comparison src/spellfile.c @ 9583:b0c7061d6439 v7.4.2069

commit https://github.com/vim/vim/commit/9ccfebddc3ff2a3c2853cf706fd4c26f639bf381 Author: Bram Moolenaar <Bram@vim.org> Date: Tue Jul 19 16:39:08 2016 +0200 patch 7.4.2069 Problem: spell.c is too big. Solution: Split it in spell file handling and spell checking.
author Christian Brabandt <cb@256bit.org>
date Tue, 19 Jul 2016 16:45:06 +0200
parents
children 4aead6a9b7a9
comparison
equal deleted inserted replaced
9582:96737caf272d 9583:b0c7061d6439
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 * spellfile.c: code for reading and writing spell files.
12 *
13 * See spell.c for information about spell checking.
14 */
15
16 /*
17 * Vim spell file format: <HEADER>
18 * <SECTIONS>
19 * <LWORDTREE>
20 * <KWORDTREE>
21 * <PREFIXTREE>
22 *
23 * <HEADER>: <fileID> <versionnr>
24 *
25 * <fileID> 8 bytes "VIMspell"
26 * <versionnr> 1 byte VIMSPELLVERSION
27 *
28 *
29 * Sections make it possible to add information to the .spl file without
30 * making it incompatible with previous versions. There are two kinds of
31 * sections:
32 * 1. Not essential for correct spell checking. E.g. for making suggestions.
33 * These are skipped when not supported.
34 * 2. Optional information, but essential for spell checking when present.
35 * E.g. conditions for affixes. When this section is present but not
36 * supported an error message is given.
37 *
38 * <SECTIONS>: <section> ... <sectionend>
39 *
40 * <section>: <sectionID> <sectionflags> <sectionlen> (section contents)
41 *
42 * <sectionID> 1 byte number from 0 to 254 identifying the section
43 *
44 * <sectionflags> 1 byte SNF_REQUIRED: this section is required for correct
45 * spell checking
46 *
47 * <sectionlen> 4 bytes length of section contents, MSB first
48 *
49 * <sectionend> 1 byte SN_END
50 *
51 *
52 * sectionID == SN_INFO: <infotext>
53 * <infotext> N bytes free format text with spell file info (version,
54 * website, etc)
55 *
56 * sectionID == SN_REGION: <regionname> ...
57 * <regionname> 2 bytes Up to 8 region names: ca, au, etc. Lower case.
58 * First <regionname> is region 1.
59 *
60 * sectionID == SN_CHARFLAGS: <charflagslen> <charflags>
61 * <folcharslen> <folchars>
62 * <charflagslen> 1 byte Number of bytes in <charflags> (should be 128).
63 * <charflags> N bytes List of flags (first one is for character 128):
64 * 0x01 word character CF_WORD
65 * 0x02 upper-case character CF_UPPER
66 * <folcharslen> 2 bytes Number of bytes in <folchars>.
67 * <folchars> N bytes Folded characters, first one is for character 128.
68 *
69 * sectionID == SN_MIDWORD: <midword>
70 * <midword> N bytes Characters that are word characters only when used
71 * in the middle of a word.
72 *
73 * sectionID == SN_PREFCOND: <prefcondcnt> <prefcond> ...
74 * <prefcondcnt> 2 bytes Number of <prefcond> items following.
75 * <prefcond> : <condlen> <condstr>
76 * <condlen> 1 byte Length of <condstr>.
77 * <condstr> N bytes Condition for the prefix.
78 *
79 * sectionID == SN_REP: <repcount> <rep> ...
80 * <repcount> 2 bytes number of <rep> items, MSB first.
81 * <rep> : <repfromlen> <repfrom> <reptolen> <repto>
82 * <repfromlen> 1 byte length of <repfrom>
83 * <repfrom> N bytes "from" part of replacement
84 * <reptolen> 1 byte length of <repto>
85 * <repto> N bytes "to" part of replacement
86 *
87 * sectionID == SN_REPSAL: <repcount> <rep> ...
88 * just like SN_REP but for soundfolded words
89 *
90 * sectionID == SN_SAL: <salflags> <salcount> <sal> ...
91 * <salflags> 1 byte flags for soundsalike conversion:
92 * SAL_F0LLOWUP
93 * SAL_COLLAPSE
94 * SAL_REM_ACCENTS
95 * <salcount> 2 bytes number of <sal> items following
96 * <sal> : <salfromlen> <salfrom> <saltolen> <salto>
97 * <salfromlen> 1 byte length of <salfrom>
98 * <salfrom> N bytes "from" part of soundsalike
99 * <saltolen> 1 byte length of <salto>
100 * <salto> N bytes "to" part of soundsalike
101 *
102 * sectionID == SN_SOFO: <sofofromlen> <sofofrom> <sofotolen> <sofoto>
103 * <sofofromlen> 2 bytes length of <sofofrom>
104 * <sofofrom> N bytes "from" part of soundfold
105 * <sofotolen> 2 bytes length of <sofoto>
106 * <sofoto> N bytes "to" part of soundfold
107 *
108 * sectionID == SN_SUGFILE: <timestamp>
109 * <timestamp> 8 bytes time in seconds that must match with .sug file
110 *
111 * sectionID == SN_NOSPLITSUGS: nothing
112 *
113 * sectionID == SN_NOCOMPOUNDSUGS: nothing
114 *
115 * sectionID == SN_WORDS: <word> ...
116 * <word> N bytes NUL terminated common word
117 *
118 * sectionID == SN_MAP: <mapstr>
119 * <mapstr> N bytes String with sequences of similar characters,
120 * separated by slashes.
121 *
122 * sectionID == SN_COMPOUND: <compmax> <compminlen> <compsylmax> <compoptions>
123 * <comppatcount> <comppattern> ... <compflags>
124 * <compmax> 1 byte Maximum nr of words in compound word.
125 * <compminlen> 1 byte Minimal word length for compounding.
126 * <compsylmax> 1 byte Maximum nr of syllables in compound word.
127 * <compoptions> 2 bytes COMP_ flags.
128 * <comppatcount> 2 bytes number of <comppattern> following
129 * <compflags> N bytes Flags from COMPOUNDRULE items, separated by
130 * slashes.
131 *
132 * <comppattern>: <comppatlen> <comppattext>
133 * <comppatlen> 1 byte length of <comppattext>
134 * <comppattext> N bytes end or begin chars from CHECKCOMPOUNDPATTERN
135 *
136 * sectionID == SN_NOBREAK: (empty, its presence is what matters)
137 *
138 * sectionID == SN_SYLLABLE: <syllable>
139 * <syllable> N bytes String from SYLLABLE item.
140 *
141 * <LWORDTREE>: <wordtree>
142 *
143 * <KWORDTREE>: <wordtree>
144 *
145 * <PREFIXTREE>: <wordtree>
146 *
147 *
148 * <wordtree>: <nodecount> <nodedata> ...
149 *
150 * <nodecount> 4 bytes Number of nodes following. MSB first.
151 *
152 * <nodedata>: <siblingcount> <sibling> ...
153 *
154 * <siblingcount> 1 byte Number of siblings in this node. The siblings
155 * follow in sorted order.
156 *
157 * <sibling>: <byte> [ <nodeidx> <xbyte>
158 * | <flags> [<flags2>] [<region>] [<affixID>]
159 * | [<pflags>] <affixID> <prefcondnr> ]
160 *
161 * <byte> 1 byte Byte value of the sibling. Special cases:
162 * BY_NOFLAGS: End of word without flags and for all
163 * regions.
164 * For PREFIXTREE <affixID> and
165 * <prefcondnr> follow.
166 * BY_FLAGS: End of word, <flags> follow.
167 * For PREFIXTREE <pflags>, <affixID>
168 * and <prefcondnr> follow.
169 * BY_FLAGS2: End of word, <flags> and <flags2>
170 * follow. Not used in PREFIXTREE.
171 * BY_INDEX: Child of sibling is shared, <nodeidx>
172 * and <xbyte> follow.
173 *
174 * <nodeidx> 3 bytes Index of child for this sibling, MSB first.
175 *
176 * <xbyte> 1 byte byte value of the sibling.
177 *
178 * <flags> 1 byte bitmask of:
179 * WF_ALLCAP word must have only capitals
180 * WF_ONECAP first char of word must be capital
181 * WF_KEEPCAP keep-case word
182 * WF_FIXCAP keep-case word, all caps not allowed
183 * WF_RARE rare word
184 * WF_BANNED bad word
185 * WF_REGION <region> follows
186 * WF_AFX <affixID> follows
187 *
188 * <flags2> 1 byte Bitmask of:
189 * WF_HAS_AFF >> 8 word includes affix
190 * WF_NEEDCOMP >> 8 word only valid in compound
191 * WF_NOSUGGEST >> 8 word not used for suggestions
192 * WF_COMPROOT >> 8 word already a compound
193 * WF_NOCOMPBEF >> 8 no compounding before this word
194 * WF_NOCOMPAFT >> 8 no compounding after this word
195 *
196 * <pflags> 1 byte bitmask of:
197 * WFP_RARE rare prefix
198 * WFP_NC non-combining prefix
199 * WFP_UP letter after prefix made upper case
200 *
201 * <region> 1 byte Bitmask for regions in which word is valid. When
202 * omitted it's valid in all regions.
203 * Lowest bit is for region 1.
204 *
205 * <affixID> 1 byte ID of affix that can be used with this word. In
206 * PREFIXTREE used for the required prefix ID.
207 *
208 * <prefcondnr> 2 bytes Prefix condition number, index in <prefcond> list
209 * from HEADER.
210 *
211 * All text characters are in 'encoding', but stored as single bytes.
212 */
213
214 /*
215 * Vim .sug file format: <SUGHEADER>
216 * <SUGWORDTREE>
217 * <SUGTABLE>
218 *
219 * <SUGHEADER>: <fileID> <versionnr> <timestamp>
220 *
221 * <fileID> 6 bytes "VIMsug"
222 * <versionnr> 1 byte VIMSUGVERSION
223 * <timestamp> 8 bytes timestamp that must match with .spl file
224 *
225 *
226 * <SUGWORDTREE>: <wordtree> (see above, no flags or region used)
227 *
228 *
229 * <SUGTABLE>: <sugwcount> <sugline> ...
230 *
231 * <sugwcount> 4 bytes number of <sugline> following
232 *
233 * <sugline>: <sugnr> ... NUL
234 *
235 * <sugnr>: X bytes word number that results in this soundfolded word,
236 * stored as an offset to the previous number in as
237 * few bytes as possible, see offset2bytes())
238 */
239
240 #include "vim.h"
241
242 #if defined(FEAT_SPELL) || defined(PROTO)
243
244 #ifndef UNIX /* it's in os_unix.h for Unix */
245 # include <time.h> /* for time_t */
246 #endif
247
248 #ifndef UNIX /* it's in os_unix.h for Unix */
249 # include <time.h> /* for time_t */
250 #endif
251
252 /* Special byte values for <byte>. Some are only used in the tree for
253 * postponed prefixes, some only in the other trees. This is a bit messy... */
254 #define BY_NOFLAGS 0 /* end of word without flags or region; for
255 * postponed prefix: no <pflags> */
256 #define BY_INDEX 1 /* child is shared, index follows */
257 #define BY_FLAGS 2 /* end of word, <flags> byte follows; for
258 * postponed prefix: <pflags> follows */
259 #define BY_FLAGS2 3 /* end of word, <flags> and <flags2> bytes
260 * follow; never used in prefix tree */
261 #define BY_SPECIAL BY_FLAGS2 /* highest special byte value */
262
263 /* Flags used in .spl file for soundsalike flags. */
264 #define SAL_F0LLOWUP 1
265 #define SAL_COLLAPSE 2
266 #define SAL_REM_ACCENTS 4
267
268 #define VIMSPELLMAGIC "VIMspell" /* string at start of Vim spell file */
269 #define VIMSPELLMAGICL 8
270 #define VIMSPELLVERSION 50
271
272 /* Section IDs. Only renumber them when VIMSPELLVERSION changes! */
273 #define SN_REGION 0 /* <regionname> section */
274 #define SN_CHARFLAGS 1 /* charflags section */
275 #define SN_MIDWORD 2 /* <midword> section */
276 #define SN_PREFCOND 3 /* <prefcond> section */
277 #define SN_REP 4 /* REP items section */
278 #define SN_SAL 5 /* SAL items section */
279 #define SN_SOFO 6 /* soundfolding section */
280 #define SN_MAP 7 /* MAP items section */
281 #define SN_COMPOUND 8 /* compound words section */
282 #define SN_SYLLABLE 9 /* syllable section */
283 #define SN_NOBREAK 10 /* NOBREAK section */
284 #define SN_SUGFILE 11 /* timestamp for .sug file */
285 #define SN_REPSAL 12 /* REPSAL items section */
286 #define SN_WORDS 13 /* common words */
287 #define SN_NOSPLITSUGS 14 /* don't split word for suggestions */
288 #define SN_INFO 15 /* info section */
289 #define SN_NOCOMPOUNDSUGS 16 /* don't compound for suggestions */
290 #define SN_END 255 /* end of sections */
291
292 #define SNF_REQUIRED 1 /* <sectionflags>: required section */
293
294 #define CF_WORD 0x01
295 #define CF_UPPER 0x02
296
297 static int set_spell_finish(spelltab_T *new_st);
298 static int write_spell_prefcond(FILE *fd, garray_T *gap);
299 static char_u *read_cnt_string(FILE *fd, int cnt_bytes, int *lenp);
300 static int read_region_section(FILE *fd, slang_T *slang, int len);
301 static int read_charflags_section(FILE *fd);
302 static int read_prefcond_section(FILE *fd, slang_T *lp);
303 static int read_rep_section(FILE *fd, garray_T *gap, short *first);
304 static int read_sal_section(FILE *fd, slang_T *slang);
305 static int read_words_section(FILE *fd, slang_T *lp, int len);
306 static int read_sofo_section(FILE *fd, slang_T *slang);
307 static int read_compound(FILE *fd, slang_T *slang, int len);
308 static int set_sofo(slang_T *lp, char_u *from, char_u *to);
309 static void set_sal_first(slang_T *lp);
310 #ifdef FEAT_MBYTE
311 static int *mb_str2wide(char_u *s);
312 #endif
313 static int spell_read_tree(FILE *fd, char_u **bytsp, idx_T **idxsp, int prefixtree, int prefixcnt);
314 static idx_T read_tree_node(FILE *fd, char_u *byts, idx_T *idxs, int maxidx, idx_T startidx, int prefixtree, int maxprefcondnr);
315 static void spell_reload_one(char_u *fname, int added_word);
316 static void set_spell_charflags(char_u *flags, int cnt, char_u *upp);
317 static int set_spell_chartab(char_u *fol, char_u *low, char_u *upp);
318 static void set_map_str(slang_T *lp, char_u *map);
319
320
321 static char *e_spell_trunc = N_("E758: Truncated spell file");
322 static char *e_afftrailing = N_("Trailing text in %s line %d: %s");
323 static char *e_affname = N_("Affix name too long in %s line %d: %s");
324 static char *e_affform = N_("E761: Format error in affix file FOL, LOW or UPP");
325 static char *e_affrange = N_("E762: Character in FOL, LOW or UPP is out of range");
326 static char *msg_compressing = N_("Compressing word tree...");
327
328 /*
329 * Load one spell file and store the info into a slang_T.
330 *
331 * This is invoked in three ways:
332 * - From spell_load_cb() to load a spell file for the first time. "lang" is
333 * the language name, "old_lp" is NULL. Will allocate an slang_T.
334 * - To reload a spell file that was changed. "lang" is NULL and "old_lp"
335 * points to the existing slang_T.
336 * - Just after writing a .spl file; it's read back to produce the .sug file.
337 * "old_lp" is NULL and "lang" is NULL. Will allocate an slang_T.
338 *
339 * Returns the slang_T the spell file was loaded into. NULL for error.
340 */
341 slang_T *
342 spell_load_file(
343 char_u *fname,
344 char_u *lang,
345 slang_T *old_lp,
346 int silent) /* no error if file doesn't exist */
347 {
348 FILE *fd;
349 char_u buf[VIMSPELLMAGICL];
350 char_u *p;
351 int i;
352 int n;
353 int len;
354 char_u *save_sourcing_name = sourcing_name;
355 linenr_T save_sourcing_lnum = sourcing_lnum;
356 slang_T *lp = NULL;
357 int c = 0;
358 int res;
359
360 fd = mch_fopen((char *)fname, "r");
361 if (fd == NULL)
362 {
363 if (!silent)
364 EMSG2(_(e_notopen), fname);
365 else if (p_verbose > 2)
366 {
367 verbose_enter();
368 smsg((char_u *)e_notopen, fname);
369 verbose_leave();
370 }
371 goto endFAIL;
372 }
373 if (p_verbose > 2)
374 {
375 verbose_enter();
376 smsg((char_u *)_("Reading spell file \"%s\""), fname);
377 verbose_leave();
378 }
379
380 if (old_lp == NULL)
381 {
382 lp = slang_alloc(lang);
383 if (lp == NULL)
384 goto endFAIL;
385
386 /* Remember the file name, used to reload the file when it's updated. */
387 lp->sl_fname = vim_strsave(fname);
388 if (lp->sl_fname == NULL)
389 goto endFAIL;
390
391 /* Check for .add.spl (_add.spl for VMS). */
392 lp->sl_add = strstr((char *)gettail(fname), SPL_FNAME_ADD) != NULL;
393 }
394 else
395 lp = old_lp;
396
397 /* Set sourcing_name, so that error messages mention the file name. */
398 sourcing_name = fname;
399 sourcing_lnum = 0;
400
401 /*
402 * <HEADER>: <fileID>
403 */
404 for (i = 0; i < VIMSPELLMAGICL; ++i)
405 buf[i] = getc(fd); /* <fileID> */
406 if (STRNCMP(buf, VIMSPELLMAGIC, VIMSPELLMAGICL) != 0)
407 {
408 EMSG(_("E757: This does not look like a spell file"));
409 goto endFAIL;
410 }
411 c = getc(fd); /* <versionnr> */
412 if (c < VIMSPELLVERSION)
413 {
414 EMSG(_("E771: Old spell file, needs to be updated"));
415 goto endFAIL;
416 }
417 else if (c > VIMSPELLVERSION)
418 {
419 EMSG(_("E772: Spell file is for newer version of Vim"));
420 goto endFAIL;
421 }
422
423
424 /*
425 * <SECTIONS>: <section> ... <sectionend>
426 * <section>: <sectionID> <sectionflags> <sectionlen> (section contents)
427 */
428 for (;;)
429 {
430 n = getc(fd); /* <sectionID> or <sectionend> */
431 if (n == SN_END)
432 break;
433 c = getc(fd); /* <sectionflags> */
434 len = get4c(fd); /* <sectionlen> */
435 if (len < 0)
436 goto truncerr;
437
438 res = 0;
439 switch (n)
440 {
441 case SN_INFO:
442 lp->sl_info = read_string(fd, len); /* <infotext> */
443 if (lp->sl_info == NULL)
444 goto endFAIL;
445 break;
446
447 case SN_REGION:
448 res = read_region_section(fd, lp, len);
449 break;
450
451 case SN_CHARFLAGS:
452 res = read_charflags_section(fd);
453 break;
454
455 case SN_MIDWORD:
456 lp->sl_midword = read_string(fd, len); /* <midword> */
457 if (lp->sl_midword == NULL)
458 goto endFAIL;
459 break;
460
461 case SN_PREFCOND:
462 res = read_prefcond_section(fd, lp);
463 break;
464
465 case SN_REP:
466 res = read_rep_section(fd, &lp->sl_rep, lp->sl_rep_first);
467 break;
468
469 case SN_REPSAL:
470 res = read_rep_section(fd, &lp->sl_repsal, lp->sl_repsal_first);
471 break;
472
473 case SN_SAL:
474 res = read_sal_section(fd, lp);
475 break;
476
477 case SN_SOFO:
478 res = read_sofo_section(fd, lp);
479 break;
480
481 case SN_MAP:
482 p = read_string(fd, len); /* <mapstr> */
483 if (p == NULL)
484 goto endFAIL;
485 set_map_str(lp, p);
486 vim_free(p);
487 break;
488
489 case SN_WORDS:
490 res = read_words_section(fd, lp, len);
491 break;
492
493 case SN_SUGFILE:
494 lp->sl_sugtime = get8ctime(fd); /* <timestamp> */
495 break;
496
497 case SN_NOSPLITSUGS:
498 lp->sl_nosplitsugs = TRUE;
499 break;
500
501 case SN_NOCOMPOUNDSUGS:
502 lp->sl_nocompoundsugs = TRUE;
503 break;
504
505 case SN_COMPOUND:
506 res = read_compound(fd, lp, len);
507 break;
508
509 case SN_NOBREAK:
510 lp->sl_nobreak = TRUE;
511 break;
512
513 case SN_SYLLABLE:
514 lp->sl_syllable = read_string(fd, len); /* <syllable> */
515 if (lp->sl_syllable == NULL)
516 goto endFAIL;
517 if (init_syl_tab(lp) == FAIL)
518 goto endFAIL;
519 break;
520
521 default:
522 /* Unsupported section. When it's required give an error
523 * message. When it's not required skip the contents. */
524 if (c & SNF_REQUIRED)
525 {
526 EMSG(_("E770: Unsupported section in spell file"));
527 goto endFAIL;
528 }
529 while (--len >= 0)
530 if (getc(fd) < 0)
531 goto truncerr;
532 break;
533 }
534 someerror:
535 if (res == SP_FORMERROR)
536 {
537 EMSG(_(e_format));
538 goto endFAIL;
539 }
540 if (res == SP_TRUNCERROR)
541 {
542 truncerr:
543 EMSG(_(e_spell_trunc));
544 goto endFAIL;
545 }
546 if (res == SP_OTHERERROR)
547 goto endFAIL;
548 }
549
550 /* <LWORDTREE> */
551 res = spell_read_tree(fd, &lp->sl_fbyts, &lp->sl_fidxs, FALSE, 0);
552 if (res != 0)
553 goto someerror;
554
555 /* <KWORDTREE> */
556 res = spell_read_tree(fd, &lp->sl_kbyts, &lp->sl_kidxs, FALSE, 0);
557 if (res != 0)
558 goto someerror;
559
560 /* <PREFIXTREE> */
561 res = spell_read_tree(fd, &lp->sl_pbyts, &lp->sl_pidxs, TRUE,
562 lp->sl_prefixcnt);
563 if (res != 0)
564 goto someerror;
565
566 /* For a new file link it in the list of spell files. */
567 if (old_lp == NULL && lang != NULL)
568 {
569 lp->sl_next = first_lang;
570 first_lang = lp;
571 }
572
573 goto endOK;
574
575 endFAIL:
576 if (lang != NULL)
577 /* truncating the name signals the error to spell_load_lang() */
578 *lang = NUL;
579 if (lp != NULL && old_lp == NULL)
580 slang_free(lp);
581 lp = NULL;
582
583 endOK:
584 if (fd != NULL)
585 fclose(fd);
586 sourcing_name = save_sourcing_name;
587 sourcing_lnum = save_sourcing_lnum;
588
589 return lp;
590 }
591
592 /*
593 * Fill in the wordcount fields for a trie.
594 * Returns the total number of words.
595 */
596 static void
597 tree_count_words(char_u *byts, idx_T *idxs)
598 {
599 int depth;
600 idx_T arridx[MAXWLEN];
601 int curi[MAXWLEN];
602 int c;
603 idx_T n;
604 int wordcount[MAXWLEN];
605
606 arridx[0] = 0;
607 curi[0] = 1;
608 wordcount[0] = 0;
609 depth = 0;
610 while (depth >= 0 && !got_int)
611 {
612 if (curi[depth] > byts[arridx[depth]])
613 {
614 /* Done all bytes at this node, go up one level. */
615 idxs[arridx[depth]] = wordcount[depth];
616 if (depth > 0)
617 wordcount[depth - 1] += wordcount[depth];
618
619 --depth;
620 fast_breakcheck();
621 }
622 else
623 {
624 /* Do one more byte at this node. */
625 n = arridx[depth] + curi[depth];
626 ++curi[depth];
627
628 c = byts[n];
629 if (c == 0)
630 {
631 /* End of word, count it. */
632 ++wordcount[depth];
633
634 /* Skip over any other NUL bytes (same word with different
635 * flags). */
636 while (byts[n + 1] == 0)
637 {
638 ++n;
639 ++curi[depth];
640 }
641 }
642 else
643 {
644 /* Normal char, go one level deeper to count the words. */
645 ++depth;
646 arridx[depth] = idxs[n];
647 curi[depth] = 1;
648 wordcount[depth] = 0;
649 }
650 }
651 }
652 }
653
654 /*
655 * Load the .sug files for languages that have one and weren't loaded yet.
656 */
657 void
658 suggest_load_files(void)
659 {
660 langp_T *lp;
661 int lpi;
662 slang_T *slang;
663 char_u *dotp;
664 FILE *fd;
665 char_u buf[MAXWLEN];
666 int i;
667 time_t timestamp;
668 int wcount;
669 int wordnr;
670 garray_T ga;
671 int c;
672
673 /* Do this for all languages that support sound folding. */
674 for (lpi = 0; lpi < curwin->w_s->b_langp.ga_len; ++lpi)
675 {
676 lp = LANGP_ENTRY(curwin->w_s->b_langp, lpi);
677 slang = lp->lp_slang;
678 if (slang->sl_sugtime != 0 && !slang->sl_sugloaded)
679 {
680 /* Change ".spl" to ".sug" and open the file. When the file isn't
681 * found silently skip it. Do set "sl_sugloaded" so that we
682 * don't try again and again. */
683 slang->sl_sugloaded = TRUE;
684
685 dotp = vim_strrchr(slang->sl_fname, '.');
686 if (dotp == NULL || fnamecmp(dotp, ".spl") != 0)
687 continue;
688 STRCPY(dotp, ".sug");
689 fd = mch_fopen((char *)slang->sl_fname, "r");
690 if (fd == NULL)
691 goto nextone;
692
693 /*
694 * <SUGHEADER>: <fileID> <versionnr> <timestamp>
695 */
696 for (i = 0; i < VIMSUGMAGICL; ++i)
697 buf[i] = getc(fd); /* <fileID> */
698 if (STRNCMP(buf, VIMSUGMAGIC, VIMSUGMAGICL) != 0)
699 {
700 EMSG2(_("E778: This does not look like a .sug file: %s"),
701 slang->sl_fname);
702 goto nextone;
703 }
704 c = getc(fd); /* <versionnr> */
705 if (c < VIMSUGVERSION)
706 {
707 EMSG2(_("E779: Old .sug file, needs to be updated: %s"),
708 slang->sl_fname);
709 goto nextone;
710 }
711 else if (c > VIMSUGVERSION)
712 {
713 EMSG2(_("E780: .sug file is for newer version of Vim: %s"),
714 slang->sl_fname);
715 goto nextone;
716 }
717
718 /* Check the timestamp, it must be exactly the same as the one in
719 * the .spl file. Otherwise the word numbers won't match. */
720 timestamp = get8ctime(fd); /* <timestamp> */
721 if (timestamp != slang->sl_sugtime)
722 {
723 EMSG2(_("E781: .sug file doesn't match .spl file: %s"),
724 slang->sl_fname);
725 goto nextone;
726 }
727
728 /*
729 * <SUGWORDTREE>: <wordtree>
730 * Read the trie with the soundfolded words.
731 */
732 if (spell_read_tree(fd, &slang->sl_sbyts, &slang->sl_sidxs,
733 FALSE, 0) != 0)
734 {
735 someerror:
736 EMSG2(_("E782: error while reading .sug file: %s"),
737 slang->sl_fname);
738 slang_clear_sug(slang);
739 goto nextone;
740 }
741
742 /*
743 * <SUGTABLE>: <sugwcount> <sugline> ...
744 *
745 * Read the table with word numbers. We use a file buffer for
746 * this, because it's so much like a file with lines. Makes it
747 * possible to swap the info and save on memory use.
748 */
749 slang->sl_sugbuf = open_spellbuf();
750 if (slang->sl_sugbuf == NULL)
751 goto someerror;
752 /* <sugwcount> */
753 wcount = get4c(fd);
754 if (wcount < 0)
755 goto someerror;
756
757 /* Read all the wordnr lists into the buffer, one NUL terminated
758 * list per line. */
759 ga_init2(&ga, 1, 100);
760 for (wordnr = 0; wordnr < wcount; ++wordnr)
761 {
762 ga.ga_len = 0;
763 for (;;)
764 {
765 c = getc(fd); /* <sugline> */
766 if (c < 0 || ga_grow(&ga, 1) == FAIL)
767 goto someerror;
768 ((char_u *)ga.ga_data)[ga.ga_len++] = c;
769 if (c == NUL)
770 break;
771 }
772 if (ml_append_buf(slang->sl_sugbuf, (linenr_T)wordnr,
773 ga.ga_data, ga.ga_len, TRUE) == FAIL)
774 goto someerror;
775 }
776 ga_clear(&ga);
777
778 /*
779 * Need to put word counts in the word tries, so that we can find
780 * a word by its number.
781 */
782 tree_count_words(slang->sl_fbyts, slang->sl_fidxs);
783 tree_count_words(slang->sl_sbyts, slang->sl_sidxs);
784
785 nextone:
786 if (fd != NULL)
787 fclose(fd);
788 STRCPY(dotp, ".spl");
789 }
790 }
791 }
792
793
794 /*
795 * Read a length field from "fd" in "cnt_bytes" bytes.
796 * Allocate memory, read the string into it and add a NUL at the end.
797 * Returns NULL when the count is zero.
798 * Sets "*cntp" to SP_*ERROR when there is an error, length of the result
799 * otherwise.
800 */
801 static char_u *
802 read_cnt_string(FILE *fd, int cnt_bytes, int *cntp)
803 {
804 int cnt = 0;
805 int i;
806 char_u *str;
807
808 /* read the length bytes, MSB first */
809 for (i = 0; i < cnt_bytes; ++i)
810 cnt = (cnt << 8) + getc(fd);
811 if (cnt < 0)
812 {
813 *cntp = SP_TRUNCERROR;
814 return NULL;
815 }
816 *cntp = cnt;
817 if (cnt == 0)
818 return NULL; /* nothing to read, return NULL */
819
820 str = read_string(fd, cnt);
821 if (str == NULL)
822 *cntp = SP_OTHERERROR;
823 return str;
824 }
825
826 /*
827 * Read SN_REGION: <regionname> ...
828 * Return SP_*ERROR flags.
