org>
parents:
2455
diff
changeset
|
11728 else if (sp->ts_fidx > 0) |
| 625 | 11729 { |
| 11730 /* Give a penalty when changing non-word char to word | |
| 11731 * char, e.g., "thes," -> "these". */ | |
| 366 | 11732 p = fword + sp->ts_fidx; |
| 625 | 11733 mb_ptr_back(fword, p); |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
11734 if (!spell_iswordp(p, curwin)) |
| 366 | 11735 { |
| 11736 p = preword + STRLEN(preword); | |
| 625 | 11737 mb_ptr_back(preword, p); |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
11738 if (spell_iswordp(p, curwin)) |
| 366 | 11739 newscore += SCORE_NONWORD; |
| 11740 } | |
| 11741 | |
| 625 | 11742 /* Give a bonus to words seen before. */ |
| 11743 score = score_wordcount_adj(slang, | |
| 11744 sp->ts_score + newscore, | |
| 11745 preword + sp->ts_prewordlen, | |
| 11746 sp->ts_prewordlen > 0); | |
| 11747 | |
| 11748 /* Add the suggestion if the score isn't too bad. */ | |
| 11749 if (score <= su->su_maxscore) | |
| 633 | 11750 { |
| 625 | 11751 add_suggestion(su, &su->su_ga, preword, |
| 11752 sp->ts_fidx - repextra, | |
| 11753 score, 0, FALSE, lp->lp_sallang, FALSE); | |
| 633 | 11754 |
| 11755 if (su->su_badflags & WF_MIXCAP) | |
| 11756 { | |
| 11757 /* We really don't know if the word should be | |
| 11758 * upper or lower case, add both. */ | |
| 11759 c = captype(preword, NULL); | |
| 11760 if (c == 0 || c == WF_ALLCAP) | |
| 11761 { | |
| 11762 make_case_word(tword + sp->ts_splitoff, | |
| 11763 preword + sp->ts_prewordlen, | |
| 11764 c == 0 ? WF_ALLCAP : 0); | |
| 11765 | |
| 11766 add_suggestion(su, &su->su_ga, preword, | |
| 11767 sp->ts_fidx - repextra, | |
| 11768 score + SCORE_ICASE, 0, FALSE, | |
| 11769 lp->lp_sallang, FALSE); | |
| 11770 } | |
| 11771 } | |
| 11772 } | |
| 625 | 11773 } |
| 11774 } | |
| 11775 | |
| 11776 /* | |
| 11777 * Try word split and/or compounding. | |
| 11778 */ | |
| 11779 if ((sp->ts_fidx >= sp->ts_fidxtry || fword_ends) | |
| 11780 #ifdef FEAT_MBYTE | |
| 11781 /* Don't split halfway a character. */ | |
| 11782 && (!has_mbyte || sp->ts_tcharlen == 0) | |
| 11783 #endif | |
| 11784 ) | |
| 11785 { | |
| 11786 int try_compound; | |
| 11787 int try_split; | |
| 11788 | |
| 11789 /* If past the end of the bad word don't try a split. | |
| 11790 * Otherwise try changing the next word. E.g., find | |
| 11791 * suggestions for "the the" where the second "the" is | |
| 11792 * different. It's done like a split. | |
| 11793 * TODO: word split for soundfold words */ | |
| 11794 try_split = (sp->ts_fidx - repextra < su->su_badlen) | |
| 11795 && !soundfold; | |
| 11796 | |
| 11797 /* Get here in several situations: | |
| 11798 * 1. The word in the tree ends: | |
| 11799 * If the word allows compounding try that. Otherwise try | |
| 11800 * a split by inserting a space. For both check that a | |
| 11801 * valid words starts at fword[sp->ts_fidx]. | |
| 11802 * For NOBREAK do like compounding to be able to check if | |
| 11803 * the next word is valid. | |
| 11804 * 2. The badword does end, but it was due to a change (e.g., | |
| 11805 * a swap). No need to split, but do check that the | |
| 11806 * following word is valid. | |
| 11807 * 3. The badword and the word in the tree end. It may still | |
| 11808 * be possible to compound another (short) word. | |
| 11809 */ | |
| 11810 try_compound = FALSE; | |
| 11811 if (!soundfold | |
| 11812 && slang->sl_compprog != NULL | |
| 11813 && ((unsigned)flags >> 24) != 0 | |
| 11814 && sp->ts_twordlen - sp->ts_splitoff | |
| 11815 >= slang->sl_compminlen | |
| 11816 #ifdef FEAT_MBYTE | |
| 11817 && (!has_mbyte | |
| 11818 || slang->sl_compminlen == 0 | |
| 11819 || mb_charlen(tword + sp->ts_splitoff) | |
| 11820 >= slang->sl_compminlen) | |
| 11821 #endif | |
| 11822 && (slang->sl_compsylmax < MAXWLEN | |
| 11823 || sp->ts_complen + 1 - sp->ts_compsplit | |
| 11824 < slang->sl_compmax) | |
| 1762 | 11825 && (can_be_compound(sp, slang, |
| 11826 compflags, ((unsigned)flags >> 24)))) | |
| 11827 | |
| 625 | 11828 { |
| 11829 try_compound = TRUE; | |
| 11830 compflags[sp->ts_complen] = ((unsigned)flags >> 24); | |
| 11831 compflags[sp->ts_complen + 1] = NUL; | |
| 11832 } | |
| 11833 | |
| 11834 /* For NOBREAK we never try splitting, it won't make any word | |
| 11835 * valid. */ | |
| 11836 if (slang->sl_nobreak) | |
| 11837 try_compound = TRUE; | |
| 11838 | |
| 11839 /* If we could add a compound word, and it's also possible to | |
| 11840 * split at this point, do the split first and set | |
| 11841 * TSF_DIDSPLIT to avoid doing it again. */ | |
| 11842 else if (!fword_ends | |
| 11843 && try_compound | |
| 11844 && (sp->ts_flags & TSF_DIDSPLIT) == 0) | |
| 11845 { | |
| 485 | 11846 try_compound = FALSE; |
| 625 | 11847 sp->ts_flags |= TSF_DIDSPLIT; |
| 11848 --sp->ts_curi; /* do the same NUL again */ | |
| 11849 compflags[sp->ts_complen] = NUL; | |
| 11850 } | |
| 11851 else | |
| 11852 sp->ts_flags &= ~TSF_DIDSPLIT; | |
| 11853 | |
| 11854 if (try_split || try_compound) | |
| 11855 { | |
| 534 | 11856 if (!try_compound && (!fword_ends || !goodword_ends)) |
| 485 | 11857 { |
| 11858 /* If we're going to split need to check that the | |
| 501 | 11859 * words so far are valid for compounding. If there |
| 11860 * is only one word it must not have the NEEDCOMPOUND | |
| 11861 * flag. */ | |
| 11862 if (sp->ts_complen == sp->ts_compsplit | |
| 11863 && (flags & WF_NEEDCOMP)) | |
| 11864 break; | |
| 490 | 11865 p = preword; |
| 11866 while (*skiptowhite(p) != NUL) | |
| 11867 p = skipwhite(skiptowhite(p)); | |
| 485 | 11868 if (sp->ts_complen > sp->ts_compsplit |
| 490 | 11869 && !can_compound(slang, p, |
| 485 | 11870 compflags + sp->ts_compsplit)) |
| 11871 break; | |
| 698 | 11872 |
| 11873 if (slang->sl_nosplitsugs) | |
| 11874 newscore += SCORE_SPLIT_NO; | |
| 11875 else | |
| 11876 newscore += SCORE_SPLIT; | |
| 625 | 11877 |
| 11878 /* Give a bonus to words seen before. */ | |
| 11879 newscore = score_wordcount_adj(slang, newscore, | |
| 11880 preword + sp->ts_prewordlen, TRUE); | |
| 482 | 11881 } |
| 11882 | |
| 625 | 11883 if (TRY_DEEPER(su, stack, depth, newscore)) |
| 323 | 11884 { |
| 625 | 11885 go_deeper(stack, depth, newscore); |
| 11886 #ifdef DEBUG_TRIEWALK | |
| 11887 if (!try_compound && !fword_ends) | |
| 11888 sprintf(changename[depth], "%.*s-%s: split", | |
| 11889 sp->ts_twordlen, tword, fword + sp->ts_fidx); | |
| 11890 else | |
| 11891 sprintf(changename[depth], "%.*s-%s: compound", | |
| 11892 sp->ts_twordlen, tword, fword + sp->ts_fidx); | |
| 11893 #endif | |
| 323 | 11894 /* Save things to be restored at STATE_SPLITUNDO. */ |
| 346 | 11895 sp->ts_save_badflags = su->su_badflags; |
| 376 | 11896 sp->ts_state = STATE_SPLITUNDO; |
| 11897 | |
| 11898 ++depth; | |
| 11899 sp = &stack[depth]; | |
| 323 | 11900 |
| 482 | 11901 /* Append a space to preword when splitting. */ |
| 11902 if (!try_compound && !fword_ends) | |
| 11903 STRCAT(preword, " "); | |
| 835 | 11904 sp->ts_prewordlen = (char_u)STRLEN(preword); |
| 483 | 11905 sp->ts_splitoff = sp->ts_twordlen; |
| 492 | 11906 sp->ts_splitfidx = sp->ts_fidx; |
| 376 | 11907 |
| 11908 /* If the badword has a non-word character at this | |
| 11909 * position skip it. That means replacing the | |
| 482 | 11910 * non-word character with a space. Always skip a |
| 534 | 11911 * character when the word ends. But only when the |
| 11912 * good word can end. */ | |
| 625 | 11913 if (((!try_compound && !spell_iswordp_nmw(fword |
| 11914 + sp->ts_fidx)) | |
| 11915 || fword_ends) | |
| 11916 && fword[sp->ts_fidx] != NUL | |
| 11917 && goodword_ends) | |
| 376 | 11918 { |
| 482 | 11919 int l; |
| 11920 | |
| 376 | 11921 #ifdef FEAT_MBYTE |
| 11922 if (has_mbyte) | |
| 482 | 11923 l = MB_BYTE2LEN(fword[sp->ts_fidx]); |
| 376 | 11924 else |
| 11925 #endif | |
| 482 | 11926 l = 1; |
| 11927 if (fword_ends) | |
| 11928 { | |
| 11929 /* Copy the skipped character to preword. */ | |
| 483 | 11930 mch_memmove(preword + sp->ts_prewordlen, |
| 482 | 11931 fword + sp->ts_fidx, l); |
| 483 | 11932 sp->ts_prewordlen += l; |
| 11933 preword[sp->ts_prewordlen] = NUL; | |
| 482 | 11934 } |
| 11935 else | |
| 11936 sp->ts_score -= SCORE_SPLIT - SCORE_SUBST; | |
| 11937 sp->ts_fidx += l; | |
| 376 | 11938 } |
| 455 | 11939 |
| 485 | 11940 /* When compounding include compound flag in |
| 11941 * compflags[] (already set above). When splitting we | |
| 11942 * may start compounding over again. */ | |
| 482 | 11943 if (try_compound) |
| 483 | 11944 ++sp->ts_complen; |
| 482 | 11945 else |
| 485 | 11946 sp->ts_compsplit = sp->ts_complen; |
| 11947 sp->ts_prefixdepth = PFD_NOPREFIX; | |
| 482 | 11948 |
| 455 | 11949 /* set su->su_badflags to the caps type at this |
| 11950 * position */ | |
| 324 | 11951 #ifdef FEAT_MBYTE |
| 11952 if (has_mbyte) | |
| 455 | 11953 n = nofold_len(fword, sp->ts_fidx, su->su_badptr); |
| 324 | 11954 else |
| 11955 #endif | |
| 455 | 11956 n = sp->ts_fidx; |
| 474 | 11957 su->su_badflags = badword_captype(su->su_badptr + n, |
| 455 | 11958 su->su_badptr + su->su_badlen); |
| 323 | 11959 |
| 11960 /* Restart at top of the tree. */ | |
| 376 | 11961 sp->ts_arridx = 0; |
| 485 | 11962 |
| 11963 /* If there are postponed prefixes, try these too. */ | |
| 11964 if (pbyts != NULL) | |
| 11965 { | |
| 11966 byts = pbyts; | |
| 11967 idxs = pidxs; | |
| 11968 sp->ts_prefixdepth = PFD_PREFIXTREE; | |
| 11969 sp->ts_state = STATE_NOPREFIX; | |
| 11970 } | |
| 323 | 11971 } |
| 11972 } | |
| 625 | 11973 } |
| 11974 break; | |
| 11975 | |
| 11976 case STATE_SPLITUNDO: | |
| 11977 /* Undo the changes done for word split or compound word. */ | |
| 11978 su->su_badflags = sp->ts_save_badflags; | |
| 11979 | |
| 11980 /* Continue looking for NUL bytes. */ | |
| 11981 sp->ts_state = STATE_START; | |
| 11982 | |
| 11983 /* In case we went into the prefix tree. */ | |
| 11984 byts = fbyts; | |
| 11985 idxs = fidxs; | |
| 11986 break; | |
| 11987 | |
| 11988 case STATE_ENDNUL: | |
| 11989 /* Past the NUL bytes in the node. */ | |
| 11990 su->su_badflags = sp->ts_save_badflags; | |
| 11991 if (fword[sp->ts_fidx] == NUL | |
| 11992 #ifdef FEAT_MBYTE | |
| 11993 && sp->ts_tcharlen == 0 | |
| 11994 #endif | |
| 11995 ) | |
| 11996 { | |
| 11997 /* The badword ends, can't use STATE_PLAIN. */ | |
| 11998 sp->ts_state = STATE_DEL; | |
| 323 | 11999 break; |
| 625 | 12000 } |
| 12001 sp->ts_state = STATE_PLAIN; | |
| 12002 /*FALLTHROUGH*/ | |
| 12003 | |
| 12004 case STATE_PLAIN: | |
| 12005 /* | |
| 12006 * Go over all possible bytes at this node, add each to tword[] | |
| 12007 * and use child node. "ts_curi" is the index. | |
| 12008 */ | |
| 12009 arridx = sp->ts_arridx; | |
| 12010 if (sp->ts_curi > byts[arridx]) | |
| 12011 { | |
| 12012 /* Done all bytes at this node, do next state. When still at | |
| 12013 * already changed bytes skip the other tricks. */ | |
| 12014 if (sp->ts_fidx >= sp->ts_fidxtry) | |
| 323 | 12015 sp->ts_state = STATE_DEL; |
| 12016 else | |
| 625 | 12017 sp->ts_state = STATE_FINAL; |
| 12018 } | |
| 12019 else | |
| 12020 { | |
| 12021 arridx += sp->ts_curi++; | |
| 12022 c = byts[arridx]; | |
| 12023 | |
| 12024 /* Normal byte, go one level deeper. If it's not equal to the | |
| 12025 * byte in the bad word adjust the score. But don't even try | |
| 12026 * when the byte was already changed. And don't try when we | |
| 12027 * just deleted this byte, accepting it is always cheaper then | |
| 12028 * delete + substitute. */ | |
| 12029 if (c == fword[sp->ts_fidx] | |
| 12030 #ifdef FEAT_MBYTE | |
| 12031 || (sp->ts_tcharlen > 0 && sp->ts_isdiff != DIFF_NONE) | |
| 12032 #endif | |
| 12033 ) | |
| 12034 newscore = 0; | |
| 12035 else | |
| 12036 newscore = SCORE_SUBST; | |
| 12037 if ((newscore == 0 | |
| 12038 || (sp->ts_fidx >= sp->ts_fidxtry | |
| 12039 && ((sp->ts_flags & TSF_DIDDEL) == 0 | |
| 12040 || c != fword[sp->ts_delidx]))) | |
| 12041 && TRY_DEEPER(su, stack, depth, newscore)) | |
| 12042 { | |
| 12043 go_deeper(stack, depth, newscore); | |
| 12044 #ifdef DEBUG_TRIEWALK | |
| 12045 if (newscore > 0) | |
| 12046 sprintf(changename[depth], "%.*s-%s: subst %c to %c", | |
| 12047 sp->ts_twordlen, tword, fword + sp->ts_fidx, | |
| 12048 fword[sp->ts_fidx], c); | |
| 323 | 12049 else |
| 625 | 12050 sprintf(changename[depth], "%.*s-%s: accept %c", |
| 12051 sp->ts_twordlen, tword, fword + sp->ts_fidx, | |
| 12052 fword[sp->ts_fidx]); | |
| 12053 #endif | |
| 323 | 12054 ++depth; |
| 625 | 12055 sp = &stack[depth]; |
| 12056 ++sp->ts_fidx; | |
| 12057 tword[sp->ts_twordlen++] = c; | |
| 12058 sp->ts_arridx = idxs[arridx]; | |
| 12059 #ifdef FEAT_MBYTE | |
| 12060 if (newscore == SCORE_SUBST) | |
| 12061 sp->ts_isdiff = DIFF_YES; | |
| 330 | 12062 if (has_mbyte) |
| 358 | 12063 { |
| 625 | 12064 /* Multi-byte characters are a bit complicated to |
| 12065 * handle: They differ when any of the bytes differ | |
| 12066 * and then their length may also differ. */ | |
| 12067 if (sp->ts_tcharlen == 0) | |
| 12068 { | |
| 12069 /* First byte. */ | |
| 12070 sp->ts_tcharidx = 0; | |
| 12071 sp->ts_tcharlen = MB_BYTE2LEN(c); | |
| 12072 sp->ts_fcharstart = sp->ts_fidx - 1; | |
| 12073 sp->ts_isdiff = (newscore != 0) | |
| 12074 ? DIFF_YES : DIFF_NONE; | |
| 12075 } | |
| 12076 else if (sp->ts_isdiff == DIFF_INSERT) | |
| 12077 /* When inserting trail bytes don't advance in the | |
| 12078 * bad word. */ | |
| 12079 --sp->ts_fidx; | |
| 12080 if (++sp->ts_tcharidx == sp->ts_tcharlen) | |
| 12081 { | |
| 12082 /* Last byte of character. */ | |
| 12083 if (sp->ts_isdiff == DIFF_YES) | |
| 12084 { | |
| 12085 /* Correct ts_fidx for the byte length of the | |
| 12086 * character (we didn't check that before). */ | |
| 12087 sp->ts_fidx = sp->ts_fcharstart | |
| 12088 + MB_BYTE2LEN( | |
| 12089 fword[sp->ts_fcharstart]); | |
| 12090 | |
| 12091 /* For changing a composing character adjust | |
| 12092 * the score from SCORE_SUBST to | |
| 12093 * SCORE_SUBCOMP. */ | |
| 12094 if (enc_utf8 | |
| 12095 && utf_iscomposing( | |
| 12096 mb_ptr2char(tword | |
| 12097 + sp->ts_twordlen | |
| 12098 - sp->ts_tcharlen)) | |
| 12099 && utf_iscomposing( | |
| 12100 mb_ptr2char(fword | |
| 12101 + sp->ts_fcharstart))) | |
| 12102 sp->ts_score -= | |
| 12103 SCORE_SUBST - SCORE_SUBCOMP; | |
| 12104 | |
| 12105 /* For a similar character adjust score from | |
| 12106 * SCORE_SUBST to SCORE_SIMILAR. */ | |
| 12107 else if (!soundfold | |
| 12108 && slang->sl_has_map | |
| 12109 && similar_chars(slang, | |
| 12110 mb_ptr2char(tword | |
| 12111 + sp->ts_twordlen | |
| 12112 - sp->ts_tcharlen), | |
| 12113 mb_ptr2char(fword | |
| 12114 + sp->ts_fcharstart))) | |
| 12115 sp->ts_score -= | |
| 12116 SCORE_SUBST - SCORE_SIMILAR; | |
| 12117 } | |
| 12118 else if (sp->ts_isdiff == DIFF_INSERT | |
| 12119 && sp->ts_twordlen > sp->ts_tcharlen) | |
| 12120 { | |
| 12121 p = tword + sp->ts_twordlen - sp->ts_tcharlen; | |
| 12122 c = mb_ptr2char(p); | |
| 12123 if (enc_utf8 && utf_iscomposing(c)) | |
| 12124 { | |
| 12125 /* Inserting a composing char doesn't | |
| 12126 * count that much. */ | |
| 12127 sp->ts_score -= SCORE_INS - SCORE_INSCOMP; | |
| 12128 } | |
| 12129 else | |
| 12130 { | |
| 12131 /* If the previous character was the same, | |
| 12132 * thus doubling a character, give a bonus | |
| 12133 * to the score. Also for the soundfold | |
| 12134 * tree (might seem illogical but does | |
| 12135 * give better scores). */ | |
| 12136 mb_ptr_back(tword, p); | |
| 12137 if (c == mb_ptr2char(p)) | |
| 12138 sp->ts_score -= SCORE_INS | |
| 12139 - SCORE_INSDUP; | |
| 12140 } | |
| 12141 } | |
| 12142 | |
| 12143 /* Starting a new char, reset the length. */ | |
| 12144 sp->ts_tcharlen = 0; | |
| 12145 } | |
| 358 | 12146 } |
| 330 | 12147 else |
| 12148 #endif | |
| 358 | 12149 { |
| 625 | 12150 /* If we found a similar char adjust the score. |
| 12151 * We do this after calling go_deeper() because | |
| 12152 * it's slow. */ | |
| 12153 if (newscore != 0 | |
| 12154 && !soundfold | |
| 12155 && slang->sl_has_map | |
| 12156 && similar_chars(slang, | |
| 12157 c, fword[sp->ts_fidx - 1])) | |
| 12158 sp->ts_score -= SCORE_SUBST - SCORE_SIMILAR; | |
| 358 | 12159 } |
| 625 | 12160 } |
| 12161 } | |
| 12162 break; | |
| 12163 | |
| 12164 case STATE_DEL: | |
| 12165 #ifdef FEAT_MBYTE | |
| 12166 /* When past the first byte of a multi-byte char don't try | |
| 12167 * delete/insert/swap a character. */ | |
| 12168 if (has_mbyte && sp->ts_tcharlen > 0) | |
| 12169 { | |
| 12170 sp->ts_state = STATE_FINAL; | |
| 12171 break; | |
| 12172 } | |
| 12173 #endif | |
| 12174 /* | |
| 12175 * Try skipping one character in the bad word (delete it). | |
| 12176 */ | |
| 12177 sp->ts_state = STATE_INS_PREP; | |
| 12178 sp->ts_curi = 1; | |
| 12179 if (soundfold && sp->ts_fidx == 0 && fword[sp->ts_fidx] == '*') | |
| 12180 /* Deleting a vowel at the start of a word counts less, see | |
| 12181 * soundalike_score(). */ | |
| 12182 newscore = 2 * SCORE_DEL / 3; | |
| 12183 else | |
| 12184 newscore = SCORE_DEL; | |
| 12185 if (fword[sp->ts_fidx] != NUL | |
| 12186 && TRY_DEEPER(su, stack, depth, newscore)) | |
| 12187 { | |
| 12188 go_deeper(stack, depth, newscore); | |
| 12189 #ifdef DEBUG_TRIEWALK | |
| 12190 sprintf(changename[depth], "%.*s-%s: delete %c", | |
| 12191 sp->ts_twordlen, tword, fword + sp->ts_fidx, | |
| 12192 fword[sp->ts_fidx]); | |
| 12193 #endif | |
| 12194 ++depth; | |
| 12195 | |
| 12196 /* Remember what character we deleted, so that we can avoid | |
| 12197 * inserting it again. */ | |
| 12198 stack[depth].ts_flags |= TSF_DIDDEL; | |
| 12199 stack[depth].ts_delidx = sp->ts_fidx; | |
| 12200 | |
| 12201 /* Advance over the character in fword[]. Give a bonus to the | |
| 12202 * score if the same character is following "nn" -> "n". It's | |
| 12203 * a bit illogical for soundfold tree but it does give better | |
| 12204 * results. */ | |
| 12205 #ifdef FEAT_MBYTE | |
| 12206 if (has_mbyte) | |
| 12207 { | |
| 12208 c = mb_ptr2char(fword + sp->ts_fidx); | |
| 12209 stack[depth].ts_fidx += MB_BYTE2LEN(fword[sp->ts_fidx]); | |
| 12210 if (enc_utf8 && utf_iscomposing(c)) | |
| 12211 stack[depth].ts_score -= SCORE_DEL - SCORE_DELCOMP; | |
| 12212 else if (c == mb_ptr2char(fword + stack[depth].ts_fidx)) | |
| 12213 stack[depth].ts_score -= SCORE_DEL - SCORE_DELDUP; | |
| 323 | 12214 } |
| 12215 else | |
| 625 | 12216 #endif |
| 12217 { | |
| 12218 ++stack[depth].ts_fidx; | |
| 12219 if (fword[sp->ts_fidx] == fword[sp->ts_fidx + 1]) | |
| 12220 stack[depth].ts_score -= SCORE_DEL - SCORE_DELDUP; | |
| 323 | 12221 } |
| 12222 break; | |
| 625 | 12223 } |
| 12224 /*FALLTHROUGH*/ | |
| 12225 | |
| 12226 case STATE_INS_PREP: | |
| 12227 if (sp->ts_flags & TSF_DIDDEL) | |
| 12228 { | |
| 12229 /* If we just deleted a byte then inserting won't make sense, | |
| 12230 * a substitute is always cheaper. */ | |
| 12231 sp->ts_state = STATE_SWAP; | |
| 12232 break; | |
| 12233 } | |
| 12234 | |
| 12235 /* skip over NUL bytes */ | |
| 12236 n = sp->ts_arridx; | |
| 12237 for (;;) | |
| 12238 { | |
| 12239 if (sp->ts_curi > byts[n]) | |
| 12240 { | |
| 12241 /* Only NUL bytes at this node, go to next state. */ | |
| 12242 sp->ts_state = STATE_SWAP; | |
| 12243 break; | |
| 12244 } | |
| 12245 if (byts[n + sp->ts_curi] != NUL) | |
| 12246 { | |
| 12247 /* Found a byte to insert. */ | |
| 12248 sp->ts_state = STATE_INS; | |
| 330 | 12249 break; |
| 12250 } | |
| 625 | 12251 ++sp->ts_curi; |
| 12252 } | |
| 12253 break; | |
| 12254 | |
| 12255 /*FALLTHROUGH*/ | |
| 12256 | |
| 12257 case STATE_INS: | |
| 12258 /* Insert one byte. Repeat this for each possible byte at this | |
| 12259 * node. */ | |
| 12260 n = sp->ts_arridx; | |
| 12261 if (sp->ts_curi > byts[n]) | |
| 12262 { | |
| 12263 /* Done all bytes at this node, go to next state. */ | |
| 12264 sp->ts_state = STATE_SWAP; | |
| 12265 break; | |
| 12266 } | |
| 12267 | |
| 12268 /* Do one more byte at this node, but: | |
| 12269 * - Skip NUL bytes. | |
| 12270 * - Skip the byte if it's equal to the byte in the word, | |
| 12271 * accepting that byte is always better. | |
| 12272 */ | |
| 12273 n += sp->ts_curi++; | |
| 12274 c = byts[n]; | |
| 12275 if (soundfold && sp->ts_twordlen == 0 && c == '*') | |
| 12276 /* Inserting a vowel at the start of a word counts less, | |
| 12277 * see soundalike_score(). */ | |
| 12278 newscore = 2 * SCORE_INS / 3; | |
| 12279 else | |
| 12280 newscore = SCORE_INS; | |
| 12281 if (c != fword[sp->ts_fidx] | |
| 12282 && TRY_DEEPER(su, stack, depth, newscore)) | |
| 12283 { | |
| 12284 go_deeper(stack, depth, newscore); | |
| 12285 #ifdef DEBUG_TRIEWALK | |
| 12286 sprintf(changename[depth], "%.*s-%s: insert %c", | |
| 12287 sp->ts_twordlen, tword, fword + sp->ts_fidx, | |
| 12288 c); | |
| 12289 #endif | |
| 12290 ++depth; | |
| 12291 sp = &stack[depth]; | |
| 12292 tword[sp->ts_twordlen++] = c; | |
| 12293 sp->ts_arridx = idxs[n]; | |
| 330 | 12294 #ifdef FEAT_MBYTE |
| 12295 if (has_mbyte) | |
| 12296 { | |
| 625 | 12297 fl = MB_BYTE2LEN(c); |
| 12298 if (fl > 1) | |
| 330 | 12299 { |
| 625 | 12300 /* There are following bytes for the same character. |
| 12301 * We must find all bytes before trying | |
| 12302 * delete/insert/swap/etc. */ | |
| 12303 sp->ts_tcharlen = fl; | |
| 12304 sp->ts_tcharidx = 1; | |
| 12305 sp->ts_isdiff = DIFF_INSERT; | |
| 330 | 12306 } |
| 323 | 12307 } |
| 12308 else | |
| 625 | 12309 fl = 1; |
| 12310 if (fl == 1) | |
| 12311 #endif | |
| 12312 { | |
| 12313 /* If the previous character was the same, thus doubling a | |
| 12314 * character, give a bonus to the score. Also for | |
| 12315 * soundfold words (illogical but does give a better | |
| 12316 * score). */ | |
| 12317 if (sp->ts_twordlen >= 2 | |
| 12318 && tword[sp->ts_twordlen - 2] == c) | |
| 12319 sp->ts_score -= SCORE_INS - SCORE_INSDUP; | |
| 12320 } | |
| 12321 } | |
| 12322 break; | |
| 12323 | |
| 12324 case STATE_SWAP: | |
| 12325 /* | |
| 12326 * Swap two bytes in the bad word: "12" -> "21". | |
| 12327 * We change "fword" here, it's changed back afterwards at | |
| 12328 * STATE_UNSWAP. | |
| 12329 */ | |
| 12330 p = fword + sp->ts_fidx; | |
| 12331 c = *p; | |
| 12332 if (c == NUL) | |
| 12333 { | |
| 12334 /* End of word, can't swap or replace. */ | |
| 12335 sp->ts_state = STATE_FINAL; | |
| 12336 break; | |
| 12337 } | |
| 12338 | |
| 12339 /* Don't swap if the first character is not a word character. | |
| 12340 * SWAP3 etc. also don't make sense then. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
12341 if (!soundfold && !spell_iswordp(p, curwin)) |
| 625 | 12342 { |
| 12343 sp->ts_state = STATE_REP_INI; | |
| 323 | 12344 break; |
| 625 | 12345 } |
| 12346 | |
| 12347 #ifdef FEAT_MBYTE | |
| 12348 if (has_mbyte) | |
| 12349 { | |
| 12350 n = mb_cptr2len(p); | |
| 12351 c = mb_ptr2char(p); | |
| 1337 | 12352 if (p[n] == NUL) |
| 12353 c2 = NUL; | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
12354 else if (!soundfold && !spell_iswordp(p + n, curwin)) |
| 625 | 12355 c2 = c; /* don't swap non-word char */ |
| 12356 else | |
| 12357 c2 = mb_ptr2char(p + n); | |
| 12358 } | |
| 12359 else | |
| 12360 #endif | |
| 12361 { | |
| 1337 | 12362 if (p[1] == NUL) |
| 12363 c2 = NUL; | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
12364 else if (!soundfold && !spell_iswordp(p + 1, curwin)) |
| 625 | 12365 c2 = c; /* don't swap non-word char */ |
| 12366 else | |
| 12367 c2 = p[1]; | |
| 12368 } | |
| 12369 | |
| 1337 | 12370 /* When the second character is NUL we can't swap. */ |
| 12371 if (c2 == NUL) | |
| 12372 { | |
| 12373 sp->ts_state = STATE_REP_INI; | |
| 12374 break; | |
| 12375 } | |
| 12376 | |
| 625 | 12377 /* When characters are identical, swap won't do anything. |
| 12378 * Also get here if the second char is not a word character. */ | |
| 12379 if (c == c2) | |
| 12380 { | |
| 12381 sp->ts_state = STATE_SWAP3; | |
| 12382 break; | |
| 12383 } | |
| 12384 if (c2 != NUL && TRY_DEEPER(su, stack, depth, SCORE_SWAP)) | |
| 12385 { | |
| 12386 go_deeper(stack, depth, SCORE_SWAP); | |
| 12387 #ifdef DEBUG_TRIEWALK | |
| 12388 sprintf(changename[depth], "%.*s-%s: swap %c and %c", | |
| 12389 sp->ts_twordlen, tword, fword + sp->ts_fidx, | |
| 12390 c, c2); | |
| 12391 #endif | |
| 12392 sp->ts_state = STATE_UNSWAP; | |
| 12393 ++depth; | |
| 330 | 12394 #ifdef FEAT_MBYTE |
| 12395 if (has_mbyte) | |
| 12396 { | |
| 625 | 12397 fl = mb_char2len(c2); |
| 12398 mch_memmove(p, p + n, fl); | |
| 12399 mb_char2bytes(c, p + fl); | |
| 12400 stack[depth].ts_fidxtry = sp->ts_fidx + n + fl; | |
| 330 | 12401 } |
| 12402 else | |
| 12403 #endif | |
| 12404 { | |
| 625 | 12405 p[0] = c2; |
| 330 | 12406 p[1] = c; |
| 625 | 12407 stack[depth].ts_fidxtry = sp->ts_fidx + 2; |
| 12408 } | |
| 12409 } | |
| 12410 else | |
| 12411 /* If this swap doesn't work then SWAP3 won't either. */ | |
| 12412 sp->ts_state = STATE_REP_INI; | |
| 12413 break; | |
| 12414 | |
| 12415 case STATE_UNSWAP: | |
| 12416 /* Undo the STATE_SWAP swap: "21" -> "12". */ | |
| 12417 p = fword + sp->ts_fidx; | |
| 12418 #ifdef FEAT_MBYTE | |
| 12419 if (has_mbyte) | |
| 12420 { | |
| 12421 n = MB_BYTE2LEN(*p); | |
| 12422 c = mb_ptr2char(p + n); | |
| 12423 mch_memmove(p + MB_BYTE2LEN(p[n]), p, n); | |
| 12424 mb_char2bytes(c, p); | |
| 12425 } | |
| 12426 else | |
| 12427 #endif | |
| 12428 { | |
| 12429 c = *p; | |
| 12430 *p = p[1]; | |
| 12431 p[1] = c; | |
| 12432 } | |
| 12433 /*FALLTHROUGH*/ | |
| 12434 | |
| 12435 case STATE_SWAP3: | |
| 12436 /* Swap two bytes, skipping one: "123" -> "321". We change | |
| 12437 * "fword" here, it's changed back afterwards at STATE_UNSWAP3. */ | |
| 12438 p = fword + sp->ts_fidx; | |
| 12439 #ifdef FEAT_MBYTE | |
| 12440 if (has_mbyte) | |
| 12441 { | |
| 12442 n = mb_cptr2len(p); | |
| 12443 c = mb_ptr2char(p); | |
| 12444 fl = mb_cptr2len(p + n); | |
| 12445 c2 = mb_ptr2char(p + n); | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
12446 if (!soundfold && !spell_iswordp(p + n + fl, curwin)) |
| 625 | 12447 c3 = c; /* don't swap non-word char */ |
| 12448 else | |
| 12449 c3 = mb_ptr2char(p + n + fl); | |
| 12450 } | |
| 12451 else | |
| 12452 #endif | |
| 12453 { | |
| 12454 c = *p; | |
| 12455 c2 = p[1]; | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
12456 if (!soundfold && !spell_iswordp(p + 2, curwin)) |
| 625 | 12457 c3 = c; /* don't swap non-word char */ |
| 12458 else | |
| 12459 c3 = p[2]; | |
| 12460 } | |
| 12461 | |
| 12462 /* When characters are identical: "121" then SWAP3 result is | |
| 12463 * identical, ROT3L result is same as SWAP: "211", ROT3L result is | |
| 12464 * same as SWAP on next char: "112". Thus skip all swapping. | |
| 12465 * Also skip when c3 is NUL. | |
| 12466 * Also get here when the third character is not a word character. | |
| 12467 * Second character may any char: "a.b" -> "b.a" */ | |
| 12468 if (c == c3 || c3 == NUL) | |
| 12469 { | |
| 12470 sp->ts_state = STATE_REP_INI; | |
| 12471 break; | |
| 12472 } | |
| 12473 if (TRY_DEEPER(su, stack, depth, SCORE_SWAP3)) | |
| 12474 { | |
| 12475 go_deeper(stack, depth, SCORE_SWAP3); | |
| 12476 #ifdef DEBUG_TRIEWALK | |
| 12477 sprintf(changename[depth], "%.*s-%s: swap3 %c and %c", | |
| 12478 sp->ts_twordlen, tword, fword + sp->ts_fidx, | |
| 12479 c, c3); | |
| 12480 #endif | |
| 12481 sp->ts_state = STATE_UNSWAP3; | |
| 12482 ++depth; | |
| 12483 #ifdef FEAT_MBYTE | |
| 12484 if (has_mbyte) | |
| 12485 { | |
| 12486 tl = mb_char2len(c3); | |
| 12487 mch_memmove(p, p + n + fl, tl); | |
| 12488 mb_char2bytes(c2, p + tl); | |
| 12489 mb_char2bytes(c, p + fl + tl); | |
| 12490 stack[depth].ts_fidxtry = sp->ts_fidx + n + fl + tl; | |
| 12491 } | |
| 12492 else | |
| 12493 #endif | |
| 12494 { | |
| 12495 p[0] = p[2]; | |
| 12496 p[2] = c; | |
| 12497 stack[depth].ts_fidxtry = sp->ts_fidx + 3; | |
| 12498 } | |
| 12499 } | |
| 12500 else | |
| 12501 sp->ts_state = STATE_REP_INI; | |
| 12502 break; | |
| 12503 | |
| 12504 case STATE_UNSWAP3: | |
| 12505 /* Undo STATE_SWAP3: "321" -> "123" */ | |
| 12506 p = fword + sp->ts_fidx; | |
| 12507 #ifdef FEAT_MBYTE | |
| 12508 if (has_mbyte) | |
| 12509 { | |
| 12510 n = MB_BYTE2LEN(*p); | |
| 12511 c2 = mb_ptr2char(p + n); | |
| 12512 fl = MB_BYTE2LEN(p[n]); | |
| 12513 c = mb_ptr2char(p + n + fl); | |
| 12514 tl = MB_BYTE2LEN(p[n + fl]); | |
| 12515 mch_memmove(p + fl + tl, p, n); | |
| 12516 mb_char2bytes(c, p); | |
| 12517 mb_char2bytes(c2, p + tl); | |
| 12518 p = p + tl; | |
| 12519 } | |
| 12520 else | |
| 12521 #endif | |
| 12522 { | |
| 12523 c = *p; | |
| 12524 *p = p[2]; | |
| 12525 p[2] = c; | |
| 12526 ++p; | |
| 12527 } | |
| 12528 | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
12529 if (!soundfold && !spell_iswordp(p, curwin)) |
| 625 | 12530 { |
| 12531 /* Middle char is not a word char, skip the rotate. First and | |
| 12532 * third char were already checked at swap and swap3. */ | |
| 12533 sp->ts_state = STATE_REP_INI; | |
| 12534 break; | |
| 12535 } | |
| 12536 | |
| 12537 /* Rotate three characters left: "123" -> "231". We change | |
| 12538 * "fword" here, it's changed back afterwards at STATE_UNROT3L. */ | |
| 12539 if (TRY_DEEPER(su, stack, depth, SCORE_SWAP3)) | |
| 12540 { | |
| 12541 go_deeper(stack, depth, SCORE_SWAP3); | |
| 12542 #ifdef DEBUG_TRIEWALK | |
| 12543 p = fword + sp->ts_fidx; | |
| 12544 sprintf(changename[depth], "%.*s-%s: rotate left %c%c%c", | |
| 12545 sp->ts_twordlen, tword, fword + sp->ts_fidx, | |
| 12546 p[0], p[1], p[2]); | |
| 12547 #endif | |
| 12548 sp->ts_state = STATE_UNROT3L; | |
| 12549 ++depth; | |
| 330 | 12550 p = fword + sp->ts_fidx; |
| 12551 #ifdef FEAT_MBYTE | |
| 12552 if (has_mbyte) | |
| 12553 { | |
| 474 | 12554 n = mb_cptr2len(p); |
| 330 | 12555 c = mb_ptr2char(p); |
| 474 | 12556 fl = mb_cptr2len(p + n); |
| 625 | 12557 fl += mb_cptr2len(p + n + fl); |
| 12558 mch_memmove(p, p + n, fl); | |
| 12559 mb_char2bytes(c, p + fl); | |
| 12560 stack[depth].ts_fidxtry = sp->ts_fidx + n + fl; | |
| 330 | 12561 } |
| 12562 else | |
| 12563 #endif | |
| 323 | 12564 { |
| 330 | 12565 c = *p; |
| 625 | 12566 *p = p[1]; |
| 12567 p[1] = p[2]; | |
| 12568 p[2] = c; | |
| 12569 stack[depth].ts_fidxtry = sp->ts_fidx + 3; | |
| 12570 } | |
| 12571 } | |
| 12572 else | |
| 12573 sp->ts_state = STATE_REP_INI; | |
| 12574 break; | |
| 12575 | |
| 12576 case STATE_UNROT3L: | |
| 12577 /* Undo ROT3L: "231" -> "123" */ | |
| 12578 p = fword + sp->ts_fidx; | |
| 12579 #ifdef FEAT_MBYTE | |
| 12580 if (has_mbyte) | |
| 12581 { | |
| 12582 n = MB_BYTE2LEN(*p); | |
| 12583 n += MB_BYTE2LEN(p[n]); | |
| 12584 c = mb_ptr2char(p + n); | |
| 12585 tl = MB_BYTE2LEN(p[n]); | |
| 12586 mch_memmove(p + tl, p, n); | |
| 12587 mb_char2bytes(c, p); | |
| 12588 } | |
| 12589 else | |
| 12590 #endif | |
| 12591 { | |
| 12592 c = p[2]; | |
| 12593 p[2] = p[1]; | |
| 12594 p[1] = *p; | |
| 12595 *p = c; | |
| 12596 } | |
| 12597 | |
| 12598 /* Rotate three bytes right: "123" -> "312". We change "fword" | |
| 12599 * here, it's changed back afterwards at STATE_UNROT3R. */ | |
| 12600 if (TRY_DEEPER(su, stack, depth, SCORE_SWAP3)) | |
| 12601 { | |
| 12602 go_deeper(stack, depth, SCORE_SWAP3); | |
| 12603 #ifdef DEBUG_TRIEWALK | |
| 12604 p = fword + sp->ts_fidx; | |
| 12605 sprintf(changename[depth], "%.*s-%s: rotate right %c%c%c", | |
| 12606 sp->ts_twordlen, tword, fword + sp->ts_fidx, | |
| 12607 p[0], p[1], p[2]); | |
| 12608 #endif | |
| 12609 sp->ts_state = STATE_UNROT3R; | |
| 12610 ++depth; | |
| 330 | 12611 p = fword + sp->ts_fidx; |
| 12612 #ifdef FEAT_MBYTE | |
| 12613 if (has_mbyte) | |
| 12614 { | |
| 625 | 12615 n = mb_cptr2len(p); |
| 12616 n += mb_cptr2len(p + n); | |
| 330 | 12617 c = mb_ptr2char(p + n); |
| 625 | 12618 tl = mb_cptr2len(p + n); |
| 330 | 12619 mch_memmove(p + tl, p, n); |
| 12620 mb_char2bytes(c, p); | |
| 625 | 12621 stack[depth].ts_fidxtry = sp->ts_fidx + n + tl; |
| 330 | 12622 } |
| 12623 else | |
| 12624 #endif | |
| 12625 { | |
| 12626 c = p[2]; | |
| 12627 p[2] = p[1]; | |
| 12628 p[1] = *p; | |
| 12629 *p = c; | |
| 625 | 12630 stack[depth].ts_fidxtry = sp->ts_fidx + 3; |
| 12631 } | |
| 12632 } | |
| 12633 else | |
| 12634 sp->ts_state = STATE_REP_INI; | |
| 12635 break; | |
| 12636 | |
| 12637 case STATE_UNROT3R: | |
| 12638 /* Undo ROT3R: "312" -> "123" */ | |
| 12639 p = fword + sp->ts_fidx; | |
| 12640 #ifdef FEAT_MBYTE | |
| 12641 if (has_mbyte) | |
| 12642 { | |
| 12643 c = mb_ptr2char(p); | |
| 12644 tl = MB_BYTE2LEN(*p); | |
| 12645 n = MB_BYTE2LEN(p[tl]); | |
| 12646 n += MB_BYTE2LEN(p[tl + n]); | |
| 12647 mch_memmove(p, p + tl, n); | |
| 12648 mb_char2bytes(c, p + n); | |
| 12649 } | |
| 12650 else | |
| 12651 #endif | |
| 12652 { | |
| 12653 c = *p; | |
| 12654 *p = p[1]; | |
| 12655 p[1] = p[2]; | |
| 12656 p[2] = c; | |
| 12657 } | |
| 12658 /*FALLTHROUGH*/ | |
| 12659 | |
| 12660 case STATE_REP_INI: | |
| 12661 /* Check if matching with REP items from the .aff file would work. | |
| 12662 * Quickly skip if: | |
| 12663 * - there are no REP items and we are not in the soundfold trie | |
| 12664 * - the score is going to be too high anyway | |
| 12665 * - already applied a REP item or swapped here */ | |
| 12666 if ((lp->lp_replang == NULL && !soundfold) | |
| 12667 || sp->ts_score + SCORE_REP >= su->su_maxscore | |
| 12668 || sp->ts_fidx < sp->ts_fidxtry) | |
| 12669 { | |
| 12670 sp->ts_state = STATE_FINAL; | |
| 323 | 12671 break; |
| 625 | 12672 } |
| 12673 | |
| 12674 /* Use the first byte to quickly find the first entry that may | |
| 12675 * match. If the index is -1 there is none. */ | |
| 12676 if (soundfold) | |
| 12677 sp->ts_curi = slang->sl_repsal_first[fword[sp->ts_fidx]]; | |
| 12678 else | |
| 12679 sp->ts_curi = lp->lp_replang->sl_rep_first[fword[sp->ts_fidx]]; | |
| 12680 | |
| 12681 if (sp->ts_curi < 0) | |
| 12682 { | |
| 12683 sp->ts_state = STATE_FINAL; | |
| 12684 break; | |
| 12685 } | |
| 12686 | |
| 12687 sp->ts_state = STATE_REP; | |
| 12688 /*FALLTHROUGH*/ | |
| 12689 | |
| 12690 case STATE_REP: | |
| 12691 /* Try matching with REP items from the .aff file. For each match | |
| 12692 * replace the characters and check if the resulting word is | |
| 12693 * valid. */ | |
| 12694 p = fword + sp->ts_fidx; | |
| 12695 | |
| 12696 if (soundfold) | |
| 12697 gap = &slang->sl_repsal; | |
| 12698 else | |
| 503 | 12699 gap = &lp->lp_replang->sl_rep; |
| 625 | 12700 while (sp->ts_curi < gap->ga_len) |
| 12701 { | |
| 12702 ftp = (fromto_T *)gap->ga_data + sp->ts_curi++; | |
| 12703 if (*ftp->ft_from != *p) | |
| 12704 { | |
| 12705 /* past possible matching entries */ | |
| 12706 sp->ts_curi = gap->ga_len; | |
| 323 | 12707 break; |
| 12708 } | |
| 625 | 12709 if (STRNCMP(ftp->ft_from, p, STRLEN(ftp->ft_from)) == 0 |
| 12710 && TRY_DEEPER(su, stack, depth, SCORE_REP)) | |
| 12711 { | |
| 12712 go_deeper(stack, depth, SCORE_REP); | |
| 12713 #ifdef DEBUG_TRIEWALK | |
| 12714 sprintf(changename[depth], "%.*s-%s: replace %s with %s", | |
| 12715 sp->ts_twordlen, tword, fword + sp->ts_fidx, | |
| 12716 ftp->ft_from, ftp->ft_to); | |
| 12717 #endif | |
| 12718 /* Need to undo this afterwards. */ | |
| 12719 sp->ts_state = STATE_REP_UNDO; | |
| 12720 | |
| 12721 /* Change the "from" to the "to" string. */ | |
| 12722 ++depth; | |
| 835 | 12723 fl = (int)STRLEN(ftp->ft_from); |
| 12724 tl = (int)STRLEN(ftp->ft_to); | |
| 625 | 12725 if (fl != tl) |
| 323 | 12726 { |
| 1619 | 12727 STRMOVE(p + tl, p + fl); |
| 625 | 12728 repextra += tl - fl; |
| 323 | 12729 } |
| 625 | 12730 mch_memmove(p, ftp->ft_to, tl); |
| 12731 stack[depth].ts_fidxtry = sp->ts_fidx + tl; | |
| 12732 #ifdef FEAT_MBYTE | |
| 12733 stack[depth].ts_tcharlen = 0; | |
| 12734 #endif | |
| 12735 break; | |
| 12736 } | |
| 12737 } | |
| 12738 | |
| 12739 if (sp->ts_curi >= gap->ga_len && sp->ts_state == STATE_REP) | |
| 12740 /* No (more) matches. */ | |
| 12741 sp->ts_state = STATE_FINAL; | |
| 12742 | |
| 12743 break; | |
| 12744 | |
| 12745 case STATE_REP_UNDO: | |
| 12746 /* Undo a REP replacement and continue with the next one. */ | |
| 12747 if (soundfold) | |
| 12748 gap = &slang->sl_repsal; | |
| 12749 else | |
| 12750 gap = &lp->lp_replang->sl_rep; | |
| 12751 ftp = (fromto_T *)gap->ga_data + sp->ts_curi - 1; | |
| 835 | 12752 fl = (int)STRLEN(ftp->ft_from); |
| 12753 tl = (int)STRLEN(ftp->ft_to); | |
| 625 | 12754 p = fword + sp->ts_fidx; |
| 12755 if (fl != tl) | |
| 12756 { | |
| 1619 | 12757 STRMOVE(p + fl, p + tl); |
| 625 | 12758 repextra -= tl - fl; |
| 12759 } | |
| 12760 mch_memmove(p, ftp->ft_from, fl); | |
| 12761 sp->ts_state = STATE_REP; | |
| 12762 break; | |
| 12763 | |
| 12764 default: | |
| 12765 /* Did all possible states at this level, go up one level. */ | |
| 12766 --depth; | |
| 12767 | |
| 12768 if (depth >= 0 && stack[depth].ts_prefixdepth == PFD_PREFIXTREE) | |
| 12769 { | |
| 12770 /* Continue in or go back to the prefix tree. */ | |
| 12771 byts = pbyts; | |
| 12772 idxs = pidxs; | |
| 12773 } | |
| 12774 | |
| 12775 /* Don't check for CTRL-C too often, it takes time. */ | |
| 12776 if (--breakcheckcount == 0) | |
| 12777 { | |
| 12778 ui_breakcheck(); | |
| 12779 breakcheckcount = 1000; | |
| 12780 } | |
| 12781 } | |
| 12782 } | |
| 12783 } | |
| 12784 | |
| 12785 | |
| 12786 /* | |
| 12787 * Go one level deeper in the tree. | |
| 12788 */ | |
| 12789 static void | |
| 12790 go_deeper(stack, depth, score_add) | |
| 323 | 12791 trystate_T *stack; |
| 12792 int depth; | |
| 12793 int score_add; | |
| 12794 { | |
| 330 | 12795 stack[depth + 1] = stack[depth]; |
| 323 | 12796 stack[depth + 1].ts_state = STATE_START; |
| 625 | 12797 stack[depth + 1].ts_score = stack[depth].ts_score + score_add; |
| 323 | 12798 stack[depth + 1].ts_curi = 1; /* start just after length byte */ |
| 485 | 12799 stack[depth + 1].ts_flags = 0; |
| 323 | 12800 } |
| 12801 | |
| 455 | 12802 #ifdef FEAT_MBYTE |
| 12803 /* | |
| 12804 * Case-folding may change the number of bytes: Count nr of chars in | |
| 12805 * fword[flen] and return the byte length of that many chars in "word". | |
| 12806 */ | |
| 12807 static int | |
| 12808 nofold_len(fword, flen, word) | |
| 12809 char_u *fword; | |
| 12810 int flen; | |
| 12811 char_u *word; | |
| 12812 { | |
| 12813 char_u *p; | |
| 12814 int i = 0; | |
| 12815 | |
| 12816 for (p = fword; p < fword + flen; mb_ptr_adv(p)) | |
| 12817 ++i; | |
| 12818 for (p = word; i > 0; mb_ptr_adv(p)) | |
| 12819 --i; | |
| 12820 return (int)(p - word); | |
| 12821 } | |
| 12822 #endif | |
| 12823 | |
| 323 | 12824 /* |
| 12825 * "fword" is a good word with case folded. Find the matching keep-case | |
| 12826 * words and put it in "kword". | |
| 12827 * Theoretically there could be several keep-case words that result in the | |