829 */
830 static int
831 read_region_section(FILE *fd, slang_T *lp, int len)
832 {
833 int i;
834
835 if (len > 16)
836 return SP_FORMERROR;
837 for (i = 0; i < len; ++i)
838 lp->sl_regions[i] = getc(fd); /* <regionname> */
839 lp->sl_regions[len] = NUL;
840 return 0;
841 }
842
843 /*
844 * Read SN_CHARFLAGS section: <charflagslen> <charflags>
845 * <folcharslen> <folchars>
846 * Return SP_*ERROR flags.
847 */
848 static int
849 read_charflags_section(FILE *fd)
850 {
851 char_u *flags;
852 char_u *fol;
853 int flagslen, follen;
854
855 /* <charflagslen> <charflags> */
856 flags = read_cnt_string(fd, 1, &flagslen);
857 if (flagslen < 0)
858 return flagslen;
859
860 /* <folcharslen> <folchars> */
861 fol = read_cnt_string(fd, 2, &follen);
862 if (follen < 0)
863 {
864 vim_free(flags);
865 return follen;
866 }
867
868 /* Set the word-char flags and fill SPELL_ISUPPER() table. */
869 if (flags != NULL && fol != NULL)
870 set_spell_charflags(flags, flagslen, fol);
871
872 vim_free(flags);
873 vim_free(fol);
874
875 /* When <charflagslen> is zero then <fcharlen> must also be zero. */
876 if ((flags == NULL) != (fol == NULL))
877 return SP_FORMERROR;
878 return 0;
879 }
880
881 /*
882 * Read SN_PREFCOND section.
883 * Return SP_*ERROR flags.
884 */
885 static int
886 read_prefcond_section(FILE *fd, slang_T *lp)
887 {
888 int cnt;
889 int i;
890 int n;
891 char_u *p;
892 char_u buf[MAXWLEN + 1];
893
894 /* <prefcondcnt> <prefcond> ... */
895 cnt = get2c(fd); /* <prefcondcnt> */
896 if (cnt <= 0)
897 return SP_FORMERROR;
898
899 lp->sl_prefprog = (regprog_T **)alloc_clear(
900 (unsigned)sizeof(regprog_T *) * cnt);
901 if (lp->sl_prefprog == NULL)
902 return SP_OTHERERROR;
903 lp->sl_prefixcnt = cnt;
904
905 for (i = 0; i < cnt; ++i)
906 {
907 /* <prefcond> : <condlen> <condstr> */
908 n = getc(fd); /* <condlen> */
909 if (n < 0 || n >= MAXWLEN)
910 return SP_FORMERROR;
911
912 /* When <condlen> is zero we have an empty condition. Otherwise
913 * compile the regexp program used to check for the condition. */
914 if (n > 0)
915 {
916 buf[0] = '^'; /* always match at one position only */
917 p = buf + 1;
918 while (n-- > 0)
919 *p++ = getc(fd); /* <condstr> */
920 *p = NUL;
921 lp->sl_prefprog[i] = vim_regcomp(buf, RE_MAGIC + RE_STRING);
922 }
923 }
924 return 0;
925 }
926
927 /*
928 * Read REP or REPSAL items section from "fd": <repcount> <rep> ...
929 * Return SP_*ERROR flags.
930 */
931 static int
932 read_rep_section(FILE *fd, garray_T *gap, short *first)
933 {
934 int cnt;
935 fromto_T *ftp;
936 int i;
937
938 cnt = get2c(fd); /* <repcount> */
939 if (cnt < 0)
940 return SP_TRUNCERROR;
941
942 if (ga_grow(gap, cnt) == FAIL)
943 return SP_OTHERERROR;
944
945 /* <rep> : <repfromlen> <repfrom> <reptolen> <repto> */
946 for (; gap->ga_len < cnt; ++gap->ga_len)
947 {
948 ftp = &((fromto_T *)gap->ga_data)[gap->ga_len];
949 ftp->ft_from = read_cnt_string(fd, 1, &i);
950 if (i < 0)
951 return i;
952 if (i == 0)
953 return SP_FORMERROR;
954 ftp->ft_to = read_cnt_string(fd, 1, &i);
955 if (i <= 0)
956 {
957 vim_free(ftp->ft_from);
958 if (i < 0)
959 return i;
960 return SP_FORMERROR;
961 }
962 }
963
964 /* Fill the first-index table. */
965 for (i = 0; i < 256; ++i)
966 first[i] = -1;
967 for (i = 0; i < gap->ga_len; ++i)
968 {
969 ftp = &((fromto_T *)gap->ga_data)[i];
970 if (first[*ftp->ft_from] == -1)
971 first[*ftp->ft_from] = i;
972 }
973 return 0;
974 }
975
976 /*
977 * Read SN_SAL section: <salflags> <salcount> <sal> ...
978 * Return SP_*ERROR flags.
979 */
980 static int
981 read_sal_section(FILE *fd, slang_T *slang)
982 {
983 int i;
984 int cnt;
985 garray_T *gap;
986 salitem_T *smp;
987 int ccnt;
988 char_u *p;
989 int c = NUL;
990
991 slang->sl_sofo = FALSE;
992
993 i = getc(fd); /* <salflags> */
994 if (i & SAL_F0LLOWUP)
995 slang->sl_followup = TRUE;
996 if (i & SAL_COLLAPSE)
997 slang->sl_collapse = TRUE;
998 if (i & SAL_REM_ACCENTS)
999 slang->sl_rem_accents = TRUE;
1000
1001 cnt = get2c(fd); /* <salcount> */
1002 if (cnt < 0)
1003 return SP_TRUNCERROR;
1004
1005 gap = &slang->sl_sal;
1006 ga_init2(gap, sizeof(salitem_T), 10);
1007 if (ga_grow(gap, cnt + 1) == FAIL)
1008 return SP_OTHERERROR;
1009
1010 /* <sal> : <salfromlen> <salfrom> <saltolen> <salto> */
1011 for (; gap->ga_len < cnt; ++gap->ga_len)
1012 {
1013 smp = &((salitem_T *)gap->ga_data)[gap->ga_len];
1014 ccnt = getc(fd); /* <salfromlen> */
1015 if (ccnt < 0)
1016 return SP_TRUNCERROR;
1017 if ((p = alloc(ccnt + 2)) == NULL)
1018 return SP_OTHERERROR;
1019 smp->sm_lead = p;
1020
1021 /* Read up to the first special char into sm_lead. */
1022 for (i = 0; i < ccnt; ++i)
1023 {
1024 c = getc(fd); /* <salfrom> */
1025 if (vim_strchr((char_u *)"0123456789(-<^$", c) != NULL)
1026 break;
1027 *p++ = c;
1028 }
1029 smp->sm_leadlen = (int)(p - smp->sm_lead);
1030 *p++ = NUL;
1031
1032 /* Put (abc) chars in sm_oneof, if any. */
1033 if (c == '(')
1034 {
1035 smp->sm_oneof = p;
1036 for (++i; i < ccnt; ++i)
1037 {
1038 c = getc(fd); /* <salfrom> */
1039 if (c == ')')
1040 break;
1041 *p++ = c;
1042 }
1043 *p++ = NUL;
1044 if (++i < ccnt)
1045 c = getc(fd);
1046 }
1047 else
1048 smp->sm_oneof = NULL;
1049
1050 /* Any following chars go in sm_rules. */
1051 smp->sm_rules = p;
1052 if (i < ccnt)
1053 /* store the char we got while checking for end of sm_lead */
1054 *p++ = c;
1055 for (++i; i < ccnt; ++i)
1056 *p++ = getc(fd); /* <salfrom> */
1057 *p++ = NUL;
1058
1059 /* <saltolen> <salto> */
1060 smp->sm_to = read_cnt_string(fd, 1, &ccnt);
1061 if (ccnt < 0)
1062 {
1063 vim_free(smp->sm_lead);
1064 return ccnt;
1065 }
1066
1067 #ifdef FEAT_MBYTE
1068 if (has_mbyte)
1069 {
1070 /* convert the multi-byte strings to wide char strings */
1071 smp->sm_lead_w = mb_str2wide(smp->sm_lead);
1072 smp->sm_leadlen = mb_charlen(smp->sm_lead);
1073 if (smp->sm_oneof == NULL)
1074 smp->sm_oneof_w = NULL;
1075 else
1076 smp->sm_oneof_w = mb_str2wide(smp->sm_oneof);
1077 if (smp->sm_to == NULL)
1078 smp->sm_to_w = NULL;
1079 else
1080 smp->sm_to_w = mb_str2wide(smp->sm_to);
1081 if (smp->sm_lead_w == NULL
1082 || (smp->sm_oneof_w == NULL && smp->sm_oneof != NULL)
1083 || (smp->sm_to_w == NULL && smp->sm_to != NULL))
1084 {
1085 vim_free(smp->sm_lead);
1086 vim_free(smp->sm_to);
1087 vim_free(smp->sm_lead_w);
1088 vim_free(smp->sm_oneof_w);
1089 vim_free(smp->sm_to_w);
1090 return SP_OTHERERROR;
1091 }
1092 }
1093 #endif
1094 }
1095
1096 if (gap->ga_len > 0)
1097 {
1098 /* Add one extra entry to mark the end with an empty sm_lead. Avoids
1099 * that we need to check the index every time. */
1100 smp = &((salitem_T *)gap->ga_data)[gap->ga_len];
1101 if ((p = alloc(1)) == NULL)
1102 return SP_OTHERERROR;
1103 p[0] = NUL;
1104 smp->sm_lead = p;
1105 smp->sm_leadlen = 0;
1106 smp->sm_oneof = NULL;
1107 smp->sm_rules = p;
1108 smp->sm_to = NULL;
1109 #ifdef FEAT_MBYTE
1110 if (has_mbyte)
1111 {
1112 smp->sm_lead_w = mb_str2wide(smp->sm_lead);
1113 smp->sm_leadlen = 0;
1114 smp->sm_oneof_w = NULL;
1115 smp->sm_to_w = NULL;
1116 }
1117 #endif
1118 ++gap->ga_len;
1119 }
1120
1121 /* Fill the first-index table. */
1122 set_sal_first(slang);
1123
1124 return 0;
1125 }
1126
1127 /*
1128 * Read SN_WORDS: <word> ...
1129 * Return SP_*ERROR flags.
1130 */
1131 static int
1132 read_words_section(FILE *fd, slang_T *lp, int len)
1133 {
1134 int done = 0;
1135 int i;
1136 int c;
1137 char_u word[MAXWLEN];
1138
1139 while (done < len)
1140 {
1141 /* Read one word at a time. */
1142 for (i = 0; ; ++i)
1143 {
1144 c = getc(fd);
1145 if (c == EOF)
1146 return SP_TRUNCERROR;
1147 word[i] = c;
1148 if (word[i] == NUL)
1149 break;
1150 if (i == MAXWLEN - 1)
1151 return SP_FORMERROR;
1152 }
1153
1154 /* Init the count to 10. */
1155 count_common_word(lp, word, -1, 10);
1156 done += i + 1;
1157 }
1158 return 0;
1159 }
1160
1161 /*
1162 * SN_SOFO: <sofofromlen> <sofofrom> <sofotolen> <sofoto>
1163 * Return SP_*ERROR flags.
1164 */
1165 static int
1166 read_sofo_section(FILE *fd, slang_T *slang)
1167 {
1168 int cnt;
1169 char_u *from, *to;
1170 int res;
1171
1172 slang->sl_sofo = TRUE;
1173
1174 /* <sofofromlen> <sofofrom> */
1175 from = read_cnt_string(fd, 2, &cnt);
1176 if (cnt < 0)
1177 return cnt;
1178
1179 /* <sofotolen> <sofoto> */
1180 to = read_cnt_string(fd, 2, &cnt);
1181 if (cnt < 0)
1182 {
1183 vim_free(from);
1184 return cnt;
1185 }
1186
1187 /* Store the info in slang->sl_sal and/or slang->sl_sal_first. */
1188 if (from != NULL && to != NULL)
1189 res = set_sofo(slang, from, to);
1190 else if (from != NULL || to != NULL)
1191 res = SP_FORMERROR; /* only one of two strings is an error */
1192 else
1193 res = 0;
1194
1195 vim_free(from);
1196 vim_free(to);
1197 return res;
1198 }
1199
1200 /*
1201 * Read the compound section from the .spl file:
1202 * <compmax> <compminlen> <compsylmax> <compoptions> <compflags>
1203 * Returns SP_*ERROR flags.
1204 */
1205 static int
1206 read_compound(FILE *fd, slang_T *slang, int len)
1207 {
1208 int todo = len;
1209 int c;
1210 int atstart;
1211 char_u *pat;
1212 char_u *pp;
1213 char_u *cp;
1214 char_u *ap;
1215 char_u *crp;
1216 int cnt;
1217 garray_T *gap;
1218
1219 if (todo < 2)
1220 return SP_FORMERROR; /* need at least two bytes */
1221
1222 --todo;
1223 c = getc(fd); /* <compmax> */
1224 if (c < 2)
1225 c = MAXWLEN;
1226 slang->sl_compmax = c;
1227
1228 --todo;
1229 c = getc(fd); /* <compminlen> */
1230 if (c < 1)
1231 c = 0;
1232 slang->sl_compminlen = c;
1233
1234 --todo;
1235 c = getc(fd); /* <compsylmax> */
1236 if (c < 1)
1237 c = MAXWLEN;
1238 slang->sl_compsylmax = c;
1239
1240 c = getc(fd); /* <compoptions> */
1241 if (c != 0)
1242 ungetc(c, fd); /* be backwards compatible with Vim 7.0b */
1243 else
1244 {
1245 --todo;
1246 c = getc(fd); /* only use the lower byte for now */
1247 --todo;
1248 slang->sl_compoptions = c;
1249
1250 gap = &slang->sl_comppat;
1251 c = get2c(fd); /* <comppatcount> */
1252 todo -= 2;
1253 ga_init2(gap, sizeof(char_u *), c);
1254 if (ga_grow(gap, c) == OK)
1255 while (--c >= 0)
1256 {
1257 ((char_u **)(gap->ga_data))[gap->ga_len++] =
1258 read_cnt_string(fd, 1, &cnt);
1259 /* <comppatlen> <comppattext> */
1260 if (cnt < 0)
1261 return cnt;
1262 todo -= cnt + 1;
1263 }
1264 }
1265 if (todo < 0)
1266 return SP_FORMERROR;
1267
1268 /* Turn the COMPOUNDRULE items into a regexp pattern:
1269 * "a[bc]/a*b+" -> "^\(a[bc]\|a*b\+\)$".
1270 * Inserting backslashes may double the length, "^\(\)$<Nul>" is 7 bytes.
1271 * Conversion to utf-8 may double the size. */
1272 c = todo * 2 + 7;
1273 #ifdef FEAT_MBYTE
1274 if (enc_utf8)
1275 c += todo * 2;
1276 #endif
1277 pat = alloc((unsigned)c);
1278 if (pat == NULL)
1279 return SP_OTHERERROR;
1280
1281 /* We also need a list of all flags that can appear at the start and one
1282 * for all flags. */
1283 cp = alloc(todo + 1);
1284 if (cp == NULL)
1285 {
1286 vim_free(pat);
1287 return SP_OTHERERROR;
1288 }
1289 slang->sl_compstartflags = cp;
1290 *cp = NUL;
1291
1292 ap = alloc(todo + 1);
1293 if (ap == NULL)
1294 {
1295 vim_free(pat);
1296 return SP_OTHERERROR;
1297 }
1298 slang->sl_compallflags = ap;
1299 *ap = NUL;
1300
1301 /* And a list of all patterns in their original form, for checking whether
1302 * compounding may work in match_compoundrule(). This is freed when we
1303 * encounter a wildcard, the check doesn't work then. */
1304 crp = alloc(todo + 1);
1305 slang->sl_comprules = crp;
1306
1307 pp = pat;
1308 *pp++ = '^';
1309 *pp++ = '\\';
1310 *pp++ = '(';
1311
1312 atstart = 1;
1313 while (todo-- > 0)
1314 {
1315 c = getc(fd); /* <compflags> */
1316 if (c == EOF)
1317 {
1318 vim_free(pat);
1319 return SP_TRUNCERROR;
1320 }
1321
1322 /* Add all flags to "sl_compallflags". */
1323 if (vim_strchr((char_u *)"?*+[]/", c) == NULL
1324 && !byte_in_str(slang->sl_compallflags, c))
1325 {
1326 *ap++ = c;
1327 *ap = NUL;
1328 }
1329
1330 if (atstart != 0)
1331 {
1332 /* At start of item: copy flags to "sl_compstartflags". For a
1333 * [abc] item set "atstart" to 2 and copy up to the ']'. */
1334 if (c == '[')
1335 atstart = 2;
1336 else if (c == ']')
1337 atstart = 0;
1338 else
1339 {
1340 if (!byte_in_str(slang->sl_compstartflags, c))
1341 {
1342 *cp++ = c;
1343 *cp = NUL;
1344 }
1345 if (atstart == 1)
1346 atstart = 0;
1347 }
1348 }
1349
1350 /* Copy flag to "sl_comprules", unless we run into a wildcard. */
1351 if (crp != NULL)
1352 {
1353 if (c == '?' || c == '+' || c == '*')
1354 {
1355 vim_free(slang->sl_comprules);
1356 slang->sl_comprules = NULL;
1357 crp = NULL;
1358 }
1359 else
1360 *crp++ = c;
1361 }
1362
1363 if (c == '/') /* slash separates two items */
1364 {
1365 *pp++ = '\\';
1366 *pp++ = '|';
1367 atstart = 1;
1368 }
1369 else /* normal char, "[abc]" and '*' are copied as-is */
1370 {
1371 if (c == '?' || c == '+' || c == '~')
1372 *pp++ = '\\'; /* "a?" becomes "a\?", "a+" becomes "a\+" */
1373 #ifdef FEAT_MBYTE
1374 if (enc_utf8)
1375 pp += mb_char2bytes(c, pp);
1376 else
1377 #endif
1378 *pp++ = c;
1379 }
1380 }
1381
1382 *pp++ = '\\';
1383 *pp++ = ')';
1384 *pp++ = '$';
1385 *pp = NUL;
1386
1387 if (crp != NULL)
1388 *crp = NUL;
1389
1390 slang->sl_compprog = vim_regcomp(pat, RE_MAGIC + RE_STRING + RE_STRICT);
1391 vim_free(pat);
1392 if (slang->sl_compprog == NULL)
1393 return SP_FORMERROR;
1394
1395 return 0;
1396 }
1397
1398 /*
1399 * Set the SOFOFROM and SOFOTO items in language "lp".
1400 * Returns SP_*ERROR flags when there is something wrong.
1401 */
1402 static int
1403 set_sofo(slang_T *lp, char_u *from, char_u *to)
1404 {
1405 int i;
1406
1407 #ifdef FEAT_MBYTE
1408 garray_T *gap;
1409 char_u *s;
1410 char_u *p;
1411 int c;
1412 int *inp;
1413
1414 if (has_mbyte)
1415 {
1416 /* Use "sl_sal" as an array with 256 pointers to a list of wide
1417 * characters. The index is the low byte of the character.
1418 * The list contains from-to pairs with a terminating NUL.
1419 * sl_sal_first[] is used for latin1 "from" characters. */
1420 gap = &lp->sl_sal;
1421 ga_init2(gap, sizeof(int *), 1);
1422 if (ga_grow(gap, 256) == FAIL)
1423 return SP_OTHERERROR;
1424 vim_memset(gap->ga_data, 0, sizeof(int *) * 256);
1425 gap->ga_len = 256;
1426
1427 /* First count the number of items for each list. Temporarily use
1428 * sl_sal_first[] for this. */
1429 for (p = from, s = to; *p != NUL && *s != NUL; )
1430 {
1431 c = mb_cptr2char_adv(&p);
1432 mb_cptr_adv(s);
1433 if (c >= 256)
1434 ++lp->sl_sal_first[c & 0xff];
1435 }
1436 if (*p != NUL || *s != NUL) /* lengths differ */
1437 return SP_FORMERROR;
1438
1439 /* Allocate the lists. */
1440 for (i = 0; i < 256; ++i)
1441 if (lp->sl_sal_first[i] > 0)
1442 {
1443 p = alloc(sizeof(int) * (lp->sl_sal_first[i] * 2 + 1));
1444 if (p == NULL)
1445 return SP_OTHERERROR;
1446 ((int **)gap->ga_data)[i] = (int *)p;
1447 *(int *)p = 0;
1448 }
1449
1450 /* Put the characters up to 255 in sl_sal_first[] the rest in a sl_sal
1451 * list. */
1452 vim_memset(lp->sl_sal_first, 0, sizeof(salfirst_T) * 256);
1453 for (p = from, s = to; *p != NUL && *s != NUL; )
1454 {
1455 c = mb_cptr2char_adv(&p);
1456 i = mb_cptr2char_adv(&s);
1457 if (c >= 256)
1458 {
1459 /* Append the from-to chars at the end of the list with
1460 * the low byte. */
1461 inp = ((int **)gap->ga_data)[c & 0xff];
1462 while (*inp != 0)
1463 ++inp;
1464 *inp++ = c; /* from char */
1465 *inp++ = i; /* to char */
1466 *inp++ = NUL; /* NUL at the end */
1467 }
1468 else
1469 /* mapping byte to char is done in sl_sal_first[] */
1470 lp->sl_sal_first[c] = i;
1471 }
1472 }
1473 else
1474 #endif
1475 {
1476 /* mapping bytes to bytes is done in sl_sal_first[] */
1477 if (STRLEN(from) != STRLEN(to))
1478 return SP_FORMERROR;
1479
1480 for (i = 0; to[i] != NUL; ++i)
1481 lp->sl_sal_first[from[i]] = to[i];
1482 lp->sl_sal.ga_len = 1; /* indicates we have soundfolding */
1483 }
1484
1485 return 0;
1486 }
1487
1488 /*
1489 * Fill the first-index table for "lp".
1490 */
1491 static void
1492 set_sal_first(slang_T *lp)
1493 {
1494 salfirst_T *sfirst;
1495 int i;
1496 salitem_T *smp;
1497 int c;
1498 garray_T *gap = &lp->sl_sal;
1499
1500 sfirst = lp->sl_sal_first;
1501 for (i = 0; i < 256; ++i)
1502 sfirst[i] = -1;
1503 smp = (salitem_T *)gap->ga_data;
1504 for (i = 0; i < gap->ga_len; ++i)
1505 {
1506 #ifdef FEAT_MBYTE
1507 if (has_mbyte)
1508 /* Use the lowest byte of the first character. For latin1 it's
1509 * the character, for other encodings it should differ for most
1510 * characters. */
1511 c = *smp[i].sm_lead_w & 0xff;
1512 else
1513 #endif
1514 c = *smp[i].sm_lead;
1515 if (sfirst[c] == -1)
1516 {
1517 sfirst[c] = i;
1518 #ifdef FEAT_MBYTE
1519 if (has_mbyte)
1520 {
1521 int n;
1522
1523 /* Make sure all entries with this byte are following each
1524 * other. Move the ones that are in the wrong position. Do
1525 * keep the same ordering! */
1526 while (i + 1 < gap->ga_len
1527 && (*smp[i + 1].sm_lead_w & 0xff) == c)
1528 /* Skip over entry with same index byte. */
1529 ++i;
1530
1531 for (n = 1; i + n < gap->ga_len; ++n)
1532 if ((*smp[i + n].sm_lead_w & 0xff) == c)
1533 {
1534 salitem_T tsal;
1535
1536 /* Move entry with same index byte after the entries
1537 * we already found. */
1538 ++i;
1539 --n;
1540 tsal = smp[i + n];
1541 mch_memmove(smp + i + 1, smp + i,
1542 sizeof(salitem_T) * n);
1543 smp[i] = tsal;
1544 }
1545 }
1546 #endif
1547 }
1548 }
1549 }
1550
1551 #ifdef FEAT_MBYTE
1552 /*
1553 * Turn a multi-byte string into a wide character string.
1554 * Return it in allocated memory (NULL for out-of-memory)
1555 */
1556 static int *
1557 mb_str2wide(char_u *s)
1558 {
1559 int *res;
1560 char_u *p;
1561 int i = 0;
1562
1563 res = (int *)alloc(sizeof(int) * (mb_charlen(s) + 1));
1564 if (res != NULL)
1565 {
1566 for (p = s; *p != NUL; )
1567 res[i++] = mb_ptr2char_adv(&p);
1568 res[i] = NUL;
1569 }
1570 return res;
1571 }
1572 #endif
1573
1574 /*
1575 * Read a tree from the .spl or .sug file.
1576 * Allocates the memory and stores pointers in "bytsp" and "idxsp".
1577 * This is skipped when the tree has zero length.
1578 * Returns zero when OK, SP_ value for an error.
1579 */
1580 static int
1581 spell_read_tree(
1582 FILE *fd,
1583 char_u **bytsp,
1584 idx_T **idxsp,
1585 int prefixtree, /* TRUE for the prefix tree */
1586 int prefixcnt) /* when "prefixtree" is TRUE: prefix count */
1587 {
1588 int len;
1589 int idx;
1590 char_u *bp;
1591 idx_T *ip;
1592
1593 /* The tree size was computed when writing the file, so that we can
1594 * allocate it as one long block. <nodecount> */
1595 len = get4c(fd);
1596 if (len < 0)
1597 return SP_TRUNCERROR;
1598 if (len > 0)
1599 {
1600 /* Allocate the byte array. */
1601 bp = lalloc((long_u)len, TRUE);
1602 if (bp == NULL)
1603 return SP_OTHERERROR;
1604 *bytsp = bp;
1605
1606 /* Allocate the index array. */
1607 ip = (idx_T *)lalloc_clear((long_u)(len * sizeof(int)), TRUE);
1608 if (ip == NULL)
1609 return SP_OTHERERROR;
1610 *idxsp = ip;
1611
1612 /* Recursively read the tree and store it in the array. */
1613 idx = read_tree_node(fd, bp, ip, len, 0, prefixtree, prefixcnt);
1614 if (idx < 0)
1615 return idx;
1616 }
1617 return 0;
1618 }
1619
1620 /*
1621 * Read one row of siblings from the spell file and store it in the byte array
1622 * "byts" and index array "idxs". Recursively read the children.
1623 *
1624 * NOTE: The code here must match put_node()!
1625 *
1626 * Returns the index (>= 0) following the siblings.
1627 * Returns SP_TRUNCERROR if the file is shorter than expected.
1628 * Returns SP_FORMERROR if there is a format error.
1629 */
1630 static idx_T
1631 read_tree_node(
1632 FILE *fd,
1633 char_u *byts,
1634 idx_T *idxs,
1635 int maxidx, /* size of arrays */
1636 idx_T startidx, /* current index in "byts" and "idxs" */
1637 int prefixtree, /* TRUE for reading PREFIXTREE */
1638 int maxprefcondnr) /* maximum for <prefcondnr> */
1639 {
1640 int len;
1641 int i;
1642 int n;
1643 idx_T idx = startidx;
1644 int c;
1645 int c2;
1646 #define SHARED_MASK 0x8000000
1647
1648 len = getc(fd); /* <siblingcount> */
1649 if (len <= 0)
1650 return SP_TRUNCERROR;
1651
1652 if (startidx + len >= maxidx)
1653 return SP_FORMERROR;
1654 byts[idx++] = len;
1655
1656 /* Read the byte values, flag/region bytes and shared indexes. */
1657 for (i = 1; i <= len; ++i)
1658 {
1659 c = getc(fd); /* <byte> */
1660 if (c < 0)
1661 return SP_TRUNCERROR;
1662 if (c <= BY_SPECIAL)
1663 {
1664 if (c == BY_NOFLAGS && !prefixtree)
1665 {
1666 /* No flags, all regions. */
1667 idxs[idx] = 0;
1668 c = 0;
1669 }
1670 else if (c != BY_INDEX)
1671 {
1672 if (prefixtree)
1673 {
1674 /* Read the optional pflags byte, the prefix ID and the
1675 * condition nr. In idxs[] store the prefix ID in the low
1676 * byte, the condition index shifted up 8 bits, the flags
1677 * shifted up 24 bits. */
1678 if (c == BY_FLAGS)
1679 c = getc(fd) << 24; /* <pflags> */
1680 else
1681 c = 0;
1682
1683 c |= getc(fd); /* <affixID> */
1684
1685 n = get2c(fd); /* <prefcondnr> */
1686 if (n >= maxprefcondnr)
1687 return SP_FORMERROR;
1688 c |= (n << 8);
1689 }
1690 else /* c must be BY_FLAGS or BY_FLAGS2 */
1691 {
1692 /* Read flags and optional region and prefix ID. In
1693 * idxs[] the flags go in the low two bytes, region above
1694 * that and prefix ID above the region. */
1695 c2 = c;
1696 c = getc(fd); /* <flags> */
1697 if (c2 == BY_FLAGS2)
1698 c = (getc(fd) << 8) + c; /* <flags2> */
1699 if (c & WF_REGION)
1700 c = (getc(fd) << 16) + c; /* <region> */
1701 if (c & WF_AFX)
1702 c = (getc(fd) << 24) + c; /* <affixID> */
1703 }
1704
1705 idxs[idx] = c;
1706 c = 0;
1707 }
1708 else /* c == BY_INDEX */
1709 {
1710 /* <nodeidx> */
1711 n = get3c(fd);
1712 if (n < 0 || n >= maxidx)
1713 return SP_FORMERROR;
1714 idxs[idx] = n + SHARED_MASK;
1715 c = getc(fd); /* <xbyte> */
1716 }
1717 }
1718 byts[idx++] = c;
1719 }
1720
1721 /* Recursively read the children for non-shared siblings.
1722 * Skip the end-of-word ones (zero byte value) and the shared ones (and
1723 * remove SHARED_MASK) */
1724 for (i = 1; i <= len; ++i)
1725 if (byts[startidx + i] != 0)
1726 {
1727 if (idxs[startidx + i] & SHARED_MASK)
1728 idxs[startidx + i] &= ~SHARED_MASK;
1729 else
1730 {
1731 idxs[startidx + i] = idx;
1732 idx = read_tree_node(fd, byts, idxs, maxidx, idx,
1733 prefixtree, maxprefcondnr);
1734 if (idx < 0)
1735 break;
1736 }
1737 }
1738
1739 return idx;
1740 }
1741
1742 /*
1743 * Reload the spell file "fname" if it's loaded.
1744 */
1745 static void
1746 spell_reload_one(
1747 char_u *fname,
1748 int added_word) /* invoked through "zg" */
1749 {
1750 slang_T *slang;
1751 int didit = FALSE;
1752
1753 for (slang = first_lang; slang != NULL; slang = slang->sl_next)
1754 {
1755 if (fullpathcmp(fname, slang->sl_fname, FALSE) == FPC_SAME)
1756 {
1757 slang_clear(slang);
1758 if (spell_load_file(fname, NULL, slang, FALSE) == NULL)
1759 /* reloading failed, clear the language */
1760 slang_clear(slang);
1761 redraw_all_later(SOME_VALID);
1762 didit = TRUE;
1763 }
1764 }
1765
1766 /* When "zg" was used and the file wasn't loaded yet, should redo
1767 * 'spelllang' to load it now. */
1768 if (added_word && !didit)
1769 did_set_spelllang(curwin);
1770 }
1771
1772
1773 /*
1774 * Functions for ":mkspell".