| 12828 * same case-folded word, but we only find one... | |
| 12829 */ | |
| 12830 static void | |
| 12831 find_keepcap_word(slang, fword, kword) | |
| 12832 slang_T *slang; | |
| 12833 char_u *fword; | |
| 12834 char_u *kword; | |
| 12835 { | |
| 12836 char_u uword[MAXWLEN]; /* "fword" in upper-case */ | |
| 12837 int depth; | |
| 324 | 12838 idx_T tryidx; |
| 323 | 12839 |
| 12840 /* The following arrays are used at each depth in the tree. */ | |
| 324 | 12841 idx_T arridx[MAXWLEN]; |
| 323 | 12842 int round[MAXWLEN]; |
| 12843 int fwordidx[MAXWLEN]; | |
| 12844 int uwordidx[MAXWLEN]; | |
| 12845 int kwordlen[MAXWLEN]; | |
| 12846 | |
| 12847 int flen, ulen; | |
| 12848 int l; | |
| 12849 int len; | |
| 12850 int c; | |
| 324 | 12851 idx_T lo, hi, m; |
| 323 | 12852 char_u *p; |
| 12853 char_u *byts = slang->sl_kbyts; /* array with bytes of the words */ | |
| 324 | 12854 idx_T *idxs = slang->sl_kidxs; /* array with indexes */ |
| 323 | 12855 |
| 12856 if (byts == NULL) | |
| 12857 { | |
| 12858 /* array is empty: "cannot happen" */ | |
| 12859 *kword = NUL; | |
| 12860 return; | |
| 12861 } | |
| 12862 | |
| 12863 /* Make an all-cap version of "fword". */ | |
| 12864 allcap_copy(fword, uword); | |
| 12865 | |
| 12866 /* | |
| 12867 * Each character needs to be tried both case-folded and upper-case. | |
| 12868 * All this gets very complicated if we keep in mind that changing case | |
| 12869 * may change the byte length of a multi-byte character... | |
| 12870 */ | |
| 12871 depth = 0; | |
| 12872 arridx[0] = 0; | |
| 12873 round[0] = 0; | |
| 12874 fwordidx[0] = 0; | |
| 12875 uwordidx[0] = 0; | |
| 12876 kwordlen[0] = 0; | |
| 12877 while (depth >= 0) | |
| 12878 { | |
| 12879 if (fword[fwordidx[depth]] == NUL) | |
| 12880 { | |
| 12881 /* We are at the end of "fword". If the tree allows a word to end | |
| 12882 * here we have found a match. */ | |
| 12883 if (byts[arridx[depth] + 1] == 0) | |
| 12884 { | |
| 12885 kword[kwordlen[depth]] = NUL; | |
| 12886 return; | |
| 12887 } | |
| 12888 | |
| 12889 /* kword is getting too long, continue one level up */ | |
| 12890 --depth; | |
| 12891 } | |
| 12892 else if (++round[depth] > 2) | |
| 12893 { | |
| 12894 /* tried both fold-case and upper-case character, continue one | |
| 12895 * level up */ | |
| 12896 --depth; | |
| 12897 } | |
| 12898 else | |
| 12899 { | |
| 12900 /* | |
| 12901 * round[depth] == 1: Try using the folded-case character. | |
| 12902 * round[depth] == 2: Try using the upper-case character. | |
| 12903 */ | |
| 12904 #ifdef FEAT_MBYTE | |
| 12905 if (has_mbyte) | |
| 12906 { | |
| 474 | 12907 flen = mb_cptr2len(fword + fwordidx[depth]); |
| 12908 ulen = mb_cptr2len(uword + uwordidx[depth]); | |
| 323 | 12909 } |
| 12910 else | |
| 12911 #endif | |
| 12912 ulen = flen = 1; | |
| 12913 if (round[depth] == 1) | |
| 12914 { | |
| 12915 p = fword + fwordidx[depth]; | |
| 12916 l = flen; | |
| 12917 } | |
| 12918 else | |
| 12919 { | |
| 12920 p = uword + uwordidx[depth]; | |
| 12921 l = ulen; | |
| 12922 } | |
| 12923 | |
| 12924 for (tryidx = arridx[depth]; l > 0; --l) | |
| 12925 { | |
| 12926 /* Perform a binary search in the list of accepted bytes. */ | |
| 12927 len = byts[tryidx++]; | |
| 12928 c = *p++; | |
| 12929 lo = tryidx; | |
| 12930 hi = tryidx + len - 1; | |
| 12931 while (lo < hi) | |
| 12932 { | |
| 12933 m = (lo + hi) / 2; | |
| 12934 if (byts[m] > c) | |
| 12935 hi = m - 1; | |
| 12936 else if (byts[m] < c) | |
| 12937 lo = m + 1; | |
| 12938 else | |
| 12939 { | |
| 12940 lo = hi = m; | |
| 12941 break; | |
| 12942 } | |
| 12943 } | |
| 12944 | |
| 12945 /* Stop if there is no matching byte. */ | |
| 12946 if (hi < lo || byts[lo] != c) | |
| 12947 break; | |
| 12948 | |
| 12949 /* Continue at the child (if there is one). */ | |
| 12950 tryidx = idxs[lo]; | |
| 12951 } | |
| 12952 | |
| 12953 if (l == 0) | |
| 12954 { | |
| 12955 /* | |
| 12956 * Found the matching char. Copy it to "kword" and go a | |
| 12957 * level deeper. | |
| 12958 */ | |
| 12959 if (round[depth] == 1) | |
| 12960 { | |
| 12961 STRNCPY(kword + kwordlen[depth], fword + fwordidx[depth], | |
| 12962 flen); | |
| 12963 kwordlen[depth + 1] = kwordlen[depth] + flen; | |
| 12964 } | |
| 12965 else | |
| 12966 { | |
| 12967 STRNCPY(kword + kwordlen[depth], uword + uwordidx[depth], | |
| 12968 ulen); | |
| 12969 kwordlen[depth + 1] = kwordlen[depth] + ulen; | |
| 12970 } | |
| 12971 fwordidx[depth + 1] = fwordidx[depth] + flen; | |
| 12972 uwordidx[depth + 1] = uwordidx[depth] + ulen; | |
| 12973 | |
| 12974 ++depth; | |
| 12975 arridx[depth] = tryidx; | |
| 12976 round[depth] = 0; | |
| 12977 } | |
| 12978 } | |
| 12979 } | |
| 12980 | |
| 12981 /* Didn't find it: "cannot happen". */ | |
| 12982 *kword = NUL; | |
| 12983 } | |
| 12984 | |
| 12985 /* | |
| 344 | 12986 * Compute the sound-a-like score for suggestions in su->su_ga and add them to |
| 12987 * su->su_sga. | |
| 12988 */ | |
| 12989 static void | |
| 12990 score_comp_sal(su) | |
| 12991 suginfo_T *su; | |
| 12992 { | |
| 12993 langp_T *lp; | |
| 12994 char_u badsound[MAXWLEN]; | |
| 12995 int i; | |
| 12996 suggest_T *stp; | |
| 12997 suggest_T *sstp; | |
| 12998 int score; | |
| 500 | 12999 int lpi; |
| 344 | 13000 |
| 13001 if (ga_grow(&su->su_sga, su->su_ga.ga_len) == FAIL) | |
| 13002 return; | |
| 13003 | |
| 13004 /* Use the sound-folding of the first language that supports it. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13005 for (lpi = 0; lpi < curwin->w_s->b_langp.ga_len; ++lpi) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13006 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13007 lp = LANGP_ENTRY(curwin->w_s->b_langp, lpi); |
| 344 | 13008 if (lp->lp_slang->sl_sal.ga_len > 0) |
| 13009 { | |
| 13010 /* soundfold the bad word */ | |
| 375 | 13011 spell_soundfold(lp->lp_slang, su->su_fbadword, TRUE, badsound); |
| 344 | 13012 |
| 13013 for (i = 0; i < su->su_ga.ga_len; ++i) | |
| 13014 { | |
| 13015 stp = &SUG(su->su_ga, i); | |
| 13016 | |
| 351 | 13017 /* Case-fold the suggested word, sound-fold it and compute the |
| 13018 * sound-a-like score. */ | |
| 13019 score = stp_sal_score(stp, su, lp->lp_slang, badsound); | |
| 344 | 13020 if (score < SCORE_MAXMAX) |
| 13021 { | |
| 13022 /* Add the suggestion. */ | |
| 13023 sstp = &SUG(su->su_sga, su->su_sga.ga_len); | |
| 13024 sstp->st_word = vim_strsave(stp->st_word); | |
| 13025 if (sstp->st_word != NULL) | |
| 13026 { | |
| 625 | 13027 sstp->st_wordlen = stp->st_wordlen; |
| 344 | 13028 sstp->st_score = score; |
| 13029 sstp->st_altscore = 0; | |
| 13030 sstp->st_orglen = stp->st_orglen; | |
| 13031 ++su->su_sga.ga_len; | |
| 13032 } | |
| 13033 } | |
| 13034 } | |
| 13035 break; | |
| 13036 } | |
| 500 | 13037 } |
| 344 | 13038 } |
| 13039 | |
| 13040 /* | |
| 13041 * Combine the list of suggestions in su->su_ga and su->su_sga. | |
| 13042 * They are intwined. | |
| 13043 */ | |
| 13044 static void | |
| 13045 score_combine(su) | |
| 13046 suginfo_T *su; | |
| 13047 { | |
| 13048 int i; | |
| 13049 int j; | |
| 13050 garray_T ga; | |
| 13051 garray_T *gap; | |
| 13052 langp_T *lp; | |
| 13053 suggest_T *stp; | |
| 13054 char_u *p; | |
| 13055 char_u badsound[MAXWLEN]; | |
| 13056 int round; | |
| 500 | 13057 int lpi; |
| 625 | 13058 slang_T *slang = NULL; |
| 344 | 13059 |
| 13060 /* Add the alternate score to su_ga. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13061 for (lpi = 0; lpi < curwin->w_s->b_langp.ga_len; ++lpi) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13062 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13063 lp = LANGP_ENTRY(curwin->w_s->b_langp, lpi); |
| 344 | 13064 if (lp->lp_slang->sl_sal.ga_len > 0) |
| 13065 { | |
| 13066 /* soundfold the bad word */ | |
| 625 | 13067 slang = lp->lp_slang; |
| 13068 spell_soundfold(slang, su->su_fbadword, TRUE, badsound); | |
| 344 | 13069 |
| 13070 for (i = 0; i < su->su_ga.ga_len; ++i) | |
| 13071 { | |
| 13072 stp = &SUG(su->su_ga, i); | |
| 625 | 13073 stp->st_altscore = stp_sal_score(stp, su, slang, badsound); |
| 344 | 13074 if (stp->st_altscore == SCORE_MAXMAX) |
| 13075 stp->st_score = (stp->st_score * 3 + SCORE_BIG) / 4; | |
| 13076 else | |
| 13077 stp->st_score = (stp->st_score * 3 | |
| 13078 + stp->st_altscore) / 4; | |
| 13079 stp->st_salscore = FALSE; | |
| 13080 } | |
| 13081 break; | |
| 13082 } | |
| 13083 } | |
| 13084 | |
| 844 | 13085 if (slang == NULL) /* Using "double" without sound folding. */ |
| 13086 { | |
| 13087 (void)cleanup_suggestions(&su->su_ga, su->su_maxscore, | |
| 13088 su->su_maxcount); | |
| 625 | 13089 return; |
| 844 | 13090 } |
| 625 | 13091 |
| 344 | 13092 /* Add the alternate score to su_sga. */ |
| 13093 for (i = 0; i < su->su_sga.ga_len; ++i) | |
| 13094 { | |
| 13095 stp = &SUG(su->su_sga, i); | |
| 625 | 13096 stp->st_altscore = spell_edit_score(slang, |
| 13097 su->su_badword, stp->st_word); | |
| 344 | 13098 if (stp->st_score == SCORE_MAXMAX) |
| 13099 stp->st_score = (SCORE_BIG * 7 + stp->st_altscore) / 8; | |
| 13100 else | |
| 13101 stp->st_score = (stp->st_score * 7 + stp->st_altscore) / 8; | |
| 13102 stp->st_salscore = TRUE; | |
| 13103 } | |
| 13104 | |
| 625 | 13105 /* Remove bad suggestions, sort the suggestions and truncate at "maxcount" |
| 13106 * for both lists. */ | |
| 13107 check_suggestions(su, &su->su_ga); | |
| 344 | 13108 (void)cleanup_suggestions(&su->su_ga, su->su_maxscore, su->su_maxcount); |
| 625 | 13109 check_suggestions(su, &su->su_sga); |
| 344 | 13110 (void)cleanup_suggestions(&su->su_sga, su->su_maxscore, su->su_maxcount); |
| 13111 | |
| 13112 ga_init2(&ga, (int)sizeof(suginfo_T), 1); | |
| 13113 if (ga_grow(&ga, su->su_ga.ga_len + su->su_sga.ga_len) == FAIL) | |
| 13114 return; | |
| 13115 | |
| 13116 stp = &SUG(ga, 0); | |
| 13117 for (i = 0; i < su->su_ga.ga_len || i < su->su_sga.ga_len; ++i) | |
| 13118 { | |
| 13119 /* round 1: get a suggestion from su_ga | |
| 13120 * round 2: get a suggestion from su_sga */ | |
| 13121 for (round = 1; round <= 2; ++round) | |
| 13122 { | |
| 13123 gap = round == 1 ? &su->su_ga : &su->su_sga; | |
| 13124 if (i < gap->ga_len) | |
| 13125 { | |
| 13126 /* Don't add a word if it's already there. */ | |
| 13127 p = SUG(*gap, i).st_word; | |
| 13128 for (j = 0; j < ga.ga_len; ++j) | |
| 13129 if (STRCMP(stp[j].st_word, p) == 0) | |
| 13130 break; | |
| 13131 if (j == ga.ga_len) | |
| 13132 stp[ga.ga_len++] = SUG(*gap, i); | |
| 13133 else | |
| 13134 vim_free(p); | |
| 13135 } | |
| 13136 } | |
| 13137 } | |
| 13138 | |
| 13139 ga_clear(&su->su_ga); | |
| 13140 ga_clear(&su->su_sga); | |
| 13141 | |
| 13142 /* Truncate the list to the number of suggestions that will be displayed. */ | |
| 13143 if (ga.ga_len > su->su_maxcount) | |
| 13144 { | |
| 13145 for (i = su->su_maxcount; i < ga.ga_len; ++i) | |
| 13146 vim_free(stp[i].st_word); | |
| 13147 ga.ga_len = su->su_maxcount; | |
| 13148 } | |
| 13149 | |
| 13150 su->su_ga = ga; | |
| 13151 } | |
| 13152 | |
| 13153 /* | |
| 351 | 13154 * For the goodword in "stp" compute the soundalike score compared to the |
| 13155 * badword. | |
| 13156 */ | |
| 13157 static int | |
| 13158 stp_sal_score(stp, su, slang, badsound) | |
| 13159 suggest_T *stp; | |
| 13160 suginfo_T *su; | |
| 13161 slang_T *slang; | |
| 13162 char_u *badsound; /* sound-folded badword */ | |
| 13163 { | |
| 13164 char_u *p; | |
| 534 | 13165 char_u *pbad; |
| 13166 char_u *pgood; | |
| 351 | 13167 char_u badsound2[MAXWLEN]; |
| 13168 char_u fword[MAXWLEN]; | |
| 13169 char_u goodsound[MAXWLEN]; | |
| 534 | 13170 char_u goodword[MAXWLEN]; |
| 13171 int lendiff; | |
| 13172 | |
| 13173 lendiff = (int)(su->su_badlen - stp->st_orglen); | |
| 13174 if (lendiff >= 0) | |
| 13175 pbad = badsound; | |
| 351 | 13176 else |
| 13177 { | |
| 13178 /* soundfold the bad word with more characters following */ | |
| 13179 (void)spell_casefold(su->su_badptr, stp->st_orglen, fword, MAXWLEN); | |
| 13180 | |
| 13181 /* When joining two words the sound often changes a lot. E.g., "t he" | |
| 13182 * sounds like "t h" while "the" sounds like "@". Avoid that by | |
| 13183 * removing the space. Don't do it when the good word also contains a | |
| 13184 * space. */ | |
| 13185 if (vim_iswhite(su->su_badptr[su->su_badlen]) | |
| 13186 && *skiptowhite(stp->st_word) == NUL) | |
| 13187 for (p = fword; *(p = skiptowhite(p)) != NUL; ) | |
| 1619 | 13188 STRMOVE(p, p + 1); |
| 351 | 13189 |
| 375 | 13190 spell_soundfold(slang, fword, TRUE, badsound2); |
| 534 | 13191 pbad = badsound2; |
| 13192 } | |
| 13193 | |
| 2768 | 13194 if (lendiff > 0 && stp->st_wordlen + lendiff < MAXWLEN) |
| 534 | 13195 { |
| 13196 /* Add part of the bad word to the good word, so that we soundfold | |
| 13197 * what replaces the bad word. */ | |
| 13198 STRCPY(goodword, stp->st_word); | |
| 625 | 13199 vim_strncpy(goodword + stp->st_wordlen, |
| 13200 su->su_badptr + su->su_badlen - lendiff, lendiff); | |
| 534 | 13201 pgood = goodword; |
| 13202 } | |
| 13203 else | |
| 13204 pgood = stp->st_word; | |
| 351 | 13205 |
| 375 | 13206 /* Sound-fold the word and compute the score for the difference. */ |
| 534 | 13207 spell_soundfold(slang, pgood, FALSE, goodsound); |
| 13208 | |
| 13209 return soundalike_score(goodsound, pbad); | |
| 351 | 13210 } |
| 13211 | |
| 625 | 13212 /* structure used to store soundfolded words that add_sound_suggest() has |
| 13213 * handled already. */ | |
| 13214 typedef struct | |
| 13215 { | |
| 13216 short sft_score; /* lowest score used */ | |
| 13217 char_u sft_word[1]; /* soundfolded word, actually longer */ | |
| 13218 } sftword_T; | |
| 13219 | |
| 13220 static sftword_T dumsft; | |
| 13221 #define HIKEY2SFT(p) ((sftword_T *)(p - (dumsft.sft_word - (char_u *)&dumsft))) | |
| 13222 #define HI2SFT(hi) HIKEY2SFT((hi)->hi_key) | |
| 13223 | |
| 13224 /* | |
| 13225 * Prepare for calling suggest_try_soundalike(). | |
| 13226 */ | |
| 13227 static void | |
| 13228 suggest_try_soundalike_prep() | |
| 13229 { | |
| 13230 langp_T *lp; | |
| 13231 int lpi; | |
| 13232 slang_T *slang; | |
| 13233 | |
| 13234 /* Do this for all languages that support sound folding and for which a | |
| 13235 * .sug file has been loaded. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13236 for (lpi = 0; lpi < curwin->w_s->b_langp.ga_len; ++lpi) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13237 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13238 lp = LANGP_ENTRY(curwin->w_s->b_langp, lpi); |
| 625 | 13239 slang = lp->lp_slang; |
| 13240 if (slang->sl_sal.ga_len > 0 && slang->sl_sbyts != NULL) | |
| 13241 /* prepare the hashtable used by add_sound_suggest() */ | |
| 13242 hash_init(&slang->sl_sounddone); | |
| 13243 } | |
| 13244 } | |
| 13245 | |
| 351 | 13246 /* |
| 323 | 13247 * Find suggestions by comparing the word in a sound-a-like form. |
| 503 | 13248 * Note: This doesn't support postponed prefixes. |
| 323 | 13249 */ |
| 13250 static void | |
| 346 | 13251 suggest_try_soundalike(su) |
| 323 | 13252 suginfo_T *su; |
| 13253 { | |
| 13254 char_u salword[MAXWLEN]; | |
| 13255 langp_T *lp; | |
| 500 | 13256 int lpi; |
| 503 | 13257 slang_T *slang; |
| 344 | 13258 |
| 625 | 13259 /* Do this for all languages that support sound folding and for which a |
| 13260 * .sug file has been loaded. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13261 for (lpi = 0; lpi < curwin->w_s->b_langp.ga_len; ++lpi) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13262 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13263 lp = LANGP_ENTRY(curwin->w_s->b_langp, lpi); |
| 625 | 13264 slang = lp->lp_slang; |
| 13265 if (slang->sl_sal.ga_len > 0 && slang->sl_sbyts != NULL) | |
| 13266 { | |
| 13267 /* soundfold the bad word */ | |
| 13268 spell_soundfold(slang, su->su_fbadword, TRUE, salword); | |
| 13269 | |
| 13270 /* try all kinds of inserts/deletes/swaps/etc. */ | |
| 13271 /* TODO: also soundfold the next words, so that we can try joining | |
| 13272 * and splitting */ | |
| 13273 suggest_trie_walk(su, lp, salword, TRUE); | |
| 13274 } | |
| 13275 } | |
| 13276 } | |
| 13277 | |
| 13278 /* | |
| 13279 * Finish up after calling suggest_try_soundalike(). | |
| 13280 */ | |
| 13281 static void | |
| 13282 suggest_try_soundalike_finish() | |
| 13283 { | |
| 13284 langp_T *lp; | |
| 13285 int lpi; | |
| 13286 slang_T *slang; | |
| 13287 int todo; | |
| 13288 hashitem_T *hi; | |
| 13289 | |
| 13290 /* Do this for all languages that support sound folding and for which a | |
| 13291 * .sug file has been loaded. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13292 for (lpi = 0; lpi < curwin->w_s->b_langp.ga_len; ++lpi) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13293 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
13294 lp = LANGP_ENTRY(curwin->w_s->b_langp, lpi); |
| 503 | 13295 slang = lp->lp_slang; |
| 625 | 13296 if (slang->sl_sal.ga_len > 0 && slang->sl_sbyts != NULL) |
| 13297 { | |
| 13298 /* Free the info about handled words. */ | |
| 835 | 13299 todo = (int)slang->sl_sounddone.ht_used; |
| 625 | 13300 for (hi = slang->sl_sounddone.ht_array; todo > 0; ++hi) |
| 13301 if (!HASHITEM_EMPTY(hi)) | |
| 13302 { | |
| 13303 vim_free(HI2SFT(hi)); | |
| 13304 --todo; | |
| 13305 } | |
| 1087 | 13306 |
| 13307 /* Clear the hashtable, it may also be used by another region. */ | |
| 625 | 13308 hash_clear(&slang->sl_sounddone); |
| 1087 | 13309 hash_init(&slang->sl_sounddone); |
| 625 | 13310 } |
| 13311 } | |
| 13312 } | |
| 13313 | |
| 13314 /* | |
| 13315 * A match with a soundfolded word is found. Add the good word(s) that | |
| 13316 * produce this soundfolded word. | |
| 13317 */ | |
| 13318 static void | |
| 13319 add_sound_suggest(su, goodword, score, lp) | |
| 13320 suginfo_T *su; | |
| 13321 char_u *goodword; | |
| 13322 int score; /* soundfold score */ | |
| 13323 langp_T *lp; | |
| 13324 { | |
| 13325 slang_T *slang = lp->lp_slang; /* language for sound folding */ | |
| 13326 int sfwordnr; | |
| 13327 char_u *nrline; | |
| 13328 int orgnr; | |
| 13329 char_u theword[MAXWLEN]; | |
| 13330 int i; | |
| 13331 int wlen; | |
| 13332 char_u *byts; | |
| 13333 idx_T *idxs; | |
| 13334 int n; | |
| 13335 int wordcount; | |
| 13336 int wc; | |
| 13337 int goodscore; | |
| 13338 hash_T hash; | |
| 13339 hashitem_T *hi; | |
| 13340 sftword_T *sft; | |
| 13341 int bc, gc; | |
| 13342 int limit; | |
| 13343 | |
| 13344 /* | |
| 13345 * It's very well possible that the same soundfold word is found several | |
| 13346 * times with different scores. Since the following is quite slow only do | |
| 13347 * the words that have a better score than before. Use a hashtable to | |
| 13348 * remember the words that have been done. | |
| 13349 */ | |
| 13350 hash = hash_hash(goodword); | |
| 13351 hi = hash_lookup(&slang->sl_sounddone, goodword, hash); | |
| 13352 if (HASHITEM_EMPTY(hi)) | |
| 13353 { | |
| 844 | 13354 sft = (sftword_T *)alloc((unsigned)(sizeof(sftword_T) |
| 13355 + STRLEN(goodword))); | |
| 625 | 13356 if (sft != NULL) |
| 13357 { | |
| 13358 sft->sft_score = score; | |
| 13359 STRCPY(sft->sft_word, goodword); | |
| 13360 hash_add_item(&slang->sl_sounddone, hi, sft->sft_word, hash); | |
| 13361 } | |
| 13362 } | |
| 13363 else | |
| 13364 { | |
| 13365 sft = HI2SFT(hi); | |
| 13366 if (score >= sft->sft_score) | |
| 13367 return; | |
| 13368 sft->sft_score = score; | |
| 13369 } | |
| 13370 | |
| 13371 /* | |
| 13372 * Find the word nr in the soundfold tree. | |
| 13373 */ | |
| 13374 sfwordnr = soundfold_find(slang, goodword); | |
| 13375 if (sfwordnr < 0) | |
| 13376 { | |
| 13377 EMSG2(_(e_intern2), "add_sound_suggest()"); | |
| 13378 return; | |
| 13379 } | |
| 13380 | |
| 13381 /* | |
| 13382 * go over the list of good words that produce this soundfold word | |
| 13383 */ | |
| 13384 nrline = ml_get_buf(slang->sl_sugbuf, (linenr_T)(sfwordnr + 1), FALSE); | |
| 13385 orgnr = 0; | |
| 13386 while (*nrline != NUL) | |
| 13387 { | |
| 13388 /* The wordnr was stored in a minimal nr of bytes as an offset to the | |
| 13389 * previous wordnr. */ | |
| 13390 orgnr += bytes2offset(&nrline); | |
| 13391 | |
| 13392 byts = slang->sl_fbyts; | |
| 13393 idxs = slang->sl_fidxs; | |
| 13394 | |
| 13395 /* Lookup the word "orgnr" one of the two tries. */ | |
| 13396 n = 0; | |
| 13397 wlen = 0; | |
| 13398 wordcount = 0; | |
| 13399 for (;;) | |
| 13400 { | |
| 13401 i = 1; | |
| 13402 if (wordcount == orgnr && byts[n + 1] == NUL) | |
| 13403 break; /* found end of word */ | |
| 13404 | |
| 13405 if (byts[n + 1] == NUL) | |
| 13406 ++wordcount; | |
| 13407 | |
| 13408 /* skip over the NUL bytes */ | |
| 13409 for ( ; byts[n + i] == NUL; ++i) | |
| 13410 if (i > byts[n]) /* safety check */ | |
| 13411 { | |
| 13412 STRCPY(theword + wlen, "BAD"); | |
| 13413 goto badword; | |
| 13414 } | |
| 13415 | |
| 13416 /* One of the siblings must have the word. */ | |
| 13417 for ( ; i < byts[n]; ++i) | |
| 13418 { | |
| 13419 wc = idxs[idxs[n + i]]; /* nr of words under this byte */ | |
| 13420 if (wordcount + wc > orgnr) | |
| 13421 break; | |
| 13422 wordcount += wc; | |
| 13423 } | |
| 13424 | |
| 13425 theword[wlen++] = byts[n + i]; | |
| 13426 n = idxs[n + i]; | |
| 13427 } | |
| 13428 badword: | |
| 13429 theword[wlen] = NUL; | |
| 13430 | |
| 13431 /* Go over the possible flags and regions. */ | |
| 13432 for (; i <= byts[n] && byts[n + i] == NUL; ++i) | |
| 13433 { | |
| 13434 char_u cword[MAXWLEN]; | |
| 13435 char_u *p; | |
| 13436 int flags = (int)idxs[n + i]; | |
| 13437 | |
| 698 | 13438 /* Skip words with the NOSUGGEST flag */ |
| 13439 if (flags & WF_NOSUGGEST) | |
| 13440 continue; | |
| 13441 | |
| 625 | 13442 if (flags & WF_KEEPCAP) |
| 13443 { | |
| 13444 /* Must find the word in the keep-case tree. */ | |
| 13445 find_keepcap_word(slang, theword, cword); | |
| 13446 p = cword; | |
| 13447 } | |
| 13448 else | |
| 13449 { | |
| 13450 flags |= su->su_badflags; | |
| 13451 if ((flags & WF_CAPMASK) != 0) | |
| 13452 { | |
| 13453 /* Need to fix case according to "flags". */ | |
| 13454 make_case_word(theword, cword, flags); | |
| 13455 p = cword; | |
| 323 | 13456 } |
| 13457 else | |
| 625 | 13458 p = theword; |
| 13459 } | |
| 13460 | |
| 13461 /* Add the suggestion. */ | |
| 13462 if (sps_flags & SPS_DOUBLE) | |
| 13463 { | |
| 13464 /* Add the suggestion if the score isn't too bad. */ | |
| 13465 if (score <= su->su_maxscore) | |
| 13466 add_suggestion(su, &su->su_sga, p, su->su_badlen, | |
| 13467 score, 0, FALSE, slang, FALSE); | |
| 13468 } | |
| 13469 else | |
| 13470 { | |
| 13471 /* Add a penalty for words in another region. */ | |
| 13472 if ((flags & WF_REGION) | |
| 13473 && (((unsigned)flags >> 16) & lp->lp_region) == 0) | |
| 13474 goodscore = SCORE_REGION; | |
| 13475 else | |
| 13476 goodscore = 0; | |
| 13477 | |
| 13478 /* Add a small penalty for changing the first letter from | |
| 13479 * lower to upper case. Helps for "tath" -> "Kath", which is | |
| 13480 * less common thatn "tath" -> "path". Don't do it when the | |
| 13481 * letter is the same, that has already been counted. */ | |
| 13482 gc = PTR2CHAR(p); | |
| 13483 if (SPELL_ISUPPER(gc)) | |
| 13484 { | |
| 13485 bc = PTR2CHAR(su->su_badword); | |
| 13486 if (!SPELL_ISUPPER(bc) | |
| 13487 && SPELL_TOFOLD(bc) != SPELL_TOFOLD(gc)) | |
| 13488 goodscore += SCORE_ICASE / 2; | |
| 13489 } | |
| 13490 | |
| 13491 /* Compute the score for the good word. This only does letter | |
| 13492 * insert/delete/swap/replace. REP items are not considered, | |
| 13493 * which may make the score a bit higher. | |
| 13494 * Use a limit for the score to make it work faster. Use | |
| 13495 * MAXSCORE(), because RESCORE() will change the score. | |
| 13496 * If the limit is very high then the iterative method is | |
| 13497 * inefficient, using an array is quicker. */ | |
| 13498 limit = MAXSCORE(su->su_sfmaxscore - goodscore, score); | |
| 13499 if (limit > SCORE_LIMITMAX) | |
| 13500 goodscore += spell_edit_score(slang, su->su_badword, p); | |
| 13501 else | |
| 13502 goodscore += spell_edit_score_limit(slang, su->su_badword, | |
| 13503 p, limit); | |
| 13504 | |
| 13505 /* When going over the limit don't bother to do the rest. */ | |
| 13506 if (goodscore < SCORE_MAXMAX) | |
| 13507 { | |
| 13508 /* Give a bonus to words seen before. */ | |
| 13509 goodscore = score_wordcount_adj(slang, goodscore, p, FALSE); | |
| 13510 | |
| 13511 /* Add the suggestion if the score isn't too bad. */ | |
| 13512 goodscore = RESCORE(goodscore, score); | |
| 13513 if (goodscore <= su->su_sfmaxscore) | |
| 13514 add_suggestion(su, &su->su_ga, p, su->su_badlen, | |
| 13515 goodscore, score, TRUE, slang, TRUE); | |
| 13516 } | |
| 13517 } | |
| 13518 } | |
| 13519 /* smsg("word %s (%d): %s (%d)", sftword, sftnr, theword, orgnr); */ | |
| 13520 } | |
| 13521 } | |
| 13522 | |
| 13523 /* | |
| 13524 * Find word "word" in fold-case tree for "slang" and return the word number. | |
| 13525 */ | |
| 13526 static int | |
| 13527 soundfold_find(slang, word) | |
| 13528 slang_T *slang; | |
| 13529 char_u *word; | |
| 13530 { | |
| 13531 idx_T arridx = 0; | |
| 13532 int len; | |
| 13533 int wlen = 0; | |
| 13534 int c; | |
| 13535 char_u *ptr = word; | |
| 13536 char_u *byts; | |
| 13537 idx_T *idxs; | |
| 13538 int wordnr = 0; | |
| 13539 | |
| 13540 byts = slang->sl_sbyts; | |
| 13541 idxs = slang->sl_sidxs; | |
| 13542 | |
| 13543 for (;;) | |
| 13544 { | |
| 13545 /* First byte is the number of possible bytes. */ | |
| 13546 len = byts[arridx++]; | |
| 13547 | |
| 13548 /* If the first possible byte is a zero the word could end here. | |
| 13549 * If the word ends we found the word. If not skip the NUL bytes. */ | |
| 13550 c = ptr[wlen]; | |
| 13551 if (byts[arridx] == NUL) | |
| 13552 { | |
| 13553 if (c == NUL) | |
| 13554 break; | |
| 13555 | |
| 13556 /* Skip over the zeros, there can be several. */ | |
| 13557 while (len > 0 && byts[arridx] == NUL) | |
| 13558 { | |
| 13559 ++arridx; | |
| 13560 --len; | |
| 13561 } | |
| 13562 if (len == 0) | |
| 13563 return -1; /* no children, word should have ended here */ | |
| 13564 ++wordnr; | |
| 13565 } | |
| 13566 | |
| 13567 /* If the word ends we didn't find it. */ | |
| 13568 if (c == NUL) | |
| 13569 return -1; | |
| 13570 | |
| 13571 /* Perform a binary search in the list of accepted bytes. */ | |
| 13572 if (c == TAB) /* <Tab> is handled like <Space> */ | |
| 13573 c = ' '; | |
| 13574 while (byts[arridx] < c) | |
| 13575 { | |
| 13576 /* The word count is in the first idxs[] entry of the child. */ | |
| 13577 wordnr += idxs[idxs[arridx]]; | |
| 13578 ++arridx; | |
| 13579 if (--len == 0) /* end of the bytes, didn't find it */ | |
| 13580 return -1; | |
| 13581 } | |
| 13582 if (byts[arridx] != c) /* didn't find the byte */ | |
| 13583 return -1; | |
| 13584 | |
| 13585 /* Continue at the child (if there is one). */ | |
| 13586 arridx = idxs[arridx]; | |
| 13587 ++wlen; | |
| 13588 | |
| 13589 /* One space in the good word may stand for several spaces in the | |
| 13590 * checked word. */ | |
| 13591 if (c == ' ') | |
| 13592 while (ptr[wlen] == ' ' || ptr[wlen] == TAB) | |
| 13593 ++wlen; | |
| 13594 } | |
| 13595 | |
| 13596 return wordnr; | |
| 323 | 13597 } |
| 13598 | |
| 13599 /* | |
| 324 | 13600 * Copy "fword" to "cword", fixing case according to "flags". |
| 323 | 13601 */ |
| 13602 static void | |
| 13603 make_case_word(fword, cword, flags) | |
| 13604 char_u *fword; | |
| 13605 char_u *cword; | |
| 13606 int flags; | |
| 13607 { | |
| 13608 if (flags & WF_ALLCAP) | |
| 13609 /* Make it all upper-case */ | |
| 13610 allcap_copy(fword, cword); | |
| 13611 else if (flags & WF_ONECAP) | |
| 13612 /* Make the first letter upper-case */ | |
| 324 | 13613 onecap_copy(fword, cword, TRUE); |
| 323 | 13614 else |
| 13615 /* Use goodword as-is. */ | |
| 13616 STRCPY(cword, fword); | |
| 13617 } | |
| 13618 | |
| 330 | 13619 /* |
| 13620 * Use map string "map" for languages "lp". | |
| 13621 */ | |
| 13622 static void | |
| 13623 set_map_str(lp, map) | |
| 13624 slang_T *lp; | |
| 13625 char_u *map; | |
| 13626 { | |
| 13627 char_u *p; | |
| 13628 int headc = 0; | |
| 13629 int c; | |
| 13630 int i; | |
| 13631 | |
| 13632 if (*map == NUL) | |
| 13633 { | |
| 13634 lp->sl_has_map = FALSE; | |
| 13635 return; | |
| 13636 } | |
| 13637 lp->sl_has_map = TRUE; | |
| 13638 | |
| 625 | 13639 /* Init the array and hash tables empty. */ |
| 330 | 13640 for (i = 0; i < 256; ++i) |
| 13641 lp->sl_map_array[i] = 0; | |
| 13642 #ifdef FEAT_MBYTE | |
| 13643 hash_init(&lp->sl_map_hash); | |
| 13644 #endif | |
| 13645 | |
| 13646 /* | |
| 13647 * The similar characters are stored separated with slashes: | |
| 13648 * "aaa/bbb/ccc/". Fill sl_map_array[c] with the character before c and | |
| 13649 * before the same slash. For characters above 255 sl_map_hash is used. | |
| 13650 */ | |
| 13651 for (p = map; *p != NUL; ) | |
| 13652 { | |
| 13653 #ifdef FEAT_MBYTE | |
| 474 | 13654 c = mb_cptr2char_adv(&p); |
| 330 | 13655 #else |
| 13656 c = *p++; | |
| 13657 #endif | |
| 13658 if (c == '/') | |
| 13659 headc = 0; | |
| 13660 else | |
| 13661 { | |
| 13662 if (headc == 0) | |
| 13663 headc = c; | |
| 13664 | |
| 13665 #ifdef FEAT_MBYTE | |
| 13666 /* Characters above 255 don't fit in sl_map_array[], put them in | |
| 13667 * the hash table. Each entry is the char, a NUL the headchar and | |
| 13668 * a NUL. */ | |
| 13669 if (c >= 256) | |
| 13670 { | |
| 13671 int cl = mb_char2len(c); | |
| 13672 int headcl = mb_char2len(headc); | |
| 13673 char_u *b; | |
| 13674 hash_T hash; | |
| 13675 hashitem_T *hi; | |
| 13676 | |
| 13677 b = alloc((unsigned)(cl + headcl + 2)); | |
| 13678 if (b == NULL) | |
| 13679 return; | |
| 13680 mb_char2bytes(c, b); | |
| 13681 b[cl] = NUL; | |
| 13682 mb_char2bytes(headc, b + cl + 1); | |
| 13683 b[cl + 1 + headcl] = NUL; | |
| 13684 hash = hash_hash(b); | |
| 13685 hi = hash_lookup(&lp->sl_map_hash, b, hash); | |
| 13686 if (HASHITEM_EMPTY(hi)) | |
| 13687 hash_add_item(&lp->sl_map_hash, hi, b, hash); | |
| 13688 else | |
| 13689 { | |
| 13690 /* This should have been checked when generating the .spl | |
| 13691 * file. */ | |
| 626 | 13692 EMSG(_("E783: duplicate char in MAP entry")); |
| 330 | 13693 vim_free(b); |
| 13694 } | |
| 13695 } | |
| 13696 else | |
| 13697 #endif | |
| 13698 lp->sl_map_array[c] = headc; | |
| 13699 } | |
| 13700 } | |
| 13701 } | |
| 13702 | |
| 323 | 13703 /* |
| 13704 * Return TRUE if "c1" and "c2" are similar characters according to the MAP | |
| 13705 * lines in the .aff file. | |
| 13706 */ | |
| 13707 static int | |
| 13708 similar_chars(slang, c1, c2) | |
| 13709 slang_T *slang; | |
| 13710 int c1; | |
| 13711 int c2; | |
| 13712 { | |
| 330 | 13713 int m1, m2; |
| 13714 #ifdef FEAT_MBYTE | |
| 13715 char_u buf[MB_MAXBYTES]; | |
| 13716 hashitem_T *hi; | |
| 13717 | |
| 13718 if (c1 >= 256) | |
| 13719 { | |
| 13720 buf[mb_char2bytes(c1, buf)] = 0; | |
| 13721 hi = hash_find(&slang->sl_map_hash, buf); | |
| 13722 if (HASHITEM_EMPTY(hi)) | |
| 13723 m1 = 0; | |
| 13724 else | |
| 13725 m1 = mb_ptr2char(hi->hi_key + STRLEN(hi->hi_key) + 1); | |
| 13726 } | |
| 13727 else | |
| 13728 #endif | |
| 13729 m1 = slang->sl_map_array[c1]; | |
| 13730 if (m1 == 0) | |
| 323 | 13731 return FALSE; |
| 330 | 13732 |
| 13733 | |
| 13734 #ifdef FEAT_MBYTE | |
| 13735 if (c2 >= 256) | |
| 13736 { | |
| 13737 buf[mb_char2bytes(c2, buf)] = 0; | |
| 13738 hi = hash_find(&slang->sl_map_hash, buf); | |
| 13739 if (HASHITEM_EMPTY(hi)) | |
| 13740 m2 = 0; | |
| 13741 else | |
| 13742 m2 = mb_ptr2char(hi->hi_key + STRLEN(hi->hi_key) + 1); | |
| 13743 } | |
| 13744 else | |
| 13745 #endif | |
| 13746 m2 = slang->sl_map_array[c2]; | |
| 13747 | |
| 13748 return m1 == m2; | |
| 323 | 13749 } |
| 13750 | |
| 13751 /* | |
| 13752 * Add a suggestion to the list of suggestions. | |
| 625 | 13753 * For a suggestion that is already in the list the lowest score is remembered. |
| 13754 */ | |
| 13755 static void | |
| 13756 add_suggestion(su, gap, goodword, badlenarg, score, altscore, had_bonus, | |
| 13757 slang, maxsf) | |
| 323 | 13758 suginfo_T *su; |
| 625 | 13759 garray_T *gap; /* either su_ga or su_sga */ |
| 323 | 13760 char_u *goodword; |
| 534 | 13761 int badlenarg; /* len of bad word replaced with "goodword" */ |
| 324 | 13762 int score; |
| 351 | 13763 int altscore; |
| 344 | 13764 int had_bonus; /* value for st_had_bonus */ |
| 503 | 13765 slang_T *slang; /* language for sound folding */ |
| 625 | 13766 int maxsf; /* su_maxscore applies to soundfold score, |
| 13767 su_sfmaxscore to the total score. */ | |
| 13768 { | |
| 13769 int goodlen; /* len of goodword changed */ | |
| 13770 int badlen; /* len of bad word changed */ | |
| 323 | 13771 suggest_T *stp; |
| 534 | 13772 suggest_T new_sug; |
| 323 | 13773 int i; |
| 534 | 13774 char_u *pgood, *pbad; |
| 532 | 13775 |
| 534 | 13776 /* Minimize "badlen" for consistency. Avoids that changing "the the" to |
| 13777 * "thee the" is added next to changing the first "the" the "thee". */ | |
| 13778 pgood = goodword + STRLEN(goodword); | |
| 625 | 13779 pbad = su->su_badptr + badlenarg; |
| 13780 for (;;) | |
| 13781 { | |
| 835 | 13782 goodlen = (int)(pgood - goodword); |
| 13783 badlen = (int)(pbad - su->su_badptr); | |
| 625 | 13784 if (goodlen <= 0 || badlen <= 0) |
| 13785 break; | |
| 534 | 13786 mb_ptr_back(goodword, pgood); |
| 13787 mb_ptr_back(su->su_badptr, pbad); | |
| 13788 #ifdef FEAT_MBYTE | |
| 13789 if (has_mbyte) | |
| 13790 { | |
| 13791 if (mb_ptr2char(pgood) != mb_ptr2char(pbad)) | |
| 13792 break; | |
| 346 | 13793 } |
| 13794 else | |
| 534 | 13795 #endif |
| 13796 if (*pgood != *pbad) | |
| 13797 break; | |
| 625 | 13798 } |
| 13799 | |
| 534 | 13800 if (badlen == 0 && goodlen == 0) |
| 13801 /* goodword doesn't change anything; may happen for "the the" changing | |
| 13802 * the first "the" to itself. */ | |
| 13803 return; | |
| 346 | 13804 |
| 944 | 13805 if (gap->ga_len == 0) |
| 13806 i = -1; | |
| 13807 else | |
| 13808 { | |
| 13809 /* Check if the word is already there. Also check the length that is | |
| 13810 * being replaced "thes," -> "these" is a different suggestion from | |
| 13811 * "thes" -> "these". */ | |
| 13812 stp = &SUG(*gap, 0); | |
| 13813 for (i = gap->ga_len; --i >= 0; ++stp) | |
| 13814 if (stp->st_wordlen == goodlen | |
| 13815 && stp->st_orglen == badlen | |
| 13816 && STRNCMP(stp->st_word, goodword, goodlen) == 0) | |
| 13817 { | |
| 13818 /* | |
| 13819 * Found it. Remember the word with the lowest score. | |
| 13820 */ | |
| 13821 if (stp->st_slang == NULL) | |
| 13822 stp->st_slang = slang; | |
| 13823 | |
| 13824 new_sug.st_score = score; | |
| 13825 new_sug.st_altscore = altscore; | |
| 13826 new_sug.st_had_bonus = had_bonus; | |
| 13827 | |
| 13828 if (stp->st_had_bonus != had_bonus) | |
| 13829 { | |
| 13830 /* Only one of the two had the soundalike score computed. | |
| 13831 * Need to do that for the other one now, otherwise the | |
| 13832 * scores can't be compared. This happens because | |
| 13833 * suggest_try_change() doesn't compute the soundalike | |
| 13834 * word to keep it fast, while some special methods set | |
| 13835 * the soundalike score to zero. */ | |
| 13836 if (had_bonus) | |
| 13837 rescore_one(su, stp); | |
| 13838 else | |
| 13839 { | |
| 13840 new_sug.st_word = stp->st_word; | |
| 13841 new_sug.st_wordlen = stp->st_wordlen; | |
| 13842 new_sug.st_slang = stp->st_slang; | |
| 13843 new_sug.st_orglen = badlen; | |
| 13844 rescore_one(su, &new_sug); | |
| 13845 } | |
| 13846 } | |
| 13847 | |
| 13848 if (stp->st_score > new_sug.st_score) | |
| 13849 { | |
| 13850 stp->st_score = new_sug.st_score; | |
| 13851 stp->st_altscore = new_sug.st_altscore; | |
| 13852 stp->st_had_bonus = new_sug.st_had_bonus; | |
| 13853 } | |
| 13854 break; | |
| 13855 } | |
| 13856 } | |
| 625 | 13857 |
| 13858 if (i < 0 && ga_grow(gap, 1) == OK) | |
| 13859 { | |
| 13860 /* Add a suggestion. */ | |
| 13861 stp = &SUG(*gap, gap->ga_len); | |
| 13862 stp->st_word = vim_strnsave(goodword, goodlen); | |
| 13863 if (stp->st_word != NULL) | |
| 13864 { | |
| 13865 stp->st_wordlen = goodlen; | |
| 13866 stp->st_score = score; | |
| 13867 stp->st_altscore = altscore; | |
| 13868 stp->st_had_bonus = had_bonus; | |
| 13869 stp->st_orglen = badlen; | |
| 13870 stp->st_slang = slang; | |
| 13871 ++gap->ga_len; | |
| 13872 | |
| 13873 /* If we have too many suggestions now, sort the list and keep | |
| 13874 * the best suggestions. */ | |
| 13875 if (gap->ga_len > SUG_MAX_COUNT(su)) | |
| 13876 { | |
| 13877 if (maxsf) | |
| 13878 su->su_sfmaxscore = cleanup_suggestions(gap, | |
| 13879 su->su_sfmaxscore, SUG_CLEAN_COUNT(su)); | |
| 13880 else | |
| 13881 su->su_maxscore = cleanup_suggestions(gap, | |
| 13882 su->su_maxscore, SUG_CLEAN_COUNT(su)); | |
| 13883 } | |
| 13884 } | |
| 13885 } | |
| 13886 } | |
| 13887 | |
| 13888 /* | |
| 13889 * Suggestions may in fact be flagged as errors. Esp. for banned words and | |
| 13890 * for split words, such as "the the". Remove these from the list here. | |
| 13891 */ | |
| 13892 static void | |
| 13893 check_suggestions(su, gap) | |
| 13894 suginfo_T *su; | |
| 13895 garray_T *gap; /* either su_ga or su_sga */ | |
| 13896 { | |
| 13897 suggest_T *stp; | |
| 13898 int i; | |
| 13899 char_u longword[MAXWLEN + 1]; | |
| 13900 int len; | |
| 13901 hlf_T attr; | |
| 13902 | |
| 13903 stp = &SUG(*gap, 0); | |
| 13904 for (i = gap->ga_len - 1; i >= 0; --i) | |
| 13905 { | |
| 13906 /* Need to append what follows to check for "the the". */ | |
| 2768 | 13907 vim_strncpy(longword, stp[i].st_word, MAXWLEN); |
| 625 | 13908 len = stp[i].st_wordlen; |
| 13909 vim_strncpy(longword + len, su->su_badptr + stp[i].st_orglen, | |
| 13910 MAXWLEN - len); | |
| 13911 attr = HLF_COUNT; | |
| 13912 (void)spell_check(curwin, longword, &attr, NULL, FALSE); | |
| 13913 if (attr != HLF_COUNT) | |
| 13914 { | |
| 13915 /* Remove this entry. */ | |
| 13916 vim_free(stp[i].st_word); | |
| 13917 --gap->ga_len; | |
| 13918 if (i < gap->ga_len) | |
| 13919 mch_memmove(stp + i, stp + i + 1, | |
| 13920 sizeof(suggest_T) * (gap->ga_len - i)); | |
| 13921 } | |
| 13922 } | |
| 13923 } | |
| 13924 | |
| 323 | 13925 |
| 13926 /* | |
| 13927 * Add a word to be banned. | |
| 13928 */ | |
| 13929 static void | |
| 13930 add_banned(su, word) | |
| 13931 suginfo_T *su; | |
| 13932 char_u *word; | |
| 13933 { | |
| 626 | 13934 char_u *s; |
| 323 | 13935 hash_T hash; |
| 13936 hashitem_T *hi; | |
| 13937 | |
| 625 | 13938 hash = hash_hash(word); |
| 13939 hi = hash_lookup(&su->su_banned, word, hash); | |
| 13940 if (HASHITEM_EMPTY(hi)) | |
| 13941 { | |
| 13942 s = vim_strsave(word); | |
| 13943 if (s != NULL) | |
| 323 | 13944 hash_add_item(&su->su_banned, hi, s, hash); |