1775 */
1776
1777 #define MAXLINELEN 500 /* Maximum length in bytes of a line in a .aff
1778 and .dic file. */
1779 /*
1780 * Main structure to store the contents of a ".aff" file.
1781 */
1782 typedef struct afffile_S
1783 {
1784 char_u *af_enc; /* "SET", normalized, alloc'ed string or NULL */
1785 int af_flagtype; /* AFT_CHAR, AFT_LONG, AFT_NUM or AFT_CAPLONG */
1786 unsigned af_rare; /* RARE ID for rare word */
1787 unsigned af_keepcase; /* KEEPCASE ID for keep-case word */
1788 unsigned af_bad; /* BAD ID for banned word */
1789 unsigned af_needaffix; /* NEEDAFFIX ID */
1790 unsigned af_circumfix; /* CIRCUMFIX ID */
1791 unsigned af_needcomp; /* NEEDCOMPOUND ID */
1792 unsigned af_comproot; /* COMPOUNDROOT ID */
1793 unsigned af_compforbid; /* COMPOUNDFORBIDFLAG ID */
1794 unsigned af_comppermit; /* COMPOUNDPERMITFLAG ID */
1795 unsigned af_nosuggest; /* NOSUGGEST ID */
1796 int af_pfxpostpone; /* postpone prefixes without chop string and
1797 without flags */
1798 int af_ignoreextra; /* IGNOREEXTRA present */
1799 hashtab_T af_pref; /* hashtable for prefixes, affheader_T */
1800 hashtab_T af_suff; /* hashtable for suffixes, affheader_T */
1801 hashtab_T af_comp; /* hashtable for compound flags, compitem_T */
1802 } afffile_T;
1803
1804 #define AFT_CHAR 0 /* flags are one character */
1805 #define AFT_LONG 1 /* flags are two characters */
1806 #define AFT_CAPLONG 2 /* flags are one or two characters */
1807 #define AFT_NUM 3 /* flags are numbers, comma separated */
1808
1809 typedef struct affentry_S affentry_T;
1810 /* Affix entry from ".aff" file. Used for prefixes and suffixes. */
1811 struct affentry_S
1812 {
1813 affentry_T *ae_next; /* next affix with same name/number */
1814 char_u *ae_chop; /* text to chop off basic word (can be NULL) */
1815 char_u *ae_add; /* text to add to basic word (can be NULL) */
1816 char_u *ae_flags; /* flags on the affix (can be NULL) */
1817 char_u *ae_cond; /* condition (NULL for ".") */
1818 regprog_T *ae_prog; /* regexp program for ae_cond or NULL */
1819 char ae_compforbid; /* COMPOUNDFORBIDFLAG found */
1820 char ae_comppermit; /* COMPOUNDPERMITFLAG found */
1821 };
1822
1823 #ifdef FEAT_MBYTE
1824 # define AH_KEY_LEN 17 /* 2 x 8 bytes + NUL */
1825 #else
1826 # define AH_KEY_LEN 7 /* 6 digits + NUL */
1827 #endif
1828
1829 /* Affix header from ".aff" file. Used for af_pref and af_suff. */
1830 typedef struct affheader_S
1831 {
1832 char_u ah_key[AH_KEY_LEN]; /* key for hashtab == name of affix */
1833 unsigned ah_flag; /* affix name as number, uses "af_flagtype" */
1834 int ah_newID; /* prefix ID after renumbering; 0 if not used */
1835 int ah_combine; /* suffix may combine with prefix */
1836 int ah_follows; /* another affix block should be following */
1837 affentry_T *ah_first; /* first affix entry */
1838 } affheader_T;
1839
1840 #define HI2AH(hi) ((affheader_T *)(hi)->hi_key)
1841
1842 /* Flag used in compound items. */
1843 typedef struct compitem_S
1844 {
1845 char_u ci_key[AH_KEY_LEN]; /* key for hashtab == name of compound */
1846 unsigned ci_flag; /* affix name as number, uses "af_flagtype" */
1847 int ci_newID; /* affix ID after renumbering. */
1848 } compitem_T;
1849
1850 #define HI2CI(hi) ((compitem_T *)(hi)->hi_key)
1851
1852 /*
1853 * Structure that is used to store the items in the word tree. This avoids
1854 * the need to keep track of each allocated thing, everything is freed all at
1855 * once after ":mkspell" is done.
1856 * Note: "sb_next" must be just before "sb_data" to make sure the alignment of
1857 * "sb_data" is correct for systems where pointers must be aligned on
1858 * pointer-size boundaries and sizeof(pointer) > sizeof(int) (e.g., Sparc).
1859 */
1860 #define SBLOCKSIZE 16000 /* size of sb_data */
1861 typedef struct sblock_S sblock_T;
1862 struct sblock_S
1863 {
1864 int sb_used; /* nr of bytes already in use */
1865 sblock_T *sb_next; /* next block in list */
1866 char_u sb_data[1]; /* data, actually longer */
1867 };
1868
1869 /*
1870 * A node in the tree.
1871 */
1872 typedef struct wordnode_S wordnode_T;
1873 struct wordnode_S
1874 {
1875 union /* shared to save space */
1876 {
1877 char_u hashkey[6]; /* the hash key, only used while compressing */
1878 int index; /* index in written nodes (valid after first
1879 round) */
1880 } wn_u1;
1881 union /* shared to save space */
1882 {
1883 wordnode_T *next; /* next node with same hash key */
1884 wordnode_T *wnode; /* parent node that will write this node */
1885 } wn_u2;
1886 wordnode_T *wn_child; /* child (next byte in word) */
1887 wordnode_T *wn_sibling; /* next sibling (alternate byte in word,
1888 always sorted) */
1889 int wn_refs; /* Nr. of references to this node. Only
1890 relevant for first node in a list of
1891 siblings, in following siblings it is
1892 always one. */
1893 char_u wn_byte; /* Byte for this node. NUL for word end */
1894
1895 /* Info for when "wn_byte" is NUL.
1896 * In PREFIXTREE "wn_region" is used for the prefcondnr.
1897 * In the soundfolded word tree "wn_flags" has the MSW of the wordnr and
1898 * "wn_region" the LSW of the wordnr. */
1899 char_u wn_affixID; /* supported/required prefix ID or 0 */
1900 short_u wn_flags; /* WF_ flags */
1901 short wn_region; /* region mask */
1902
1903 #ifdef SPELL_PRINTTREE
1904 int wn_nr; /* sequence nr for printing */
1905 #endif
1906 };
1907
1908 #define WN_MASK 0xffff /* mask relevant bits of "wn_flags" */
1909
1910 #define HI2WN(hi) (wordnode_T *)((hi)->hi_key)
1911
1912 /*
1913 * Info used while reading the spell files.
1914 */
1915 typedef struct spellinfo_S
1916 {
1917 wordnode_T *si_foldroot; /* tree with case-folded words */
1918 long si_foldwcount; /* nr of words in si_foldroot */
1919
1920 wordnode_T *si_keeproot; /* tree with keep-case words */
1921 long si_keepwcount; /* nr of words in si_keeproot */
1922
1923 wordnode_T *si_prefroot; /* tree with postponed prefixes */
1924
1925 long si_sugtree; /* creating the soundfolding trie */
1926
1927 sblock_T *si_blocks; /* memory blocks used */
1928 long si_blocks_cnt; /* memory blocks allocated */
1929 int si_did_emsg; /* TRUE when ran out of memory */
1930
1931 long si_compress_cnt; /* words to add before lowering
1932 compression limit */
1933 wordnode_T *si_first_free; /* List of nodes that have been freed during
1934 compression, linked by "wn_child" field. */
1935 long si_free_count; /* number of nodes in si_first_free */
1936 #ifdef SPELL_PRINTTREE
1937 int si_wordnode_nr; /* sequence nr for nodes */
1938 #endif
1939 buf_T *si_spellbuf; /* buffer used to store soundfold word table */
1940
1941 int si_ascii; /* handling only ASCII words */
1942 int si_add; /* addition file */
1943 int si_clear_chartab; /* when TRUE clear char tables */
1944 int si_region; /* region mask */
1945 vimconv_T si_conv; /* for conversion to 'encoding' */
1946 int si_memtot; /* runtime memory used */
1947 int si_verbose; /* verbose messages */
1948 int si_msg_count; /* number of words added since last message */
1949 char_u *si_info; /* info text chars or NULL */
1950 int si_region_count; /* number of regions supported (1 when there
1951 are no regions) */
1952 char_u si_region_name[17]; /* region names; used only if
1953 * si_region_count > 1) */
1954
1955 garray_T si_rep; /* list of fromto_T entries from REP lines */
1956 garray_T si_repsal; /* list of fromto_T entries from REPSAL lines */
1957 garray_T si_sal; /* list of fromto_T entries from SAL lines */
1958 char_u *si_sofofr; /* SOFOFROM text */
1959 char_u *si_sofoto; /* SOFOTO text */
1960 int si_nosugfile; /* NOSUGFILE item found */
1961 int si_nosplitsugs; /* NOSPLITSUGS item found */
1962 int si_nocompoundsugs; /* NOCOMPOUNDSUGS item found */
1963 int si_followup; /* soundsalike: ? */
1964 int si_collapse; /* soundsalike: ? */
1965 hashtab_T si_commonwords; /* hashtable for common words */
1966 time_t si_sugtime; /* timestamp for .sug file */
1967 int si_rem_accents; /* soundsalike: remove accents */
1968 garray_T si_map; /* MAP info concatenated */
1969 char_u *si_midword; /* MIDWORD chars or NULL */
1970 int si_compmax; /* max nr of words for compounding */
1971 int si_compminlen; /* minimal length for compounding */
1972 int si_compsylmax; /* max nr of syllables for compounding */
1973 int si_compoptions; /* COMP_ flags */
1974 garray_T si_comppat; /* CHECKCOMPOUNDPATTERN items, each stored as
1975 a string */
1976 char_u *si_compflags; /* flags used for compounding */
1977 char_u si_nobreak; /* NOBREAK */
1978 char_u *si_syllable; /* syllable string */
1979 garray_T si_prefcond; /* table with conditions for postponed
1980 * prefixes, each stored as a string */
1981 int si_newprefID; /* current value for ah_newID */
1982 int si_newcompID; /* current value for compound ID */
1983 } spellinfo_T;
1984
1985 static afffile_T *spell_read_aff(spellinfo_T *spin, char_u *fname);
1986 static int is_aff_rule(char_u **items, int itemcnt, char *rulename, int mincount);
1987 static void aff_process_flags(afffile_T *affile, affentry_T *entry);
1988 static int spell_info_item(char_u *s);
1989 static unsigned affitem2flag(int flagtype, char_u *item, char_u *fname, int lnum);
1990 static unsigned get_affitem(int flagtype, char_u **pp);
1991 static void process_compflags(spellinfo_T *spin, afffile_T *aff, char_u *compflags);
1992 static void check_renumber(spellinfo_T *spin);
1993 static int flag_in_afflist(int flagtype, char_u *afflist, unsigned flag);
1994 static void aff_check_number(int spinval, int affval, char *name);
1995 static void aff_check_string(char_u *spinval, char_u *affval, char *name);
1996 static int str_equal(char_u *s1, char_u *s2);
1997 static void add_fromto(spellinfo_T *spin, garray_T *gap, char_u *from, char_u *to);
1998 static int sal_to_bool(char_u *s);
1999 static void spell_free_aff(afffile_T *aff);
2000 static int spell_read_dic(spellinfo_T *spin, char_u *fname, afffile_T *affile);
2001 static int get_affix_flags(afffile_T *affile, char_u *afflist);
2002 static int get_pfxlist(afffile_T *affile, char_u *afflist, char_u *store_afflist);
2003 static void get_compflags(afffile_T *affile, char_u *afflist, char_u *store_afflist);
2004 static int store_aff_word(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);
2005 static int spell_read_wordfile(spellinfo_T *spin, char_u *fname);
2006 static void *getroom(spellinfo_T *spin, size_t len, int align);
2007 static char_u *getroom_save(spellinfo_T *spin, char_u *s);
2008 static void free_blocks(sblock_T *bl);
2009 static wordnode_T *wordtree_alloc(spellinfo_T *spin);
2010 static int store_word(spellinfo_T *spin, char_u *word, int flags, int region, char_u *pfxlist, int need_affix);
2011 static int tree_add_word(spellinfo_T *spin, char_u *word, wordnode_T *tree, int flags, int region, int affixID);
2012 static wordnode_T *get_wordnode(spellinfo_T *spin);
2013 static int deref_wordnode(spellinfo_T *spin, wordnode_T *node);
2014 static void free_wordnode(spellinfo_T *spin, wordnode_T *n);
2015 static void wordtree_compress(spellinfo_T *spin, wordnode_T *root);
2016 static int node_compress(spellinfo_T *spin, wordnode_T *node, hashtab_T *ht, int *tot);
2017 static int node_equal(wordnode_T *n1, wordnode_T *n2);
2018 static int write_vim_spell(spellinfo_T *spin, char_u *fname);
2019 static void clear_node(wordnode_T *node);
2020 static int put_node(FILE *fd, wordnode_T *node, int idx, int regionmask, int prefixtree);
2021 static void spell_make_sugfile(spellinfo_T *spin, char_u *wfname);
2022 static int sug_filltree(spellinfo_T *spin, slang_T *slang);
2023 static int sug_maketable(spellinfo_T *spin);
2024 static int sug_filltable(spellinfo_T *spin, wordnode_T *node, int startwordnr, garray_T *gap);
2025 static int offset2bytes(int nr, char_u *buf);
2026 static void sug_write(spellinfo_T *spin, char_u *fname);
2027 static void spell_message(spellinfo_T *spin, char_u *str);
2028 static void init_spellfile(void);
2029
2030 /* In the postponed prefixes tree wn_flags is used to store the WFP_ flags,
2031 * but it must be negative to indicate the prefix tree to tree_add_word().
2032 * Use a negative number with the lower 8 bits zero. */
2033 #define PFX_FLAGS -256
2034
2035 /* flags for "condit" argument of store_aff_word() */
2036 #define CONDIT_COMB 1 /* affix must combine */
2037 #define CONDIT_CFIX 2 /* affix must have CIRCUMFIX flag */
2038 #define CONDIT_SUF 4 /* add a suffix for matching flags */
2039 #define CONDIT_AFF 8 /* word already has an affix */
2040
2041 /*
2042 * Tunable parameters for when the tree is compressed. See 'mkspellmem'.
2043 */
2044 static long compress_start = 30000; /* memory / SBLOCKSIZE */
2045 static long compress_inc = 100; /* memory / SBLOCKSIZE */
2046 static long compress_added = 500000; /* word count */
2047
2048 /*
2049 * Check the 'mkspellmem' option. Return FAIL if it's wrong.
2050 * Sets "sps_flags".
2051 */
2052 int
2053 spell_check_msm(void)
2054 {
2055 char_u *p = p_msm;
2056 long start = 0;
2057 long incr = 0;
2058 long added = 0;
2059
2060 if (!VIM_ISDIGIT(*p))
2061 return FAIL;
2062 /* block count = (value * 1024) / SBLOCKSIZE (but avoid overflow)*/
2063 start = (getdigits(&p) * 10) / (SBLOCKSIZE / 102);
2064 if (*p != ',')
2065 return FAIL;
2066 ++p;
2067 if (!VIM_ISDIGIT(*p))
2068 return FAIL;
2069 incr = (getdigits(&p) * 102) / (SBLOCKSIZE / 10);
2070 if (*p != ',')
2071 return FAIL;
2072 ++p;
2073 if (!VIM_ISDIGIT(*p))
2074 return FAIL;
2075 added = getdigits(&p) * 1024;
2076 if (*p != NUL)
2077 return FAIL;
2078
2079 if (start == 0 || incr == 0 || added == 0 || incr > start)
2080 return FAIL;
2081
2082 compress_start = start;
2083 compress_inc = incr;
2084 compress_added = added;
2085 return OK;
2086 }
2087
2088 #ifdef SPELL_PRINTTREE
2089 /*
2090 * For debugging the tree code: print the current tree in a (more or less)
2091 * readable format, so that we can see what happens when adding a word and/or
2092 * compressing the tree.
2093 * Based on code from Olaf Seibert.
2094 */
2095 #define PRINTLINESIZE 1000
2096 #define PRINTWIDTH 6
2097
2098 #define PRINTSOME(l, depth, fmt, a1, a2) vim_snprintf(l + depth * PRINTWIDTH, \
2099 PRINTLINESIZE - PRINTWIDTH * depth, fmt, a1, a2)
2100
2101 static char line1[PRINTLINESIZE];
2102 static char line2[PRINTLINESIZE];
2103 static char line3[PRINTLINESIZE];
2104
2105 static void
2106 spell_clear_flags(wordnode_T *node)
2107 {
2108 wordnode_T *np;
2109
2110 for (np = node; np != NULL; np = np->wn_sibling)
2111 {
2112 np->wn_u1.index = FALSE;
2113 spell_clear_flags(np->wn_child);
2114 }
2115 }
2116
2117 static void
2118 spell_print_node(wordnode_T *node, int depth)
2119 {
2120 if (node->wn_u1.index)
2121 {
2122 /* Done this node before, print the reference. */
2123 PRINTSOME(line1, depth, "(%d)", node->wn_nr, 0);
2124 PRINTSOME(line2, depth, " ", 0, 0);
2125 PRINTSOME(line3, depth, " ", 0, 0);
2126 msg((char_u *)line1);
2127 msg((char_u *)line2);
2128 msg((char_u *)line3);
2129 }
2130 else
2131 {
2132 node->wn_u1.index = TRUE;
2133
2134 if (node->wn_byte != NUL)
2135 {
2136 if (node->wn_child != NULL)
2137 PRINTSOME(line1, depth, " %c -> ", node->wn_byte, 0);
2138 else
2139 /* Cannot happen? */
2140 PRINTSOME(line1, depth, " %c ???", node->wn_byte, 0);
2141 }
2142 else
2143 PRINTSOME(line1, depth, " $ ", 0, 0);
2144
2145 PRINTSOME(line2, depth, "%d/%d ", node->wn_nr, node->wn_refs);
2146
2147 if (node->wn_sibling != NULL)
2148 PRINTSOME(line3, depth, " | ", 0, 0);
2149 else
2150 PRINTSOME(line3, depth, " ", 0, 0);
2151
2152 if (node->wn_byte == NUL)
2153 {
2154 msg((char_u *)line1);
2155 msg((char_u *)line2);
2156 msg((char_u *)line3);
2157 }
2158
2159 /* do the children */
2160 if (node->wn_byte != NUL && node->wn_child != NULL)
2161 spell_print_node(node->wn_child, depth + 1);
2162
2163 /* do the siblings */
2164 if (node->wn_sibling != NULL)
2165 {
2166 /* get rid of all parent details except | */
2167 STRCPY(line1, line3);
2168 STRCPY(line2, line3);
2169 spell_print_node(node->wn_sibling, depth);
2170 }
2171 }
2172 }
2173
2174 static void
2175 spell_print_tree(wordnode_T *root)
2176 {
2177 if (root != NULL)
2178 {
2179 /* Clear the "wn_u1.index" fields, used to remember what has been
2180 * done. */
2181 spell_clear_flags(root);
2182
2183 /* Recursively print the tree. */
2184 spell_print_node(root, 0);
2185 }
2186 }
2187 #endif /* SPELL_PRINTTREE */
2188
2189 /*
2190 * Read the affix file "fname".
2191 * Returns an afffile_T, NULL for complete failure.
2192 */
2193 static afffile_T *
2194 spell_read_aff(spellinfo_T *spin, char_u *fname)
2195 {
2196 FILE *fd;
2197 afffile_T *aff;
2198 char_u rline[MAXLINELEN];
2199 char_u *line;
2200 char_u *pc = NULL;
2201 #define MAXITEMCNT 30
2202 char_u *(items[MAXITEMCNT]);
2203 int itemcnt;
2204 char_u *p;
2205 int lnum = 0;
2206 affheader_T *cur_aff = NULL;
2207 int did_postpone_prefix = FALSE;
2208 int aff_todo = 0;
2209 hashtab_T *tp;
2210 char_u *low = NULL;
2211 char_u *fol = NULL;
2212 char_u *upp = NULL;
2213 int do_rep;
2214 int do_repsal;
2215 int do_sal;
2216 int do_mapline;
2217 int found_map = FALSE;
2218 hashitem_T *hi;
2219 int l;
2220 int compminlen = 0; /* COMPOUNDMIN value */
2221 int compsylmax = 0; /* COMPOUNDSYLMAX value */
2222 int compoptions = 0; /* COMP_ flags */
2223 int compmax = 0; /* COMPOUNDWORDMAX value */
2224 char_u *compflags = NULL; /* COMPOUNDFLAG and COMPOUNDRULE
2225 concatenated */
2226 char_u *midword = NULL; /* MIDWORD value */
2227 char_u *syllable = NULL; /* SYLLABLE value */
2228 char_u *sofofrom = NULL; /* SOFOFROM value */
2229 char_u *sofoto = NULL; /* SOFOTO value */
2230
2231 /*
2232 * Open the file.
2233 */
2234 fd = mch_fopen((char *)fname, "r");
2235 if (fd == NULL)
2236 {
2237 EMSG2(_(e_notopen), fname);
2238 return NULL;
2239 }
2240
2241 vim_snprintf((char *)IObuff, IOSIZE, _("Reading affix file %s ..."), fname);
2242 spell_message(spin, IObuff);
2243
2244 /* Only do REP lines when not done in another .aff file already. */
2245 do_rep = spin->si_rep.ga_len == 0;
2246
2247 /* Only do REPSAL lines when not done in another .aff file already. */
2248 do_repsal = spin->si_repsal.ga_len == 0;
2249
2250 /* Only do SAL lines when not done in another .aff file already. */
2251 do_sal = spin->si_sal.ga_len == 0;
2252
2253 /* Only do MAP lines when not done in another .aff file already. */
2254 do_mapline = spin->si_map.ga_len == 0;
2255
2256 /*
2257 * Allocate and init the afffile_T structure.
2258 */
2259 aff = (afffile_T *)getroom(spin, sizeof(afffile_T), TRUE);
2260 if (aff == NULL)
2261 {
2262 fclose(fd);
2263 return NULL;
2264 }
2265 hash_init(&aff->af_pref);
2266 hash_init(&aff->af_suff);
2267 hash_init(&aff->af_comp);
2268
2269 /*
2270 * Read all the lines in the file one by one.