| 625 | 13945 } |
| 323 | 13946 } |
| 13947 | |
| 324 | 13948 /* |
| 534 | 13949 * Recompute the score for all suggestions if sound-folding is possible. This |
| 13950 * is slow, thus only done for the final results. | |
| 324 | 13951 */ |
| 13952 static void | |
| 13953 rescore_suggestions(su) | |
| 13954 suginfo_T *su; | |
| 13955 { | |
| 13956 int i; | |
| 534 | 13957 |
| 13958 if (su->su_sallang != NULL) | |
| 503 | 13959 for (i = 0; i < su->su_ga.ga_len; ++i) |
| 534 | 13960 rescore_one(su, &SUG(su->su_ga, i)); |
| 13961 } | |
| 13962 | |
| 13963 /* | |
| 13964 * Recompute the score for one suggestion if sound-folding is possible. | |
| 13965 */ | |
| 13966 static void | |
| 13967 rescore_one(su, stp) | |
| 535 | 13968 suginfo_T *su; |
| 13969 suggest_T *stp; | |
| 534 | 13970 { |
| 13971 slang_T *slang = stp->st_slang; | |
| 13972 char_u sal_badword[MAXWLEN]; | |
| 535 | 13973 char_u *p; |
| 534 | 13974 |
| 13975 /* Only rescore suggestions that have no sal score yet and do have a | |
| 13976 * language. */ | |
| 13977 if (slang != NULL && slang->sl_sal.ga_len > 0 && !stp->st_had_bonus) | |
| 13978 { | |
| 13979 if (slang == su->su_sallang) | |
| 535 | 13980 p = su->su_sal_badword; |
| 534 | 13981 else |
| 13982 { | |
| 13983 spell_soundfold(slang, su->su_fbadword, TRUE, sal_badword); | |
| 535 | 13984 p = sal_badword; |
| 13985 } | |
| 13986 | |
| 13987 stp->st_altscore = stp_sal_score(stp, su, slang, p); | |
| 534 | 13988 if (stp->st_altscore == SCORE_MAXMAX) |
| 13989 stp->st_altscore = SCORE_BIG; | |
| 13990 stp->st_score = RESCORE(stp->st_score, stp->st_altscore); | |
| 13991 stp->st_had_bonus = TRUE; | |
| 324 | 13992 } |
| 13993 } | |
| 13994 | |
| 323 | 13995 static int |
| 13996 #ifdef __BORLANDC__ | |
| 13997 _RTLENTRYF | |
| 13998 #endif | |
| 13999 sug_compare __ARGS((const void *s1, const void *s2)); | |
| 14000 | |
| 14001 /* | |
| 14002 * Function given to qsort() to sort the suggestions on st_score. | |
| 529 | 14003 * First on "st_score", then "st_altscore" then alphabetically. |
| 323 | 14004 */ |
| 14005 static int | |
| 14006 #ifdef __BORLANDC__ | |
| 14007 _RTLENTRYF | |
| 14008 #endif | |
| 14009 sug_compare(s1, s2) | |
| 14010 const void *s1; | |
| 14011 const void *s2; | |
| 14012 { | |
| 14013 suggest_T *p1 = (suggest_T *)s1; | |
| 14014 suggest_T *p2 = (suggest_T *)s2; | |
| 344 | 14015 int n = p1->st_score - p2->st_score; |
| 14016 | |
| 14017 if (n == 0) | |
| 529 | 14018 { |
| 14019 n = p1->st_altscore - p2->st_altscore; | |
| 14020 if (n == 0) | |
| 14021 n = STRICMP(p1->st_word, p2->st_word); | |
| 14022 } | |
| 344 | 14023 return n; |
| 323 | 14024 } |
| 14025 | |
| 14026 /* | |
| 14027 * Cleanup the suggestions: | |
| 14028 * - Sort on score. | |
| 14029 * - Remove words that won't be displayed. | |
| 344 | 14030 * Returns the maximum score in the list or "maxscore" unmodified. |
| 323 | 14031 */ |
| 344 | 14032 static int |
| 14033 cleanup_suggestions(gap, maxscore, keep) | |
| 14034 garray_T *gap; | |
| 14035 int maxscore; | |
| 324 | 14036 int keep; /* nr of suggestions to keep */ |
| 323 | 14037 { |
| 344 | 14038 suggest_T *stp = &SUG(*gap, 0); |
| 323 | 14039 int i; |
| 14040 | |
| 14041 /* Sort the list. */ | |
| 344 | 14042 qsort(gap->ga_data, (size_t)gap->ga_len, sizeof(suggest_T), sug_compare); |
| 323 | 14043 |
| 14044 /* Truncate the list to the number of suggestions that will be displayed. */ | |
| 344 | 14045 if (gap->ga_len > keep) |
| 323 | 14046 { |
| 344 | 14047 for (i = keep; i < gap->ga_len; ++i) |
| 323 | 14048 vim_free(stp[i].st_word); |
| 344 | 14049 gap->ga_len = keep; |
| 14050 return stp[keep - 1].st_score; | |
| 323 | 14051 } |
| 344 | 14052 return maxscore; |
| 323 | 14053 } |
| 14054 | |
| 372 | 14055 #if defined(FEAT_EVAL) || defined(PROTO) |
| 14056 /* | |
| 14057 * Soundfold a string, for soundfold(). | |
| 14058 * Result is in allocated memory, NULL for an error. | |
| 14059 */ | |
| 14060 char_u * | |
| 14061 eval_soundfold(word) | |
| 14062 char_u *word; | |
| 14063 { | |
| 14064 langp_T *lp; | |
| 14065 char_u sound[MAXWLEN]; | |
| 500 | 14066 int lpi; |
| 372 | 14067 |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
14068 if (curwin->w_p_spell && *curwin->w_s->b_p_spl != NUL) |
| 372 | 14069 /* Use the sound-folding of the first language that supports it. */ |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
14070 for (lpi = 0; lpi < curwin->w_s->b_langp.ga_len; ++lpi) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
14071 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
14072 lp = LANGP_ENTRY(curwin->w_s->b_langp, lpi); |
| 372 | 14073 if (lp->lp_slang->sl_sal.ga_len > 0) |
| 14074 { | |
| 14075 /* soundfold the word */ | |
| 375 | 14076 spell_soundfold(lp->lp_slang, word, FALSE, sound); |
| 372 | 14077 return vim_strsave(sound); |
| 14078 } | |
| 500 | 14079 } |
| 372 | 14080 |
| 14081 /* No language with sound folding, return word as-is. */ | |
| 14082 return vim_strsave(word); | |
| 14083 } | |
| 14084 #endif | |
| 14085 | |
| 323 | 14086 /* |
| 14087 * Turn "inword" into its sound-a-like equivalent in "res[MAXWLEN]". | |
| 485 | 14088 * |
| 14089 * There are many ways to turn a word into a sound-a-like representation. The | |
| 14090 * oldest is Soundex (1918!). A nice overview can be found in "Approximate | |
| 14091 * swedish name matching - survey and test of different algorithms" by Klas | |
| 14092 * Erikson. | |
| 14093 * | |
| 14094 * We support two methods: | |
| 14095 * 1. SOFOFROM/SOFOTO do a simple character mapping. | |
| 14096 * 2. SAL items define a more advanced sound-folding (and much slower). | |
| 323 | 14097 */ |
| 14098 static void | |
| 375 | 14099 spell_soundfold(slang, inword, folded, res) |
| 14100 slang_T *slang; | |
| 14101 char_u *inword; | |
| 14102 int folded; /* "inword" is already case-folded */ | |
| 14103 char_u *res; | |
| 14104 { | |
| 14105 char_u fword[MAXWLEN]; | |
| 14106 char_u *word; | |
| 14107 | |
| 14108 if (slang->sl_sofo) | |
| 14109 /* SOFOFROM and SOFOTO used */ | |
| 14110 spell_soundfold_sofo(slang, inword, res); | |
| 14111 else | |
| 14112 { | |
| 14113 /* SAL items used. Requires the word to be case-folded. */ | |
| 14114 if (folded) | |
| 14115 word = inword; | |
| 14116 else | |
| 14117 { | |
| 835 | 14118 (void)spell_casefold(inword, (int)STRLEN(inword), fword, MAXWLEN); |
| 375 | 14119 word = fword; |
| 14120 } | |
| 14121 | |
| 14122 #ifdef FEAT_MBYTE | |
| 14123 if (has_mbyte) | |
| 14124 spell_soundfold_wsal(slang, word, res); | |
| 14125 else | |
| 14126 #endif | |
| 14127 spell_soundfold_sal(slang, word, res); | |
| 14128 } | |
| 14129 } | |
| 14130 | |
| 14131 /* | |
| 14132 * Perform sound folding of "inword" into "res" according to SOFOFROM and | |
| 14133 * SOFOTO lines. | |
| 14134 */ | |
| 14135 static void | |
| 14136 spell_soundfold_sofo(slang, inword, res) | |
| 14137 slang_T *slang; | |
| 14138 char_u *inword; | |
| 14139 char_u *res; | |
| 14140 { | |
| 14141 char_u *s; | |
| 14142 int ri = 0; | |
| 14143 int c; | |
| 14144 | |
| 14145 #ifdef FEAT_MBYTE | |
| 14146 if (has_mbyte) | |
| 14147 { | |
| 14148 int prevc = 0; | |
| 14149 int *ip; | |
| 14150 | |
| 14151 /* The sl_sal_first[] table contains the translation for chars up to | |
| 14152 * 255, sl_sal the rest. */ | |
| 14153 for (s = inword; *s != NUL; ) | |
| 14154 { | |
| 474 | 14155 c = mb_cptr2char_adv(&s); |
| 375 | 14156 if (enc_utf8 ? utf_class(c) == 0 : vim_iswhite(c)) |
| 14157 c = ' '; | |
| 14158 else if (c < 256) | |
| 14159 c = slang->sl_sal_first[c]; | |
| 14160 else | |
| 14161 { | |
| 14162 ip = ((int **)slang->sl_sal.ga_data)[c & 0xff]; | |
| 14163 if (ip == NULL) /* empty list, can't match */ | |
| 14164 c = NUL; | |
| 14165 else | |
| 14166 for (;;) /* find "c" in the list */ | |
| 14167 { | |
| 14168 if (*ip == 0) /* not found */ | |
| 14169 { | |
| 14170 c = NUL; | |
| 14171 break; | |
| 14172 } | |
| 14173 if (*ip == c) /* match! */ | |
| 14174 { | |
| 14175 c = ip[1]; | |
| 14176 break; | |
| 14177 } | |
| 14178 ip += 2; | |
| 14179 } | |
| 14180 } | |
| 14181 | |
| 14182 if (c != NUL && c != prevc) | |
| 14183 { | |
| 14184 ri += mb_char2bytes(c, res + ri); | |
| 14185 if (ri + MB_MAXBYTES > MAXWLEN) | |
| 14186 break; | |
| 14187 prevc = c; | |
| 14188 } | |
| 14189 } | |
| 14190 } | |
| 14191 else | |
| 14192 #endif | |
| 14193 { | |
| 14194 /* The sl_sal_first[] table contains the translation. */ | |
| 14195 for (s = inword; (c = *s) != NUL; ++s) | |
| 14196 { | |
| 14197 if (vim_iswhite(c)) | |
| 14198 c = ' '; | |
| 14199 else | |
| 14200 c = slang->sl_sal_first[c]; | |
| 14201 if (c != NUL && (ri == 0 || res[ri - 1] != c)) | |
| 14202 res[ri++] = c; | |
| 14203 } | |
| 14204 } | |
| 14205 | |
| 14206 res[ri] = NUL; | |
| 14207 } | |
| 14208 | |
| 14209 static void | |
| 14210 spell_soundfold_sal(slang, inword, res) | |
| 323 | 14211 slang_T *slang; |
| 14212 char_u *inword; | |
| 14213 char_u *res; | |
| 14214 { | |
| 344 | 14215 salitem_T *smp; |
| 323 | 14216 char_u word[MAXWLEN]; |
| 375 | 14217 char_u *s = inword; |
| 323 | 14218 char_u *t; |
| 344 | 14219 char_u *pf; |
| 323 | 14220 int i, j, z; |
| 344 | 14221 int reslen; |
| 323 | 14222 int n, k = 0; |
| 14223 int z0; | |
| 14224 int k0; | |
| 14225 int n0; | |
| 14226 int c; | |
| 14227 int pri; | |
| 14228 int p0 = -333; | |
| 14229 int c0; | |
| 14230 | |
| 324 | 14231 /* Remove accents, if wanted. We actually remove all non-word characters. |
| 375 | 14232 * But keep white space. We need a copy, the word may be changed here. */ |
| 323 | 14233 if (slang->sl_rem_accents) |
| 14234 { | |
| 14235 t = word; | |
| 375 | 14236 while (*s != NUL) |
| 323 | 14237 { |
| 324 | 14238 if (vim_iswhite(*s)) |
| 344 | 14239 { |
| 14240 *t++ = ' '; | |
| 14241 s = skipwhite(s); | |
| 14242 } | |
| 323 | 14243 else |
| 14244 { | |
| 376 | 14245 if (spell_iswordp_nmw(s)) |
| 323 | 14246 *t++ = *s; |
| 14247 ++s; | |
| 14248 } | |
| 14249 } | |
| 14250 *t = NUL; | |
| 14251 } | |
| 14252 else | |
| 2768 | 14253 vim_strncpy(word, s, MAXWLEN - 1); |
| 323 | 14254 |
| 344 | 14255 smp = (salitem_T *)slang->sl_sal.ga_data; |
| 323 | 14256 |
| 14257 /* | |
| 14258 * This comes from Aspell phonet.cpp. Converted from C++ to C. | |
| 324 | 14259 * Changed to keep spaces. |
| 323 | 14260 */ |
| 344 | 14261 i = reslen = z = 0; |
| 323 | 14262 while ((c = word[i]) != NUL) |
| 14263 { | |
| 344 | 14264 /* Start with the first rule that has the character in the word. */ |
| 323 | 14265 n = slang->sl_sal_first[c]; |
| 14266 z0 = 0; | |
| 14267 | |
| 14268 if (n >= 0) | |
| 14269 { | |
| 14270 /* check all rules for the same letter */ | |
| 344 | 14271 for (; (s = smp[n].sm_lead)[0] == c; ++n) |
| 323 | 14272 { |
| 344 | 14273 /* Quickly skip entries that don't match the word. Most |
| 14274 * entries are less then three chars, optimize for that. */ | |
| 14275 k = smp[n].sm_leadlen; | |
| 14276 if (k > 1) | |
| 323 | 14277 { |
| 344 | 14278 if (word[i + 1] != s[1]) |
| 14279 continue; | |
| 14280 if (k > 2) | |
| 14281 { | |
| 14282 for (j = 2; j < k; ++j) | |
| 14283 if (word[i + j] != s[j]) | |
| 14284 break; | |
| 14285 if (j < k) | |
| 14286 continue; | |
| 14287 } | |
| 323 | 14288 } |
| 14289 | |
| 375 | 14290 if ((pf = smp[n].sm_oneof) != NULL) |
| 323 | 14291 { |
| 375 | 14292 /* Check for match with one of the chars in "sm_oneof". */ |
| 344 | 14293 while (*pf != NUL && *pf != word[i + k]) |
| 14294 ++pf; | |
| 14295 if (*pf == NUL) | |
| 14296 continue; | |
| 14297 ++k; | |
| 323 | 14298 } |
| 344 | 14299 s = smp[n].sm_rules; |
| 14300 pri = 5; /* default priority */ | |
| 323 | 14301 |
| 14302 p0 = *s; | |
| 14303 k0 = k; | |
| 14304 while (*s == '-' && k > 1) | |
| 14305 { | |
| 14306 k--; | |
| 14307 s++; | |
| 14308 } | |
| 14309 if (*s == '<') | |
| 14310 s++; | |
| 344 | 14311 if (VIM_ISDIGIT(*s)) |
| 323 | 14312 { |
| 14313 /* determine priority */ | |
| 14314 pri = *s - '0'; | |
| 14315 s++; | |
| 14316 } | |
| 14317 if (*s == '^' && *(s + 1) == '^') | |
| 14318 s++; | |
| 14319 | |
| 14320 if (*s == NUL | |
| 14321 || (*s == '^' | |
| 324 | 14322 && (i == 0 || !(word[i - 1] == ' ' |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
14323 || spell_iswordp(word + i - 1, curwin))) |
| 323 | 14324 && (*(s + 1) != '$' |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
14325 || (!spell_iswordp(word + i + k0, curwin)))) |
| 323 | 14326 || (*s == '$' && i > 0 |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
14327 && spell_iswordp(word + i - 1, curwin) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
14328 && (!spell_iswordp(word + i + k0, curwin)))) |
| 323 | 14329 { |
| 14330 /* search for followup rules, if: */ | |
| 14331 /* followup and k > 1 and NO '-' in searchstring */ | |
| 14332 c0 = word[i + k - 1]; | |
| 14333 n0 = slang->sl_sal_first[c0]; | |
| 14334 | |
| 14335 if (slang->sl_followup && k > 1 && n0 >= 0 | |
| 344 | 14336 && p0 != '-' && word[i + k] != NUL) |
| 323 | 14337 { |
| 14338 /* test follow-up rule for "word[i + k]" */ | |
| 344 | 14339 for ( ; (s = smp[n0].sm_lead)[0] == c0; ++n0) |
| 323 | 14340 { |
| 344 | 14341 /* Quickly skip entries that don't match the word. |
| 14342 * */ | |
| 14343 k0 = smp[n0].sm_leadlen; | |
| 14344 if (k0 > 1) | |
| 323 | 14345 { |
| 344 | 14346 if (word[i + k] != s[1]) |
| 14347 continue; | |
| 14348 if (k0 > 2) | |
| 14349 { | |
| 14350 pf = word + i + k + 1; | |
| 14351 for (j = 2; j < k0; ++j) | |
| 14352 if (*pf++ != s[j]) | |
| 14353 break; | |
| 14354 if (j < k0) | |
| 14355 continue; | |
| 14356 } | |
| 323 | 14357 } |
| 344 | 14358 k0 += k - 1; |
| 14359 | |
| 375 | 14360 if ((pf = smp[n0].sm_oneof) != NULL) |
| 323 | 14361 { |
| 344 | 14362 /* Check for match with one of the chars in |
| 375 | 14363 * "sm_oneof". */ |
| 344 | 14364 while (*pf != NUL && *pf != word[i + k0]) |
| 14365 ++pf; | |
| 14366 if (*pf == NUL) | |
| 14367 continue; | |
| 14368 ++k0; | |
| 323 | 14369 } |
| 344 | 14370 |
| 14371 p0 = 5; | |
| 14372 s = smp[n0].sm_rules; | |
| 323 | 14373 while (*s == '-') |
| 14374 { | |
| 344 | 14375 /* "k0" gets NOT reduced because |
| 14376 * "if (k0 == k)" */ | |
| 323 | 14377 s++; |
| 14378 } | |
| 14379 if (*s == '<') | |
| 14380 s++; | |
| 344 | 14381 if (VIM_ISDIGIT(*s)) |
| 323 | 14382 { |
| 14383 p0 = *s - '0'; | |
| 14384 s++; | |
| 14385 } | |
| 14386 | |
| 14387 if (*s == NUL | |
| 14388 /* *s == '^' cuts */ | |
| 14389 || (*s == '$' | |
| 376 | 14390 && !spell_iswordp(word + i + k0, |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
14391 curwin))) |
| 323 | 14392 { |
| 14393 if (k0 == k) | |
| 14394 /* this is just a piece of the string */ | |
| 14395 continue; | |
| 14396 | |
| 14397 if (p0 < pri) | |
| 14398 /* priority too low */ | |
| 14399 continue; | |
| 14400 /* rule fits; stop search */ | |
| 14401 break; | |
| 14402 } | |
| 14403 } | |
| 14404 | |
| 344 | 14405 if (p0 >= pri && smp[n0].sm_lead[0] == c0) |
| 323 | 14406 continue; |
| 14407 } | |
| 14408 | |
| 14409 /* replace string */ | |
| 344 | 14410 s = smp[n].sm_to; |
| 389 | 14411 if (s == NULL) |
| 14412 s = (char_u *)""; | |
| 344 | 14413 pf = smp[n].sm_rules; |
| 14414 p0 = (vim_strchr(pf, '<') != NULL) ? 1 : 0; | |
| 323 | 14415 if (p0 == 1 && z == 0) |
| 14416 { | |
| 14417 /* rule with '<' is used */ | |
| 344 | 14418 if (reslen > 0 && *s != NUL && (res[reslen - 1] == c |
| 14419 || res[reslen - 1] == *s)) | |
| 14420 reslen--; | |
| 323 | 14421 z0 = 1; |
| 14422 z = 1; | |
| 14423 k0 = 0; | |
| 372 | 14424 while (*s != NUL && word[i + k0] != NUL) |
| 323 | 14425 { |
| 14426 word[i + k0] = *s; | |
| 14427 k0++; | |
| 14428 s++; | |
| 14429 } | |
| 14430 if (k > k0) | |
| 1619 | 14431 STRMOVE(word + i + k0, word + i + k); |
| 323 | 14432 |
| 14433 /* new "actual letter" */ | |
| 14434 c = word[i]; | |
| 14435 } | |
| 14436 else | |
| 14437 { | |
| 14438 /* no '<' rule used */ | |
| 14439 i += k - 1; | |
| 14440 z = 0; | |
| 344 | 14441 while (*s != NUL && s[1] != NUL && reslen < MAXWLEN) |
| 323 | 14442 { |
| 344 | 14443 if (reslen == 0 || res[reslen - 1] != *s) |
| 372 | 14444 res[reslen++] = *s; |
| 323 | 14445 s++; |
| 14446 } | |
| 14447 /* new "actual letter" */ | |
| 14448 c = *s; | |
| 344 | 14449 if (strstr((char *)pf, "^^") != NULL) |
| 323 | 14450 { |
| 14451 if (c != NUL) | |
| 372 | 14452 res[reslen++] = c; |
| 1619 | 14453 STRMOVE(word, word + i + 1); |
| 323 | 14454 i = 0; |
| 14455 z0 = 1; | |
| 14456 } | |
| 14457 } | |
| 14458 break; | |
| 14459 } | |
| 14460 } | |
| 14461 } | |
| 324 | 14462 else if (vim_iswhite(c)) |
| 14463 { | |
| 14464 c = ' '; | |
| 14465 k = 1; | |
| 14466 } | |
| 323 | 14467 |
| 14468 if (z0 == 0) | |
| 14469 { | |
| 344 | 14470 if (k && !p0 && reslen < MAXWLEN && c != NUL |
| 14471 && (!slang->sl_collapse || reslen == 0 | |
| 14472 || res[reslen - 1] != c)) | |
| 323 | 14473 /* condense only double letters */ |
| 372 | 14474 res[reslen++] = c; |
| 323 | 14475 |
| 14476 i++; | |
| 14477 z = 0; | |
| 14478 k = 0; | |
| 14479 } | |
| 14480 } | |
| 14481 | |
| 344 | 14482 res[reslen] = NUL; |
| 323 | 14483 } |
| 14484 | |
| 372 | 14485 #ifdef FEAT_MBYTE |
| 14486 /* | |
| 14487 * Turn "inword" into its sound-a-like equivalent in "res[MAXWLEN]". | |
| 14488 * Multi-byte version of spell_soundfold(). | |
| 14489 */ | |
| 14490 static void | |
| 375 | 14491 spell_soundfold_wsal(slang, inword, res) |
| 372 | 14492 slang_T *slang; |
| 14493 char_u *inword; | |
| 14494 char_u *res; | |
| 14495 { | |
| 375 | 14496 salitem_T *smp = (salitem_T *)slang->sl_sal.ga_data; |
| 372 | 14497 int word[MAXWLEN]; |
| 14498 int wres[MAXWLEN]; | |
| 14499 int l; | |
| 14500 char_u *s; | |
| 14501 int *ws; | |
| 14502 char_u *t; | |
| 14503 int *pf; | |
| 14504 int i, j, z; | |
| 14505 int reslen; | |
| 14506 int n, k = 0; | |
| 14507 int z0; | |
| 14508 int k0; | |
| 14509 int n0; | |
| 14510 int c; | |
| 14511 int pri; | |
| 14512 int p0 = -333; | |
| 14513 int c0; | |
| 14514 int did_white = FALSE; | |
| 14515 | |
| 14516 /* | |
| 14517 * Convert the multi-byte string to a wide-character string. | |
| 14518 * Remove accents, if wanted. We actually remove all non-word characters. | |
| 14519 * But keep white space. | |
| 14520 */ | |
| 14521 n = 0; | |
| 14522 for (s = inword; *s != NUL; ) | |
| 14523 { | |
| 14524 t = s; | |
| 474 | 14525 c = mb_cptr2char_adv(&s); |
| 372 | 14526 if (slang->sl_rem_accents) |
| 14527 { | |
| 14528 if (enc_utf8 ? utf_class(c) == 0 : vim_iswhite(c)) | |
| 14529 { | |
| 14530 if (did_white) | |
| 14531 continue; | |
| 14532 c = ' '; | |
| 14533 did_white = TRUE; | |
| 14534 } | |
| 14535 else | |
| 14536 { | |
| 14537 did_white = FALSE; | |
| 376 | 14538 if (!spell_iswordp_nmw(t)) |
| 372 | 14539 continue; |
| 14540 } | |
| 14541 } | |
| 14542 word[n++] = c; | |
| 14543 } | |
| 14544 word[n] = NUL; | |
| 14545 | |
| 14546 /* | |
| 14547 * This comes from Aspell phonet.cpp. | |
| 14548 * Converted from C++ to C. Added support for multi-byte chars. | |
| 14549 * Changed to keep spaces. | |
| 14550 */ | |
| 14551 i = reslen = z = 0; | |
| 14552 while ((c = word[i]) != NUL) | |
| 14553 { | |
| 14554 /* Start with the first rule that has the character in the word. */ | |
| 14555 n = slang->sl_sal_first[c & 0xff]; | |
| 14556 z0 = 0; | |
| 14557 | |
| 14558 if (n >= 0) | |
| 14559 { | |
|
2455
9367de3e2e1b
Fix: crash in spell checking with a 0x300 character.
Bram Moolenaar <bram@vim.org>
parents:
2454
diff
changeset
|
14560 /* Check all rules for the same index byte. |
|
9367de3e2e1b
Fix: crash in spell checking with a 0x300 character.
Bram Moolenaar <bram@vim.org>
parents:
2454
diff
changeset
|
14561 * If c is 0x300 need extra check for the end of the array, as |
|
9367de3e2e1b
Fix: crash in spell checking with a 0x300 character.
Bram Moolenaar <bram@vim.org>
parents:
2454
diff
changeset
|
14562 * (c & 0xff) is NUL. */ |
|
9367de3e2e1b
Fix: crash in spell checking with a 0x300 character.
Bram Moolenaar <bram@vim.org>
parents:
2454
diff
changeset
|
14563 for (; ((ws = smp[n].sm_lead_w)[0] & 0xff) == (c & 0xff) |
|
9367de3e2e1b
Fix: crash in spell checking with a 0x300 character.
Bram Moolenaar <bram@vim.org>
parents:
2454
diff
changeset
|
14564 && ws[0] != NUL; ++n) |
| 372 | 14565 { |
| 14566 /* Quickly skip entries that don't match the word. Most | |
| 14567 * entries are less then three chars, optimize for that. */ | |
| 375 | 14568 if (c != ws[0]) |
| 14569 continue; | |
| 372 | 14570 k = smp[n].sm_leadlen; |
| 14571 if (k > 1) | |
| 14572 { | |
| 14573 if (word[i + 1] != ws[1]) | |
| 14574 continue; | |
| 14575 if (k > 2) | |
| 14576 { | |
| 14577 for (j = 2; j < k; ++j) | |
| 14578 if (word[i + j] != ws[j]) | |
| 14579 break; | |
| 14580 if (j < k) | |
| 14581 continue; | |
| 14582 } | |
| 14583 } | |
| 14584 | |
| 375 | 14585 if ((pf = smp[n].sm_oneof_w) != NULL) |
| 372 | 14586 { |
| 375 | 14587 /* Check for match with one of the chars in "sm_oneof". */ |
| 372 | 14588 while (*pf != NUL && *pf != word[i + k]) |
| 14589 ++pf; | |
| 14590 if (*pf == NUL) | |
| 14591 continue; | |
| 14592 ++k; | |
| 14593 } | |
| 14594 s = smp[n].sm_rules; | |
| 14595 pri = 5; /* default priority */ | |
| 14596 | |
| 14597 p0 = *s; | |
| 14598 k0 = k; | |
| 14599 while (*s == '-' && k > 1) | |
| 14600 { | |
| 14601 k--; | |
| 14602 s++; | |
| 14603 } | |
| 14604 if (*s == '<') | |
| 14605 s++; | |
| 14606 if (VIM_ISDIGIT(*s)) | |
| 14607 { | |
| 14608 /* determine priority */ | |
| 14609 pri = *s - '0'; | |
| 14610 s++; | |
| 14611 } | |
| 14612 if (*s == '^' && *(s + 1) == '^') | |
| 14613 s++; | |
| 14614 | |
| 14615 if (*s == NUL | |
| 14616 || (*s == '^' | |
| 14617 && (i == 0 || !(word[i - 1] == ' ' | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
14618 || spell_iswordp_w(word + i - 1, curwin))) |
| 372 | 14619 && (*(s + 1) != '$' |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
14620 || (!spell_iswordp_w(word + i + k0, curwin)))) |
| 372 | 14621 || (*s == '$' && i > 0 |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
14622 && spell_iswordp_w(word + i - 1, curwin) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
14623 && (!spell_iswordp_w(word + i + k0, curwin)))) |
| 372 | 14624 { |
| 14625 /* search for followup rules, if: */ | |
| 14626 /* followup and k > 1 and NO '-' in searchstring */ | |
| 14627 c0 = word[i + k - 1]; | |
| 14628 n0 = slang->sl_sal_first[c0 & 0xff]; | |
| 14629 | |
| 14630 if (slang->sl_followup && k > 1 && n0 >= 0 | |
| 14631 && p0 != '-' && word[i + k] != NUL) | |
| 14632 { | |
| 375 | 14633 /* Test follow-up rule for "word[i + k]"; loop over |
| 14634 * all entries with the same index byte. */ | |
| 372 | 14635 for ( ; ((ws = smp[n0].sm_lead_w)[0] & 0xff) |
| 14636 == (c0 & 0xff); ++n0) | |
| 14637 { | |
| 14638 /* Quickly skip entries that don't match the word. | |
| 375 | 14639 */ |
| 14640 if (c0 != ws[0]) | |
| 14641 continue; | |
| 372 | 14642 k0 = smp[n0].sm_leadlen; |
| 14643 if (k0 > 1) | |
| 14644 { | |
| 14645 if (word[i + k] != ws[1]) | |
| 14646 continue; | |
| 14647 if (k0 > 2) | |
| 14648 { | |
| 14649 pf = word + i + k + 1; | |
| 14650 for (j = 2; j < k0; ++j) | |
| 14651 if (*pf++ != ws[j]) | |
| 14652 break; | |
| 14653 if (j < k0) | |
| 14654 continue; | |
| 14655 } | |
| 14656 } | |
| 14657 k0 += k - 1; | |
| 14658 | |
| 375 | 14659 if ((pf = smp[n0].sm_oneof_w) != NULL) |
| 372 | 14660 { |
| 14661 /* Check for match with one of the chars in | |
| 375 | 14662 * "sm_oneof". */ |
| 372 | 14663 while (*pf != NUL && *pf != word[i + k0]) |
| 14664 ++pf; | |
| 14665 if (*pf == NUL) | |
| 14666 continue; | |
| 14667 ++k0; | |
| 14668 } | |
| 14669 | |
| 14670 p0 = 5; | |
| 14671 s = smp[n0].sm_rules; | |
| 14672 while (*s == '-') | |
| 14673 { | |
| 14674 /* "k0" gets NOT reduced because | |
| 14675 * "if (k0 == k)" */ | |
| 14676 s++; | |
| 14677 } | |
| 14678 if (*s == '<') | |
| 14679 s++; | |
| 14680 if (VIM_ISDIGIT(*s)) | |
| 14681 { | |
| 14682 p0 = *s - '0'; | |
| 14683 s++; | |
| 14684 } | |
| 14685 | |
| 14686 if (*s == NUL | |
| 14687 /* *s == '^' cuts */ | |
| 14688 || (*s == '$' | |
| 376 | 14689 && !spell_iswordp_w(word + i + k0, |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
14690 curwin))) |
| 372 | 14691 { |
| 14692 if (k0 == k) | |
| 14693 /* this is just a piece of the string */ | |
| 14694 continue; | |
| 14695 | |
| 14696 if (p0 < pri) | |
| 14697 /* priority too low */ | |
| 14698 continue; | |
| 14699 /* rule fits; stop search */ | |
| 14700 break; | |
| 14701 } | |
| 14702 } | |
| 14703 | |
| 14704 if (p0 >= pri && (smp[n0].sm_lead_w[0] & 0xff) | |
| 14705 == (c0 & 0xff)) | |
| 14706 continue; | |
| 14707 } | |
| 14708 | |
| 14709 /* replace string */ | |
| 14710 ws = smp[n].sm_to_w; | |
| 14711 s = smp[n].sm_rules; | |
| 14712 p0 = (vim_strchr(s, '<') != NULL) ? 1 : 0; | |
| 14713 if (p0 == 1 && z == 0) | |
| 14714 { | |
| 14715 /* rule with '<' is used */ | |
| 389 | 14716 if (reslen > 0 && ws != NULL && *ws != NUL |
| 14717 && (wres[reslen - 1] == c | |
| 372 | 14718 || wres[reslen - 1] == *ws)) |
| 14719 reslen--; | |
| 14720 z0 = 1; | |
| 14721 z = 1; | |
| 14722 k0 = 0; | |
| 389 | 14723 if (ws != NULL) |
| 14724 while (*ws != NUL && word[i + k0] != NUL) | |
| 14725 { | |
| 14726 word[i + k0] = *ws; | |
| 14727 k0++; | |
| 14728 ws++; | |
| 14729 } | |
| 372 | 14730 if (k > k0) |
| 14731 mch_memmove(word + i + k0, word + i + k, | |
| 14732 sizeof(int) * (STRLEN(word + i + k) + 1)); | |
| 14733 | |
| 14734 /* new "actual letter" */ | |
| 14735 c = word[i]; | |
| 14736 } | |
| 14737 else | |
| 14738 { | |
| 14739 /* no '<' rule used */ | |
| 14740 i += k - 1; | |
| 14741 z = 0; | |
| 389 | 14742 if (ws != NULL) |
| 14743 while (*ws != NUL && ws[1] != NUL | |
| 14744 && reslen < MAXWLEN) | |
| 14745 { | |
| 14746 if (reslen == 0 || wres[reslen - 1] != *ws) | |
| 14747 wres[reslen++] = *ws; | |
| 14748 ws++; | |
| 14749 } | |
| 372 | 14750 /* new "actual letter" */ |
| 389 | 14751 if (ws == NULL) |
| 14752 c = NUL; | |
| 14753 else | |
| 14754 c = *ws; | |
| 372 | 14755 if (strstr((char *)s, "^^") != NULL) |
| 14756 { | |
| 14757 if (c != NUL) | |
| 14758 wres[reslen++] = c; | |
| 14759 mch_memmove(word, word + i + 1, | |
| 14760 sizeof(int) * (STRLEN(word + i + 1) + 1)); | |
| 14761 i = 0; | |
| 14762 z0 = 1; | |
| 14763 } | |
| 14764 } | |
| 14765 break; | |
| 14766 } | |
| 14767 } | |
| 14768 } | |
| 14769 else if (vim_iswhite(c)) | |
| 14770 { | |
| 14771 c = ' '; | |
| 14772 k = 1; | |
| 14773 } | |
| 14774 | |
| 14775 if (z0 == 0) | |
| 14776 { | |
| 14777 if (k && !p0 && reslen < MAXWLEN && c != NUL | |
| 14778 && (!slang->sl_collapse || reslen == 0 | |
| 14779 || wres[reslen - 1] != c)) | |
| 14780 /* condense only double letters */ | |
| 14781 wres[reslen++] = c; | |
| 14782 | |
| 14783 i++; | |
| 14784 z = 0; | |
| 14785 k = 0; | |
| 14786 } | |
| 14787 } | |
| 14788 | |
| 14789 /* Convert wide characters in "wres" to a multi-byte string in "res". */ | |
| 14790 l = 0; | |
| 14791 for (n = 0; n < reslen; ++n) | |
| 14792 { | |
| 14793 l += mb_char2bytes(wres[n], res + l); | |
| 14794 if (l + MB_MAXBYTES > MAXWLEN) | |
| 14795 break; | |
| 14796 } | |
| 14797 res[l] = NUL; | |
| 14798 } | |
| 14799 #endif | |
| 14800 | |
| 324 | 14801 /* |
| 344 | 14802 * Compute a score for two sound-a-like words. |
| 14803 * This permits up to two inserts/deletes/swaps/etc. to keep things fast. | |
| 14804 * Instead of a generic loop we write out the code. That keeps it fast by | |
| 14805 * avoiding checks that will not be possible. | |
| 14806 */ | |
| 14807 static int | |
| 351 | 14808 soundalike_score(goodstart, badstart) |
| 14809 char_u *goodstart; /* sound-folded good word */ | |
| 14810 char_u *badstart; /* sound-folded bad word */ | |
| 344 | 14811 { |
| 351 | 14812 char_u *goodsound = goodstart; |
| 14813 char_u *badsound = badstart; | |
| 14814 int goodlen; | |
| 14815 int badlen; | |
| 344 | 14816 int n; |
| 14817 char_u *pl, *ps; | |
| 14818 char_u *pl2, *ps2; | |
| 351 | 14819 int score = 0; |
| 14820 | |
|
2454
91fec91a522c
Fix: uninitialized memory access. (Dominique Pelle)
Bram Moolenaar <bram@vim.org>
parents:
2445
diff
changeset
|
14821 /* Adding/inserting "*" at the start (word starts with vowel) shouldn't be |
| 351 | 14822 * counted so much, vowels halfway the word aren't counted at all. */ |
| 14823 if ((*badsound == '*' || *goodsound == '*') && *badsound != *goodsound) | |
| 14824 { | |
|
2454
91fec91a522c
Fix: uninitialized memory access. (Dominique Pelle)
Bram Moolenaar <bram@vim.org>
parents:
2445
diff
changeset
|
14825 if ((badsound[0] == NUL && goodsound[1] == NUL) |
|
91fec91a522c
Fix: uninitialized memory access. (Dominique Pelle)
Bram Moolenaar <bram@vim.org>
parents:
2445
diff
changeset
|
14826 || (goodsound[0] == NUL && badsound[1] == NUL)) |
|
91fec91a522c
Fix: uninitialized memory access. (Dominique Pelle)
Bram Moolenaar <bram@vim.org>
parents:
2445
diff
changeset
|
14827 /* changing word with vowel to word without a sound */ |
|
91fec91a522c
Fix: uninitialized memory access. (Dominique Pelle)
Bram Moolenaar <bram@vim.org>
parents:
2445
diff
changeset
|
14828 return SCORE_DEL; |
|
91fec91a522c
Fix: uninitialized memory access. (Dominique Pelle)
Bram Moolenaar <bram@vim.org>
parents:
2445
diff
changeset
|
14829 if (badsound[0] == NUL || goodsound[0] == NUL) |
|
91fec91a522c
Fix: uninitialized memory access. (Dominique Pelle)
Bram Moolenaar <bram@vim.org>
parents:
2445
diff
changeset
|
14830 /* more than two changes */ |
|
91fec91a522c
Fix: uninitialized memory access. (Dominique Pelle)
Bram Moolenaar <bram@vim.org>
parents:
2445
diff
changeset
|
14831 return SCORE_MAXMAX; |
|
91fec91a522c
Fix: uninitialized memory access. (Dominique Pelle)
Bram Moolenaar <bram@vim.org>
parents:
2445
diff
changeset
|
14832 |
| 625 | 14833 if (badsound[1] == goodsound[1] |
| 14834 || (badsound[1] != NUL | |
| 14835 && goodsound[1] != NUL | |
| 14836 && badsound[2] == goodsound[2])) | |
| 14837 { | |
| 14838 /* handle like a substitute */ | |
| 14839 } | |
| 351 | 14840 else |
| 625 | 14841 { |
| 14842 score = 2 * SCORE_DEL / 3; | |
| 14843 if (*badsound == '*') | |
| 14844 ++badsound; | |
| 14845 else | |
| 14846 ++goodsound; | |
| 14847 } | |
| 351 | 14848 } |
| 14849 | |
| 835 | 14850 goodlen = (int)STRLEN(goodsound); |
| 14851 badlen = (int)STRLEN(badsound); | |
| 344 | 14852 |
| 1188 | 14853 /* Return quickly if the lengths are too different to be fixed by two |
| 344 | 14854 * changes. */ |
| 14855 n = goodlen - badlen; | |
| 14856 if (n < -2 || n > 2) | |
| 14857 return SCORE_MAXMAX; | |
| 14858 | |
| 14859 if (n > 0) | |
| 14860 { | |
| 351 | 14861 pl = goodsound; /* goodsound is longest */ |
| 344 | 14862 ps = badsound; |
| 14863 } | |
| 14864 else | |
| 14865 { | |
| 351 | 14866 pl = badsound; /* badsound is longest */ |
| 344 | 14867 ps = goodsound; |
| 14868 } | |
| 14869 | |
| 14870 /* Skip over the identical part. */ | |
| 14871 while (*pl == *ps && *pl != NUL) | |
| 14872 { | |
| 14873 ++pl; | |
| 14874 ++ps; | |
| 14875 } | |
| 14876 | |
| 14877 switch (n) | |
| 14878 { | |
| 14879 case -2: | |
| 14880 case 2: | |
| 14881 /* | |
| 14882 * Must delete two characters from "pl". | |
| 14883 */ | |
| 14884 ++pl; /* first delete */ | |
| 14885 while (*pl == *ps) | |
| 14886 { | |
| 14887 ++pl; | |
| 14888 ++ps; | |
| 14889 } | |
| 14890 /* strings must be equal after second delete */ | |
| 14891 if (STRCMP(pl + 1, ps) == 0) | |
| 351 | 14892 return score + SCORE_DEL * 2; |
| 344 | 14893 |
| 14894 /* Failed to compare. */ | |
| 14895 break; | |
| 14896 | |
| 14897 case -1: | |
| 14898 case 1: | |
| 14899 /* | |
| 14900 * Minimal one delete from "pl" required. | |
| 14901 */ | |
| 14902 | |
| 14903 /* 1: delete */ | |
| 14904 pl2 = pl + 1; | |
| 14905 ps2 = ps; | |
| 14906 while (*pl2 == *ps2) | |
| 14907 { | |
| 14908 if (*pl2 == NUL) /* reached the end */ | |
| 351 | 14909 return score + SCORE_DEL; |
| 344 | 14910 ++pl2; |
| 14911 ++ps2; | |
| 14912 } | |
| 14913 | |
| 14914 /* 2: delete then swap, then rest must be equal */ | |
| 14915 if (pl2[0] == ps2[1] && pl2[1] == ps2[0] | |
| 14916 && STRCMP(pl2 + 2, ps2 + 2) == 0) | |
| 351 | 14917 return score + SCORE_DEL + SCORE_SWAP; |
| 344 | 14918 |
| 14919 /* 3: delete then substitute, then the rest must be equal */ | |
| 14920 if (STRCMP(pl2 + 1, ps2 + 1) == 0) | |
| 351 | 14921 return score + SCORE_DEL + SCORE_SUBST; |
| 344 | 14922 |
| 14923 /* 4: first swap then delete */ | |
| 14924 if (pl[0] == ps[1] && pl[1] == ps[0]) | |
| 14925 { | |
| 14926 pl2 = pl + 2; /* swap, skip two chars */ | |
| 14927 ps2 = ps + 2; | |
| 14928 while (*pl2 == *ps2) | |
| 14929 { | |
| 14930 ++pl2; | |
| 14931 ++ps2; | |
| 14932 } | |
| 14933 /* delete a char and then strings must be equal */ | |
| 14934 if (STRCMP(pl2 + 1, ps2) == 0) | |
| 351 | 14935 return score + SCORE_SWAP + SCORE_DEL; |
| 344 | 14936 } |
| 14937 | |
| 14938 /* 5: first substitute then delete */ | |
| 14939 pl2 = pl + 1; /* substitute, skip one char */ | |
| 14940 ps2 = ps + 1; | |
| 14941 while (*pl2 == *ps2) | |
| 14942 { | |
| 14943 ++pl2; | |
| 14944 ++ps2; | |
| 14945 } | |
| 14946 /* delete a char and then strings must be equal */ | |
| 14947 if (STRCMP(pl2 + 1, ps2) == 0) | |
| 351 | 14948 return score + SCORE_SUBST + SCORE_DEL; |
| 344 | 14949 |
| 14950 /* Failed to compare. */ | |
| 14951 break; | |
| 14952 | |
| 14953 case 0: | |
| 14954 /* | |
| 1808 | 14955 * Lengths are equal, thus changes must result in same length: An |
| 344 | 14956 * insert is only possible in combination with a delete. |
| 14957 * 1: check if for identical strings | |
| 14958 */ | |
| 14959 if (*pl == NUL) | |
| 351 | 14960 return score; |
| 344 | 14961 |
| 14962 /* 2: swap */ | |
| 14963 if (pl[0] == ps[1] && pl[1] == ps[0]) | |
| 14964 { | |
| 14965 pl2 = pl + 2; /* swap, skip two chars */ | |
| 14966 ps2 = ps + 2; | |
| 14967 while (*pl2 == *ps2) | |
| 14968 { | |
| 14969 if (*pl2 == NUL) /* reached the end */ | |
| 351 | 14970 return score + SCORE_SWAP; |
| 344 | 14971 ++pl2; |
| 14972 ++ps2; | |
| 14973 } | |
| 14974 /* 3: swap and swap again */ | |
| 14975 if (pl2[0] == ps2[1] && pl2[1] == ps2[0] | |
| 14976 && STRCMP(pl2 + 2, ps2 + 2) == 0) | |
| 351 | 14977 return score + SCORE_SWAP + SCORE_SWAP; |
| 344 | 14978 |
| 14979 /* 4: swap and substitute */ | |
| 14980 if (STRCMP(pl2 + 1, ps2 + 1) == 0) | |
| 351 | 14981 return score + SCORE_SWAP + SCORE_SUBST; |
| 344 | 14982 } |
| 14983 | |
| 14984 /* 5: substitute */ | |
| 14985 pl2 = pl + 1; | |
| 14986 ps2 = ps + 1; | |
| 14987 while (*pl2 == *ps2) | |
| 14988 { | |
| 14989 if (*pl2 == NUL) /* reached the end */ | |
| 351 | 14990 return score + SCORE_SUBST; |
| 344 | 14991 ++pl2; |
| 14992 ++ps2; | |
| 14993 } | |
| 14994 | |
| 14995 /* 6: substitute and swap */ | |
| 14996 if (pl2[0] == ps2[1] && pl2[1] == ps2[0] | |
| 14997 && STRCMP(pl2 + 2, ps2 + 2) == 0) | |
| 351 | 14998 return score + SCORE_SUBST + SCORE_SWAP; |
| 344 | 14999 |
| 15000 /* 7: substitute and substitute */ | |
| 15001 if (STRCMP(pl2 + 1, ps2 + 1) == 0) | |
| 351 | 15002 return score + SCORE_SUBST + SCORE_SUBST; |
| 344 | 15003 |
| 15004 /* 8: insert then delete */ | |
| 15005 pl2 = pl; | |
| 15006 ps2 = ps + 1; | |
| 15007 while (*pl2 == *ps2) | |
| 15008 { | |
| 15009 ++pl2; | |
| 15010 ++ps2; | |
| 15011 } | |
| 15012 if (STRCMP(pl2 + 1, ps2) == 0) | |
| 351 | 15013 return score + SCORE_INS + SCORE_DEL; |
| 344 | 15014 |
| 15015 /* 9: delete then insert */ | |
| 15016 pl2 = pl + 1; | |
| 15017 ps2 = ps; | |
| 15018 while (*pl2 == *ps2) | |
| 15019 { | |
| 15020 ++pl2; | |
| 15021 ++ps2; | |
| 15022 } | |
| 15023 if (STRCMP(pl2, ps2 + 1) == 0) | |
| 351 | 15024 return score + SCORE_INS + SCORE_DEL; |
| 344 | 15025 |
| 15026 /* Failed to compare. */ | |
| 15027 break; | |
| 15028 } | |
| 15029 | |
| 15030 return SCORE_MAXMAX; | |
| 15031 } | |
| 324 | 15032 |
| 323 | 15033 /* |
| 15034 * Compute the "edit distance" to turn "badword" into "goodword". The less | |
| 344 | 15035 * deletes/inserts/substitutes/swaps are required the lower the score. |
| 324 | 15036 * |
| 485 | 15037 * The algorithm is described by Du and Chang, 1992. |
| 15038 * The implementation of the algorithm comes from Aspell editdist.cpp, | |
| 15039 * edit_distance(). It has been converted from C++ to C and modified to | |
| 15040 * support multi-byte characters. | |
| 323 | 15041 */ |
| 15042 static int | |
| 625 | 15043 spell_edit_score(slang, badword, goodword) |
| 15044 slang_T *slang; | |
| 323 | 15045 char_u *badword; |
| 15046 char_u *goodword; | |
| 15047 { | |
| 15048 int *cnt; | |
| 1188 | 15049 int badlen, goodlen; /* lengths including NUL */ |
| 323 | 15050 int j, i; |
| 15051 int t; | |
| 15052 int bc, gc; | |
| 324 | 15053 int pbc, pgc; |
| 15054 #ifdef FEAT_MBYTE | |
| 15055 char_u *p; | |
| 15056 int wbadword[MAXWLEN]; | |
| 15057 int wgoodword[MAXWLEN]; | |
| 15058 | |
| 15059 if (has_mbyte) | |
| 15060 { | |
| 15061 /* Get the characters from the multi-byte strings and put them in an | |
| 15062 * int array for easy access. */ | |
| 15063 for (p = badword, badlen = 0; *p != NUL; ) | |
| 474 | 15064 wbadword[badlen++] = mb_cptr2char_adv(&p); |
| 398 | 15065 wbadword[badlen++] = 0; |