2271 */
2272 while (!vim_fgets(rline, MAXLINELEN, fd) && !got_int)
2273 {
2274 line_breakcheck();
2275 ++lnum;
2276
2277 /* Skip comment lines. */
2278 if (*rline == '#')
2279 continue;
2280
2281 /* Convert from "SET" to 'encoding' when needed. */
2282 vim_free(pc);
2283 #ifdef FEAT_MBYTE
2284 if (spin->si_conv.vc_type != CONV_NONE)
2285 {
2286 pc = string_convert(&spin->si_conv, rline, NULL);
2287 if (pc == NULL)
2288 {
2289 smsg((char_u *)_("Conversion failure for word in %s line %d: %s"),
2290 fname, lnum, rline);
2291 continue;
2292 }
2293 line = pc;
2294 }
2295 else
2296 #endif
2297 {
2298 pc = NULL;
2299 line = rline;
2300 }
2301
2302 /* Split the line up in white separated items. Put a NUL after each
2303 * item. */
2304 itemcnt = 0;
2305 for (p = line; ; )
2306 {
2307 while (*p != NUL && *p <= ' ') /* skip white space and CR/NL */
2308 ++p;
2309 if (*p == NUL)
2310 break;
2311 if (itemcnt == MAXITEMCNT) /* too many items */
2312 break;
2313 items[itemcnt++] = p;
2314 /* A few items have arbitrary text argument, don't split them. */
2315 if (itemcnt == 2 && spell_info_item(items[0]))
2316 while (*p >= ' ' || *p == TAB) /* skip until CR/NL */
2317 ++p;
2318 else
2319 while (*p > ' ') /* skip until white space or CR/NL */
2320 ++p;
2321 if (*p == NUL)
2322 break;
2323 *p++ = NUL;
2324 }
2325
2326 /* Handle non-empty lines. */
2327 if (itemcnt > 0)
2328 {
2329 if (is_aff_rule(items, itemcnt, "SET", 2) && aff->af_enc == NULL)
2330 {
2331 #ifdef FEAT_MBYTE
2332 /* Setup for conversion from "ENC" to 'encoding'. */
2333 aff->af_enc = enc_canonize(items[1]);
2334 if (aff->af_enc != NULL && !spin->si_ascii
2335 && convert_setup(&spin->si_conv, aff->af_enc,
2336 p_enc) == FAIL)
2337 smsg((char_u *)_("Conversion in %s not supported: from %s to %s"),
2338 fname, aff->af_enc, p_enc);
2339 spin->si_conv.vc_fail = TRUE;
2340 #else
2341 smsg((char_u *)_("Conversion in %s not supported"), fname);
2342 #endif
2343 }
2344 else if (is_aff_rule(items, itemcnt, "FLAG", 2)
2345 && aff->af_flagtype == AFT_CHAR)
2346 {
2347 if (STRCMP(items[1], "long") == 0)
2348 aff->af_flagtype = AFT_LONG;
2349 else if (STRCMP(items[1], "num") == 0)
2350 aff->af_flagtype = AFT_NUM;
2351 else if (STRCMP(items[1], "caplong") == 0)
2352 aff->af_flagtype = AFT_CAPLONG;
2353 else
2354 smsg((char_u *)_("Invalid value for FLAG in %s line %d: %s"),
2355 fname, lnum, items[1]);
2356 if (aff->af_rare != 0
2357 || aff->af_keepcase != 0
2358 || aff->af_bad != 0
2359 || aff->af_needaffix != 0
2360 || aff->af_circumfix != 0
2361 || aff->af_needcomp != 0
2362 || aff->af_comproot != 0
2363 || aff->af_nosuggest != 0
2364 || compflags != NULL
2365 || aff->af_suff.ht_used > 0
2366 || aff->af_pref.ht_used > 0)
2367 smsg((char_u *)_("FLAG after using flags in %s line %d: %s"),
2368 fname, lnum, items[1]);
2369 }
2370 else if (spell_info_item(items[0]))
2371 {
2372 p = (char_u *)getroom(spin,
2373 (spin->si_info == NULL ? 0 : STRLEN(spin->si_info))
2374 + STRLEN(items[0])
2375 + STRLEN(items[1]) + 3, FALSE);
2376 if (p != NULL)
2377 {
2378 if (spin->si_info != NULL)
2379 {
2380 STRCPY(p, spin->si_info);
2381 STRCAT(p, "\n");
2382 }
2383 STRCAT(p, items[0]);
2384 STRCAT(p, " ");
2385 STRCAT(p, items[1]);
2386 spin->si_info = p;
2387 }
2388 }
2389 else if (is_aff_rule(items, itemcnt, "MIDWORD", 2)
2390 && midword == NULL)
2391 {
2392 midword = getroom_save(spin, items[1]);
2393 }
2394 else if (is_aff_rule(items, itemcnt, "TRY", 2))
2395 {
2396 /* ignored, we look in the tree for what chars may appear */
2397 }
2398 /* TODO: remove "RAR" later */
2399 else if ((is_aff_rule(items, itemcnt, "RAR", 2)
2400 || is_aff_rule(items, itemcnt, "RARE", 2))
2401 && aff->af_rare == 0)
2402 {
2403 aff->af_rare = affitem2flag(aff->af_flagtype, items[1],
2404 fname, lnum);
2405 }
2406 /* TODO: remove "KEP" later */
2407 else if ((is_aff_rule(items, itemcnt, "KEP", 2)
2408 || is_aff_rule(items, itemcnt, "KEEPCASE", 2))
2409 && aff->af_keepcase == 0)
2410 {
2411 aff->af_keepcase = affitem2flag(aff->af_flagtype, items[1],
2412 fname, lnum);
2413 }
2414 else if ((is_aff_rule(items, itemcnt, "BAD", 2)
2415 || is_aff_rule(items, itemcnt, "FORBIDDENWORD", 2))
2416 && aff->af_bad == 0)
2417 {
2418 aff->af_bad = affitem2flag(aff->af_flagtype, items[1],
2419 fname, lnum);
2420 }
2421 else if (is_aff_rule(items, itemcnt, "NEEDAFFIX", 2)
2422 && aff->af_needaffix == 0)
2423 {
2424 aff->af_needaffix = affitem2flag(aff->af_flagtype, items[1],
2425 fname, lnum);
2426 }
2427 else if (is_aff_rule(items, itemcnt, "CIRCUMFIX", 2)
2428 && aff->af_circumfix == 0)
2429 {
2430 aff->af_circumfix = affitem2flag(aff->af_flagtype, items[1],
2431 fname, lnum);
2432 }
2433 else if (is_aff_rule(items, itemcnt, "NOSUGGEST", 2)
2434 && aff->af_nosuggest == 0)
2435 {
2436 aff->af_nosuggest = affitem2flag(aff->af_flagtype, items[1],
2437 fname, lnum);
2438 }
2439 else if ((is_aff_rule(items, itemcnt, "NEEDCOMPOUND", 2)
2440 || is_aff_rule(items, itemcnt, "ONLYINCOMPOUND", 2))
2441 && aff->af_needcomp == 0)
2442 {
2443 aff->af_needcomp = affitem2flag(aff->af_flagtype, items[1],
2444 fname, lnum);
2445 }
2446 else if (is_aff_rule(items, itemcnt, "COMPOUNDROOT", 2)
2447 && aff->af_comproot == 0)
2448 {
2449 aff->af_comproot = affitem2flag(aff->af_flagtype, items[1],
2450 fname, lnum);
2451 }
2452 else if (is_aff_rule(items, itemcnt, "COMPOUNDFORBIDFLAG", 2)
2453 && aff->af_compforbid == 0)
2454 {
2455 aff->af_compforbid = affitem2flag(aff->af_flagtype, items[1],
2456 fname, lnum);
2457 if (aff->af_pref.ht_used > 0)
2458 smsg((char_u *)_("Defining COMPOUNDFORBIDFLAG after PFX item may give wrong results in %s line %d"),
2459 fname, lnum);
2460 }
2461 else if (is_aff_rule(items, itemcnt, "COMPOUNDPERMITFLAG", 2)
2462 && aff->af_comppermit == 0)
2463 {
2464 aff->af_comppermit = affitem2flag(aff->af_flagtype, items[1],
2465 fname, lnum);
2466 if (aff->af_pref.ht_used > 0)
2467 smsg((char_u *)_("Defining COMPOUNDPERMITFLAG after PFX item may give wrong results in %s line %d"),
2468 fname, lnum);
2469 }
2470 else if (is_aff_rule(items, itemcnt, "COMPOUNDFLAG", 2)
2471 && compflags == NULL)
2472 {
2473 /* Turn flag "c" into COMPOUNDRULE compatible string "c+",
2474 * "Na" into "Na+", "1234" into "1234+". */
2475 p = getroom(spin, STRLEN(items[1]) + 2, FALSE);
2476 if (p != NULL)
2477 {
2478 STRCPY(p, items[1]);
2479 STRCAT(p, "+");
2480 compflags = p;
2481 }
2482 }
2483 else if (is_aff_rule(items, itemcnt, "COMPOUNDRULES", 2))
2484 {
2485 /* We don't use the count, but do check that it's a number and
2486 * not COMPOUNDRULE mistyped. */
2487 if (atoi((char *)items[1]) == 0)
2488 smsg((char_u *)_("Wrong COMPOUNDRULES value in %s line %d: %s"),
2489 fname, lnum, items[1]);
2490 }
2491 else if (is_aff_rule(items, itemcnt, "COMPOUNDRULE", 2))
2492 {
2493 /* Don't use the first rule if it is a number. */
2494 if (compflags != NULL || *skipdigits(items[1]) != NUL)
2495 {
2496 /* Concatenate this string to previously defined ones,
2497 * using a slash to separate them. */
2498 l = (int)STRLEN(items[1]) + 1;
2499 if (compflags != NULL)
2500 l += (int)STRLEN(compflags) + 1;
2501 p = getroom(spin, l, FALSE);
2502 if (p != NULL)
2503 {
2504 if (compflags != NULL)
2505 {
2506 STRCPY(p, compflags);
2507 STRCAT(p, "/");
2508 }
2509 STRCAT(p, items[1]);
2510 compflags = p;
2511 }
2512 }
2513 }
2514 else if (is_aff_rule(items, itemcnt, "COMPOUNDWORDMAX", 2)
2515 && compmax == 0)
2516 {
2517 compmax = atoi((char *)items[1]);
2518 if (compmax == 0)
2519 smsg((char_u *)_("Wrong COMPOUNDWORDMAX value in %s line %d: %s"),
2520 fname, lnum, items[1]);
2521 }
2522 else if (is_aff_rule(items, itemcnt, "COMPOUNDMIN", 2)
2523 && compminlen == 0)
2524 {
2525 compminlen = atoi((char *)items[1]);
2526 if (compminlen == 0)
2527 smsg((char_u *)_("Wrong COMPOUNDMIN value in %s line %d: %s"),
2528 fname, lnum, items[1]);
2529 }
2530 else if (is_aff_rule(items, itemcnt, "COMPOUNDSYLMAX", 2)
2531 && compsylmax == 0)
2532 {
2533 compsylmax = atoi((char *)items[1]);
2534 if (compsylmax == 0)
2535 smsg((char_u *)_("Wrong COMPOUNDSYLMAX value in %s line %d: %s"),
2536 fname, lnum, items[1]);
2537 }
2538 else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDDUP", 1))
2539 {
2540 compoptions |= COMP_CHECKDUP;
2541 }
2542 else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDREP", 1))
2543 {
2544 compoptions |= COMP_CHECKREP;
2545 }
2546 else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDCASE", 1))
2547 {
2548 compoptions |= COMP_CHECKCASE;
2549 }
2550 else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDTRIPLE", 1))
2551 {
2552 compoptions |= COMP_CHECKTRIPLE;
2553 }
2554 else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDPATTERN", 2))
2555 {
2556 if (atoi((char *)items[1]) == 0)
2557 smsg((char_u *)_("Wrong CHECKCOMPOUNDPATTERN value in %s line %d: %s"),
2558 fname, lnum, items[1]);
2559 }
2560 else if (is_aff_rule(items, itemcnt, "CHECKCOMPOUNDPATTERN", 3))
2561 {
2562 garray_T *gap = &spin->si_comppat;
2563 int i;
2564
2565 /* Only add the couple if it isn't already there. */
2566 for (i = 0; i < gap->ga_len - 1; i += 2)
2567 if (STRCMP(((char_u **)(gap->ga_data))[i], items[1]) == 0
2568 && STRCMP(((char_u **)(gap->ga_data))[i + 1],
2569 items[2]) == 0)
2570 break;
2571 if (i >= gap->ga_len && ga_grow(gap, 2) == OK)
2572 {
2573 ((char_u **)(gap->ga_data))[gap->ga_len++]
2574 = getroom_save(spin, items[1]);
2575 ((char_u **)(gap->ga_data))[gap->ga_len++]
2576 = getroom_save(spin, items[2]);
2577 }
2578 }
2579 else if (is_aff_rule(items, itemcnt, "SYLLABLE", 2)
2580 && syllable == NULL)
2581 {
2582 syllable = getroom_save(spin, items[1]);
2583 }
2584 else if (is_aff_rule(items, itemcnt, "NOBREAK", 1))
2585 {
2586 spin->si_nobreak = TRUE;
2587 }
2588 else if (is_aff_rule(items, itemcnt, "NOSPLITSUGS", 1))
2589 {
2590 spin->si_nosplitsugs = TRUE;
2591 }
2592 else if (is_aff_rule(items, itemcnt, "NOCOMPOUNDSUGS", 1))
2593 {
2594 spin->si_nocompoundsugs = TRUE;
2595 }
2596 else if (is_aff_rule(items, itemcnt, "NOSUGFILE", 1))
2597 {
2598 spin->si_nosugfile = TRUE;
2599 }
2600 else if (is_aff_rule(items, itemcnt, "PFXPOSTPONE", 1))
2601 {
2602 aff->af_pfxpostpone = TRUE;
2603 }
2604 else if (is_aff_rule(items, itemcnt, "IGNOREEXTRA", 1))
2605 {
2606 aff->af_ignoreextra = TRUE;
2607 }
2608 else if ((STRCMP(items[0], "PFX") == 0
2609 || STRCMP(items[0], "SFX") == 0)
2610 && aff_todo == 0
2611 && itemcnt >= 4)
2612 {
2613 int lasti = 4;
2614 char_u key[AH_KEY_LEN];
2615
2616 if (*items[0] == 'P')
2617 tp = &aff->af_pref;
2618 else
2619 tp = &aff->af_suff;
2620
2621 /* Myspell allows the same affix name to be used multiple
2622 * times. The affix files that do this have an undocumented
2623 * "S" flag on all but the last block, thus we check for that
2624 * and store it in ah_follows. */
2625 vim_strncpy(key, items[1], AH_KEY_LEN - 1);
2626 hi = hash_find(tp, key);
2627 if (!HASHITEM_EMPTY(hi))
2628 {
2629 cur_aff = HI2AH(hi);
2630 if (cur_aff->ah_combine != (*items[2] == 'Y'))
2631 smsg((char_u *)_("Different combining flag in continued affix block in %s line %d: %s"),
2632 fname, lnum, items[1]);
2633 if (!cur_aff->ah_follows)
2634 smsg((char_u *)_("Duplicate affix in %s line %d: %s"),
2635 fname, lnum, items[1]);
2636 }
2637 else
2638 {
2639 /* New affix letter. */
2640 cur_aff = (affheader_T *)getroom(spin,
2641 sizeof(affheader_T), TRUE);
2642 if (cur_aff == NULL)
2643 break;
2644 cur_aff->ah_flag = affitem2flag(aff->af_flagtype, items[1],
2645 fname, lnum);
2646 if (cur_aff->ah_flag == 0 || STRLEN(items[1]) >= AH_KEY_LEN)
2647 break;
2648 if (cur_aff->ah_flag == aff->af_bad
2649 || cur_aff->ah_flag == aff->af_rare
2650 || cur_aff->ah_flag == aff->af_keepcase
2651 || cur_aff->ah_flag == aff->af_needaffix
2652 || cur_aff->ah_flag == aff->af_circumfix
2653 || cur_aff->ah_flag == aff->af_nosuggest
2654 || cur_aff->ah_flag == aff->af_needcomp
2655 || cur_aff->ah_flag == aff->af_comproot)
2656 smsg((char_u *)_("Affix also used for BAD/RARE/KEEPCASE/NEEDAFFIX/NEEDCOMPOUND/NOSUGGEST in %s line %d: %s"),
2657 fname, lnum, items[1]);
2658 STRCPY(cur_aff->ah_key, items[1]);
2659 hash_add(tp, cur_aff->ah_key);
2660
2661 cur_aff->ah_combine = (*items[2] == 'Y');
2662 }
2663
2664 /* Check for the "S" flag, which apparently means that another
2665 * block with the same affix name is following. */
2666 if (itemcnt > lasti && STRCMP(items[lasti], "S") == 0)
2667 {
2668 ++lasti;
2669 cur_aff->ah_follows = TRUE;
2670 }
2671 else
2672 cur_aff->ah_follows = FALSE;
2673
2674 /* Myspell allows extra text after the item, but that might
2675 * mean mistakes go unnoticed. Require a comment-starter. */
2676 if (itemcnt > lasti && *items[lasti] != '#')
2677 smsg((char_u *)_(e_afftrailing), fname, lnum, items[lasti]);
2678
2679 if (STRCMP(items[2], "Y") != 0 && STRCMP(items[2], "N") != 0)
2680 smsg((char_u *)_("Expected Y or N in %s line %d: %s"),
2681 fname, lnum, items[2]);
2682
2683 if (*items[0] == 'P' && aff->af_pfxpostpone)
2684 {
2685 if (cur_aff->ah_newID == 0)
2686 {
2687 /* Use a new number in the .spl file later, to be able
2688 * to handle multiple .aff files. */
2689 check_renumber(spin);
2690 cur_aff->ah_newID = ++spin->si_newprefID;
2691
2692 /* We only really use ah_newID if the prefix is
2693 * postponed. We know that only after handling all
2694 * the items. */
2695 did_postpone_prefix = FALSE;
2696 }
2697 else
2698 /* Did use the ID in a previous block. */
2699 did_postpone_prefix = TRUE;
2700 }
2701
2702 aff_todo = atoi((char *)items[3]);
2703 }
2704 else if ((STRCMP(items[0], "PFX") == 0
2705 || STRCMP(items[0], "SFX") == 0)
2706 && aff_todo > 0
2707 && STRCMP(cur_aff->ah_key, items[1]) == 0
2708 && itemcnt >= 5)
2709 {
2710 affentry_T *aff_entry;
2711 int upper = FALSE;
2712 int lasti = 5;
2713
2714 /* Myspell allows extra text after the item, but that might
2715 * mean mistakes go unnoticed. Require a comment-starter,
2716 * unless IGNOREEXTRA is used. Hunspell uses a "-" item. */
2717 if (itemcnt > lasti
2718 && !aff->af_ignoreextra
2719 && *items[lasti] != '#'
2720 && (STRCMP(items[lasti], "-") != 0
2721 || itemcnt != lasti + 1))
2722 smsg((char_u *)_(e_afftrailing), fname, lnum, items[lasti]);
2723
2724 /* New item for an affix letter. */
2725 --aff_todo;
2726 aff_entry = (affentry_T *)getroom(spin,
2727 sizeof(affentry_T), TRUE);
2728 if (aff_entry == NULL)
2729 break;
2730
2731 if (STRCMP(items[2], "0") != 0)
2732 aff_entry->ae_chop = getroom_save(spin, items[2]);
2733 if (STRCMP(items[3], "0") != 0)
2734 {
2735 aff_entry->ae_add = getroom_save(spin, items[3]);
2736
2737 /* Recognize flags on the affix: abcd/XYZ */
2738 aff_entry->ae_flags = vim_strchr(aff_entry->ae_add, '/');
2739 if (aff_entry->ae_flags != NULL)
2740 {
2741 *aff_entry->ae_flags++ = NUL;
2742 aff_process_flags(aff, aff_entry);
2743 }
2744 }
2745
2746 /* Don't use an affix entry with non-ASCII characters when
2747 * "spin->si_ascii" is TRUE. */
2748 if (!spin->si_ascii || !(has_non_ascii(aff_entry->ae_chop)
2749 || has_non_ascii(aff_entry->ae_add)))
2750 {
2751 aff_entry->ae_next = cur_aff->ah_first;
2752 cur_aff->ah_first = aff_entry;
2753
2754 if (STRCMP(items[4], ".") != 0)
2755 {
2756 char_u buf[MAXLINELEN];
2757
2758 aff_entry->ae_cond = getroom_save(spin, items[4]);
2759 if (*items[0] == 'P')
2760 sprintf((char *)buf, "^%s", items[4]);
2761 else
2762 sprintf((char *)buf, "%s$", items[4]);
2763 aff_entry->ae_prog = vim_regcomp(buf,
2764 RE_MAGIC + RE_STRING + RE_STRICT);
2765 if (aff_entry->ae_prog == NULL)
2766 smsg((char_u *)_("Broken condition in %s line %d: %s"),
2767 fname, lnum, items[4]);
2768 }
2769
2770 /* For postponed prefixes we need an entry in si_prefcond
2771 * for the condition. Use an existing one if possible.
2772 * Can't be done for an affix with flags, ignoring
2773 * COMPOUNDFORBIDFLAG and COMPOUNDPERMITFLAG. */
2774 if (*items[0] == 'P' && aff->af_pfxpostpone
2775 && aff_entry->ae_flags == NULL)
2776 {
2777 /* When the chop string is one lower-case letter and
2778 * the add string ends in the upper-case letter we set
2779 * the "upper" flag, clear "ae_chop" and remove the
2780 * letters from "ae_add". The condition must either
2781 * be empty or start with the same letter. */
2782 if (aff_entry->ae_chop != NULL
2783 && aff_entry->ae_add != NULL
2784 #ifdef FEAT_MBYTE
2785 && aff_entry->ae_chop[(*mb_ptr2len)(
2786 aff_entry->ae_chop)] == NUL
2787 #else
2788 && aff_entry->ae_chop[1] == NUL
2789 #endif
2790 )
2791 {
2792 int c, c_up;
2793
2794 c = PTR2CHAR(aff_entry->ae_chop);
2795 c_up = SPELL_TOUPPER(c);
2796 if (c_up != c
2797 && (aff_entry->ae_cond == NULL
2798 || PTR2CHAR(aff_entry->ae_cond) == c))
2799 {
2800 p = aff_entry->ae_add
2801 + STRLEN(aff_entry->ae_add);
2802 mb_ptr_back(aff_entry->ae_add, p);
2803 if (PTR2CHAR(p) == c_up)
2804 {
2805 upper = TRUE;
2806 aff_entry->ae_chop = NULL;
2807 *p = NUL;
2808
2809 /* The condition is matched with the
2810 * actual word, thus must check for the
2811 * upper-case letter. */
2812 if (aff_entry->ae_cond != NULL)
2813 {
2814 char_u buf[MAXLINELEN];
2815 #ifdef FEAT_MBYTE
2816 if (has_mbyte)
2817 {
2818 onecap_copy(items[4], buf, TRUE);
2819 aff_entry->ae_cond = getroom_save(
2820 spin, buf);
2821 }
2822 else
2823 #endif
2824 *aff_entry->ae_cond = c_up;
2825 if (aff_entry->ae_cond != NULL)
2826 {
2827 sprintf((char *)buf, "^%s",
2828 aff_entry->ae_cond);
2829 vim_regfree(aff_entry->ae_prog);
2830 aff_entry->ae_prog = vim_regcomp(
2831 buf, RE_MAGIC + RE_STRING);
2832 }
2833 }
2834 }
2835 }
2836 }
2837
2838 if (aff_entry->ae_chop == NULL
2839 && aff_entry->ae_flags == NULL)
2840 {
2841 int idx;
2842 char_u **pp;
2843 int n;
2844
2845 /* Find a previously used condition. */
2846 for (idx = spin->si_prefcond.ga_len - 1; idx >= 0;
2847 --idx)
2848 {
2849 p = ((char_u **)spin->si_prefcond.ga_data)[idx];
2850 if (str_equal(p, aff_entry->ae_cond))
2851 break;
2852 }
2853 if (idx < 0 && ga_grow(&spin->si_prefcond, 1) == OK)
2854 {
2855 /* Not found, add a new condition. */
2856 idx = spin->si_prefcond.ga_len++;
2857 pp = ((char_u **)spin->si_prefcond.ga_data)
2858 + idx;
2859 if (aff_entry->ae_cond == NULL)
2860 *pp = NULL;
2861 else
2862 *pp = getroom_save(spin,
2863 aff_entry->ae_cond);
2864 }
2865
2866 /* Add the prefix to the prefix tree. */
2867 if (aff_entry->ae_add == NULL)
2868 p = (char_u *)"";
2869 else
2870 p = aff_entry->ae_add;
2871
2872 /* PFX_FLAGS is a negative number, so that
2873 * tree_add_word() knows this is the prefix tree. */
2874 n = PFX_FLAGS;
2875 if (!cur_aff->ah_combine)
2876 n |= WFP_NC;
2877 if (upper)
2878 n |= WFP_UP;
2879 if (aff_entry->ae_comppermit)
2880 n |= WFP_COMPPERMIT;
2881 if (aff_entry->ae_compforbid)
2882 n |= WFP_COMPFORBID;
2883 tree_add_word(spin, p, spin->si_prefroot, n,
2884 idx, cur_aff->ah_newID);
2885 did_postpone_prefix = TRUE;
2886 }
2887
2888 /* Didn't actually use ah_newID, backup si_newprefID. */
2889 if (aff_todo == 0 && !did_postpone_prefix)
2890 {
2891 --spin->si_newprefID;
2892 cur_aff->ah_newID = 0;
2893 }
2894 }
2895 }
2896 }
2897 else if (is_aff_rule(items, itemcnt, "FOL", 2) && fol == NULL)
2898 {
2899 fol = vim_strsave(items[1]);
2900 }
2901 else if (is_aff_rule(items, itemcnt, "LOW", 2) && low == NULL)
2902 {
2903 low = vim_strsave(items[1]);
2904 }
2905 else if (is_aff_rule(items, itemcnt, "UPP", 2) && upp == NULL)
2906 {
2907 upp = vim_strsave(items[1]);
2908 }
2909 else if (is_aff_rule(items, itemcnt, "REP", 2)
2910 || is_aff_rule(items, itemcnt, "REPSAL", 2))
2911 {
2912 /* Ignore REP/REPSAL count */;
2913 if (!isdigit(*items[1]))
2914 smsg((char_u *)_("Expected REP(SAL) count in %s line %d"),
2915 fname, lnum);
2916 }
2917 else if ((STRCMP(items[0], "REP") == 0
2918 || STRCMP(items[0], "REPSAL") == 0)
2919 && itemcnt >= 3)
2920 {
2921 /* REP/REPSAL item */
2922 /* Myspell ignores extra arguments, we require it starts with
2923 * # to detect mistakes. */
2924 if (itemcnt > 3 && items[3][0] != '#')
2925 smsg((char_u *)_(e_afftrailing), fname, lnum, items[3]);
2926 if (items[0][3] == 'S' ? do_repsal : do_rep)
2927 {
2928 /* Replace underscore with space (can't include a space
2929 * directly). */
2930 for (p = items[1]; *p != NUL; mb_ptr_adv(p))
2931 if (*p == '_')
2932 *p = ' ';
2933 for (p = items[2]; *p != NUL; mb_ptr_adv(p))
2934 if (*p == '_')
2935 *p = ' ';
2936 add_fromto(spin, items[0][3] == 'S'
2937 ? &spin->si_repsal
2938 : &spin->si_rep, items[1], items[2]);
2939 }
2940 }
2941 else if (is_aff_rule(items, itemcnt, "MAP", 2))
2942 {
2943 /* MAP item or count */
2944 if (!found_map)
2945 {
2946 /* First line contains the count. */
2947 found_map = TRUE;
2948 if (!isdigit(*items[1]))
2949 smsg((char_u *)_("Expected MAP count in %s line %d"),
2950 fname, lnum);
2951 }
2952 else if (do_mapline)
2953 {
2954 int c;
2955
2956 /* Check that every character appears only once. */
2957 for (p = items[1]; *p != NUL; )
2958 {
2959 #ifdef FEAT_MBYTE
2960 c = mb_ptr2char_adv(&p);
2961 #else
2962 c = *p++;
2963 #endif
2964 if ((spin->si_map.ga_len > 0
2965 && vim_strchr(spin->si_map.ga_data, c)
2966 != NULL)
2967 || vim_strchr(p, c) != NULL)
2968 smsg((char_u *)_("Duplicate character in MAP in %s line %d"),
2969 fname, lnum);
2970 }
2971
2972 /* We simply concatenate all the MAP strings, separated by
2973 * slashes. */
2974 ga_concat(&spin->si_map, items[1]);
2975 ga_append(&spin->si_map, '/');
2976 }
2977 }
2978 /* Accept "SAL from to" and "SAL from to #comment". */
2979 else if (is_aff_rule(items, itemcnt, "SAL", 3))
2980 {
2981 if (do_sal)
2982 {
2983 /* SAL item (sounds-a-like)
2984 * Either one of the known keys or a from-to pair. */
2985 if (STRCMP(items[1], "followup") == 0)
2986 spin->si_followup = sal_to_bool(items[2]);
2987 else if (STRCMP(items[1], "collapse_result") == 0)
2988 spin->si_collapse = sal_to_bool(items[2]);
2989 else if (STRCMP(items[1], "remove_accents") == 0)
2990 spin->si_rem_accents = sal_to_bool(items[2]);
2991 else
2992 /* when "to" is "_" it means empty */
2993 add_fromto(spin, &spin->si_sal, items[1],
2994 STRCMP(items[2], "_") == 0 ? (char_u *)""
2995 : items[2]);
2996 }
2997 }
2998 else if (is_aff_rule(items, itemcnt, "SOFOFROM", 2)
2999 && sofofrom == NULL)
3000 {
3001 sofofrom = getroom_save(spin, items[1]);
3002 }
3003 else if (is_aff_rule(items, itemcnt, "SOFOTO", 2)
3004 && sofoto == NULL)
3005 {
3006 sofoto = getroom_save(spin, items[1]);
3007 }
3008 else if (STRCMP(items[0], "COMMON") == 0)
3009 {
3010 int i;
3011
3012 for (i = 1; i < itemcnt; ++i)
3013 {
3014 if (HASHITEM_EMPTY(hash_find(&spin->si_commonwords,
3015 items[i])))
3016 {
3017 p = vim_strsave(items[i]);
3018 if (p == NULL)
3019 break;
3020 hash_add(&spin->si_commonwords, p);
3021 }
3022 }
3023 }
3024 else
3025 smsg((char_u *)_("Unrecognized or duplicate item in %s line %d: %s"),
3026 fname, lnum, items[0]);
3027 }
3028 }
3029
3030 if (fol != NULL || low != NULL || upp != NULL)
3031 {
3032 if (spin->si_clear_chartab)
3033 {
3034 /* Clear the char type tables, don't want to use any of the
3035 * currently used spell properties. */
3036 init_spell_chartab();
3037 spin->si_clear_chartab = FALSE;
3038 }
3039
3040 /*
3041 * Don't write a word table for an ASCII file, so that we don't check
3042 * for conflicts with a word table that matches 'encoding'.
3043 * Don't write one for utf-8 either, we use utf_*() and
3044 * mb_get_class(), the list of chars in the file will be incomplete.
3045 */
3046 if (!spin->si_ascii
3047 #ifdef FEAT_MBYTE
3048 && !enc_utf8
3049 #endif
3050 )
3051 {
3052 if (fol == NULL || low == NULL || upp == NULL)
3053 smsg((char_u *)_("Missing FOL/LOW/UPP line in %s"), fname);
3054 else
3055 (void)set_spell_chartab(fol, low, upp);
3056 }
3057
3058 vim_free(fol);
3059 vim_free(low);
3060 vim_free(upp);
3061 }
3062
3063 /* Use compound specifications of the .aff file for the spell info. */
3064 if (compmax != 0)
3065 {
3066 aff_check_number(spin->si_compmax, compmax, "COMPOUNDWORDMAX");
3067 spin->si_compmax = compmax;
3068 }
3069
3070 if (compminlen != 0)
3071 {
3072 aff_check_number(spin->si_compminlen, compminlen, "COMPOUNDMIN");
3073 spin->si_compminlen = compminlen;
3074 }
3075
3076 if (compsylmax != 0)
3077 {
3078 if (syllable == NULL)
3079 smsg((char_u *)_("COMPOUNDSYLMAX used without SYLLABLE"));
3080 aff_check_number(spin->si_compsylmax, compsylmax, "COMPOUNDSYLMAX");
3081 spin->si_compsylmax = compsylmax;
3082 }
3083
3084 if (compoptions != 0)
3085 {
3086 aff_check_number(spin->si_compoptions, compoptions, "COMPOUND options");
3087 spin->si_compoptions |= compoptions;
3088 }
3089
3090 if (compflags != NULL)
3091 process_compflags(spin, aff, compflags);
3092
3093 /* Check that we didn't use too many renumbered flags. */
3094 if (spin->si_newcompID < spin->si_newprefID)
3095 {
3096 if (spin->si_newcompID == 127 || spin->si_newcompID == 255)
3097 MSG(_("Too many postponed prefixes"));
3098 else if (spin->si_newprefID == 0 || spin->si_newprefID == 127)
3099 MSG(_("Too many compound flags"));
3100 else
3101 MSG(_("Too many postponed prefixes and/or compound flags"));
3102 }
3103
3104 if (syllable != NULL)
3105 {
3106 aff_check_string(spin->si_syllable, syllable, "SYLLABLE");
3107 spin->si_syllable = syllable;
3108 }
3109
3110 if (sofofrom != NULL || sofoto != NULL)
3111 {
3112 if (sofofrom == NULL || sofoto == NULL)
3113 smsg((char_u *)_("Missing SOFO%s line in %s"),
3114 sofofrom == NULL ? "FROM" : "TO", fname);
3115 else if (spin->si_sal.ga_len > 0)
3116 smsg((char_u *)_("Both SAL and SOFO lines in %s"), fname);
3117 else
3118 {
3119 aff_check_string(spin->si_sofofr, sofofrom, "SOFOFROM");
3120 aff_check_string(spin->si_sofoto, sofoto, "SOFOTO");
3121 spin->si_sofofr = sofofrom;
3122 spin->si_sofoto = sofoto;
3123 }
3124 }
3125
3126 if (midword != NULL)
3127 {
3128 aff_check_string(spin->si_midword, midword, "MIDWORD");
3129 spin->si_midword = midword;
3130 }
3131
3132 vim_free(pc);
3133 fclose(fd);
3134 return aff;
3135 }
3136
3137 /*
3138 * Return TRUE when items[0] equals "rulename", there are "mincount" items or
3139 * a comment is following after item "mincount".
3140 */
3141 static int
3142 is_aff_rule(
3143 char_u **items,
3144 int itemcnt,
3145 char *rulename,
3146 int mincount)
3147 {
3148 return (STRCMP(items[0], rulename) == 0
3149 && (itemcnt == mincount
3150 || (itemcnt > mincount && items[mincount][0] == '#')));
3151 }
3152
3153 /*
3154 * For affix "entry" move COMPOUNDFORBIDFLAG and COMPOUNDPERMITFLAG from
3155 * ae_flags to ae_comppermit and ae_compforbid.
3156 */
3157 static void
3158 aff_process_flags(afffile_T *affile, affentry_T *entry)
3159 {
3160 char_u *p;
3161 char_u *prevp;
3162 unsigned flag;
3163
3164 if (entry->ae_flags != NULL
3165 && (affile->af_compforbid != 0 || affile->af_comppermit != 0))
3166 {
3167 for (p = entry->ae_flags; *p != NUL; )
3168 {
3169 prevp = p;
3170 flag = get_affitem(affile->af_flagtype, &p);
3171 if (flag == affile->af_comppermit || flag == affile->af_compforbid)
3172 {
3173 STRMOVE(prevp, p);
3174 p = prevp;
3175 if (flag == affile->af_comppermit)
3176 entry->ae_comppermit = TRUE;
3177 else
3178 entry->ae_compforbid = TRUE;
3179 }
3180 if (affile->af_flagtype == AFT_NUM && *p == ',')
3181 ++p;
3182 }
3183 if (*entry->ae_flags == NUL)
3184 entry->ae_flags = NULL; /* nothing left */
3185 }
3186 }
3187
3188 /*
3189 * Return TRUE if "s" is the name of an info item in the affix file.
3190 */
3191 static int
3192 spell_info_item(char_u *s)
3193 {
3194 return STRCMP(s, "NAME") == 0
3195 || STRCMP(s, "HOME") == 0
3196 || STRCMP(s, "VERSION") == 0
3197 || STRCMP(s, "AUTHOR") == 0
3198 || STRCMP(s, "EMAIL") == 0
3199 || STRCMP(s, "COPYRIGHT") == 0;
3200 }
3201
3202 /*
3203 * Turn an affix flag name into a number, according to the FLAG type.
3204 * returns zero for failure.
3205 */
3206 static unsigned
3207 affitem2flag(
3208 int flagtype,
3209 char_u *item,
3210 char_u *fname,
3211 int lnum)
3212 {
3213 unsigned res;
3214 char_u *p = item;
3215
3216 res = get_affitem(flagtype, &p);
3217 if (res == 0)
3218 {
3219 if (flagtype == AFT_NUM)
3220 smsg((char_u *)_("Flag is not a number in %s line %d: %s"),
3221 fname, lnum, item);
3222 else
3223 smsg((char_u *)_("Illegal flag in %s line %d: %s"),
3224 fname, lnum, item);
3225 }
3226 if (*p != NUL)
3227 {
3228 smsg((char_u *)_(e_affname), fname, lnum, item);
3229 return 0;
3230 }
3231
3232 return res;
3233 }
3234
3235 /*
3236 * Get one affix name from "*pp" and advance the pointer.
3237 * Returns zero for an error, still advances the pointer then.
3238 */
3239 static unsigned
3240 get_affitem(int flagtype, char_u **pp)
3241 {
3242 int res;
3243
3244 if (flagtype == AFT_NUM)
3245 {
3246 if (!VIM_ISDIGIT(**pp))
3247 {
3248 ++*pp; /* always advance, avoid getting stuck */
3249 return 0;
3250 }
3251 res = getdigits(pp);
3252 }
3253 else
3254 {
3255 #ifdef FEAT_MBYTE
3256 res = mb_ptr2char_adv(pp);
3257 #else
3258 res = *(*pp)++;
3259 #endif
3260 if (flagtype == AFT_LONG || (flagtype == AFT_CAPLONG
3261 && res >= 'A' && res <= 'Z'))
3262 {
3263 if (**pp == NUL)
3264 return 0;
3265 #ifdef FEAT_MBYTE
3266 res = mb_ptr2char_adv(pp) + (res << 16);
3267 #else
3268 res = *(*pp)++ + (res << 16);
3269 #endif
3270 }
3271 }
3272 return res;
3273 }
3274
3275 /*
3276 * Process the "compflags" string used in an affix file and append it to
3277 * spin->si_compflags.
3278 * The processing involves changing the affix names to ID numbers, so that
3279 * they fit in one byte.
3280 */
3281 static void
3282 process_compflags(
3283 spellinfo_T *spin,
3284 afffile_T *aff,
3285 char_u *compflags)
3286 {
3287 char_u *p;
3288 char_u *prevp;
3289 unsigned flag;
3290 compitem_T *ci;
3291 int id;
3292 int len;
3293 char_u *tp;
3294 char_u key[AH_KEY_LEN];
3295 hashitem_T *hi;
3296
3297 /* Make room for the old and the new compflags, concatenated with a / in
3298 * between. Processing it makes it shorter, but we don't know by how
3299 * much, thus allocate the maximum. */
3300 len = (int)STRLEN(compflags) + 1;
3301 if (spin->si_compflags != NULL)
3302 len += (int)STRLEN(spin->si_compflags) + 1;
3303 p = getroom(spin, len, FALSE);
3304 if (p == NULL)
3305 return;
3306 if (spin->si_compflags != NULL)
3307 {
3308 STRCPY(p, spin->si_compflags);
3309 STRCAT(p, "/");
3310 }
3311 spin->si_compflags = p;
3312 tp = p + STRLEN(p);
3313
3314 for (p = compflags; *p != NUL; )
3315 {
3316 if (vim_strchr((char_u *)"/?*+[]", *p) != NULL)
3317 /* Copy non-flag characters directly. */
3318 *tp++ = *p++;
3319 else
3320 {
3321 /* First get the flag number, also checks validity. */
3322 prevp = p;
3323 flag = get_affitem(aff->af_flagtype, &p);
3324 if (flag != 0)
3325 {
3326 /* Find the flag in the hashtable. If it was used before, use
3327 * the existing ID. Otherwise add a new entry. */
3328 vim_strncpy(key, prevp, p - prevp);
3329 hi = hash_find(&aff->af_comp, key);
3330 if (!HASHITEM_EMPTY(hi))
3331 id = HI2CI(hi)->ci_newID;
3332 else
3333 {
3334 ci = (compitem_T *)getroom(spin, sizeof(compitem_T), TRUE);
3335 if (ci == NULL)
3336 break;
3337 STRCPY(ci->ci_key, key);
3338 ci->ci_flag = flag;
3339 /* Avoid using a flag ID that has a special meaning in a
3340 * regexp (also inside []). */
3341 do
3342 {
3343 check_renumber(spin);
3344 id = spin->si_newcompID--;
3345 } while (vim_strchr((char_u *)"/?*+[]\\-^", id) != NULL);
3346 ci->ci_newID = id;
3347 hash_add(&aff->af_comp, ci->ci_key);
3348 }
3349 *tp++ = id;
3350 }
3351 if (aff->af_flagtype == AFT_NUM && *p == ',')
3352 ++p;
3353 }
3354 }
3355
3356 *tp = NUL;
3357 }
3358
3359 /*
3360 * Check that the new IDs for postponed affixes and compounding don't overrun
3361 * each other. We have almost 255 available, but start at 0-127 to avoid
3362 * using two bytes for utf-8. When the 0-127 range is used up go to 128-255.
3363 * When that is used up an error message is given.
3364 */
3365 static void
3366 check_renumber(spellinfo_T *spin)
3367 {
3368 if (spin->si_newprefID == spin->si_newcompID && spin->si_newcompID < 128)
3369 {
3370 spin->si_newprefID = 127;
3371 spin->si_newcompID = 255;
3372 }
3373 }
3374
3375 /*
3376 * Return TRUE if flag "flag" appears in affix list "afflist".
3377 */
3378 static int
3379 flag_in_afflist(int flagtype, char_u *afflist, unsigned flag)
3380 {
3381 char_u *p;
3382 unsigned n;
3383
3384 switch (flagtype)
3385 {
3386 case AFT_CHAR:
3387 return vim_strchr(afflist, flag) != NULL;
3388
3389 case AFT_CAPLONG:
3390 case AFT_LONG:
3391 for (p = afflist; *p != NUL; )
3392 {
3393 #ifdef FEAT_MBYTE
3394 n = mb_ptr2char_adv(&p);
3395 #else
3396 n = *p++;
3397 #endif
3398 if ((flagtype == AFT_LONG || (n >= 'A' && n <= 'Z'))
3399 && *p != NUL)
3400 #ifdef FEAT_MBYTE
3401 n = mb_ptr2char_adv(&p) + (n << 16);
3402 #else
3403 n = *p++ + (n << 16);
3404 #endif
3405 if (n == flag)
3406 return TRUE;
3407 }
3408 break;
3409
3410 case AFT_NUM:
3411 for (p = afflist; *p != NUL; )
3412 {
3413 n = getdigits(&p);
3414 if (n == flag)
3415 return TRUE;
3416 if (*p != NUL) /* skip over comma */
3417 ++p;
3418 }
3419 break;
3420 }
3421 return FALSE;
3422 }
3423
3424 /*
3425 * Give a warning when "spinval" and "affval" numbers are set and not the same.
3426 */
3427 static void
3428 aff_check_number(int spinval, int affval, char *name)
3429 {
3430 if (spinval != 0 && spinval != affval)
3431 smsg((char_u *)_("%s value differs from what is used in another .aff file"), name);
3432 }
3433
3434 /*
3435 * Give a warning when "spinval" and "affval" strings are set and not the same.
3436 */
3437 static void
3438 aff_check_string(char_u *spinval, char_u *affval, char *name)
3439 {
3440 if (spinval != NULL && STRCMP(spinval, affval) != 0)
3441 smsg((char_u *)_("%s value differs from what is used in another .aff file"), name);
3442 }
3443
3444 /*
3445 * Return TRUE if strings "s1" and "s2" are equal. Also consider both being
3446 * NULL as equal.
3447 */
3448 static int
3449 str_equal(char_u *s1, char_u *s2)
3450 {
3451 if (s1 == NULL || s2 == NULL)
3452 return s1 == s2;
3453 return STRCMP(s1, s2) == 0;
3454 }
3455
3456 /*
3457 * Add a from-to item to "gap". Used for REP and SAL items.
3458 * They are stored case-folded.
3459 */
3460 static void
3461 add_fromto(
3462 spellinfo_T *spin,
3463 garray_T *gap,
3464 char_u *from,
3465 char_u *to)
3466 {
3467 fromto_T *ftp;
3468 char_u word[MAXWLEN];
3469
3470 if (ga_grow(gap, 1) == OK)
3471 {
3472 ftp = ((fromto_T *)gap->ga_data) + gap->ga_len;
3473 (void)spell_casefold(from, (int)STRLEN(from), word, MAXWLEN);
3474 ftp->ft_from = getroom_save(spin, word);
3475 (void)spell_casefold(to, (int)STRLEN(to), word, MAXWLEN);
3476 ftp->ft_to = getroom_save(spin, word);
3477 ++gap->ga_len;
3478 }
3479 }
3480
3481 /*
3482 * Convert a boolean argument in a SAL line to TRUE or FALSE;
3483 */
3484 static int
3485 sal_to_bool(char_u *s)
3486 {
3487 return STRCMP(s, "1") == 0 || STRCMP(s, "true") == 0;
3488 }
3489
3490 /*
3491 * Free the structure filled by spell_read_aff().
3492 */
3493 static void
3494 spell_free_aff(afffile_T *aff)
3495 {
3496 hashtab_T *ht;
3497 hashitem_T *hi;
3498 int todo;
3499 affheader_T *ah;
3500 affentry_T *ae;
3501
3502 vim_free(aff->af_enc);
3503
3504 /* All this trouble to free the "ae_prog" items... */
3505 for (ht = &aff->af_pref; ; ht = &aff->af_suff)
3506 {
3507 todo = (int)ht->ht_used;
3508 for (hi = ht->ht_array; todo > 0; ++hi)
3509 {
3510 if (!HASHITEM_EMPTY(hi))
3511 {
3512 --todo;
3513 ah = HI2AH(hi);
3514 for (ae = ah->ah_first; ae != NULL; ae = ae->ae_next)
3515 vim_regfree(ae->ae_prog);
3516 }
3517 }
3518 if (ht == &aff->af_suff)
3519 break;
3520 }
3521
3522 hash_clear(&aff->af_pref);
3523 hash_clear(&aff->af_suff);
3524 hash_clear(&aff->af_comp);
3525 }
3526
3527 /*
3528 * Read dictionary file "fname".
3529 * Returns OK or FAIL;
3530 */
3531 static int
3532 spell_read_dic(spellinfo_T *spin, char_u *fname, afffile_T *affile)
3533 {
3534 hashtab_T ht;
3535 char_u line[MAXLINELEN];
3536 char_u *p;
3537 char_u *afflist;
3538 char_u store_afflist[MAXWLEN];
3539 int pfxlen;
3540 int need_affix;
3541 char_u *dw;
3542 char_u *pc;
3543 char_u *w;
3544 int l;
3545 hash_T hash;
3546 hashitem_T *hi;
3547 FILE *fd;
3548 int lnum = 1;
3549 int non_ascii = 0;
3550 int retval = OK;
3551 char_u message[MAXLINELEN + MAXWLEN];
3552 int flags;
3553 int duplicate = 0;
3554
3555 /*
3556 * Open the file.
3557 */
3558 fd = mch_fopen((char *)fname, "r");
3559 if (fd == NULL)
3560 {
3561 EMSG2(_(e_notopen), fname);
3562 return FAIL;
3563 }
3564
3565 /* The hashtable is only used to detect duplicated words. */
3566 hash_init(&ht);
3567
3568 vim_snprintf((char *)IObuff, IOSIZE,
3569 _("Reading dictionary file %s ..."), fname);
3570 spell_message(spin, IObuff);
3571
3572 /* start with a message for the first line */
3573 spin->si_msg_count = 999999;
3574
3575 /* Read and ignore the first line: word count. */
3576 (void)vim_fgets(line, MAXLINELEN, fd);
3577 if (!vim_isdigit(*skipwhite(line)))
3578 EMSG2(_("E760: No word count in %s"), fname);
3579
3580 /*
3581 * Read all the lines in the file one by one.
3582 * The words are converted to 'encoding' here, before being added to
3583 * the hashtable.
3584 */
3585 while (!vim_fgets(line, MAXLINELEN, fd) && !got_int)
3586 {
3587 line_breakcheck();
3588 ++lnum;
3589 if (line[0] == '#' || line[0] == '/')
3590 continue; /* comment line */
3591
3592 /* Remove CR, LF and white space from the end. White space halfway
3593 * the word is kept to allow e.g., "et al.". */
3594 l = (int)STRLEN(line);
3595 while (l > 0 && line[l - 1] <= ' ')
3596 --l;
3597 if (l == 0)
3598 continue; /* empty line */
3599 line[l] = NUL;
3600
3601 #ifdef FEAT_MBYTE
3602 /* Convert from "SET" to 'encoding' when needed. */
3603 if (spin->si_conv.vc_type != CONV_NONE)
3604 {
3605 pc = string_convert(&spin->si_conv, line, NULL);
3606 if (pc == NULL)
3607 {
3608 smsg((char_u *)_("Conversion failure for word in %s line %d: %s"),
3609 fname, lnum, line);
3610 continue;
3611 }
3612 w = pc;
3613 }
3614 else
3615 #endif
3616 {
3617 pc = NULL;
3618 w = line;
3619 }
3620
3621 /* Truncate the word at the "/", set "afflist" to what follows.
3622 * Replace "\/" by "/" and "\\" by "\". */
3623 afflist = NULL;
3624 for (p = w; *p != NUL; mb_ptr_adv(p))
3625 {
3626 if (*p == '\\' && (p[1] == '\\' || p[1] == '/'))
3627 STRMOVE(p, p + 1);
3628 else if (*p == '/')
3629 {
3630 *p = NUL;
3631 afflist = p + 1;
3632 break;
3633 }
3634 }
3635
3636 /* Skip non-ASCII words when "spin->si_ascii" is TRUE. */
3637 if (spin->si_ascii && has_non_ascii(w))
3638 {
3639 ++non_ascii;
3640 vim_free(pc);
3641 continue;
3642 }
3643
3644 /* This takes time, print a message every 10000 words. */
3645 if (spin->si_verbose && spin->si_msg_count > 10000)
3646 {
3647 spin->si_msg_count = 0;
3648 vim_snprintf((char *)message, sizeof(message),
3649 _("line %6d, word %6d - %s"),
3650 lnum, spin->si_foldwcount + spin->si_keepwcount, w);
3651 msg_start();
3652 msg_puts_long_attr(message, 0);
3653 msg_clr_eos();
3654 msg_didout = FALSE;
3655 msg_col = 0;
3656 out_flush();
3657 }
3658
3659 /* Store the word in the hashtable to be able to find duplicates. */
3660 dw = (char_u *)getroom_save(spin, w);
3661 if (dw == NULL)
3662 {
3663 retval = FAIL;
3664 vim_free(pc);
3665 break;
3666 }
3667
3668 hash = hash_hash(dw);
3669 hi = hash_lookup(&ht, dw, hash);
3670 if (!HASHITEM_EMPTY(hi))
3671 {
3672 if (p_verbose > 0)
3673 smsg((char_u *)_("Duplicate word in %s line %d: %s"),
3674 fname, lnum, dw);
3675 else if (duplicate == 0)
3676 smsg((char_u *)_("First duplicate word in %s line %d: %s"),
3677 fname, lnum, dw);
3678 ++duplicate;
3679 }
3680 else
3681 hash_add_item(&ht, hi, dw, hash);
3682
3683 flags = 0;
3684 store_afflist[0] = NUL;
3685 pfxlen = 0;
3686 need_affix = FALSE;
3687 if (afflist != NULL)
3688 {
3689 /* Extract flags from the affix list. */
3690 flags |= get_affix_flags(affile, afflist);
3691
3692 if (affile->af_needaffix != 0 && flag_in_afflist(
3693 affile->af_flagtype, afflist, affile->af_needaffix))
3694 need_affix = TRUE;
3695
3696 if (affile->af_pfxpostpone)
3697 /* Need to store the list of prefix IDs with the word. */
3698 pfxlen = get_pfxlist(affile, afflist, store_afflist);
3699
3700 if (spin->si_compflags != NULL)
3701 /* Need to store the list of compound flags with the word.
3702 * Concatenate them to the list of prefix IDs. */
3703 get_compflags(affile, afflist, store_afflist + pfxlen);
3704 }
3705
3706 /* Add the word to the word tree(s). */
3707 if (store_word(spin, dw, flags, spin->si_region,
3708 store_afflist, need_affix) == FAIL)
3709 retval = FAIL;
3710
3711 if (afflist != NULL)
3712 {
3713 /* Find all matching suffixes and add the resulting words.
3714 * Additionally do matching prefixes that combine. */
3715 if (store_aff_word(spin, dw, afflist, affile,
3716 &affile->af_suff, &affile->af_pref,
3717 CONDIT_SUF, flags, store_afflist, pfxlen) == FAIL)
3718 retval = FAIL;
3719
3720 /* Find all matching prefixes and add the resulting words. */
3721 if (store_aff_word(spin, dw, afflist, affile,
3722 &affile->af_pref, NULL,
3723 CONDIT_SUF, flags, store_afflist, pfxlen) == FAIL)
3724 retval = FAIL;
3725 }
3726
3727 vim_free(pc);
3728 }
3729
3730 if (duplicate > 0)
3731 smsg((char_u *)_("%d duplicate word(s) in %s"), duplicate, fname);
3732 if (spin->si_ascii && non_ascii > 0)
3733 smsg((char_u *)_("Ignored %d word(s) with non-ASCII characters in %s"),
3734 non_ascii, fname);
3735 hash_clear(&ht);
3736
3737 fclose(fd);
3738 return retval;
3739 }
3740
3741 /*
3742 * Check for affix flags in "afflist" that are turned into word flags.
3743 * Return WF_ flags.
3744 */
3745 static int
3746 get_affix_flags(afffile_T *affile, char_u *afflist)
3747 {
3748 int flags = 0;
3749
3750 if (affile->af_keepcase != 0 && flag_in_afflist(
3751 affile->af_flagtype, afflist, affile->af_keepcase))
3752 flags |= WF_KEEPCAP | WF_FIXCAP;
3753 if (affile->af_rare != 0 && flag_in_afflist(
3754 affile->af_flagtype, afflist, affile->af_rare))
3755 flags |= WF_RARE;
3756 if (affile->af_bad != 0 && flag_in_afflist(
3757 affile->af_flagtype, afflist, affile->af_bad))
3758 flags |= WF_BANNED;
3759 if (affile->af_needcomp != 0 && flag_in_afflist(
3760 affile->af_flagtype, afflist, affile->af_needcomp))
3761 flags |= WF_NEEDCOMP;
3762 if (affile->af_comproot != 0 && flag_in_afflist(
3763 affile->af_flagtype, afflist, affile->af_comproot))
3764 flags |= WF_COMPROOT;
3765 if (affile->af_nosuggest != 0 && flag_in_afflist(
3766 affile->af_flagtype, afflist, affile->af_nosuggest))
3767 flags |= WF_NOSUGGEST;
3768 return flags;
3769 }
3770
3771 /*
3772 * Get the list of prefix IDs from the affix list "afflist".
3773 * Used for PFXPOSTPONE.
3774 * Put the resulting flags in "store_afflist[MAXWLEN]" with a terminating NUL
3775 * and return the number of affixes.
3776 */
3777 static int
3778 get_pfxlist(
3779 afffile_T *affile,
3780 char_u *afflist,
3781 char_u *store_afflist)
3782 {
3783 char_u *p;
3784 char_u *prevp;
3785 int cnt = 0;
3786 int id;
3787 char_u key[AH_KEY_LEN];
3788 hashitem_T *hi;
3789
3790 for (p = afflist; *p != NUL; )
3791 {
3792 prevp = p;
3793 if (get_affitem(affile->af_flagtype, &p) != 0)
3794 {
3795 /* A flag is a postponed prefix flag if it appears in "af_pref"
3796 * and it's ID is not zero. */
3797 vim_strncpy(key, prevp, p - prevp);
3798 hi = hash_find(&affile->af_pref, key);
3799 if (!HASHITEM_EMPTY(hi))
3800 {
3801 id = HI2AH(hi)->ah_newID;
3802 if (id != 0)
3803 store_afflist[cnt++] = id;
3804 }
3805 }
3806 if (affile->af_flagtype == AFT_NUM && *p == ',')
3807 ++p;
3808 }
3809
3810 store_afflist[cnt] = NUL;
3811 return cnt;
3812 }
3813
3814 /*
3815 * Get the list of compound IDs from the affix list "afflist" that are used
3816 * for compound words.
3817 * Puts the flags in "store_afflist[]".
3818 */
3819 static void
3820 get_compflags(
3821 afffile_T *affile,
3822 char_u *afflist,
3823 char_u *store_afflist)
3824 {
3825 char_u *p;
3826 char_u *prevp;
3827 int cnt = 0;
3828 char_u key[AH_KEY_LEN];
3829 hashitem_T *hi;
3830
3831 for (p = afflist; *p != NUL; )
3832 {
3833 prevp = p;
3834 if (get_affitem(affile->af_flagtype, &p) != 0)
3835 {
3836 /* A flag is a compound flag if it appears in "af_comp". */
3837 vim_strncpy(key, prevp, p - prevp);
3838 hi = hash_find(&affile->af_comp, key);
3839 if (!HASHITEM_EMPTY(hi))
3840 store_afflist[cnt++] = HI2CI(hi)->ci_newID;
3841 }
3842 if (affile->af_flagtype == AFT_NUM && *p == ',')
3843 ++p;
3844 }
3845
3846 store_afflist[cnt] = NUL;
3847 }
3848
3849 /*
3850 * Apply affixes to a word and store the resulting words.
3851 * "ht" is the hashtable with affentry_T that need to be applied, either
3852 * prefixes or suffixes.
3853 * "xht", when not NULL, is the prefix hashtable, to be used additionally on
3854 * the resulting words for combining affixes.
3855 *
3856 * Returns FAIL when out of memory.
3857 */
3858 static int
3859 store_aff_word(
3860 spellinfo_T *spin, /* spell info */
3861 char_u *word, /* basic word start */
3862 char_u *afflist, /* list of names of supported affixes */
3863 afffile_T *affile,
3864 hashtab_T *ht,
3865 hashtab_T *xht,
3866 int condit, /* CONDIT_SUF et al. */
3867 int flags, /* flags for the word */
3868 char_u *pfxlist, /* list of prefix IDs */
3869 int pfxlen) /* nr of flags in "pfxlist" for prefixes, rest
3870 * is compound flags */
3871 {
3872 int todo;
3873 hashitem_T *hi;
3874 affheader_T *ah;
3875 affentry_T *ae;
3876 char_u newword[MAXWLEN];
3877 int retval = OK;
3878 int i, j;
3879 char_u *p;
3880 int use_flags;
3881 char_u *use_pfxlist;
3882 int use_pfxlen;
3883 int need_affix;
3884 char_u store_afflist[MAXWLEN];
3885 char_u pfx_pfxlist[MAXWLEN];
3886 size_t wordlen = STRLEN(word);
3887 int use_condit;
3888
3889 todo = (int)ht->ht_used;
3890 for (hi = ht->ht_array; todo > 0 && retval == OK; ++hi)
3891 {
3892 if (!HASHITEM_EMPTY(hi))
3893 {
3894 --todo;
3895 ah = HI2AH(hi);
3896
3897 /* Check that the affix combines, if required, and that the word
3898 * supports this affix. */
3899 if (((condit & CONDIT_COMB) == 0 || ah->ah_combine)
3900 && flag_in_afflist(affile->af_flagtype, afflist,
3901 ah->ah_flag))
3902 {
3903 /* Loop over all affix entries with this name. */
3904 for (ae = ah->ah_first; ae != NULL; ae = ae->ae_next)
3905 {
3906 /* Check the condition. It's not logical to match case
3907 * here, but it is required for compatibility with
3908 * Myspell.
3909 * Another requirement from Myspell is that the chop
3910 * string is shorter than the word itself.
3911 * For prefixes, when "PFXPOSTPONE" was used, only do
3912 * prefixes with a chop string and/or flags.
3913 * When a previously added affix had CIRCUMFIX this one
3914 * must have it too, if it had not then this one must not
3915 * have one either. */
3916 if ((xht != NULL || !affile->af_pfxpostpone
3917 || ae->ae_chop != NULL
3918 || ae->ae_flags != NULL)
3919 && (ae->ae_chop == NULL
3920 || STRLEN(ae->ae_chop) < wordlen)
3921 && (ae->ae_prog == NULL
3922 || vim_regexec_prog(&ae->ae_prog, FALSE,
3923 word, (colnr_T)0))
3924 && (((condit & CONDIT_CFIX) == 0)
3925 == ((condit & CONDIT_AFF) == 0
3926 || ae->ae_flags == NULL
3927 || !flag_in_afflist(affile->af_flagtype,
3928 ae->ae_flags, affile->af_circumfix))))
3929 {
3930 /* Match. Remove the chop and add the affix. */
3931 if (xht == NULL)
3932 {
3933 /* prefix: chop/add at the start of the word */
3934 if (ae->ae_add == NULL)
3935 *newword = NUL;
3936 else
3937 vim_strncpy(newword, ae->ae_add, MAXWLEN - 1);
3938 p = word;
3939 if (ae->ae_chop != NULL)
3940 {
3941 /* Skip chop string. */
3942 #ifdef FEAT_MBYTE
3943 if (has_mbyte)
3944 {
3945 i = mb_charlen(ae->ae_chop);
3946 for ( ; i > 0; --i)
3947 mb_ptr_adv(p);
3948 }
3949 else
3950 #endif
3951 p += STRLEN(ae->ae_chop);
3952 }
3953 STRCAT(newword, p);
3954 }
3955 else
3956 {
3957 /* suffix: chop/add at the end of the word */
3958 vim_strncpy(newword, word, MAXWLEN - 1);
3959 if (ae->ae_chop != NULL)
3960 {
3961 /* Remove chop string. */
3962 p = newword + STRLEN(newword);
3963 i = (int)MB_CHARLEN(ae->ae_chop);
3964 for ( ; i > 0; --i)
3965 mb_ptr_back(newword, p);
3966 *p = NUL;
3967 }
3968 if (ae->ae_add != NULL)
3969 STRCAT(newword, ae->ae_add);
3970 }
3971
3972 use_flags = flags;
3973 use_pfxlist = pfxlist;
3974 use_pfxlen = pfxlen;
3975 need_affix = FALSE;
3976 use_condit = condit | CONDIT_COMB | CONDIT_AFF;
3977 if (ae->ae_flags != NULL)
3978 {
3979 /* Extract flags from the affix list. */
3980 use_flags |= get_affix_flags(affile, ae->ae_flags);
3981
3982 if (affile->af_needaffix != 0 && flag_in_afflist(
3983 affile->af_flagtype, ae->ae_flags,
3984 affile->af_needaffix))
3985 need_affix = TRUE;
3986
3987 /* When there is a CIRCUMFIX flag the other affix
3988 * must also have it and we don't add the word
3989 * with one affix. */
3990 if (affile->af_circumfix != 0 && flag_in_afflist(
3991 affile->af_flagtype, ae->ae_flags,
3992 affile->af_circumfix))
3993 {
3994 use_condit |= CONDIT_CFIX;
3995 if ((condit & CONDIT_CFIX) == 0)
3996 need_affix = TRUE;
3997 }
3998
3999 if (affile->af_pfxpostpone
4000 || spin->si_compflags != NULL)
4001 {
4002 if (affile->af_pfxpostpone)
4003 /* Get prefix IDS from the affix list. */
4004 use_pfxlen = get_pfxlist(affile,
4005 ae->ae_flags, store_afflist);
4006 else
4007 use_pfxlen = 0;
4008 use_pfxlist = store_afflist;
4009
4010 /* Combine the prefix IDs. Avoid adding the
4011 * same ID twice. */
4012 for (i = 0; i < pfxlen; ++i)
4013 {
4014 for (j = 0; j < use_pfxlen; ++j)
4015 if (pfxlist[i] == use_pfxlist[j])
4016 break;
4017 if (j == use_pfxlen)
4018 use_pfxlist[use_pfxlen++] = pfxlist[i];
4019 }
4020
4021 if (spin->si_compflags != NULL)
4022 /* Get compound IDS from the affix list. */
4023 get_compflags(affile, ae->ae_flags,
4024 use_pfxlist + use_pfxlen);
4025
4026 /* Combine the list of compound flags.
4027 * Concatenate them to the prefix IDs list.
4028 * Avoid adding the same ID twice. */
4029 for (i = pfxlen; pfxlist[i] != NUL; ++i)
4030 {
4031 for (j = use_pfxlen;
4032 use_pfxlist[j] != NUL; ++j)
4033 if (pfxlist[i] == use_pfxlist[j])
4034 break;
4035 if (use_pfxlist[j] == NUL)
4036 {
4037 use_pfxlist[j++] = pfxlist[i];
4038 use_pfxlist[j] = NUL;
4039 }
4040 }
4041 }
4042 }
4043
4044 /* Obey a "COMPOUNDFORBIDFLAG" of the affix: don't
4045 * use the compound flags. */
4046 if (use_pfxlist != NULL && ae->ae_compforbid)
4047 {
4048 vim_strncpy(pfx_pfxlist, use_pfxlist, use_pfxlen);
4049 use_pfxlist = pfx_pfxlist;
4050 }
4051
4052 /* When there are postponed prefixes... */
4053 if (spin->si_prefroot != NULL
4054 && spin->si_prefroot->wn_sibling != NULL)
4055 {
4056 /* ... add a flag to indicate an affix was used. */
4057 use_flags |= WF_HAS_AFF;
4058
4059 /* ... don't use a prefix list if combining
4060 * affixes is not allowed. But do use the
4061 * compound flags after them. */
4062 if (!ah->ah_combine && use_pfxlist != NULL)
4063 use_pfxlist += use_pfxlen;
4064 }
4065
4066 /* When compounding is supported and there is no
4067 * "COMPOUNDPERMITFLAG" then forbid compounding on the
4068 * side where the affix is applied. */
4069 if (spin->si_compflags != NULL && !ae->ae_comppermit)
4070 {
4071 if (xht != NULL)
4072 use_flags |= WF_NOCOMPAFT;
4073 else
4074 use_flags |= WF_NOCOMPBEF;
4075 }
4076
4077 /* Store the modified word. */
4078 if (store_word(spin, newword, use_flags,
4079 spin->si_region, use_pfxlist,
4080 need_affix) == FAIL)
4081 retval = FAIL;
4082
4083 /* When added a prefix or a first suffix and the affix
4084 * has flags may add a(nother) suffix. RECURSIVE! */
4085 if ((condit & CONDIT_SUF) && ae->ae_flags != NULL)
4086 if (store_aff_word(spin, newword, ae->ae_flags,
4087 affile, &affile->af_suff, xht,
4088 use_condit & (xht == NULL
4089 ? ~0 : ~CONDIT_SUF),
4090 use_flags, use_pfxlist, pfxlen) == FAIL)
4091 retval = FAIL;
4092
4093 /* When added a suffix and combining is allowed also
4094 * try adding a prefix additionally. Both for the
4095 * word flags and for the affix flags. RECURSIVE! */
4096 if (xht != NULL && ah->ah_combine)
4097 {
4098 if (store_aff_word(spin, newword,
4099 afflist, affile,
4100 xht, NULL, use_condit,
4101 use_flags, use_pfxlist,
4102 pfxlen) == FAIL
4103 || (ae->ae_flags != NULL
4104 && store_aff_word(spin, newword,
4105 ae->ae_flags, affile,
4106 xht, NULL, use_condit,
4107 use_flags, use_pfxlist,
4108 pfxlen) == FAIL))
4109 retval = FAIL;
4110 }
4111 }
4112 }
4113 }
4114 }
4115 }
4116
4117 return retval;
4118 }
4119
4120 /*
4121 * Read a file with a list of words.
4122 */
4123 static int
4124 spell_read_wordfile(spellinfo_T *spin, char_u *fname)
4125 {
4126 FILE *fd;
4127 long lnum = 0;
4128 char_u rline[MAXLINELEN];
4129 char_u *line;
4130 char_u *pc = NULL;
4131 char_u *p;
4132 int l;
4133 int retval = OK;
4134 int did_word = FALSE;
4135 int non_ascii = 0;
4136 int flags;
4137 int regionmask;
4138
4139 /*
4140 * Open the file.