| 324 | 15066 for (p = goodword, goodlen = 0; *p != NUL; ) |
| 474 | 15067 wgoodword[goodlen++] = mb_cptr2char_adv(&p); |
| 398 | 15068 wgoodword[goodlen++] = 0; |
| 324 | 15069 } |
| 15070 else | |
| 15071 #endif | |
| 15072 { | |
| 835 | 15073 badlen = (int)STRLEN(badword) + 1; |
| 15074 goodlen = (int)STRLEN(goodword) + 1; | |
| 324 | 15075 } |
| 323 | 15076 |
| 15077 /* We use "cnt" as an array: CNT(badword_idx, goodword_idx). */ | |
| 15078 #define CNT(a, b) cnt[(a) + (b) * (badlen + 1)] | |
| 15079 cnt = (int *)lalloc((long_u)(sizeof(int) * (badlen + 1) * (goodlen + 1)), | |
| 15080 TRUE); | |
| 324 | 15081 if (cnt == NULL) |
| 15082 return 0; /* out of memory */ | |
| 323 | 15083 |
| 15084 CNT(0, 0) = 0; | |
| 15085 for (j = 1; j <= goodlen; ++j) | |
| 625 | 15086 CNT(0, j) = CNT(0, j - 1) + SCORE_INS; |
| 323 | 15087 |
| 15088 for (i = 1; i <= badlen; ++i) | |
| 15089 { | |
| 625 | 15090 CNT(i, 0) = CNT(i - 1, 0) + SCORE_DEL; |
| 323 | 15091 for (j = 1; j <= goodlen; ++j) |
| 15092 { | |
| 324 | 15093 #ifdef FEAT_MBYTE |
| 15094 if (has_mbyte) | |
| 15095 { | |
| 15096 bc = wbadword[i - 1]; | |
| 15097 gc = wgoodword[j - 1]; | |
| 15098 } | |
| 15099 else | |
| 15100 #endif | |
| 15101 { | |
| 15102 bc = badword[i - 1]; | |
| 15103 gc = goodword[j - 1]; | |
| 15104 } | |
| 323 | 15105 if (bc == gc) |
| 15106 CNT(i, j) = CNT(i - 1, j - 1); | |
| 15107 else | |
| 15108 { | |
| 15109 /* Use a better score when there is only a case difference. */ | |
| 324 | 15110 if (SPELL_TOFOLD(bc) == SPELL_TOFOLD(gc)) |
| 323 | 15111 CNT(i, j) = SCORE_ICASE + CNT(i - 1, j - 1); |
| 15112 else | |
| 625 | 15113 { |
| 15114 /* For a similar character use SCORE_SIMILAR. */ | |
| 15115 if (slang != NULL | |
| 15116 && slang->sl_has_map | |
| 15117 && similar_chars(slang, gc, bc)) | |
| 15118 CNT(i, j) = SCORE_SIMILAR + CNT(i - 1, j - 1); | |
| 15119 else | |
| 15120 CNT(i, j) = SCORE_SUBST + CNT(i - 1, j - 1); | |
| 15121 } | |
| 323 | 15122 |
| 324 | 15123 if (i > 1 && j > 1) |
| 323 | 15124 { |
| 324 | 15125 #ifdef FEAT_MBYTE |
| 15126 if (has_mbyte) | |
| 15127 { | |
| 15128 pbc = wbadword[i - 2]; | |
| 15129 pgc = wgoodword[j - 2]; | |
| 15130 } | |
| 15131 else | |
| 15132 #endif | |
| 15133 { | |
| 15134 pbc = badword[i - 2]; | |
| 15135 pgc = goodword[j - 2]; | |
| 15136 } | |
| 15137 if (bc == pgc && pbc == gc) | |
| 15138 { | |
| 15139 t = SCORE_SWAP + CNT(i - 2, j - 2); | |
| 15140 if (t < CNT(i, j)) | |
| 15141 CNT(i, j) = t; | |
| 15142 } | |
| 323 | 15143 } |
| 15144 t = SCORE_DEL + CNT(i - 1, j); | |
| 15145 if (t < CNT(i, j)) | |
| 15146 CNT(i, j) = t; | |
| 15147 t = SCORE_INS + CNT(i, j - 1); | |
| 15148 if (t < CNT(i, j)) | |
| 15149 CNT(i, j) = t; | |
| 15150 } | |
| 15151 } | |
| 15152 } | |
| 344 | 15153 |
| 15154 i = CNT(badlen - 1, goodlen - 1); | |
| 15155 vim_free(cnt); | |
| 15156 return i; | |
| 323 | 15157 } |
| 307 | 15158 |
| 625 | 15159 typedef struct |
| 15160 { | |
| 15161 int badi; | |
| 15162 int goodi; | |
| 15163 int score; | |
| 15164 } limitscore_T; | |
| 15165 | |
| 15166 /* | |
| 15167 * Like spell_edit_score(), but with a limit on the score to make it faster. | |
| 15168 * May return SCORE_MAXMAX when the score is higher than "limit". | |
| 15169 * | |
| 15170 * This uses a stack for the edits still to be tried. | |
| 15171 * The idea comes from Aspell leditdist.cpp. Rewritten in C and added support | |
| 15172 * for multi-byte characters. | |
| 15173 */ | |
| 15174 static int | |
| 15175 spell_edit_score_limit(slang, badword, goodword, limit) | |
| 15176 slang_T *slang; | |
| 15177 char_u *badword; | |
| 15178 char_u *goodword; | |
| 15179 int limit; | |
| 15180 { | |
| 15181 limitscore_T stack[10]; /* allow for over 3 * 2 edits */ | |
| 15182 int stackidx; | |
| 15183 int bi, gi; | |
| 15184 int bi2, gi2; | |
| 15185 int bc, gc; | |
| 15186 int score; | |
| 15187 int score_off; | |
| 15188 int minscore; | |
| 15189 int round; | |
| 15190 | |
| 15191 #ifdef FEAT_MBYTE | |
| 15192 /* Multi-byte characters require a bit more work, use a different function | |
| 15193 * to avoid testing "has_mbyte" quite often. */ | |
| 15194 if (has_mbyte) | |
| 15195 return spell_edit_score_limit_w(slang, badword, goodword, limit); | |
| 15196 #endif | |
| 15197 | |
| 15198 /* | |
| 15199 * The idea is to go from start to end over the words. So long as | |
| 15200 * characters are equal just continue, this always gives the lowest score. | |
| 15201 * When there is a difference try several alternatives. Each alternative | |
| 15202 * increases "score" for the edit distance. Some of the alternatives are | |
| 15203 * pushed unto a stack and tried later, some are tried right away. At the | |
| 15204 * end of the word the score for one alternative is known. The lowest | |
| 15205 * possible score is stored in "minscore". | |
| 15206 */ | |
| 15207 stackidx = 0; | |
| 15208 bi = 0; | |
| 15209 gi = 0; | |
| 15210 score = 0; | |
| 15211 minscore = limit + 1; | |
| 15212 | |
| 15213 for (;;) | |
| 15214 { | |
| 15215 /* Skip over an equal part, score remains the same. */ | |
| 15216 for (;;) | |
| 15217 { | |
| 15218 bc = badword[bi]; | |
| 15219 gc = goodword[gi]; | |
| 15220 if (bc != gc) /* stop at a char that's different */ | |
| 15221 break; | |
| 15222 if (bc == NUL) /* both words end */ | |
| 15223 { | |
| 15224 if (score < minscore) | |
| 15225 minscore = score; | |
| 15226 goto pop; /* do next alternative */ | |
| 15227 } | |
| 15228 ++bi; | |
| 15229 ++gi; | |
| 15230 } | |
| 15231 | |
| 15232 if (gc == NUL) /* goodword ends, delete badword chars */ | |
| 15233 { | |
| 15234 do | |
| 15235 { | |
| 15236 if ((score += SCORE_DEL) >= minscore) | |
| 15237 goto pop; /* do next alternative */ | |
| 15238 } while (badword[++bi] != NUL); | |
| 15239 minscore = score; | |
| 15240 } | |
| 15241 else if (bc == NUL) /* badword ends, insert badword chars */ | |
| 15242 { | |
| 15243 do | |
| 15244 { | |
| 15245 if ((score += SCORE_INS) >= minscore) | |
| 15246 goto pop; /* do next alternative */ | |
| 15247 } while (goodword[++gi] != NUL); | |
| 15248 minscore = score; | |
| 15249 } | |
| 15250 else /* both words continue */ | |
| 15251 { | |
| 15252 /* If not close to the limit, perform a change. Only try changes | |
| 15253 * that may lead to a lower score than "minscore". | |
| 15254 * round 0: try deleting a char from badword | |
| 15255 * round 1: try inserting a char in badword */ | |
| 15256 for (round = 0; round <= 1; ++round) | |
| 15257 { | |
| 15258 score_off = score + (round == 0 ? SCORE_DEL : SCORE_INS); | |
| 15259 if (score_off < minscore) | |
| 15260 { | |
| 15261 if (score_off + SCORE_EDIT_MIN >= minscore) | |
| 15262 { | |
| 15263 /* Near the limit, rest of the words must match. We | |
| 15264 * can check that right now, no need to push an item | |
| 15265 * onto the stack. */ | |
| 15266 bi2 = bi + 1 - round; | |
| 15267 gi2 = gi + round; | |
| 15268 while (goodword[gi2] == badword[bi2]) | |
| 15269 { | |
| 15270 if (goodword[gi2] == NUL) | |
| 15271 { | |
| 15272 minscore = score_off; | |
| 15273 break; | |
| 15274 } | |
| 15275 ++bi2; | |
| 15276 ++gi2; | |
| 15277 } | |
| 15278 } | |
| 15279 else | |
| 15280 { | |
| 15281 /* try deleting/inserting a character later */ | |
| 15282 stack[stackidx].badi = bi + 1 - round; | |
| 15283 stack[stackidx].goodi = gi + round; | |
| 15284 stack[stackidx].score = score_off; | |
| 15285 ++stackidx; | |
| 15286 } | |
| 15287 } | |
| 15288 } | |
| 15289 | |
| 15290 if (score + SCORE_SWAP < minscore) | |
| 15291 { | |
| 15292 /* If swapping two characters makes a match then the | |
| 15293 * substitution is more expensive, thus there is no need to | |
| 15294 * try both. */ | |
| 15295 if (gc == badword[bi + 1] && bc == goodword[gi + 1]) | |
| 15296 { | |
| 15297 /* Swap two characters, that is: skip them. */ | |
| 15298 gi += 2; | |
| 15299 bi += 2; | |
| 15300 score += SCORE_SWAP; | |
| 15301 continue; | |
| 15302 } | |
| 15303 } | |
| 15304 | |
| 15305 /* Substitute one character for another which is the same | |
| 15306 * thing as deleting a character from both goodword and badword. | |
| 15307 * Use a better score when there is only a case difference. */ | |
| 15308 if (SPELL_TOFOLD(bc) == SPELL_TOFOLD(gc)) | |
| 15309 score += SCORE_ICASE; | |
| 15310 else | |
| 15311 { | |
| 15312 /* For a similar character use SCORE_SIMILAR. */ | |
| 15313 if (slang != NULL | |
| 15314 && slang->sl_has_map | |
| 15315 && similar_chars(slang, gc, bc)) | |
| 15316 score += SCORE_SIMILAR; | |
| 15317 else | |
| 15318 score += SCORE_SUBST; | |
| 15319 } | |
| 15320 | |
| 15321 if (score < minscore) | |
| 15322 { | |
| 15323 /* Do the substitution. */ | |
| 15324 ++gi; | |
| 15325 ++bi; | |
| 15326 continue; | |
| 15327 } | |
| 15328 } | |
| 15329 pop: | |
| 15330 /* | |
| 15331 * Get here to try the next alternative, pop it from the stack. | |
| 15332 */ | |
| 15333 if (stackidx == 0) /* stack is empty, finished */ | |
| 15334 break; | |
| 15335 | |
| 15336 /* pop an item from the stack */ | |
| 15337 --stackidx; | |
| 15338 gi = stack[stackidx].goodi; | |
| 15339 bi = stack[stackidx].badi; | |
| 15340 score = stack[stackidx].score; | |
| 15341 } | |
| 15342 | |
| 15343 /* When the score goes over "limit" it may actually be much higher. | |
| 15344 * Return a very large number to avoid going below the limit when giving a | |
| 15345 * bonus. */ | |
| 15346 if (minscore > limit) | |
| 15347 return SCORE_MAXMAX; | |
| 15348 return minscore; | |
| 15349 } | |
| 15350 | |
| 15351 #ifdef FEAT_MBYTE | |
| 15352 /* | |
| 15353 * Multi-byte version of spell_edit_score_limit(). | |
| 15354 * Keep it in sync with the above! | |
| 15355 */ | |
| 15356 static int | |
| 15357 spell_edit_score_limit_w(slang, badword, goodword, limit) | |
| 15358 slang_T *slang; | |
| 15359 char_u *badword; | |
| 15360 char_u *goodword; | |
| 15361 int limit; | |
| 15362 { | |
| 15363 limitscore_T stack[10]; /* allow for over 3 * 2 edits */ | |
| 15364 int stackidx; | |
| 15365 int bi, gi; | |
| 15366 int bi2, gi2; | |
| 15367 int bc, gc; | |
| 15368 int score; | |
| 15369 int score_off; | |
| 15370 int minscore; | |
| 15371 int round; | |
| 15372 char_u *p; | |
| 15373 int wbadword[MAXWLEN]; | |
| 15374 int wgoodword[MAXWLEN]; | |
| 15375 | |
| 15376 /* Get the characters from the multi-byte strings and put them in an | |
| 15377 * int array for easy access. */ | |
| 15378 bi = 0; | |
| 15379 for (p = badword; *p != NUL; ) | |
| 15380 wbadword[bi++] = mb_cptr2char_adv(&p); | |
| 15381 wbadword[bi++] = 0; | |
| 15382 gi = 0; | |
| 15383 for (p = goodword; *p != NUL; ) | |
| 15384 wgoodword[gi++] = mb_cptr2char_adv(&p); | |
| 15385 wgoodword[gi++] = 0; | |
| 15386 | |
| 15387 /* | |
| 15388 * The idea is to go from start to end over the words. So long as | |
| 15389 * characters are equal just continue, this always gives the lowest score. | |
| 15390 * When there is a difference try several alternatives. Each alternative | |
| 15391 * increases "score" for the edit distance. Some of the alternatives are | |
| 15392 * pushed unto a stack and tried later, some are tried right away. At the | |
| 15393 * end of the word the score for one alternative is known. The lowest | |
| 15394 * possible score is stored in "minscore". | |
| 15395 */ | |
| 15396 stackidx = 0; | |
| 15397 bi = 0; | |
| 15398 gi = 0; | |
| 15399 score = 0; | |
| 15400 minscore = limit + 1; | |
| 15401 | |
| 15402 for (;;) | |
| 15403 { | |
| 15404 /* Skip over an equal part, score remains the same. */ | |
| 15405 for (;;) | |
| 15406 { | |
| 15407 bc = wbadword[bi]; | |
| 15408 gc = wgoodword[gi]; | |
| 15409 | |
| 15410 if (bc != gc) /* stop at a char that's different */ | |
| 15411 break; | |
| 15412 if (bc == NUL) /* both words end */ | |
| 15413 { | |
| 15414 if (score < minscore) | |
| 15415 minscore = score; | |
| 15416 goto pop; /* do next alternative */ | |
| 15417 } | |
| 15418 ++bi; | |
| 15419 ++gi; | |
| 15420 } | |
| 15421 | |
| 15422 if (gc == NUL) /* goodword ends, delete badword chars */ | |
| 15423 { | |
| 15424 do | |
| 15425 { | |
| 15426 if ((score += SCORE_DEL) >= minscore) | |
| 15427 goto pop; /* do next alternative */ | |
| 15428 } while (wbadword[++bi] != NUL); | |
| 15429 minscore = score; | |
| 15430 } | |
| 15431 else if (bc == NUL) /* badword ends, insert badword chars */ | |
| 15432 { | |
| 15433 do | |
| 15434 { | |
| 15435 if ((score += SCORE_INS) >= minscore) | |
| 15436 goto pop; /* do next alternative */ | |
| 15437 } while (wgoodword[++gi] != NUL); | |
| 15438 minscore = score; | |
| 15439 } | |
| 15440 else /* both words continue */ | |
| 15441 { | |
| 15442 /* If not close to the limit, perform a change. Only try changes | |
| 15443 * that may lead to a lower score than "minscore". | |
| 15444 * round 0: try deleting a char from badword | |
| 15445 * round 1: try inserting a char in badword */ | |
| 15446 for (round = 0; round <= 1; ++round) | |
| 15447 { | |
| 15448 score_off = score + (round == 0 ? SCORE_DEL : SCORE_INS); | |
| 15449 if (score_off < minscore) | |
| 15450 { | |
| 15451 if (score_off + SCORE_EDIT_MIN >= minscore) | |
| 15452 { | |
| 15453 /* Near the limit, rest of the words must match. We | |
| 15454 * can check that right now, no need to push an item | |
| 15455 * onto the stack. */ | |
| 15456 bi2 = bi + 1 - round; | |
| 15457 gi2 = gi + round; | |
| 15458 while (wgoodword[gi2] == wbadword[bi2]) | |
| 15459 { | |
| 15460 if (wgoodword[gi2] == NUL) | |
| 15461 { | |
| 15462 minscore = score_off; | |
| 15463 break; | |
| 15464 } | |
| 15465 ++bi2; | |
| 15466 ++gi2; | |
| 15467 } | |
| 15468 } | |
| 15469 else | |
| 15470 { | |
| 15471 /* try deleting a character from badword later */ | |
| 15472 stack[stackidx].badi = bi + 1 - round; | |
| 15473 stack[stackidx].goodi = gi + round; | |
| 15474 stack[stackidx].score = score_off; | |
| 15475 ++stackidx; | |
| 15476 } | |
| 15477 } | |
| 15478 } | |
| 15479 | |
| 15480 if (score + SCORE_SWAP < minscore) | |
| 15481 { | |
| 15482 /* If swapping two characters makes a match then the | |
| 15483 * substitution is more expensive, thus there is no need to | |
| 15484 * try both. */ | |
| 15485 if (gc == wbadword[bi + 1] && bc == wgoodword[gi + 1]) | |
| 15486 { | |
| 15487 /* Swap two characters, that is: skip them. */ | |
| 15488 gi += 2; | |
| 15489 bi += 2; | |
| 15490 score += SCORE_SWAP; | |
| 15491 continue; | |
| 15492 } | |
| 15493 } | |
| 15494 | |
| 15495 /* Substitute one character for another which is the same | |
| 15496 * thing as deleting a character from both goodword and badword. | |
| 15497 * Use a better score when there is only a case difference. */ | |
| 15498 if (SPELL_TOFOLD(bc) == SPELL_TOFOLD(gc)) | |
| 15499 score += SCORE_ICASE; | |
| 15500 else | |
| 15501 { | |
| 15502 /* For a similar character use SCORE_SIMILAR. */ | |
| 15503 if (slang != NULL | |
| 15504 && slang->sl_has_map | |
| 15505 && similar_chars(slang, gc, bc)) | |
| 15506 score += SCORE_SIMILAR; | |
| 15507 else | |
| 15508 score += SCORE_SUBST; | |
| 15509 } | |
| 15510 | |
| 15511 if (score < minscore) | |
| 15512 { | |
| 15513 /* Do the substitution. */ | |
| 15514 ++gi; | |
| 15515 ++bi; | |
| 15516 continue; | |
| 15517 } | |
| 15518 } | |
| 15519 pop: | |
| 15520 /* | |
| 15521 * Get here to try the next alternative, pop it from the stack. | |
| 15522 */ | |
| 15523 if (stackidx == 0) /* stack is empty, finished */ | |
| 15524 break; | |
| 15525 | |
| 15526 /* pop an item from the stack */ | |
| 15527 --stackidx; | |
| 15528 gi = stack[stackidx].goodi; | |
| 15529 bi = stack[stackidx].badi; | |
| 15530 score = stack[stackidx].score; | |
| 15531 } | |
| 15532 | |
| 15533 /* When the score goes over "limit" it may actually be much higher. | |
| 15534 * Return a very large number to avoid going below the limit when giving a | |
| 15535 * bonus. */ | |
| 15536 if (minscore > limit) | |
| 15537 return SCORE_MAXMAX; | |
| 15538 return minscore; | |
| 15539 } | |
| 15540 #endif | |
| 15541 | |
| 714 | 15542 /* |
| 15543 * ":spellinfo" | |
| 15544 */ | |
| 15545 void | |
| 15546 ex_spellinfo(eap) | |
| 1883 | 15547 exarg_T *eap UNUSED; |
| 714 | 15548 { |
| 15549 int lpi; | |
| 15550 langp_T *lp; | |
| 15551 char_u *p; | |
| 15552 | |
| 15553 if (no_spell_checking(curwin)) | |
| 15554 return; | |
| 15555 | |
| 15556 msg_start(); | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
15557 for (lpi = 0; lpi < curwin->w_s->b_langp.ga_len && !got_int; ++lpi) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
15558 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
15559 lp = LANGP_ENTRY(curwin->w_s->b_langp, lpi); |
| 714 | 15560 msg_puts((char_u *)"file: "); |
| 15561 msg_puts(lp->lp_slang->sl_fname); | |
| 15562 msg_putchar('\n'); | |
| 15563 p = lp->lp_slang->sl_info; | |
| 15564 if (p != NULL) | |
| 15565 { | |
| 15566 msg_puts(p); | |
| 15567 msg_putchar('\n'); | |
| 15568 } | |
| 15569 } | |
| 15570 msg_end(); | |
| 15571 } | |
| 15572 | |
| 625 | 15573 #define DUMPFLAG_KEEPCASE 1 /* round 2: keep-case tree */ |
| 15574 #define DUMPFLAG_COUNT 2 /* include word count */ | |
| 701 | 15575 #define DUMPFLAG_ICASE 4 /* ignore case when finding matches */ |
| 710 | 15576 #define DUMPFLAG_ONECAP 8 /* pattern starts with capital */ |
| 15577 #define DUMPFLAG_ALLCAP 16 /* pattern is all capitals */ | |
| 625 | 15578 |
| 351 | 15579 /* |
| 15580 * ":spelldump" | |
| 15581 */ | |
| 15582 void | |
| 15583 ex_spelldump(eap) | |
| 15584 exarg_T *eap; | |
| 15585 { | |
| 701 | 15586 if (no_spell_checking(curwin)) |
| 15587 return; | |
| 15588 | |
| 15589 /* Create a new empty buffer by splitting the window. */ | |
| 15590 do_cmdline_cmd((char_u *)"new"); | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
15591 if (!bufempty() || !buf_valid(curbuf)) |
| 701 | 15592 return; |