4141 */
4142 fd = mch_fopen((char *)fname, "r");
4143 if (fd == NULL)
4144 {
4145 EMSG2(_(e_notopen), fname);
4146 return FAIL;
4147 }
4148
4149 vim_snprintf((char *)IObuff, IOSIZE, _("Reading word file %s ..."), fname);
4150 spell_message(spin, IObuff);
4151
4152 /*
4153 * Read all the lines in the file one by one.
4154 */
4155 while (!vim_fgets(rline, MAXLINELEN, fd) && !got_int)
4156 {
4157 line_breakcheck();
4158 ++lnum;
4159
4160 /* Skip comment lines. */
4161 if (*rline == '#')
4162 continue;
4163
4164 /* Remove CR, LF and white space from the end. */
4165 l = (int)STRLEN(rline);
4166 while (l > 0 && rline[l - 1] <= ' ')
4167 --l;
4168 if (l == 0)
4169 continue; /* empty or blank line */
4170 rline[l] = NUL;
4171
4172 /* Convert from "/encoding={encoding}" to 'encoding' when needed. */
4173 vim_free(pc);
4174 #ifdef FEAT_MBYTE
4175 if (spin->si_conv.vc_type != CONV_NONE)
4176 {
4177 pc = string_convert(&spin->si_conv, rline, NULL);
4178 if (pc == NULL)
4179 {
4180 smsg((char_u *)_("Conversion failure for word in %s line %d: %s"),
4181 fname, lnum, rline);
4182 continue;
4183 }
4184 line = pc;
4185 }
4186 else
4187 #endif
4188 {
4189 pc = NULL;
4190 line = rline;
4191 }
4192
4193 if (*line == '/')
4194 {
4195 ++line;
4196 if (STRNCMP(line, "encoding=", 9) == 0)
4197 {
4198 if (spin->si_conv.vc_type != CONV_NONE)
4199 smsg((char_u *)_("Duplicate /encoding= line ignored in %s line %d: %s"),
4200 fname, lnum, line - 1);
4201 else if (did_word)
4202 smsg((char_u *)_("/encoding= line after word ignored in %s line %d: %s"),
4203 fname, lnum, line - 1);
4204 else
4205 {
4206 #ifdef FEAT_MBYTE
4207 char_u *enc;
4208
4209 /* Setup for conversion to 'encoding'. */
4210 line += 9;
4211 enc = enc_canonize(line);
4212 if (enc != NULL && !spin->si_ascii
4213 && convert_setup(&spin->si_conv, enc,
4214 p_enc) == FAIL)
4215 smsg((char_u *)_("Conversion in %s not supported: from %s to %s"),
4216 fname, line, p_enc);
4217 vim_free(enc);
4218 spin->si_conv.vc_fail = TRUE;
4219 #else
4220 smsg((char_u *)_("Conversion in %s not supported"), fname);
4221 #endif
4222 }
4223 continue;
4224 }
4225
4226 if (STRNCMP(line, "regions=", 8) == 0)
4227 {
4228 if (spin->si_region_count > 1)
4229 smsg((char_u *)_("Duplicate /regions= line ignored in %s line %d: %s"),
4230 fname, lnum, line);
4231 else
4232 {
4233 line += 8;
4234 if (STRLEN(line) > 16)
4235 smsg((char_u *)_("Too many regions in %s line %d: %s"),
4236 fname, lnum, line);
4237 else
4238 {
4239 spin->si_region_count = (int)STRLEN(line) / 2;
4240 STRCPY(spin->si_region_name, line);
4241
4242 /* Adjust the mask for a word valid in all regions. */
4243 spin->si_region = (1 << spin->si_region_count) - 1;
4244 }
4245 }
4246 continue;
4247 }
4248
4249 smsg((char_u *)_("/ line ignored in %s line %d: %s"),
4250 fname, lnum, line - 1);
4251 continue;
4252 }
4253
4254 flags = 0;
4255 regionmask = spin->si_region;
4256
4257 /* Check for flags and region after a slash. */
4258 p = vim_strchr(line, '/');
4259 if (p != NULL)
4260 {
4261 *p++ = NUL;
4262 while (*p != NUL)
4263 {
4264 if (*p == '=') /* keep-case word */
4265 flags |= WF_KEEPCAP | WF_FIXCAP;
4266 else if (*p == '!') /* Bad, bad, wicked word. */
4267 flags |= WF_BANNED;
4268 else if (*p == '?') /* Rare word. */
4269 flags |= WF_RARE;
4270 else if (VIM_ISDIGIT(*p)) /* region number(s) */
4271 {
4272 if ((flags & WF_REGION) == 0) /* first one */
4273 regionmask = 0;
4274 flags |= WF_REGION;
4275
4276 l = *p - '0';
4277 if (l > spin->si_region_count)
4278 {
4279 smsg((char_u *)_("Invalid region nr in %s line %d: %s"),
4280 fname, lnum, p);
4281 break;
4282 }
4283 regionmask |= 1 << (l - 1);
4284 }
4285 else
4286 {
4287 smsg((char_u *)_("Unrecognized flags in %s line %d: %s"),
4288 fname, lnum, p);
4289 break;
4290 }
4291 ++p;
4292 }
4293 }
4294
4295 /* Skip non-ASCII words when "spin->si_ascii" is TRUE. */
4296 if (spin->si_ascii && has_non_ascii(line))
4297 {
4298 ++non_ascii;
4299 continue;
4300 }
4301
4302 /* Normal word: store it. */
4303 if (store_word(spin, line, flags, regionmask, NULL, FALSE) == FAIL)
4304 {
4305 retval = FAIL;
4306 break;
4307 }
4308 did_word = TRUE;
4309 }
4310
4311 vim_free(pc);
4312 fclose(fd);
4313
4314 if (spin->si_ascii && non_ascii > 0)
4315 {
4316 vim_snprintf((char *)IObuff, IOSIZE,
4317 _("Ignored %d words with non-ASCII characters"), non_ascii);
4318 spell_message(spin, IObuff);
4319 }
4320
4321 return retval;
4322 }
4323
4324 /*
4325 * Get part of an sblock_T, "len" bytes long.
4326 * This avoids calling free() for every little struct we use (and keeping
4327 * track of them).
4328 * The memory is cleared to all zeros.
4329 * Returns NULL when out of memory.
4330 */
4331 static void *
4332 getroom(
4333 spellinfo_T *spin,
4334 size_t len, /* length needed */
4335 int align) /* align for pointer */
4336 {
4337 char_u *p;
4338 sblock_T *bl = spin->si_blocks;
4339
4340 if (align && bl != NULL)
4341 /* Round size up for alignment. On some systems structures need to be
4342 * aligned to the size of a pointer (e.g., SPARC). */
4343 bl->sb_used = (bl->sb_used + sizeof(char *) - 1)
4344 & ~(sizeof(char *) - 1);
4345
4346 if (bl == NULL || bl->sb_used + len > SBLOCKSIZE)
4347 {
4348 if (len >= SBLOCKSIZE)
4349 bl = NULL;
4350 else
4351 /* Allocate a block of memory. It is not freed until much later. */
4352 bl = (sblock_T *)alloc_clear(
4353 (unsigned)(sizeof(sblock_T) + SBLOCKSIZE));
4354 if (bl == NULL)
4355 {
4356 if (!spin->si_did_emsg)
4357 {
4358 EMSG(_("E845: Insufficient memory, word list will be incomplete"));
4359 spin->si_did_emsg = TRUE;
4360 }
4361 return NULL;
4362 }
4363 bl->sb_next = spin->si_blocks;
4364 spin->si_blocks = bl;
4365 bl->sb_used = 0;
4366 ++spin->si_blocks_cnt;
4367 }
4368
4369 p = bl->sb_data + bl->sb_used;
4370 bl->sb_used += (int)len;
4371
4372 return p;
4373 }
4374
4375 /*
4376 * Make a copy of a string into memory allocated with getroom().
4377 * Returns NULL when out of memory.
4378 */
4379 static char_u *
4380 getroom_save(spellinfo_T *spin, char_u *s)
4381 {
4382 char_u *sc;
4383
4384 sc = (char_u *)getroom(spin, STRLEN(s) + 1, FALSE);
4385 if (sc != NULL)
4386 STRCPY(sc, s);
4387 return sc;
4388 }
4389
4390
4391 /*
4392 * Free the list of allocated sblock_T.
4393 */
4394 static void
4395 free_blocks(sblock_T *bl)
4396 {
4397 sblock_T *next;
4398
4399 while (bl != NULL)
4400 {
4401 next = bl->sb_next;
4402 vim_free(bl);
4403 bl = next;
4404 }
4405 }
4406
4407 /*
4408 * Allocate the root of a word tree.
4409 * Returns NULL when out of memory.
4410 */
4411 static wordnode_T *
4412 wordtree_alloc(spellinfo_T *spin)
4413 {
4414 return (wordnode_T *)getroom(spin, sizeof(wordnode_T), TRUE);
4415 }
4416
4417 /*
4418 * Store a word in the tree(s).
4419 * Always store it in the case-folded tree. For a keep-case word this is
4420 * useful when the word can also be used with all caps (no WF_FIXCAP flag) and
4421 * used to find suggestions.
4422 * For a keep-case word also store it in the keep-case tree.
4423 * When "pfxlist" is not NULL store the word for each postponed prefix ID and
4424 * compound flag.
4425 */
4426 static int
4427 store_word(
4428 spellinfo_T *spin,
4429 char_u *word,
4430 int flags, /* extra flags, WF_BANNED */
4431 int region, /* supported region(s) */
4432 char_u *pfxlist, /* list of prefix IDs or NULL */
4433 int need_affix) /* only store word with affix ID */
4434 {
4435 int len = (int)STRLEN(word);
4436 int ct = captype(word, word + len);
4437 char_u foldword[MAXWLEN];
4438 int res = OK;
4439 char_u *p;
4440
4441 (void)spell_casefold(word, len, foldword, MAXWLEN);
4442 for (p = pfxlist; res == OK; ++p)
4443 {
4444 if (!need_affix || (p != NULL && *p != NUL))
4445 res = tree_add_word(spin, foldword, spin->si_foldroot, ct | flags,
4446 region, p == NULL ? 0 : *p);
4447 if (p == NULL || *p == NUL)
4448 break;
4449 }
4450 ++spin->si_foldwcount;
4451
4452 if (res == OK && (ct == WF_KEEPCAP || (flags & WF_KEEPCAP)))
4453 {
4454 for (p = pfxlist; res == OK; ++p)
4455 {
4456 if (!need_affix || (p != NULL && *p != NUL))
4457 res = tree_add_word(spin, word, spin->si_keeproot, flags,
4458 region, p == NULL ? 0 : *p);
4459 if (p == NULL || *p == NUL)
4460 break;
4461 }
4462 ++spin->si_keepwcount;
4463 }
4464 return res;
4465 }
4466
4467 /*
4468 * Add word "word" to a word tree at "root".
4469 * When "flags" < 0 we are adding to the prefix tree where "flags" is used for
4470 * "rare" and "region" is the condition nr.
4471 * Returns FAIL when out of memory.
4472 */
4473 static int
4474 tree_add_word(
4475 spellinfo_T *spin,
4476 char_u *word,
4477 wordnode_T *root,
4478 int flags,
4479 int region,
4480 int affixID)
4481 {
4482 wordnode_T *node = root;
4483 wordnode_T *np;
4484 wordnode_T *copyp, **copyprev;
4485 wordnode_T **prev = NULL;
4486 int i;
4487
4488 /* Add each byte of the word to the tree, including the NUL at the end. */
4489 for (i = 0; ; ++i)
4490 {
4491 /* When there is more than one reference to this node we need to make
4492 * a copy, so that we can modify it. Copy the whole list of siblings
4493 * (we don't optimize for a partly shared list of siblings). */
4494 if (node != NULL && node->wn_refs > 1)
4495 {
4496 --node->wn_refs;
4497 copyprev = prev;
4498 for (copyp = node; copyp != NULL; copyp = copyp->wn_sibling)
4499 {
4500 /* Allocate a new node and copy the info. */
4501 np = get_wordnode(spin);
4502 if (np == NULL)
4503 return FAIL;
4504 np->wn_child = copyp->wn_child;
4505 if (np->wn_child != NULL)
4506 ++np->wn_child->wn_refs; /* child gets extra ref */
4507 np->wn_byte = copyp->wn_byte;
4508 if (np->wn_byte == NUL)
4509 {
4510 np->wn_flags = copyp->wn_flags;
4511 np->wn_region = copyp->wn_region;
4512 np->wn_affixID = copyp->wn_affixID;
4513 }
4514
4515 /* Link the new node in the list, there will be one ref. */
4516 np->wn_refs = 1;
4517 if (copyprev != NULL)
4518 *copyprev = np;
4519 copyprev = &np->wn_sibling;
4520
4521 /* Let "node" point to the head of the copied list. */
4522 if (copyp == node)
4523 node = np;
4524 }
4525 }
4526
4527 /* Look for the sibling that has the same character. They are sorted
4528 * on byte value, thus stop searching when a sibling is found with a
4529 * higher byte value. For zero bytes (end of word) the sorting is
4530 * done on flags and then on affixID. */
4531 while (node != NULL
4532 && (node->wn_byte < word[i]
4533 || (node->wn_byte == NUL
4534 && (flags < 0
4535 ? node->wn_affixID < (unsigned)affixID
4536 : (node->wn_flags < (unsigned)(flags & WN_MASK)
4537 || (node->wn_flags == (flags & WN_MASK)
4538 && (spin->si_sugtree
4539 ? (node->wn_region & 0xffff) < region
4540 : node->wn_affixID
4541 < (unsigned)affixID)))))))
4542 {
4543 prev = &node->wn_sibling;
4544 node = *prev;
4545 }
4546 if (node == NULL
4547 || node->wn_byte != word[i]
4548 || (word[i] == NUL
4549 && (flags < 0
4550 || spin->si_sugtree
4551 || node->wn_flags != (flags & WN_MASK)
4552 || node->wn_affixID != affixID)))
4553 {
4554 /* Allocate a new node. */
4555 np = get_wordnode(spin);
4556 if (np == NULL)
4557 return FAIL;
4558 np->wn_byte = word[i];
4559
4560 /* If "node" is NULL this is a new child or the end of the sibling
4561 * list: ref count is one. Otherwise use ref count of sibling and
4562 * make ref count of sibling one (matters when inserting in front
4563 * of the list of siblings). */
4564 if (node == NULL)
4565 np->wn_refs = 1;
4566 else
4567 {
4568 np->wn_refs = node->wn_refs;
4569 node->wn_refs = 1;
4570 }
4571 if (prev != NULL)
4572 *prev = np;
4573 np->wn_sibling = node;
4574 node = np;
4575 }
4576
4577 if (word[i] == NUL)
4578 {
4579 node->wn_flags = flags;
4580 node->wn_region |= region;
4581 node->wn_affixID = affixID;
4582 break;
4583 }
4584 prev = &node->wn_child;
4585 node = *prev;
4586 }
4587 #ifdef SPELL_PRINTTREE
4588 smsg((char_u *)"Added \"%s\"", word);
4589 spell_print_tree(root->wn_sibling);
4590 #endif
4591
4592 /* count nr of words added since last message */
4593 ++spin->si_msg_count;
4594
4595 if (spin->si_compress_cnt > 1)
4596 {
4597 if (--spin->si_compress_cnt == 1)
4598 /* Did enough words to lower the block count limit. */
4599 spin->si_blocks_cnt += compress_inc;
4600 }
4601
4602 /*
4603 * When we have allocated lots of memory we need to compress the word tree
4604 * to free up some room. But compression is slow, and we might actually
4605 * need that room, thus only compress in the following situations:
4606 * 1. When not compressed before (si_compress_cnt == 0): when using
4607 * "compress_start" blocks.
4608 * 2. When compressed before and used "compress_inc" blocks before
4609 * adding "compress_added" words (si_compress_cnt > 1).
4610 * 3. When compressed before, added "compress_added" words
4611 * (si_compress_cnt == 1) and the number of free nodes drops below the
4612 * maximum word length.
4613 */
4614 #ifndef SPELL_COMPRESS_ALLWAYS
4615 if (spin->si_compress_cnt == 1
4616 ? spin->si_free_count < MAXWLEN
4617 : spin->si_blocks_cnt >= compress_start)
4618 #endif
4619 {
4620 /* Decrement the block counter. The effect is that we compress again
4621 * when the freed up room has been used and another "compress_inc"
4622 * blocks have been allocated. Unless "compress_added" words have
4623 * been added, then the limit is put back again. */
4624 spin->si_blocks_cnt -= compress_inc;
4625 spin->si_compress_cnt = compress_added;
4626
4627 if (spin->si_verbose)
4628 {
4629 msg_start();
4630 msg_puts((char_u *)_(msg_compressing));
4631 msg_clr_eos();
4632 msg_didout = FALSE;
4633 msg_col = 0;
4634 out_flush();
4635 }
4636
4637 /* Compress both trees. Either they both have many nodes, which makes
4638 * compression useful, or one of them is small, which means
4639 * compression goes fast. But when filling the soundfold word tree
4640 * there is no keep-case tree. */
4641 wordtree_compress(spin, spin->si_foldroot);
4642 if (affixID >= 0)
4643 wordtree_compress(spin, spin->si_keeproot);
4644 }
4645
4646 return OK;
4647 }
4648
4649 /*
4650 * Get a wordnode_T, either from the list of previously freed nodes or
4651 * allocate a new one.
4652 * Returns NULL when out of memory.
4653 */
4654 static wordnode_T *
4655 get_wordnode(spellinfo_T *spin)
4656 {
4657 wordnode_T *n;
4658
4659 if (spin->si_first_free == NULL)
4660 n = (wordnode_T *)getroom(spin, sizeof(wordnode_T), TRUE);
4661 else
4662 {
4663 n = spin->si_first_free;
4664 spin->si_first_free = n->wn_child;
4665 vim_memset(n, 0, sizeof(wordnode_T));
4666 --spin->si_free_count;
4667 }
4668 #ifdef SPELL_PRINTTREE
4669 if (n != NULL)
4670 n->wn_nr = ++spin->si_wordnode_nr;
4671 #endif
4672 return n;
4673 }
4674
4675 /*
4676 * Decrement the reference count on a node (which is the head of a list of
4677 * siblings). If the reference count becomes zero free the node and its
4678 * siblings.
4679 * Returns the number of nodes actually freed.
4680 */
4681 static int
4682 deref_wordnode(spellinfo_T *spin, wordnode_T *node)
4683 {
4684 wordnode_T *np;
4685 int cnt = 0;
4686
4687 if (--node->wn_refs == 0)
4688 {
4689 for (np = node; np != NULL; np = np->wn_sibling)
4690 {
4691 if (np->wn_child != NULL)
4692 cnt += deref_wordnode(spin, np->wn_child);
4693 free_wordnode(spin, np);
4694 ++cnt;
4695 }
4696 ++cnt; /* length field */
4697 }
4698 return cnt;
4699 }
4700
4701 /*
4702 * Free a wordnode_T for re-use later.
4703 * Only the "wn_child" field becomes invalid.
4704 */
4705 static void
4706 free_wordnode(spellinfo_T *spin, wordnode_T *n)
4707 {
4708 n->wn_child = spin->si_first_free;
4709 spin->si_first_free = n;
4710 ++spin->si_free_count;
4711 }
4712
4713 /*
4714 * Compress a tree: find tails that are identical and can be shared.
4715 */
4716 static void
4717 wordtree_compress(spellinfo_T *spin, wordnode_T *root)
4718 {
4719 hashtab_T ht;
4720 int n;
4721 int tot = 0;
4722 int perc;
4723
4724 /* Skip the root itself, it's not actually used. The first sibling is the
4725 * start of the tree. */
4726 if (root->wn_sibling != NULL)
4727 {
4728 hash_init(&ht);
4729 n = node_compress(spin, root->wn_sibling, &ht, &tot);
4730
4731 #ifndef SPELL_PRINTTREE
4732 if (spin->si_verbose || p_verbose > 2)
4733 #endif
4734 {
4735 if (tot > 1000000)
4736 perc = (tot - n) / (tot / 100);
4737 else if (tot == 0)
4738 perc = 0;
4739 else
4740 perc = (tot - n) * 100 / tot;
4741 vim_snprintf((char *)IObuff, IOSIZE,
4742 _("Compressed %d of %d nodes; %d (%d%%) remaining"),
4743 n, tot, tot - n, perc);
4744 spell_message(spin, IObuff);
4745 }
4746 #ifdef SPELL_PRINTTREE
4747 spell_print_tree(root->wn_sibling);
4748 #endif
4749 hash_clear(&ht);
4750 }
4751 }
4752
4753 /*
4754 * Compress a node, its siblings and its children, depth first.
4755 * Returns the number of compressed nodes.
4756 */
4757 static int
4758 node_compress(
4759 spellinfo_T *spin,
4760 wordnode_T *node,
4761 hashtab_T *ht,
4762 int *tot) /* total count of nodes before compressing,
4763 incremented while going through the tree */
4764 {
4765 wordnode_T *np;
4766 wordnode_T *tp;
4767 wordnode_T *child;
4768 hash_T hash;
4769 hashitem_T *hi;
4770 int len = 0;
4771 unsigned nr, n;
4772 int compressed = 0;
4773
4774 /*
4775 * Go through the list of siblings. Compress each child and then try
4776 * finding an identical child to replace it.
4777 * Note that with "child" we mean not just the node that is pointed to,
4778 * but the whole list of siblings of which the child node is the first.
4779 */
4780 for (np = node; np != NULL && !got_int; np = np->wn_sibling)
4781 {
4782 ++len;
4783 if ((child = np->wn_child) != NULL)
4784 {
4785 /* Compress the child first. This fills hashkey. */
4786 compressed += node_compress(spin, child, ht, tot);
4787
4788 /* Try to find an identical child. */
4789 hash = hash_hash(child->wn_u1.hashkey);
4790 hi = hash_lookup(ht, child->wn_u1.hashkey, hash);
4791 if (!HASHITEM_EMPTY(hi))
4792 {
4793 /* There are children we encountered before with a hash value
4794 * identical to the current child. Now check if there is one
4795 * that is really identical. */
4796 for (tp = HI2WN(hi); tp != NULL; tp = tp->wn_u2.next)
4797 if (node_equal(child, tp))
4798 {
4799 /* Found one! Now use that child in place of the
4800 * current one. This means the current child and all
4801 * its siblings is unlinked from the tree. */
4802 ++tp->wn_refs;
4803 compressed += deref_wordnode(spin, child);
4804 np->wn_child = tp;
4805 break;
4806 }
4807 if (tp == NULL)
4808 {
4809 /* No other child with this hash value equals the child of
4810 * the node, add it to the linked list after the first
4811 * item. */
4812 tp = HI2WN(hi);
4813 child->wn_u2.next = tp->wn_u2.next;
4814 tp->wn_u2.next = child;
4815 }
4816 }
4817 else
4818 /* No other child has this hash value, add it to the
4819 * hashtable. */
4820 hash_add_item(ht, hi, child->wn_u1.hashkey, hash);
4821 }
4822 }
4823 *tot += len + 1; /* add one for the node that stores the length */
4824
4825 /*
4826 * Make a hash key for the node and its siblings, so that we can quickly
4827 * find a lookalike node. This must be done after compressing the sibling
4828 * list, otherwise the hash key would become invalid by the compression.
4829 */
4830 node->wn_u1.hashkey[0] = len;
4831 nr = 0;
4832 for (np = node; np != NULL; np = np->wn_sibling)
4833 {
4834 if (np->wn_byte == NUL)
4835 /* end node: use wn_flags, wn_region and wn_affixID */
4836 n = np->wn_flags + (np->wn_region << 8) + (np->wn_affixID << 16);
4837 else
4838 /* byte node: use the byte value and the child pointer */
4839 n = (unsigned)(np->wn_byte + ((long_u)np->wn_child << 8));
4840 nr = nr * 101 + n;
4841 }
4842
4843 /* Avoid NUL bytes, it terminates the hash key. */
4844 n = nr & 0xff;
4845 node->wn_u1.hashkey[1] = n == 0 ? 1 : n;
4846 n = (nr >> 8) & 0xff;
4847 node->wn_u1.hashkey[2] = n == 0 ? 1 : n;
4848 n = (nr >> 16) & 0xff;
4849 node->wn_u1.hashkey[3] = n == 0 ? 1 : n;
4850 n = (nr >> 24) & 0xff;
4851 node->wn_u1.hashkey[4] = n == 0 ? 1 : n;
4852 node->wn_u1.hashkey[5] = NUL;
4853
4854 /* Check for CTRL-C pressed now and then. */
4855 fast_breakcheck();
4856
4857 return compressed;
4858 }
4859
4860 /*
4861 * Return TRUE when two nodes have identical siblings and children.
4862 */
4863 static int
4864 node_equal(wordnode_T *n1, wordnode_T *n2)
4865 {
4866 wordnode_T *p1;
4867 wordnode_T *p2;
4868
4869 for (p1 = n1, p2 = n2; p1 != NULL && p2 != NULL;
4870 p1 = p1->wn_sibling, p2 = p2->wn_sibling)
4871 if (p1->wn_byte != p2->wn_byte
4872 || (p1->wn_byte == NUL
4873 ? (p1->wn_flags != p2->wn_flags
4874 || p1->wn_region != p2->wn_region
4875 || p1->wn_affixID != p2->wn_affixID)
4876 : (p1->wn_child != p2->wn_child)))
4877 break;
4878
4879 return p1 == NULL && p2 == NULL;
4880 }
4881
4882 static int
4883 #ifdef __BORLANDC__
4884 _RTLENTRYF
4885 #endif
4886 rep_compare(const void *s1, const void *s2);
4887
4888 /*
4889 * Function given to qsort() to sort the REP items on "from" string.
4890 */
4891 static int
4892 #ifdef __BORLANDC__
4893 _RTLENTRYF
4894 #endif
4895 rep_compare(const void *s1, const void *s2)
4896 {
4897 fromto_T *p1 = (fromto_T *)s1;
4898 fromto_T *p2 = (fromto_T *)s2;
4899
4900 return STRCMP(p1->ft_from, p2->ft_from);
4901 }
4902
4903 /*
4904 * Write the Vim .spl file "fname".
4905 * Return FAIL or OK;
4906 */
4907 static int
4908 write_vim_spell(spellinfo_T *spin, char_u *fname)
4909 {
4910 FILE *fd;
4911 int regionmask;
4912 int round;
4913 wordnode_T *tree;
4914 int nodecount;
4915 int i;
4916 int l;
4917 garray_T *gap;
4918 fromto_T *ftp;
4919 char_u *p;
4920 int rr;
4921 int retval = OK;
4922 size_t fwv = 1; /* collect return value of fwrite() to avoid
4923 warnings from picky compiler */
4924
4925 fd = mch_fopen((char *)fname, "w");
4926 if (fd == NULL)
4927 {
4928 EMSG2(_(e_notopen), fname);
4929 return FAIL;
4930 }
4931
4932 /* <HEADER>: <fileID> <versionnr> */
4933 /* <fileID> */
4934 fwv &= fwrite(VIMSPELLMAGIC, VIMSPELLMAGICL, (size_t)1, fd);
4935 if (fwv != (size_t)1)
4936 /* Catch first write error, don't try writing more. */
4937 goto theend;
4938
4939 putc(VIMSPELLVERSION, fd); /* <versionnr> */
4940
4941 /*
4942 * <SECTIONS>: <section> ... <sectionend>
4943 */
4944
4945 /* SN_INFO: <infotext> */
4946 if (spin->si_info != NULL)
4947 {
4948 putc(SN_INFO, fd); /* <sectionID> */
4949 putc(0, fd); /* <sectionflags> */
4950
4951 i = (int)STRLEN(spin->si_info);
4952 put_bytes(fd, (long_u)i, 4); /* <sectionlen> */
4953 fwv &= fwrite(spin->si_info, (size_t)i, (size_t)1, fd); /* <infotext> */
4954 }
4955
4956 /* SN_REGION: <regionname> ...
4957 * Write the region names only if there is more than one. */
4958 if (spin->si_region_count > 1)
4959 {
4960 putc(SN_REGION, fd); /* <sectionID> */
4961 putc(SNF_REQUIRED, fd); /* <sectionflags> */
4962 l = spin->si_region_count * 2;
4963 put_bytes(fd, (long_u)l, 4); /* <sectionlen> */
4964 fwv &= fwrite(spin->si_region_name, (size_t)l, (size_t)1, fd);
4965 /* <regionname> ... */
4966 regionmask = (1 << spin->si_region_count) - 1;
4967 }
4968 else
4969 regionmask = 0;
4970
4971 /* SN_CHARFLAGS: <charflagslen> <charflags> <folcharslen> <folchars>
4972 *
4973 * The table with character flags and the table for case folding.
4974 * This makes sure the same characters are recognized as word characters
4975 * when generating an when using a spell file.
4976 * Skip this for ASCII, the table may conflict with the one used for
4977 * 'encoding'.
4978 * Also skip this for an .add.spl file, the main spell file must contain
4979 * the table (avoids that it conflicts). File is shorter too.
4980 */
4981 if (!spin->si_ascii && !spin->si_add)
4982 {
4983 char_u folchars[128 * 8];
4984 int flags;
4985
4986 putc(SN_CHARFLAGS, fd); /* <sectionID> */
4987 putc(SNF_REQUIRED, fd); /* <sectionflags> */
4988
4989 /* Form the <folchars> string first, we need to know its length. */
4990 l = 0;
4991 for (i = 128; i < 256; ++i)
4992 {
4993 #ifdef FEAT_MBYTE
4994 if (has_mbyte)
4995 l += mb_char2bytes(spelltab.st_fold[i], folchars + l);
4996 else
4997 #endif
4998 folchars[l++] = spelltab.st_fold[i];
4999 }
5000 put_bytes(fd, (long_u)(1 + 128 + 2 + l), 4); /* <sectionlen> */
5001
5002 fputc(128, fd); /* <charflagslen> */
5003 for (i = 128; i < 256; ++i)
5004 {
5005 flags = 0;
5006 if (spelltab.st_isw[i])
5007 flags |= CF_WORD;
5008 if (spelltab.st_isu[i])
5009 flags |= CF_UPPER;
5010 fputc(flags, fd); /* <charflags> */
5011 }
5012
5013 put_bytes(fd, (long_u)l, 2); /* <folcharslen> */
5014 fwv &= fwrite(folchars, (size_t)l, (size_t)1, fd); /* <folchars> */
5015 }
5016
5017 /* SN_MIDWORD: <midword> */
5018 if (spin->si_midword != NULL)
5019 {
5020 putc(SN_MIDWORD, fd); /* <sectionID> */
5021 putc(SNF_REQUIRED, fd); /* <sectionflags> */
5022
5023 i = (int)STRLEN(spin->si_midword);
5024 put_bytes(fd, (long_u)i, 4); /* <sectionlen> */
5025 fwv &= fwrite(spin->si_midword, (size_t)i, (size_t)1, fd);
5026 /* <midword> */
5027 }
5028
5029 /* SN_PREFCOND: <prefcondcnt> <prefcond> ... */
5030 if (spin->si_prefcond.ga_len > 0)
5031 {
5032 putc(SN_PREFCOND, fd); /* <sectionID> */
5033 putc(SNF_REQUIRED, fd); /* <sectionflags> */
5034
5035 l = write_spell_prefcond(NULL, &spin->si_prefcond);
5036 put_bytes(fd, (long_u)l, 4); /* <sectionlen> */
5037
5038 write_spell_prefcond(fd, &spin->si_prefcond);
5039 }
5040
5041 /* SN_REP: <repcount> <rep> ...