| 15593 | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
15594 spell_dump_compl(NULL, 0, NULL, eap->forceit ? DUMPFLAG_COUNT : 0); |
| 701 | 15595 |
| 15596 /* Delete the empty line that we started with. */ | |
| 15597 if (curbuf->b_ml.ml_line_count > 1) | |
| 15598 ml_delete(curbuf->b_ml.ml_line_count, FALSE); | |
| 15599 | |
| 15600 redraw_later(NOT_VALID); | |
| 15601 } | |
| 15602 | |
| 15603 /* | |
| 15604 * Go through all possible words and: | |
| 15605 * 1. When "pat" is NULL: dump a list of all words in the current buffer. | |
| 15606 * "ic" and "dir" are not used. | |
| 15607 * 2. When "pat" is not NULL: add matching words to insert mode completion. | |
| 15608 */ | |
| 15609 void | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
15610 spell_dump_compl(pat, ic, dir, dumpflags_arg) |
| 701 | 15611 char_u *pat; /* leading part of the word */ |
| 15612 int ic; /* ignore case */ | |
| 15613 int *dir; /* direction for adding matches */ | |
| 15614 int dumpflags_arg; /* DUMPFLAG_* */ | |
| 15615 { | |
| 351 | 15616 langp_T *lp; |
| 15617 slang_T *slang; | |
| 15618 idx_T arridx[MAXWLEN]; | |
| 15619 int curi[MAXWLEN]; | |
| 15620 char_u word[MAXWLEN]; | |
| 15621 int c; | |
| 15622 char_u *byts; | |
| 15623 idx_T *idxs; | |
| 15624 linenr_T lnum = 0; | |
| 15625 int round; | |
| 15626 int depth; | |
| 15627 int n; | |
| 15628 int flags; | |
| 381 | 15629 char_u *region_names = NULL; /* region names being used */ |
| 15630 int do_region = TRUE; /* dump region names and numbers */ | |
| 15631 char_u *p; | |
| 500 | 15632 int lpi; |
| 701 | 15633 int dumpflags = dumpflags_arg; |
| 15634 int patlen; | |
| 15635 | |
| 710 | 15636 /* When ignoring case or when the pattern starts with capital pass this on |
| 15637 * to dump_word(). */ | |
| 15638 if (pat != NULL) | |
| 15639 { | |
| 15640 if (ic) | |
| 15641 dumpflags |= DUMPFLAG_ICASE; | |
| 15642 else | |
| 15643 { | |
| 15644 n = captype(pat, NULL); | |
| 15645 if (n == WF_ONECAP) | |
| 15646 dumpflags |= DUMPFLAG_ONECAP; | |
| 15647 else if (n == WF_ALLCAP | |
| 15648 #ifdef FEAT_MBYTE | |
| 714 | 15649 && (int)STRLEN(pat) > mb_ptr2len(pat) |
| 710 | 15650 #else |
| 714 | 15651 && (int)STRLEN(pat) > 1 |
| 710 | 15652 #endif |
| 15653 ) | |
| 15654 dumpflags |= DUMPFLAG_ALLCAP; | |
| 15655 } | |
| 15656 } | |
| 351 | 15657 |
| 381 | 15658 /* Find out if we can support regions: All languages must support the same |
| 15659 * regions or none at all. */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
15660 for (lpi = 0; lpi < curwin->w_s->b_langp.ga_len; ++lpi) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
15661 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
15662 lp = LANGP_ENTRY(curwin->w_s->b_langp, lpi); |
| 381 | 15663 p = lp->lp_slang->sl_regions; |
| 15664 if (p[0] != 0) | |
| 15665 { | |
| 15666 if (region_names == NULL) /* first language with regions */ | |
| 15667 region_names = p; | |
| 15668 else if (STRCMP(region_names, p) != 0) | |
| 15669 { | |
| 15670 do_region = FALSE; /* region names are different */ | |
| 15671 break; | |
| 15672 } | |
| 15673 } | |
| 15674 } | |
| 15675 | |
| 15676 if (do_region && region_names != NULL) | |
| 15677 { | |
| 701 | 15678 if (pat == NULL) |
| 15679 { | |
| 15680 vim_snprintf((char *)IObuff, IOSIZE, "/regions=%s", region_names); | |
| 15681 ml_append(lnum++, IObuff, (colnr_T)0, FALSE); | |
| 15682 } | |
| 381 | 15683 } |
| 15684 else | |
| 15685 do_region = FALSE; | |
| 15686 | |
| 15687 /* | |
| 15688 * Loop over all files loaded for the entries in 'spelllang'. | |
| 15689 */ | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
15690 for (lpi = 0; lpi < curwin->w_s->b_langp.ga_len; ++lpi) |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
15691 { |
|
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
15692 lp = LANGP_ENTRY(curwin->w_s->b_langp, lpi); |
| 351 | 15693 slang = lp->lp_slang; |
| 500 | 15694 if (slang->sl_fbyts == NULL) /* reloading failed */ |
| 15695 continue; | |
| 351 | 15696 |
| 701 | 15697 if (pat == NULL) |
| 15698 { | |
| 15699 vim_snprintf((char *)IObuff, IOSIZE, "# file: %s", slang->sl_fname); | |
| 15700 ml_append(lnum++, IObuff, (colnr_T)0, FALSE); | |
| 15701 } | |
| 15702 | |
| 15703 /* When matching with a pattern and there are no prefixes only use | |
| 15704 * parts of the tree that match "pat". */ | |
| 15705 if (pat != NULL && slang->sl_pbyts == NULL) | |
| 835 | 15706 patlen = (int)STRLEN(pat); |
| 701 | 15707 else |
| 840 | 15708 patlen = -1; |
| 351 | 15709 |
| 15710 /* round 1: case-folded tree | |
| 15711 * round 2: keep-case tree */ | |
| 15712 for (round = 1; round <= 2; ++round) | |
| 15713 { | |
| 15714 if (round == 1) | |
| 15715 { | |
| 701 | 15716 dumpflags &= ~DUMPFLAG_KEEPCASE; |
| 351 | 15717 byts = slang->sl_fbyts; |
| 15718 idxs = slang->sl_fidxs; | |
| 15719 } | |
| 15720 else | |
| 15721 { | |
| 701 | 15722 dumpflags |= DUMPFLAG_KEEPCASE; |
| 351 | 15723 byts = slang->sl_kbyts; |
| 15724 idxs = slang->sl_kidxs; | |
| 15725 } | |
| 15726 if (byts == NULL) | |
| 15727 continue; /* array is empty */ | |
| 15728 | |
| 15729 depth = 0; | |
| 15730 arridx[0] = 0; | |
| 15731 curi[0] = 1; | |
| 701 | 15732 while (depth >= 0 && !got_int |
| 15733 && (pat == NULL || !compl_interrupted)) | |
| 351 | 15734 { |
| 15735 if (curi[depth] > byts[arridx[depth]]) | |
| 15736 { | |
| 15737 /* Done all bytes at this node, go up one level. */ | |
| 15738 --depth; | |
| 15739 line_breakcheck(); | |
| 715 | 15740 ins_compl_check_keys(50); |
| 351 | 15741 } |
| 15742 else | |
| 15743 { | |
| 15744 /* Do one more byte at this node. */ | |
| 15745 n = arridx[depth] + curi[depth]; | |
| 15746 ++curi[depth]; | |
| 15747 c = byts[n]; | |
| 15748 if (c == 0) | |
| 15749 { | |
| 15750 /* End of word, deal with the word. | |
| 15751 * Don't use keep-case words in the fold-case tree, | |
| 15752 * they will appear in the keep-case tree. | |
| 15753 * Only use the word when the region matches. */ | |
| 15754 flags = (int)idxs[n]; | |
| 15755 if ((round == 2 || (flags & WF_KEEPCAP) == 0) | |
| 500 | 15756 && (flags & WF_NEEDCOMP) == 0 |
| 381 | 15757 && (do_region |
| 15758 || (flags & WF_REGION) == 0 | |
| 390 | 15759 || (((unsigned)flags >> 16) |
| 351 | 15760 & lp->lp_region) != 0)) |
| 15761 { | |
| 15762 word[depth] = NUL; | |
| 381 | 15763 if (!do_region) |
| 15764 flags &= ~WF_REGION; | |
| 355 | 15765 |
| 15766 /* Dump the basic word if there is no prefix or | |
| 15767 * when it's the first one. */ | |
| 390 | 15768 c = (unsigned)flags >> 24; |
| 355 | 15769 if (c == 0 || curi[depth] == 2) |
| 701 | 15770 { |
| 15771 dump_word(slang, word, pat, dir, | |
| 15772 dumpflags, flags, lnum); | |
| 15773 if (pat == NULL) | |
| 15774 ++lnum; | |
| 15775 } | |
| 351 | 15776 |
| 15777 /* Apply the prefix, if there is one. */ | |
| 355 | 15778 if (c != 0) |
| 701 | 15779 lnum = dump_prefixes(slang, word, pat, dir, |
| 15780 dumpflags, flags, lnum); | |
| 351 | 15781 } |
| 15782 } | |
| 15783 else | |
| 15784 { | |
| 15785 /* Normal char, go one level deeper. */ | |
| 15786 word[depth++] = c; | |
| 15787 arridx[depth] = idxs[n]; | |
| 15788 curi[depth] = 1; | |
| 701 | 15789 |
| 15790 /* Check if this characters matches with the pattern. | |
| 15791 * If not skip the whole tree below it. | |
| 710 | 15792 * Always ignore case here, dump_word() will check |
| 15793 * proper case later. This isn't exactly right when | |
| 15794 * length changes for multi-byte characters with | |
| 15795 * ignore case... */ | |
| 15796 if (depth <= patlen | |
| 15797 && MB_STRNICMP(word, pat, depth) != 0) | |
| 701 | 15798 --depth; |
| 351 | 15799 } |
| 15800 } | |
| 15801 } | |
| 15802 } | |
| 15803 } | |
| 15804 } | |
| 15805 | |
| 15806 /* | |
| 15807 * Dump one word: apply case modifications and append a line to the buffer. | |
| 701 | 15808 * When "lnum" is zero add insert mode completion. |
| 15809 */ | |
| 15810 static void | |
| 710 | 15811 dump_word(slang, word, pat, dir, dumpflags, wordflags, lnum) |
| 625 | 15812 slang_T *slang; |
| 351 | 15813 char_u *word; |
| 701 | 15814 char_u *pat; |
| 15815 int *dir; | |
| 625 | 15816 int dumpflags; |
| 710 | 15817 int wordflags; |
| 351 | 15818 linenr_T lnum; |
| 15819 { | |
| 15820 int keepcap = FALSE; | |
| 15821 char_u *p; | |
| 625 | 15822 char_u *tw; |
| 351 | 15823 char_u cword[MAXWLEN]; |
| 381 | 15824 char_u badword[MAXWLEN + 10]; |
| 15825 int i; | |
| 710 | 15826 int flags = wordflags; |
| 15827 | |
| 15828 if (dumpflags & DUMPFLAG_ONECAP) | |
| 15829 flags |= WF_ONECAP; | |
| 15830 if (dumpflags & DUMPFLAG_ALLCAP) | |
| 15831 flags |= WF_ALLCAP; | |
| 351 | 15832 |
| 625 | 15833 if ((dumpflags & DUMPFLAG_KEEPCASE) == 0 && (flags & WF_CAPMASK) != 0) |
| 351 | 15834 { |
| 15835 /* Need to fix case according to "flags". */ | |
| 15836 make_case_word(word, cword, flags); | |
| 15837 p = cword; | |
| 15838 } | |
| 15839 else | |
| 15840 { | |
| 15841 p = word; | |
| 625 | 15842 if ((dumpflags & DUMPFLAG_KEEPCASE) |
| 15843 && ((captype(word, NULL) & WF_KEEPCAP) == 0 | |
| 389 | 15844 || (flags & WF_FIXCAP) != 0)) |
| 351 | 15845 keepcap = TRUE; |
| 15846 } | |
| 625 | 15847 tw = p; |
| 351 | 15848 |
| 701 | 15849 if (pat == NULL) |
| 15850 { | |
| 15851 /* Add flags and regions after a slash. */ | |
| 15852 if ((flags & (WF_BANNED | WF_RARE | WF_REGION)) || keepcap) | |
| 15853 { | |
| 15854 STRCPY(badword, p); | |
| 15855 STRCAT(badword, "/"); | |
| 15856 if (keepcap) | |
| 15857 STRCAT(badword, "="); | |
| 15858 if (flags & WF_BANNED) | |
| 15859 STRCAT(badword, "!"); | |
| 15860 else if (flags & WF_RARE) | |
| 15861 STRCAT(badword, "?"); | |
| 15862 if (flags & WF_REGION) | |
| 15863 for (i = 0; i < 7; ++i) | |
| 15864 if (flags & (0x10000 << i)) | |
| 15865 sprintf((char *)badword + STRLEN(badword), "%d", i + 1); | |
| 15866 p = badword; | |
| 15867 } | |
| 15868 | |
| 15869 if (dumpflags & DUMPFLAG_COUNT) | |
| 15870 { | |
| 15871 hashitem_T *hi; | |
| 15872 | |
| 15873 /* Include the word count for ":spelldump!". */ | |
| 15874 hi = hash_find(&slang->sl_wordcount, tw); | |
| 15875 if (!HASHITEM_EMPTY(hi)) | |
| 15876 { | |
| 15877 vim_snprintf((char *)IObuff, IOSIZE, "%s\t%d", | |
| 625 | 15878 tw, HI2WC(hi)->wc_count); |
| 701 | 15879 p = IObuff; |
| 15880 } | |
| 15881 } | |
| 15882 | |
| 15883 ml_append(lnum, p, (colnr_T)0, FALSE); | |
| 15884 } | |
| 710 | 15885 else if (((dumpflags & DUMPFLAG_ICASE) |
| 15886 ? MB_STRNICMP(p, pat, STRLEN(pat)) == 0 | |
| 15887 : STRNCMP(p, pat, STRLEN(pat)) == 0) | |
| 701 | 15888 && ins_compl_add_infercase(p, (int)STRLEN(p), |
| 942 | 15889 p_ic, NULL, *dir, 0) == OK) |
| 710 | 15890 /* if dir was BACKWARD then honor it just once */ |
| 15891 *dir = FORWARD; | |
| 351 | 15892 } |
| 15893 | |
| 15894 /* | |
| 372 | 15895 * For ":spelldump": Find matching prefixes for "word". Prepend each to |
| 15896 * "word" and append a line to the buffer. | |
| 701 | 15897 * When "lnum" is zero add insert mode completion. |
| 351 | 15898 * Return the updated line number. |
| 15899 */ | |
| 15900 static linenr_T | |
| 701 | 15901 dump_prefixes(slang, word, pat, dir, dumpflags, flags, startlnum) |
| 351 | 15902 slang_T *slang; |
| 15903 char_u *word; /* case-folded word */ | |
| 701 | 15904 char_u *pat; |
| 15905 int *dir; | |
| 625 | 15906 int dumpflags; |
| 351 | 15907 int flags; /* flags with prefix ID */ |
| 15908 linenr_T startlnum; | |
| 15909 { | |
| 15910 idx_T arridx[MAXWLEN]; | |
| 15911 int curi[MAXWLEN]; | |
| 15912 char_u prefix[MAXWLEN]; | |
| 455 | 15913 char_u word_up[MAXWLEN]; |
| 15914 int has_word_up = FALSE; | |
| 351 | 15915 int c; |
| 15916 char_u *byts; | |
| 15917 idx_T *idxs; | |
| 15918 linenr_T lnum = startlnum; | |
| 15919 int depth; | |
| 15920 int n; | |
| 15921 int len; | |
| 15922 int i; | |
| 15923 | |
| 626 | 15924 /* If the word starts with a lower-case letter make the word with an |
| 455 | 15925 * upper-case letter in word_up[]. */ |
| 15926 c = PTR2CHAR(word); | |
| 15927 if (SPELL_TOUPPER(c) != c) | |
| 15928 { | |
| 15929 onecap_copy(word, word_up, TRUE); | |
| 15930 has_word_up = TRUE; | |
| 15931 } | |
| 15932 | |
| 351 | 15933 byts = slang->sl_pbyts; |
| 15934 idxs = slang->sl_pidxs; | |
| 15935 if (byts != NULL) /* array not is empty */ | |
| 15936 { | |
| 15937 /* | |
| 15938 * Loop over all prefixes, building them byte-by-byte in prefix[]. | |
| 390 | 15939 * When at the end of a prefix check that it supports "flags". |
| 351 | 15940 */ |
| 15941 depth = 0; | |
| 15942 arridx[0] = 0; | |
| 15943 curi[0] = 1; | |
| 15944 while (depth >= 0 && !got_int) | |
| 15945 { | |
| 390 | 15946 n = arridx[depth]; |
| 15947 len = byts[n]; | |
| 15948 if (curi[depth] > len) | |
| 351 | 15949 { |
| 15950 /* Done all bytes at this node, go up one level. */ | |
| 15951 --depth; | |
| 15952 line_breakcheck(); | |
| 15953 } | |
| 15954 else | |
| 15955 { | |
| 15956 /* Do one more byte at this node. */ | |
| 390 | 15957 n += curi[depth]; |
| 351 | 15958 ++curi[depth]; |
| 15959 c = byts[n]; | |
| 15960 if (c == 0) | |
| 15961 { | |
| 15962 /* End of prefix, find out how many IDs there are. */ | |
| 15963 for (i = 1; i < len; ++i) | |
| 15964 if (byts[n + i] != 0) | |
| 15965 break; | |
| 15966 curi[depth] += i - 1; | |
| 15967 | |
| 455 | 15968 c = valid_word_prefix(i, n, flags, word, slang, FALSE); |
| 15969 if (c != 0) | |
| 351 | 15970 { |
| 376 | 15971 vim_strncpy(prefix + depth, word, MAXWLEN - depth - 1); |
| 701 | 15972 dump_word(slang, prefix, pat, dir, dumpflags, |
| 455 | 15973 (c & WF_RAREPFX) ? (flags | WF_RARE) |
| 701 | 15974 : flags, lnum); |
| 15975 if (lnum != 0) | |
| 15976 ++lnum; | |
| 351 | 15977 } |
| 455 | 15978 |
| 15979 /* Check for prefix that matches the word when the | |
| 15980 * first letter is upper-case, but only if the prefix has | |
| 15981 * a condition. */ | |
| 15982 if (has_word_up) | |
| 15983 { | |
| 15984 c = valid_word_prefix(i, n, flags, word_up, slang, | |
| 15985 TRUE); | |
| 15986 if (c != 0) | |
| 15987 { | |
| 15988 vim_strncpy(prefix + depth, word_up, | |
| 15989 MAXWLEN - depth - 1); | |
| 701 | 15990 dump_word(slang, prefix, pat, dir, dumpflags, |
| 455 | 15991 (c & WF_RAREPFX) ? (flags | WF_RARE) |
| 701 | 15992 : flags, lnum); |
| 15993 if (lnum != 0) | |
| 15994 ++lnum; | |
| 455 | 15995 } |
| 15996 } | |
| 351 | 15997 } |
| 15998 else | |
| 15999 { | |
| 16000 /* Normal char, go one level deeper. */ | |
| 16001 prefix[depth++] = c; | |
| 16002 arridx[depth] = idxs[n]; | |
| 16003 curi[depth] = 1; | |
| 16004 } | |
| 16005 } | |
| 16006 } | |
| 16007 } | |
| 16008 | |
| 16009 return lnum; | |
| 16010 } | |
| 16011 | |
| 498 | 16012 /* |
| 626 | 16013 * Move "p" to the end of word "start". |
| 16014 * Uses the spell-checking word characters. | |
| 498 | 16015 */ |
| 16016 char_u * | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
16017 spell_to_word_end(start, win) |
| 498 | 16018 char_u *start; |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
16019 win_T *win; |
| 498 | 16020 { |
| 16021 char_u *p = start; | |
| 16022 | |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
16023 while (*p != NUL && spell_iswordp(p, win)) |
| 498 | 16024 mb_ptr_adv(p); |
| 16025 return p; | |
| 16026 } | |
| 16027 | |
| 475 | 16028 #if defined(FEAT_INS_EXPAND) || defined(PROTO) |
| 16029 /* | |
| 626 | 16030 * For Insert mode completion CTRL-X s: |
| 16031 * Find start of the word in front of column "startcol". | |
| 16032 * We don't check if it is badly spelled, with completion we can only change | |
| 16033 * the word in front of the cursor. | |
| 475 | 16034 * Returns the column number of the word. |
| 16035 */ | |
| 16036 int | |
| 16037 spell_word_start(startcol) | |
| 16038 int startcol; | |
| 16039 { | |
| 16040 char_u *line; | |
| 16041 char_u *p; | |
| 16042 int col = 0; | |
| 16043 | |
| 498 | 16044 if (no_spell_checking(curwin)) |
| 475 | 16045 return startcol; |
| 16046 | |
| 16047 /* Find a word character before "startcol". */ | |
| 16048 line = ml_get_curline(); | |
| 16049 for (p = line + startcol; p > line; ) | |
| 16050 { | |
| 16051 mb_ptr_back(line, p); | |
| 16052 if (spell_iswordp_nmw(p)) | |
| 16053 break; | |
| 16054 } | |
| 16055 | |
| 16056 /* Go back to start of the word. */ | |
| 16057 while (p > line) | |
| 16058 { | |
| 835 | 16059 col = (int)(p - line); |
| 475 | 16060 mb_ptr_back(line, p); |
|
2250
1bac28a53fae
Add the conceal patch from Vince Negri.
Bram Moolenaar <bram@vim.org>
parents:
2229
diff
changeset
|
16061 if (!spell_iswordp(p, curwin)) |
| 475 | 16062 break; |
| 16063 col = 0; | |
| 16064 } | |
| 16065 | |
| 535 | 16066 return col; |
| 16067 } | |
| 16068 | |
| 16069 /* | |
| 16070 * Need to check for 'spellcapcheck' now, the word is removed before | |
| 16071 * expand_spelling() is called. Therefore the ugly global variable. | |
| 16072 */ | |
| 16073 static int spell_expand_need_cap; | |
| 16074 | |
| 16075 void | |
| 16076 spell_expand_check_cap(col) | |
| 16077 colnr_T col; | |
| 16078 { | |
| 475 | 16079 spell_expand_need_cap = check_need_cap(curwin->w_cursor.lnum, col); |
| 16080 } | |
| 16081 | |
| 16082 /* | |
| 16083 * Get list of spelling suggestions. | |
| 16084 * Used for Insert mode completion CTRL-X ?. | |
| 16085 * Returns the number of matches. The matches are in "matchp[]", array of | |
| 16086 * allocated strings. | |
| 16087 */ | |
| 16088 int | |
| 1872 | 16089 expand_spelling(lnum, pat, matchp) |
| 1883 | 16090 linenr_T lnum UNUSED; |
| 475 | 16091 char_u *pat; |
| 16092 char_u ***matchp; | |
| 16093 { | |
| 16094 garray_T ga; | |
| 16095 | |
| 625 | 16096 spell_suggest_list(&ga, pat, 100, spell_expand_need_cap, TRUE); |
| 475 | 16097 *matchp = ga.ga_data; |
| 16098 return ga.ga_len; | |
| 16099 } | |
| 16100 #endif | |
| 16101 | |
| 737 | 16102 #endif /* FEAT_SPELL */ |