5042 * SN_SAL: <salflags> <salcount> <sal> ...
5043 * SN_REPSAL: <repcount> <rep> ... */
5044
5045 /* round 1: SN_REP section
5046 * round 2: SN_SAL section (unless SN_SOFO is used)
5047 * round 3: SN_REPSAL section */
5048 for (round = 1; round <= 3; ++round)
5049 {
5050 if (round == 1)
5051 gap = &spin->si_rep;
5052 else if (round == 2)
5053 {
5054 /* Don't write SN_SAL when using a SN_SOFO section */
5055 if (spin->si_sofofr != NULL && spin->si_sofoto != NULL)
5056 continue;
5057 gap = &spin->si_sal;
5058 }
5059 else
5060 gap = &spin->si_repsal;
5061
5062 /* Don't write the section if there are no items. */
5063 if (gap->ga_len == 0)
5064 continue;
5065
5066 /* Sort the REP/REPSAL items. */
5067 if (round != 2)
5068 qsort(gap->ga_data, (size_t)gap->ga_len,
5069 sizeof(fromto_T), rep_compare);
5070
5071 i = round == 1 ? SN_REP : (round == 2 ? SN_SAL : SN_REPSAL);
5072 putc(i, fd); /* <sectionID> */
5073
5074 /* This is for making suggestions, section is not required. */
5075 putc(0, fd); /* <sectionflags> */
5076
5077 /* Compute the length of what follows. */
5078 l = 2; /* count <repcount> or <salcount> */
5079 for (i = 0; i < gap->ga_len; ++i)
5080 {
5081 ftp = &((fromto_T *)gap->ga_data)[i];
5082 l += 1 + (int)STRLEN(ftp->ft_from); /* count <*fromlen> and <*from> */
5083 l += 1 + (int)STRLEN(ftp->ft_to); /* count <*tolen> and <*to> */
5084 }
5085 if (round == 2)
5086 ++l; /* count <salflags> */
5087 put_bytes(fd, (long_u)l, 4); /* <sectionlen> */
5088
5089 if (round == 2)
5090 {
5091 i = 0;
5092 if (spin->si_followup)
5093 i |= SAL_F0LLOWUP;
5094 if (spin->si_collapse)
5095 i |= SAL_COLLAPSE;
5096 if (spin->si_rem_accents)
5097 i |= SAL_REM_ACCENTS;
5098 putc(i, fd); /* <salflags> */
5099 }
5100
5101 put_bytes(fd, (long_u)gap->ga_len, 2); /* <repcount> or <salcount> */
5102 for (i = 0; i < gap->ga_len; ++i)
5103 {
5104 /* <rep> : <repfromlen> <repfrom> <reptolen> <repto> */
5105 /* <sal> : <salfromlen> <salfrom> <saltolen> <salto> */
5106 ftp = &((fromto_T *)gap->ga_data)[i];
5107 for (rr = 1; rr <= 2; ++rr)
5108 {
5109 p = rr == 1 ? ftp->ft_from : ftp->ft_to;
5110 l = (int)STRLEN(p);
5111 putc(l, fd);
5112 if (l > 0)
5113 fwv &= fwrite(p, l, (size_t)1, fd);
5114 }
5115 }
5116
5117 }
5118
5119 /* SN_SOFO: <sofofromlen> <sofofrom> <sofotolen> <sofoto>
5120 * This is for making suggestions, section is not required. */
5121 if (spin->si_sofofr != NULL && spin->si_sofoto != NULL)
5122 {
5123 putc(SN_SOFO, fd); /* <sectionID> */
5124 putc(0, fd); /* <sectionflags> */
5125
5126 l = (int)STRLEN(spin->si_sofofr);
5127 put_bytes(fd, (long_u)(l + STRLEN(spin->si_sofoto) + 4), 4);
5128 /* <sectionlen> */
5129
5130 put_bytes(fd, (long_u)l, 2); /* <sofofromlen> */
5131 fwv &= fwrite(spin->si_sofofr, l, (size_t)1, fd); /* <sofofrom> */
5132
5133 l = (int)STRLEN(spin->si_sofoto);
5134 put_bytes(fd, (long_u)l, 2); /* <sofotolen> */
5135 fwv &= fwrite(spin->si_sofoto, l, (size_t)1, fd); /* <sofoto> */
5136 }
5137
5138 /* SN_WORDS: <word> ...
5139 * This is for making suggestions, section is not required. */
5140 if (spin->si_commonwords.ht_used > 0)
5141 {
5142 putc(SN_WORDS, fd); /* <sectionID> */
5143 putc(0, fd); /* <sectionflags> */
5144
5145 /* round 1: count the bytes
5146 * round 2: write the bytes */
5147 for (round = 1; round <= 2; ++round)
5148 {
5149 int todo;
5150 int len = 0;
5151 hashitem_T *hi;
5152
5153 todo = (int)spin->si_commonwords.ht_used;
5154 for (hi = spin->si_commonwords.ht_array; todo > 0; ++hi)
5155 if (!HASHITEM_EMPTY(hi))
5156 {
5157 l = (int)STRLEN(hi->hi_key) + 1;
5158 len += l;
5159 if (round == 2) /* <word> */
5160 fwv &= fwrite(hi->hi_key, (size_t)l, (size_t)1, fd);
5161 --todo;
5162 }
5163 if (round == 1)
5164 put_bytes(fd, (long_u)len, 4); /* <sectionlen> */
5165 }
5166 }
5167
5168 /* SN_MAP: <mapstr>
5169 * This is for making suggestions, section is not required. */
5170 if (spin->si_map.ga_len > 0)
5171 {
5172 putc(SN_MAP, fd); /* <sectionID> */
5173 putc(0, fd); /* <sectionflags> */
5174 l = spin->si_map.ga_len;
5175 put_bytes(fd, (long_u)l, 4); /* <sectionlen> */
5176 fwv &= fwrite(spin->si_map.ga_data, (size_t)l, (size_t)1, fd);
5177 /* <mapstr> */
5178 }
5179
5180 /* SN_SUGFILE: <timestamp>
5181 * This is used to notify that a .sug file may be available and at the
5182 * same time allows for checking that a .sug file that is found matches
5183 * with this .spl file. That's because the word numbers must be exactly
5184 * right. */
5185 if (!spin->si_nosugfile
5186 && (spin->si_sal.ga_len > 0
5187 || (spin->si_sofofr != NULL && spin->si_sofoto != NULL)))
5188 {
5189 putc(SN_SUGFILE, fd); /* <sectionID> */
5190 putc(0, fd); /* <sectionflags> */
5191 put_bytes(fd, (long_u)8, 4); /* <sectionlen> */
5192
5193 /* Set si_sugtime and write it to the file. */
5194 spin->si_sugtime = time(NULL);
5195 put_time(fd, spin->si_sugtime); /* <timestamp> */
5196 }
5197
5198 /* SN_NOSPLITSUGS: nothing
5199 * This is used to notify that no suggestions with word splits are to be
5200 * made. */
5201 if (spin->si_nosplitsugs)
5202 {
5203 putc(SN_NOSPLITSUGS, fd); /* <sectionID> */
5204 putc(0, fd); /* <sectionflags> */
5205 put_bytes(fd, (long_u)0, 4); /* <sectionlen> */
5206 }
5207
5208 /* SN_NOCOMPUNDSUGS: nothing
5209 * This is used to notify that no suggestions with compounds are to be
5210 * made. */
5211 if (spin->si_nocompoundsugs)
5212 {
5213 putc(SN_NOCOMPOUNDSUGS, fd); /* <sectionID> */
5214 putc(0, fd); /* <sectionflags> */
5215 put_bytes(fd, (long_u)0, 4); /* <sectionlen> */
5216 }
5217
5218 /* SN_COMPOUND: compound info.
5219 * We don't mark it required, when not supported all compound words will
5220 * be bad words. */
5221 if (spin->si_compflags != NULL)
5222 {
5223 putc(SN_COMPOUND, fd); /* <sectionID> */
5224 putc(0, fd); /* <sectionflags> */
5225
5226 l = (int)STRLEN(spin->si_compflags);
5227 for (i = 0; i < spin->si_comppat.ga_len; ++i)
5228 l += (int)STRLEN(((char_u **)(spin->si_comppat.ga_data))[i]) + 1;
5229 put_bytes(fd, (long_u)(l + 7), 4); /* <sectionlen> */
5230
5231 putc(spin->si_compmax, fd); /* <compmax> */
5232 putc(spin->si_compminlen, fd); /* <compminlen> */
5233 putc(spin->si_compsylmax, fd); /* <compsylmax> */
5234 putc(0, fd); /* for Vim 7.0b compatibility */
5235 putc(spin->si_compoptions, fd); /* <compoptions> */
5236 put_bytes(fd, (long_u)spin->si_comppat.ga_len, 2);
5237 /* <comppatcount> */
5238 for (i = 0; i < spin->si_comppat.ga_len; ++i)
5239 {
5240 p = ((char_u **)(spin->si_comppat.ga_data))[i];
5241 putc((int)STRLEN(p), fd); /* <comppatlen> */
5242 fwv &= fwrite(p, (size_t)STRLEN(p), (size_t)1, fd);
5243 /* <comppattext> */
5244 }
5245 /* <compflags> */
5246 fwv &= fwrite(spin->si_compflags, (size_t)STRLEN(spin->si_compflags),
5247 (size_t)1, fd);
5248 }
5249
5250 /* SN_NOBREAK: NOBREAK flag */
5251 if (spin->si_nobreak)
5252 {
5253 putc(SN_NOBREAK, fd); /* <sectionID> */
5254 putc(0, fd); /* <sectionflags> */
5255
5256 /* It's empty, the presence of the section flags the feature. */
5257 put_bytes(fd, (long_u)0, 4); /* <sectionlen> */
5258 }
5259
5260 /* SN_SYLLABLE: syllable info.
5261 * We don't mark it required, when not supported syllables will not be
5262 * counted. */
5263 if (spin->si_syllable != NULL)
5264 {
5265 putc(SN_SYLLABLE, fd); /* <sectionID> */
5266 putc(0, fd); /* <sectionflags> */
5267
5268 l = (int)STRLEN(spin->si_syllable);
5269 put_bytes(fd, (long_u)l, 4); /* <sectionlen> */
5270 fwv &= fwrite(spin->si_syllable, (size_t)l, (size_t)1, fd);
5271 /* <syllable> */
5272 }
5273
5274 /* end of <SECTIONS> */
5275 putc(SN_END, fd); /* <sectionend> */
5276
5277
5278 /*
5279 * <LWORDTREE> <KWORDTREE> <PREFIXTREE>
5280 */
5281 spin->si_memtot = 0;
5282 for (round = 1; round <= 3; ++round)
5283 {
5284 if (round == 1)
5285 tree = spin->si_foldroot->wn_sibling;
5286 else if (round == 2)
5287 tree = spin->si_keeproot->wn_sibling;
5288 else
5289 tree = spin->si_prefroot->wn_sibling;
5290
5291 /* Clear the index and wnode fields in the tree. */
5292 clear_node(tree);
5293
5294 /* Count the number of nodes. Needed to be able to allocate the
5295 * memory when reading the nodes. Also fills in index for shared
5296 * nodes. */
5297 nodecount = put_node(NULL, tree, 0, regionmask, round == 3);
5298
5299 /* number of nodes in 4 bytes */
5300 put_bytes(fd, (long_u)nodecount, 4); /* <nodecount> */
5301 spin->si_memtot += nodecount + nodecount * sizeof(int);
5302
5303 /* Write the nodes. */
5304 (void)put_node(fd, tree, 0, regionmask, round == 3);
5305 }
5306
5307 /* Write another byte to check for errors (file system full). */
5308 if (putc(0, fd) == EOF)
5309 retval = FAIL;
5310 theend:
5311 if (fclose(fd) == EOF)
5312 retval = FAIL;
5313
5314 if (fwv != (size_t)1)
5315 retval = FAIL;
5316 if (retval == FAIL)
5317 EMSG(_(e_write));
5318
5319 return retval;
5320 }
5321
5322 /*
5323 * Clear the index and wnode fields of "node", it siblings and its
5324 * children. This is needed because they are a union with other items to save
5325 * space.
5326 */
5327 static void
5328 clear_node(wordnode_T *node)
5329 {
5330 wordnode_T *np;
5331
5332 if (node != NULL)
5333 for (np = node; np != NULL; np = np->wn_sibling)
5334 {
5335 np->wn_u1.index = 0;
5336 np->wn_u2.wnode = NULL;
5337
5338 if (np->wn_byte != NUL)
5339 clear_node(np->wn_child);
5340 }
5341 }
5342
5343
5344 /*
5345 * Dump a word tree at node "node".
5346 *
5347 * This first writes the list of possible bytes (siblings). Then for each
5348 * byte recursively write the children.
5349 *
5350 * NOTE: The code here must match the code in read_tree_node(), since
5351 * assumptions are made about the indexes (so that we don't have to write them
5352 * in the file).
5353 *
5354 * Returns the number of nodes used.
5355 */
5356 static int
5357 put_node(
5358 FILE *fd, /* NULL when only counting */
5359 wordnode_T *node,
5360 int idx,
5361 int regionmask,
5362 int prefixtree) /* TRUE for PREFIXTREE */
5363 {
5364 int newindex = idx;
5365 int siblingcount = 0;
5366 wordnode_T *np;
5367 int flags;
5368
5369 /* If "node" is zero the tree is empty. */
5370 if (node == NULL)
5371 return 0;
5372
5373 /* Store the index where this node is written. */
5374 node->wn_u1.index = idx;
5375
5376 /* Count the number of siblings. */
5377 for (np = node; np != NULL; np = np->wn_sibling)
5378 ++siblingcount;
5379
5380 /* Write the sibling count. */
5381 if (fd != NULL)
5382 putc(siblingcount, fd); /* <siblingcount> */
5383
5384 /* Write each sibling byte and optionally extra info. */
5385 for (np = node; np != NULL; np = np->wn_sibling)
5386 {
5387 if (np->wn_byte == 0)
5388 {
5389 if (fd != NULL)
5390 {
5391 /* For a NUL byte (end of word) write the flags etc. */
5392 if (prefixtree)
5393 {
5394 /* In PREFIXTREE write the required affixID and the
5395 * associated condition nr (stored in wn_region). The
5396 * byte value is misused to store the "rare" and "not
5397 * combining" flags */
5398 if (np->wn_flags == (short_u)PFX_FLAGS)
5399 putc(BY_NOFLAGS, fd); /* <byte> */
5400 else
5401 {
5402 putc(BY_FLAGS, fd); /* <byte> */
5403 putc(np->wn_flags, fd); /* <pflags> */
5404 }
5405 putc(np->wn_affixID, fd); /* <affixID> */
5406 put_bytes(fd, (long_u)np->wn_region, 2); /* <prefcondnr> */
5407 }
5408 else
5409 {
5410 /* For word trees we write the flag/region items. */
5411 flags = np->wn_flags;
5412 if (regionmask != 0 && np->wn_region != regionmask)
5413 flags |= WF_REGION;
5414 if (np->wn_affixID != 0)
5415 flags |= WF_AFX;
5416 if (flags == 0)
5417 {
5418 /* word without flags or region */
5419 putc(BY_NOFLAGS, fd); /* <byte> */
5420 }
5421 else
5422 {
5423 if (np->wn_flags >= 0x100)
5424 {
5425 putc(BY_FLAGS2, fd); /* <byte> */
5426 putc(flags, fd); /* <flags> */
5427 putc((unsigned)flags >> 8, fd); /* <flags2> */
5428 }
5429 else
5430 {
5431 putc(BY_FLAGS, fd); /* <byte> */
5432 putc(flags, fd); /* <flags> */
5433 }
5434 if (flags & WF_REGION)
5435 putc(np->wn_region, fd); /* <region> */
5436 if (flags & WF_AFX)
5437 putc(np->wn_affixID, fd); /* <affixID> */
5438 }
5439 }
5440 }
5441 }
5442 else
5443 {
5444 if (np->wn_child->wn_u1.index != 0
5445 && np->wn_child->wn_u2.wnode != node)
5446 {
5447 /* The child is written elsewhere, write the reference. */
5448 if (fd != NULL)
5449 {
5450 putc(BY_INDEX, fd); /* <byte> */
5451 /* <nodeidx> */
5452 put_bytes(fd, (long_u)np->wn_child->wn_u1.index, 3);
5453 }
5454 }
5455 else if (np->wn_child->wn_u2.wnode == NULL)
5456 /* We will write the child below and give it an index. */
5457 np->wn_child->wn_u2.wnode = node;
5458
5459 if (fd != NULL)
5460 if (putc(np->wn_byte, fd) == EOF) /* <byte> or <xbyte> */
5461 {
5462 EMSG(_(e_write));
5463 return 0;
5464 }
5465 }
5466 }
5467
5468 /* Space used in the array when reading: one for each sibling and one for
5469 * the count. */
5470 newindex += siblingcount + 1;
5471
5472 /* Recursively dump the children of each sibling. */
5473 for (np = node; np != NULL; np = np->wn_sibling)
5474 if (np->wn_byte != 0 && np->wn_child->wn_u2.wnode == node)
5475 newindex = put_node(fd, np->wn_child, newindex, regionmask,
5476 prefixtree);
5477
5478 return newindex;
5479 }
5480
5481
5482 /*
5483 * ":mkspell [-ascii] outfile infile ..."
5484 * ":mkspell [-ascii] addfile"
5485 */
5486 void
5487 ex_mkspell(exarg_T *eap)
5488 {
5489 int fcount;
5490 char_u **fnames;
5491 char_u *arg = eap->arg;
5492 int ascii = FALSE;
5493
5494 if (STRNCMP(arg, "-ascii", 6) == 0)
5495 {
5496 ascii = TRUE;
5497 arg = skipwhite(arg + 6);
5498 }
5499
5500 /* Expand all the remaining arguments (e.g., $VIMRUNTIME). */
5501 if (get_arglist_exp(arg, &fcount, &fnames, FALSE) == OK)
5502 {
5503 mkspell(fcount, fnames, ascii, eap->forceit, FALSE);
5504 FreeWild(fcount, fnames);
5505 }
5506 }
5507
5508 /*
5509 * Create the .sug file.
5510 * Uses the soundfold info in "spin".
5511 * Writes the file with the name "wfname", with ".spl" changed to ".sug".
5512 */
5513 static void
5514 spell_make_sugfile(spellinfo_T *spin, char_u *wfname)
5515 {
5516 char_u *fname = NULL;
5517 int len;
5518 slang_T *slang;
5519 int free_slang = FALSE;
5520
5521 /*
5522 * Read back the .spl file that was written. This fills the required
5523 * info for soundfolding. This also uses less memory than the
5524 * pointer-linked version of the trie. And it avoids having two versions
5525 * of the code for the soundfolding stuff.
5526 * It might have been done already by spell_reload_one().
5527 */
5528 for (slang = first_lang; slang != NULL; slang = slang->sl_next)
5529 if (fullpathcmp(wfname, slang->sl_fname, FALSE) == FPC_SAME)
5530 break;
5531 if (slang == NULL)
5532 {
5533 spell_message(spin, (char_u *)_("Reading back spell file..."));
5534 slang = spell_load_file(wfname, NULL, NULL, FALSE);
5535 if (slang == NULL)
5536 return;
5537 free_slang = TRUE;
5538 }
5539
5540 /*
5541 * Clear the info in "spin" that is used.
5542 */
5543 spin->si_blocks = NULL;
5544 spin->si_blocks_cnt = 0;
5545 spin->si_compress_cnt = 0; /* will stay at 0 all the time*/
5546 spin->si_free_count = 0;
5547 spin->si_first_free = NULL;
5548 spin->si_foldwcount = 0;
5549
5550 /*
5551 * Go through the trie of good words, soundfold each word and add it to
5552 * the soundfold trie.
5553 */
5554 spell_message(spin, (char_u *)_("Performing soundfolding..."));
5555 if (sug_filltree(spin, slang) == FAIL)
5556 goto theend;
5557
5558 /*
5559 * Create the table which links each soundfold word with a list of the
5560 * good words it may come from. Creates buffer "spin->si_spellbuf".
5561 * This also removes the wordnr from the NUL byte entries to make
5562 * compression possible.
5563 */
5564 if (sug_maketable(spin) == FAIL)
5565 goto theend;
5566
5567 smsg((char_u *)_("Number of words after soundfolding: %ld"),
5568 (long)spin->si_spellbuf->b_ml.ml_line_count);
5569
5570 /*
5571 * Compress the soundfold trie.
5572 */
5573 spell_message(spin, (char_u *)_(msg_compressing));
5574 wordtree_compress(spin, spin->si_foldroot);
5575
5576 /*
5577 * Write the .sug file.
5578 * Make the file name by changing ".spl" to ".sug".
5579 */
5580 fname = alloc(MAXPATHL);
5581 if (fname == NULL)
5582 goto theend;
5583 vim_strncpy(fname, wfname, MAXPATHL - 1);
5584 len = (int)STRLEN(fname);
5585 fname[len - 2] = 'u';
5586 fname[len - 1] = 'g';
5587 sug_write(spin, fname);
5588
5589 theend:
5590 vim_free(fname);
5591 if (free_slang)
5592 slang_free(slang);
5593 free_blocks(spin->si_blocks);
5594 close_spellbuf(spin->si_spellbuf);
5595 }
5596
5597 /*
5598 * Build the soundfold trie for language "slang".
5599 */
5600 static int
5601 sug_filltree(spellinfo_T *spin, slang_T *slang)
5602 {
5603 char_u *byts;
5604 idx_T *idxs;
5605 int depth;
5606 idx_T arridx[MAXWLEN];
5607 int curi[MAXWLEN];
5608 char_u tword[MAXWLEN];
5609 char_u tsalword[MAXWLEN];
5610 int c;
5611 idx_T n;
5612 unsigned words_done = 0;
5613 int wordcount[MAXWLEN];
5614
5615 /* We use si_foldroot for the soundfolded trie. */
5616 spin->si_foldroot = wordtree_alloc(spin);
5617 if (spin->si_foldroot == NULL)
5618 return FAIL;
5619
5620 /* let tree_add_word() know we're adding to the soundfolded tree */
5621 spin->si_sugtree = TRUE;
5622
5623 /*
5624 * Go through the whole case-folded tree, soundfold each word and put it
5625 * in the trie.
5626 */
5627 byts = slang->sl_fbyts;
5628 idxs = slang->sl_fidxs;
5629
5630 arridx[0] = 0;
5631 curi[0] = 1;
5632 wordcount[0] = 0;
5633
5634 depth = 0;
5635 while (depth >= 0 && !got_int)
5636 {
5637 if (curi[depth] > byts[arridx[depth]])
5638 {
5639 /* Done all bytes at this node, go up one level. */
5640 idxs[arridx[depth]] = wordcount[depth];
5641 if (depth > 0)
5642 wordcount[depth - 1] += wordcount[depth];
5643
5644 --depth;
5645 line_breakcheck();
5646 }
5647 else
5648 {
5649
5650 /* Do one more byte at this node. */
5651 n = arridx[depth] + curi[depth];
5652 ++curi[depth];
5653
5654 c = byts[n];
5655 if (c == 0)
5656 {
5657 /* Sound-fold the word. */
5658 tword[depth] = NUL;
5659 spell_soundfold(slang, tword, TRUE, tsalword);
5660
5661 /* We use the "flags" field for the MSB of the wordnr,
5662 * "region" for the LSB of the wordnr. */
5663 if (tree_add_word(spin, tsalword, spin->si_foldroot,
5664 words_done >> 16, words_done & 0xffff,
5665 0) == FAIL)
5666 return FAIL;
5667
5668 ++words_done;
5669 ++wordcount[depth];
5670
5671 /* Reset the block count each time to avoid compression
5672 * kicking in. */
5673 spin->si_blocks_cnt = 0;
5674
5675 /* Skip over any other NUL bytes (same word with different
5676 * flags). */
5677 while (byts[n + 1] == 0)
5678 {
5679 ++n;
5680 ++curi[depth];
5681 }
5682 }
5683 else
5684 {
5685 /* Normal char, go one level deeper. */
5686 tword[depth++] = c;
5687 arridx[depth] = idxs[n];
5688 curi[depth] = 1;
5689 wordcount[depth] = 0;
5690 }
5691 }
5692 }
5693
5694 smsg((char_u *)_("Total number of words: %d"), words_done);
5695
5696 return OK;
5697 }
5698
5699 /*
5700 * Make the table that links each word in the soundfold trie to the words it
5701 * can be produced from.
5702 * This is not unlike lines in a file, thus use a memfile to be able to access
5703 * the table efficiently.
5704 * Returns FAIL when out of memory.
5705 */
5706 static int
5707 sug_maketable(spellinfo_T *spin)
5708 {
5709 garray_T ga;
5710 int res = OK;
5711
5712 /* Allocate a buffer, open a memline for it and create the swap file
5713 * (uses a temp file, not a .swp file). */
5714 spin->si_spellbuf = open_spellbuf();
5715 if (spin->si_spellbuf == NULL)
5716 return FAIL;
5717
5718 /* Use a buffer to store the line info, avoids allocating many small
5719 * pieces of memory. */
5720 ga_init2(&ga, 1, 100);
5721
5722 /* recursively go through the tree */
5723 if (sug_filltable(spin, spin->si_foldroot->wn_sibling, 0, &ga) == -1)
5724 res = FAIL;
5725
5726 ga_clear(&ga);
5727 return res;
5728 }
5729
5730 /*
5731 * Fill the table for one node and its children.
5732 * Returns the wordnr at the start of the node.
5733 * Returns -1 when out of memory.
5734 */
5735 static int
5736 sug_filltable(
5737 spellinfo_T *spin,
5738 wordnode_T *node,
5739 int startwordnr,
5740 garray_T *gap) /* place to store line of numbers */
5741 {
5742 wordnode_T *p, *np;
5743 int wordnr = startwordnr;
5744 int nr;
5745 int prev_nr;
5746
5747 for (p = node; p != NULL; p = p->wn_sibling)
5748 {
5749 if (p->wn_byte == NUL)
5750 {
5751 gap->ga_len = 0;
5752 prev_nr = 0;
5753 for (np = p; np != NULL && np->wn_byte == NUL; np = np->wn_sibling)
5754 {
5755 if (ga_grow(gap, 10) == FAIL)
5756 return -1;
5757
5758 nr = (np->wn_flags << 16) + (np->wn_region & 0xffff);
5759 /* Compute the offset from the previous nr and store the
5760 * offset in a way that it takes a minimum number of bytes.
5761 * It's a bit like utf-8, but without the need to mark
5762 * following bytes. */
5763 nr -= prev_nr;
5764 prev_nr += nr;
5765 gap->ga_len += offset2bytes(nr,
5766 (char_u *)gap->ga_data + gap->ga_len);
5767 }
5768
5769 /* add the NUL byte */
5770 ((char_u *)gap->ga_data)[gap->ga_len++] = NUL;
5771
5772 if (ml_append_buf(spin->si_spellbuf, (linenr_T)wordnr,
5773 gap->ga_data, gap->ga_len, TRUE) == FAIL)
5774 return -1;
5775 ++wordnr;
5776
5777 /* Remove extra NUL entries, we no longer need them. We don't
5778 * bother freeing the nodes, the won't be reused anyway. */
5779 while (p->wn_sibling != NULL && p->wn_sibling->wn_byte == NUL)
5780 p->wn_sibling = p->wn_sibling->wn_sibling;
5781
5782 /* Clear the flags on the remaining NUL node, so that compression
5783 * works a lot better. */
5784 p->wn_flags = 0;
5785 p->wn_region = 0;
5786 }
5787 else
5788 {
5789 wordnr = sug_filltable(spin, p->wn_child, wordnr, gap);
5790 if (wordnr == -1)
5791 return -1;
5792 }
5793 }
5794 return wordnr;
5795 }
5796
5797 /*
5798 * Convert an offset into a minimal number of bytes.
5799 * Similar to utf_char2byters, but use 8 bits in followup bytes and avoid NUL
5800 * bytes.
5801 */
5802 static int
5803 offset2bytes(int nr, char_u *buf)
5804 {
5805 int rem;
5806 int b1, b2, b3, b4;
5807
5808 /* Split the number in parts of base 255. We need to avoid NUL bytes. */
5809 b1 = nr % 255 + 1;
5810 rem = nr / 255;
5811 b2 = rem % 255 + 1;
5812 rem = rem / 255;
5813 b3 = rem % 255 + 1;
5814 b4 = rem / 255 + 1;
5815
5816 if (b4 > 1 || b3 > 0x1f) /* 4 bytes */
5817 {
5818 buf[0] = 0xe0 + b4;
5819 buf[1] = b3;
5820 buf[2] = b2;
5821 buf[3] = b1;
5822 return 4;
5823 }
5824 if (b3 > 1 || b2 > 0x3f ) /* 3 bytes */
5825 {
5826 buf[0] = 0xc0 + b3;
5827 buf[1] = b2;
5828 buf[2] = b1;
5829 return 3;
5830 }
5831 if (b2 > 1 || b1 > 0x7f ) /* 2 bytes */
5832 {
5833 buf[0] = 0x80 + b2;
5834 buf[1] = b1;
5835 return 2;
5836 }
5837 /* 1 byte */
5838 buf[0] = b1;
5839 return 1;
5840 }
5841
5842 /*
5843 * Write the .sug file in "fname".
5844 */
5845 static void
5846 sug_write(spellinfo_T *spin, char_u *fname)
5847 {
5848 FILE *fd;
5849 wordnode_T *tree;
5850 int nodecount;
5851 int wcount;
5852 char_u *line;
5853 linenr_T lnum;
5854 int len;
5855
5856 /* Create the file. Note that an existing file is silently overwritten! */
5857 fd = mch_fopen((char *)fname, "w");
5858 if (fd == NULL)
5859 {
5860 EMSG2(_(e_notopen), fname);
5861 return;
5862 }
5863
5864 vim_snprintf((char *)IObuff, IOSIZE,
5865 _("Writing suggestion file %s ..."), fname);
5866 spell_message(spin, IObuff);
5867
5868 /*
5869 * <SUGHEADER>: <fileID> <versionnr> <timestamp>
5870 */
5871 if (fwrite(VIMSUGMAGIC, VIMSUGMAGICL, (size_t)1, fd) != 1) /* <fileID> */
5872 {
5873 EMSG(_(e_write));
5874 goto theend;
5875 }
5876 putc(VIMSUGVERSION, fd); /* <versionnr> */
5877
5878 /* Write si_sugtime to the file. */
5879 put_time(fd, spin->si_sugtime); /* <timestamp> */
5880
5881 /*
5882 * <SUGWORDTREE>
5883 */
5884 spin->si_memtot = 0;
5885 tree = spin->si_foldroot->wn_sibling;
5886
5887 /* Clear the index and wnode fields in the tree. */
5888 clear_node(tree);
5889
5890 /* Count the number of nodes. Needed to be able to allocate the
5891 * memory when reading the nodes. Also fills in index for shared
5892 * nodes. */
5893 nodecount = put_node(NULL, tree, 0, 0, FALSE);
5894
5895 /* number of nodes in 4 bytes */
5896 put_bytes(fd, (long_u)nodecount, 4); /* <nodecount> */
5897 spin->si_memtot += nodecount + nodecount * sizeof(int);
5898
5899 /* Write the nodes. */
5900 (void)put_node(fd, tree, 0, 0, FALSE);
5901
5902 /*
5903 * <SUGTABLE>: <sugwcount> <sugline> ...
5904 */
5905 wcount = spin->si_spellbuf->b_ml.ml_line_count;
5906 put_bytes(fd, (long_u)wcount, 4); /* <sugwcount> */
5907
5908 for (lnum = 1; lnum <= (linenr_T)wcount; ++lnum)
5909 {
5910 /* <sugline>: <sugnr> ... NUL */
5911 line = ml_get_buf(spin->si_spellbuf, lnum, FALSE);
5912 len = (int)STRLEN(line) + 1;
5913 if (fwrite(line, (size_t)len, (size_t)1, fd) == 0)
5914 {
5915 EMSG(_(e_write));
5916 goto theend;
5917 }
5918 spin->si_memtot += len;
5919 }
5920
5921 /* Write another byte to check for errors. */
5922 if (putc(0, fd) == EOF)
5923 EMSG(_(e_write));
5924
5925 vim_snprintf((char *)IObuff, IOSIZE,
5926 _("Estimated runtime memory use: %d bytes"), spin->si_memtot);
5927 spell_message(spin, IObuff);
5928
5929 theend:
5930 /* close the file */
5931 fclose(fd);
5932 }
5933
5934
5935 /*
5936 * Create a Vim spell file from one or more word lists.
5937 * "fnames[0]" is the output file name.
5938 * "fnames[fcount - 1]" is the last input file name.
5939 * Exception: when "fnames[0]" ends in ".add" it's used as the input file name
5940 * and ".spl" is appended to make the output file name.
5941 */
5942 void
5943 mkspell(
5944 int fcount,
5945 char_u **fnames,
5946 int ascii, /* -ascii argument given */
5947 int over_write, /* overwrite existing output file */
5948 int added_word) /* invoked through "zg" */
5949 {
5950 char_u *fname = NULL;
5951 char_u *wfname;
5952 char_u **innames;
5953 int incount;
5954 afffile_T *(afile[8]);
5955 int i;
5956 int len;
5957 stat_T st;
5958 int error = FALSE;
5959 spellinfo_T spin;
5960
5961 vim_memset(&spin, 0, sizeof(spin));
5962 spin.si_verbose = !added_word;
5963 spin.si_ascii = ascii;
5964 spin.si_followup = TRUE;
5965 spin.si_rem_accents = TRUE;
5966 ga_init2(&spin.si_rep, (int)sizeof(fromto_T), 20);
5967 ga_init2(&spin.si_repsal, (int)sizeof(fromto_T), 20);
5968 ga_init2(&spin.si_sal, (int)sizeof(fromto_T), 20);
5969 ga_init2(&spin.si_map, (int)sizeof(char_u), 100);
5970 ga_init2(&spin.si_comppat, (int)sizeof(char_u *), 20);
5971 ga_init2(&spin.si_prefcond, (int)sizeof(char_u *), 50);
5972 hash_init(&spin.si_commonwords);
5973 spin.si_newcompID = 127; /* start compound ID at first maximum */
5974
5975 /* default: fnames[0] is output file, following are input files */
5976 innames = &fnames[1];
5977 incount = fcount - 1;
5978
5979 wfname = alloc(MAXPATHL);
5980 if (wfname == NULL)
5981 return;
5982
5983 if (fcount >= 1)
5984 {
5985 len = (int)STRLEN(fnames[0]);
5986 if (fcount == 1 && len > 4 && STRCMP(fnames[0] + len - 4, ".add") == 0)
5987 {
5988 /* For ":mkspell path/en.latin1.add" output file is
5989 * "path/en.latin1.add.spl". */
5990 innames = &fnames[0];
5991 incount = 1;
5992 vim_snprintf((char *)wfname, MAXPATHL, "%s.spl", fnames[0]);
5993 }
5994 else if (fcount == 1)
5995 {
5996 /* For ":mkspell path/vim" output file is "path/vim.latin1.spl". */
5997 innames = &fnames[0];
5998 incount = 1;
5999 vim_snprintf((char *)wfname, MAXPATHL, SPL_FNAME_TMPL,
6000 fnames[0], spin.si_ascii ? (char_u *)"ascii" : spell_enc());
6001 }
6002 else if (len > 4 && STRCMP(fnames[0] + len - 4, ".spl") == 0)
6003 {
6004 /* Name ends in ".spl", use as the file name. */
6005 vim_strncpy(wfname, fnames[0], MAXPATHL - 1);
6006 }
6007 else
6008 /* Name should be language, make the file name from it. */
6009 vim_snprintf((char *)wfname, MAXPATHL, SPL_FNAME_TMPL,
6010 fnames[0], spin.si_ascii ? (char_u *)"ascii" : spell_enc());
6011
6012 /* Check for .ascii.spl. */
6013 if (strstr((char *)gettail(wfname), SPL_FNAME_ASCII) != NULL)
6014 spin.si_ascii = TRUE;
6015
6016 /* Check for .add.spl. */
6017 if (strstr((char *)gettail(wfname), SPL_FNAME_ADD) != NULL)
6018 spin.si_add = TRUE;
6019 }
6020
6021 if (incount <= 0)
6022 EMSG(_(e_invarg)); /* need at least output and input names */
6023 else if (vim_strchr(gettail(wfname), '_') != NULL)
6024 EMSG(_("E751: Output file name must not have region name"));
6025 else if (incount > 8)
6026 EMSG(_("E754: Only up to 8 regions supported"));
6027 else
6028 {
6029 /* Check for overwriting before doing things that may take a lot of
6030 * time. */
6031 if (!over_write && mch_stat((char *)wfname, &st) >= 0)
6032 {
6033 EMSG(_(e_exists));
6034 goto theend;
6035 }
6036 if (mch_isdir(wfname))
6037 {
6038 EMSG2(_(e_isadir2), wfname);
6039 goto theend;
6040 }
6041
6042 fname = alloc(MAXPATHL);
6043 if (fname == NULL)
6044 goto theend;
6045
6046 /*
6047 * Init the aff and dic pointers.
6048 * Get the region names if there are more than 2 arguments.
6049 */
6050 for (i = 0; i < incount; ++i)
6051 {
6052 afile[i] = NULL;
6053
6054 if (incount > 1)
6055 {
6056 len = (int)STRLEN(innames[i]);
6057 if (STRLEN(gettail(innames[i])) < 5
6058 || innames[i][len - 3] != '_')
6059 {
6060 EMSG2(_("E755: Invalid region in %s"), innames[i]);
6061 goto theend;
6062 }
6063 spin.si_region_name[i * 2] = TOLOWER_ASC(innames[i][len - 2]);
6064 spin.si_region_name[i * 2 + 1] =
6065 TOLOWER_ASC(innames[i][len - 1]);
6066 }
6067 }
6068 spin.si_region_count = incount;
6069
6070 spin.si_foldroot = wordtree_alloc(&spin);
6071 spin.si_keeproot = wordtree_alloc(&spin);
6072 spin.si_prefroot = wordtree_alloc(&spin);
6073 if (spin.si_foldroot == NULL
6074 || spin.si_keeproot == NULL
6075 || spin.si_prefroot == NULL)
6076 {
6077 free_blocks(spin.si_blocks);
6078 goto theend;
6079 }
6080
6081 /* When not producing a .add.spl file clear the character table when
6082 * we encounter one in the .aff file. This means we dump the current
6083 * one in the .spl file if the .aff file doesn't define one. That's
6084 * better than guessing the contents, the table will match a
6085 * previously loaded spell file. */
6086 if (!spin.si_add)
6087 spin.si_clear_chartab = TRUE;
6088
6089 /*
6090 * Read all the .aff and .dic files.
6091 * Text is converted to 'encoding'.
6092 * Words are stored in the case-folded and keep-case trees.
6093 */
6094 for (i = 0; i < incount && !error; ++i)
6095 {
6096 spin.si_conv.vc_type = CONV_NONE;
6097 spin.si_region = 1 << i;
6098
6099 vim_snprintf((char *)fname, MAXPATHL, "%s.aff", innames[i]);
6100 if (mch_stat((char *)fname, &st) >= 0)
6101 {
6102 /* Read the .aff file. Will init "spin->si_conv" based on the
6103 * "SET" line. */
6104 afile[i] = spell_read_aff(&spin, fname);
6105 if (afile[i] == NULL)
6106 error = TRUE;
6107 else
6108 {
6109 /* Read the .dic file and store the words in the trees. */
6110 vim_snprintf((char *)fname, MAXPATHL, "%s.dic",
6111 innames[i]);
6112 if (spell_read_dic(&spin, fname, afile[i]) == FAIL)
6113 error = TRUE;
6114 }
6115 }
6116 else
6117 {
6118 /* No .aff file, try reading the file as a word list. Store
6119 * the words in the trees. */
6120 if (spell_read_wordfile(&spin, innames[i]) == FAIL)
6121 error = TRUE;
6122 }
6123
6124 #ifdef FEAT_MBYTE
6125 /* Free any conversion stuff. */
6126 convert_setup(&spin.si_conv, NULL, NULL);
6127 #endif
6128 }
6129
6130 if (spin.si_compflags != NULL && spin.si_nobreak)
6131 MSG(_("Warning: both compounding and NOBREAK specified"));
6132
6133 if (!error && !got_int)
6134 {
6135 /*
6136 * Combine tails in the tree.
6137 */
6138 spell_message(&spin, (char_u *)_(msg_compressing));
6139 wordtree_compress(&spin, spin.si_foldroot);
6140 wordtree_compress(&spin, spin.si_keeproot);
6141 wordtree_compress(&spin, spin.si_prefroot);
6142 }
6143
6144 if (!error && !got_int)
6145 {
6146 /*
6147 * Write the info in the spell file.
6148 */
6149 vim_snprintf((char *)IObuff, IOSIZE,
6150 _("Writing spell file %s ..."), wfname);
6151 spell_message(&spin, IObuff);
6152
6153 error = write_vim_spell(&spin, wfname) == FAIL;
6154
6155 spell_message(&spin, (char_u *)_("Done!"));
6156 vim_snprintf((char *)IObuff, IOSIZE,
6157 _("Estimated runtime memory use: %d bytes"), spin.si_memtot);
6158 spell_message(&spin, IObuff);
6159
6160 /*
6161 * If the file is loaded need to reload it.
6162 */
6163 if (!error)
6164 spell_reload_one(wfname, added_word);
6165 }
6166
6167 /* Free the allocated memory. */
6168 ga_clear(&spin.si_rep);
6169 ga_clear(&spin.si_repsal);
6170 ga_clear(&spin.si_sal);
6171 ga_clear(&spin.si_map);
6172 ga_clear(&spin.si_comppat);
6173 ga_clear(&spin.si_prefcond);
6174 hash_clear_all(&spin.si_commonwords, 0);
6175
6176 /* Free the .aff file structures. */
6177 for (i = 0; i < incount; ++i)
6178 if (afile[i] != NULL)
6179 spell_free_aff(afile[i]);
6180
6181 /* Free all the bits and pieces at once. */
6182 free_blocks(spin.si_blocks);
6183
6184 /*
6185 * If there is soundfolding info and no NOSUGFILE item create the
6186 * .sug file with the soundfolded word trie.
6187 */
6188 if (spin.si_sugtime != 0 && !error && !got_int)
6189 spell_make_sugfile(&spin, wfname);
6190
6191 }
6192
6193 theend:
6194 vim_free(fname);
6195 vim_free(wfname);
6196 }
6197
6198 /*
6199 * Display a message for spell file processing when 'verbose' is set or using
6200 * ":mkspell". "str" can be IObuff.
6201 */
6202 static void
6203 spell_message(spellinfo_T *spin, char_u *str)
6204 {
6205 if (spin->si_verbose || p_verbose > 2)
6206 {
6207 if (!spin->si_verbose)
6208 verbose_enter();
6209 MSG(str);
6210 out_flush();
6211 if (!spin->si_verbose)
6212 verbose_leave();
6213 }
6214 }
6215
6216 /*
6217 * ":[count]spellgood {word}"
6218 * ":[count]spellwrong {word}"
6219 * ":[count]spellundo {word}"
6220 */
6221 void
6222 ex_spell(exarg_T *eap)
6223 {
6224 spell_add_word(eap->arg, (int)STRLEN(eap->arg), eap->cmdidx == CMD_spellwrong,
6225 eap->forceit ? 0 : (int)eap->line2,
6226 eap->cmdidx == CMD_spellundo);
6227 }
6228
6229 /*
6230 * Add "word[len]" to 'spellfile' as a good or bad word.
6231 */
6232 void
6233 spell_add_word(
6234 char_u *word,
6235 int len,
6236 int bad,
6237 int idx, /* "zG" and "zW": zero, otherwise index in
6238 'spellfile' */
6239 int undo) /* TRUE for "zug", "zuG", "zuw" and "zuW" */
6240 {
6241 FILE *fd = NULL;
6242 buf_T *buf = NULL;
6243 int new_spf = FALSE;
6244 char_u *fname;
6245 char_u *fnamebuf = NULL;
6246 char_u line[MAXWLEN * 2];
6247 long fpos, fpos_next = 0;
6248 int i;
6249 char_u *spf;
6250
6251 if (idx == 0) /* use internal wordlist */
6252 {
6253 if (int_wordlist == NULL)
6254 {
6255 int_wordlist = vim_tempname('s', FALSE);
6256 if (int_wordlist == NULL)
6257 return;
6258 }
6259 fname = int_wordlist;
6260 }
6261 else
6262 {
6263 /* If 'spellfile' isn't set figure out a good default value. */
6264 if (*curwin->w_s->b_p_spf == NUL)
6265 {
6266 init_spellfile();
6267 new_spf = TRUE;
6268 }
6269
6270 if (*curwin->w_s->b_p_spf == NUL)
6271 {
6272 EMSG2(_(e_notset), "spellfile");
6273 return;
6274 }
6275 fnamebuf = alloc(MAXPATHL);
6276 if (fnamebuf == NULL)
6277 return;
6278
6279 for (spf = curwin->w_s->b_p_spf, i = 1; *spf != NUL; ++i)
6280 {
6281 copy_option_part(&spf, fnamebuf, MAXPATHL, ",");
6282 if (i == idx)
6283 break;
6284 if (*spf == NUL)
6285 {
6286 EMSGN(_("E765: 'spellfile' does not have %ld entries"), idx);
6287 vim_free(fnamebuf);
6288 return;
6289 }
6290 }
6291
6292 /* Check that the user isn't editing the .add file somewhere. */
6293 buf = buflist_findname_exp(fnamebuf);
6294 if (buf != NULL && buf->b_ml.ml_mfp == NULL)
6295 buf = NULL;
6296 if (buf != NULL && bufIsChanged(buf))
6297 {
6298 EMSG(_(e_bufloaded));
6299 vim_free(fnamebuf);
6300 return;
6301 }
6302
6303 fname = fnamebuf;
6304 }
6305
6306 if (bad || undo)
6307 {
6308 /* When the word appears as good word we need to remove that one,
6309 * since its flags sort before the one with WF_BANNED. */
6310 fd = mch_fopen((char *)fname, "r");
6311 if (fd != NULL)
6312 {
6313 while (!vim_fgets(line, MAXWLEN * 2, fd))
6314 {
6315 fpos = fpos_next;
6316 fpos_next = ftell(fd);
6317 if (STRNCMP(word, line, len) == 0
6318 && (line[len] == '/' || line[len] < ' '))
6319 {
6320 /* Found duplicate word. Remove it by writing a '#' at
6321 * the start of the line. Mixing reading and writing
6322 * doesn't work for all systems, close the file first. */
6323 fclose(fd);
6324 fd = mch_fopen((char *)fname, "r+");
6325 if (fd == NULL)
6326 break;
6327 if (fseek(fd, fpos, SEEK_SET) == 0)
6328 {
6329 fputc('#', fd);
6330 if (undo)
6331 {
6332 home_replace(NULL, fname, NameBuff, MAXPATHL, TRUE);
6333 smsg((char_u *)_("Word '%.*s' removed from %s"),
6334 len, word, NameBuff);
6335 }
6336 }
6337 fseek(fd, fpos_next, SEEK_SET);
6338 }
6339 }
6340 if (fd != NULL)
6341 fclose(fd);
6342 }
6343 }
6344
6345 if (!undo)
6346 {
6347 fd = mch_fopen((char *)fname, "a");
6348 if (fd == NULL && new_spf)
6349 {
6350 char_u *p;
6351
6352 /* We just initialized the 'spellfile' option and can't open the
6353 * file. We may need to create the "spell" directory first. We
6354 * already checked the runtime directory is writable in
6355 * init_spellfile(). */
6356 if (!dir_of_file_exists(fname) && (p = gettail_sep(fname)) != fname)
6357 {
6358 int c = *p;
6359
6360 /* The directory doesn't exist. Try creating it and opening
6361 * the file again. */
6362 *p = NUL;
6363 vim_mkdir(fname, 0755);
6364 *p = c;
6365 fd = mch_fopen((char *)fname, "a");
6366 }
6367 }
6368
6369 if (fd == NULL)
6370 EMSG2(_(e_notopen), fname);
6371 else
6372 {
6373 if (bad)
6374 fprintf(fd, "%.*s/!\n", len, word);
6375 else
6376 fprintf(fd, "%.*s\n", len, word);
6377 fclose(fd);
6378
6379 home_replace(NULL, fname, NameBuff, MAXPATHL, TRUE);
6380 smsg((char_u *)_("Word '%.*s' added to %s"), len, word, NameBuff);
6381 }
6382 }
6383
6384 if (fd != NULL)
6385 {
6386 /* Update the .add.spl file. */
6387 mkspell(1, &fname, FALSE, TRUE, TRUE);
6388
6389 /* If the .add file is edited somewhere, reload it. */
6390 if (buf != NULL)
6391 buf_reload(buf, buf->b_orig_mode);
6392
6393 redraw_all_later(SOME_VALID);
6394 }
6395 vim_free(fnamebuf);
6396 }
6397
6398 /*
6399 * Initialize 'spellfile' for the current buffer.
6400 */
6401 static void
6402 init_spellfile(void)
6403 {
6404 char_u *buf;
6405 int l;
6406 char_u *fname;
6407 char_u *rtp;
6408 char_u *lend;
6409 int aspath = FALSE;
6410 char_u *lstart = curbuf->b_s.b_p_spl;
6411
6412 if (*curwin->w_s->b_p_spl != NUL && curwin->w_s->b_langp.ga_len > 0)
6413 {
6414 buf = alloc(MAXPATHL);
6415 if (buf == NULL)
6416 return;
6417
6418 /* Find the end of the language name. Exclude the region. If there
6419 * is a path separator remember the start of the tail. */
6420 for (lend = curwin->w_s->b_p_spl; *lend != NUL
6421 && vim_strchr((char_u *)",._", *lend) == NULL; ++lend)
6422 if (vim_ispathsep(*lend))
6423 {
6424 aspath = TRUE;
6425 lstart = lend + 1;
6426 }
6427
6428 /* Loop over all entries in 'runtimepath'. Use the first one where we
6429 * are allowed to write. */
6430 rtp = p_rtp;
6431 while (*rtp != NUL)
6432 {
6433 if (aspath)
6434 /* Use directory of an entry with path, e.g., for
6435 * "/dir/lg.utf-8.spl" use "/dir". */
6436 vim_strncpy(buf, curbuf->b_s.b_p_spl,
6437 lstart - curbuf->b_s.b_p_spl - 1);
6438 else
6439 /* Copy the path from 'runtimepath' to buf[]. */
6440 copy_option_part(&rtp, buf, MAXPATHL, ",");
6441 if (filewritable(buf) == 2)
6442 {
6443 /* Use the first language name from 'spelllang' and the
6444 * encoding used in the first loaded .spl file. */
6445 if (aspath)
6446 vim_strncpy(buf, curbuf->b_s.b_p_spl,
6447 lend - curbuf->b_s.b_p_spl);
6448 else
6449 {
6450 /* Create the "spell" directory if it doesn't exist yet. */
6451 l = (int)STRLEN(buf);
6452 vim_snprintf((char *)buf + l, MAXPATHL - l, "/spell");
6453 if (filewritable(buf) != 2)
6454 vim_mkdir(buf, 0755);
6455
6456 l = (int)STRLEN(buf);
6457 vim_snprintf((char *)buf + l, MAXPATHL - l,
6458 "/%.*s", (int)(lend - lstart), lstart);
6459 }
6460 l = (int)STRLEN(buf);
6461 fname = LANGP_ENTRY(curwin->w_s->b_langp, 0)
6462 ->lp_slang->sl_fname;
6463 vim_snprintf((char *)buf + l, MAXPATHL - l, ".%s.add",
6464 fname != NULL
6465 && strstr((char *)gettail(fname), ".ascii.") != NULL
6466 ? (char_u *)"ascii" : spell_enc());
6467 set_option_value((char_u *)"spellfile", 0L, buf, OPT_LOCAL);
6468 break;
6469 }
6470 aspath = FALSE;
6471 }
6472
6473 vim_free(buf);
6474 }
6475 }
6476
6477
6478
6479 /*
6480 * Set the spell character tables from strings in the affix file.
6481 */
6482 static int
6483 set_spell_chartab(char_u *fol, char_u *low, char_u *upp)
6484 {
6485 /* We build the new tables here first, so that we can compare with the
6486 * previous one. */
6487 spelltab_T new_st;
6488 char_u *pf = fol, *pl = low, *pu = upp;
6489 int f, l, u;
6490
6491 clear_spell_chartab(&new_st);
6492
6493 while (*pf != NUL)
6494 {
6495 if (*pl == NUL || *pu == NUL)
6496 {
6497 EMSG(_(e_affform));
6498 return FAIL;
6499 }
6500 #ifdef FEAT_MBYTE
6501 f = mb_ptr2char_adv(&pf);
6502 l = mb_ptr2char_adv(&pl);
6503 u = mb_ptr2char_adv(&pu);
6504 #else
6505 f = *pf++;
6506 l = *pl++;
6507 u = *pu++;
6508 #endif
6509 /* Every character that appears is a word character. */
6510 if (f < 256)
6511 new_st.st_isw[f] = TRUE;
6512 if (l < 256)
6513 new_st.st_isw[l] = TRUE;
6514 if (u < 256)
6515 new_st.st_isw[u] = TRUE;
6516
6517 /* if "LOW" and "FOL" are not the same the "LOW" char needs
6518 * case-folding */
6519 if (l < 256 && l != f)
6520 {
6521 if (f >= 256)
6522 {
6523 EMSG(_(e_affrange));
6524 return FAIL;
6525 }
6526 new_st.st_fold[l] = f;
6527 }
6528
6529 /* if "UPP" and "FOL" are not the same the "UPP" char needs
6530 * case-folding, it's upper case and the "UPP" is the upper case of
6531 * "FOL" . */
6532 if (u < 256 && u != f)
6533 {
6534 if (f >= 256)
6535 {
6536 EMSG(_(e_affrange));
6537 return FAIL;
6538 }
6539 new_st.st_fold[u] = f;
6540 new_st.st_isu[u] = TRUE;
6541 new_st.st_upper[f] = u;
6542 }
6543 }
6544
6545 if (*pl != NUL || *pu != NUL)
6546 {
6547 EMSG(_(e_affform));
6548 return FAIL;
6549 }
6550
6551 return set_spell_finish(&new_st);
6552 }
6553
6554 /*
6555 * Set the spell character tables from strings in the .spl file.
6556 */
6557 static void
6558 set_spell_charflags(
6559 char_u *flags,
6560 int cnt, /* length of "flags" */
6561 char_u *fol)
6562 {
6563 /* We build the new tables here first, so that we can compare with the
6564 * previous one. */
6565 spelltab_T new_st;
6566 int i;
6567 char_u *p = fol;
6568 int c;
6569
6570 clear_spell_chartab(&new_st);
6571
6572 for (i = 0; i < 128; ++i)
6573 {
6574 if (i < cnt)
6575 {
6576 new_st.st_isw[i + 128] = (flags[i] & CF_WORD) != 0;
6577 new_st.st_isu[i + 128] = (flags[i] & CF_UPPER) != 0;
6578 }
6579
6580 if (*p != NUL)
6581 {
6582 #ifdef FEAT_MBYTE
6583 c = mb_ptr2char_adv(&p);
6584 #else
6585 c = *p++;
6586 #endif
6587 new_st.st_fold[i + 128] = c;
6588 if (i + 128 != c && new_st.st_isu[i + 128] && c < 256)
6589 new_st.st_upper[c] = i + 128;
6590 }
6591 }
6592
6593 (void)set_spell_finish(&new_st);
6594 }
6595
6596 static int
6597 set_spell_finish(spelltab_T *new_st)
6598 {
6599 int i;
6600
6601 if (did_set_spelltab)
6602 {
6603 /* check that it's the same table */
6604 for (i = 0; i < 256; ++i)
6605 {
6606 if (spelltab.st_isw[i] != new_st->st_isw[i]
6607 || spelltab.st_isu[i] != new_st->st_isu[i]
6608 || spelltab.st_fold[i] != new_st->st_fold[i]
6609 || spelltab.st_upper[i] != new_st->st_upper[i])
6610 {
6611 EMSG(_("E763: Word characters differ between spell files"));
6612 return FAIL;
6613 }
6614 }
6615 }
6616 else
6617 {
6618 /* copy the new spelltab into the one being used */
6619 spelltab = *new_st;
6620 did_set_spelltab = TRUE;
6621 }
6622
6623 return OK;
6624 }
6625
6626 /*
6627 * Write the table with prefix conditions to the .spl file.
6628 * When "fd" is NULL only count the length of what is written.
6629 */
6630 static int
6631 write_spell_prefcond(FILE *fd, garray_T *gap)
6632 {
6633 int i;
6634 char_u *p;
6635 int len;
6636 int totlen;
6637 size_t x = 1; /* collect return value of fwrite() */
6638
6639 if (fd != NULL)
6640 put_bytes(fd, (long_u)gap->ga_len, 2); /* <prefcondcnt> */
6641
6642 totlen = 2 + gap->ga_len; /* length of <prefcondcnt> and <condlen> bytes */
6643
6644 for (i = 0; i < gap->ga_len; ++i)
6645 {
6646 /* <prefcond> : <condlen> <condstr> */
6647 p = ((char_u **)gap->ga_data)[i];
6648 if (p != NULL)
6649 {
6650 len = (int)STRLEN(p);
6651 if (fd != NULL)
6652 {
6653 fputc(len, fd);
6654 x &= fwrite(p, (size_t)len, (size_t)1, fd);
6655 }
6656 totlen += len;
6657 }
6658 else if (fd != NULL)
6659 fputc(0, fd);
6660 }
6661
6662 return totlen;
6663 }
6664
6665
6666 /*
6667 * Use map string "map" for languages "lp".
6668 */
6669 static void
6670 set_map_str(slang_T *lp, char_u *map)
6671 {
6672 char_u *p;
6673 int headc = 0;
6674 int c;
6675 int i;
6676
6677 if (*map == NUL)
6678 {
6679 lp->sl_has_map = FALSE;
6680 return;
6681 }
6682 lp->sl_has_map = TRUE;
6683
6684 /* Init the array and hash tables empty. */
6685 for (i = 0; i < 256; ++i)
6686 lp->sl_map_array[i] = 0;
6687 #ifdef FEAT_MBYTE
6688 hash_init(&lp->sl_map_hash);
6689 #endif
6690
6691 /*
6692 * The similar characters are stored separated with slashes:
6693 * "aaa/bbb/ccc/". Fill sl_map_array[c] with the character before c and
6694 * before the same slash. For characters above 255 sl_map_hash is used.
6695 */
6696 for (p = map; *p != NUL; )
6697 {
6698 #ifdef FEAT_MBYTE
6699 c = mb_cptr2char_adv(&p);
6700 #else
6701 c = *p++;
6702 #endif
6703 if (c == '/')
6704 headc = 0;
6705 else
6706 {
6707 if (headc == 0)
6708 headc = c;
6709
6710 #ifdef FEAT_MBYTE
6711 /* Characters above 255 don't fit in sl_map_array[], put them in
6712 * the hash table. Each entry is the char, a NUL the headchar and
6713 * a NUL. */
6714 if (c >= 256)
6715 {
6716 int cl = mb_char2len(c);
6717 int headcl = mb_char2len(headc);
6718 char_u *b;
6719 hash_T hash;
6720 hashitem_T *hi;
6721
6722 b = alloc((unsigned)(cl + headcl + 2));
6723 if (b == NULL)
6724 return;
6725 mb_char2bytes(c, b);
6726 b[cl] = NUL;
6727 mb_char2bytes(headc, b + cl + 1);
6728 b[cl + 1 + headcl] = NUL;
6729 hash = hash_hash(b);
6730 hi = hash_lookup(&lp->sl_map_hash, b, hash);
6731 if (HASHITEM_EMPTY(hi))
6732 hash_add_item(&lp->sl_map_hash, hi, b, hash);
6733 else
6734 {
6735 /* This should have been checked when generating the .spl
6736 * file. */
6737 EMSG(_("E783: duplicate char in MAP entry"));
6738 vim_free(b);
6739 }
6740 }
6741 else
6742 #endif
6743 lp->sl_map_array[c] = headc;
6744 }
6745 }
6746 }
6747
6748
6749 #endif /* FEAT_SPELL */