Mercurial > vim
annotate src/regexp_nfa.c @ 4533:6a2005efa02b v7.3.1014
updated for version 7.3.1014
Problem: New regexp state dump is hard to read.
Solution: Make the state dump more pretty. (Taro Muraoka)
| author | Bram Moolenaar <bram@vim.org> |
|---|---|
| date | Sat, 25 May 2013 12:28:11 +0200 |
| parents | 1be43c095aff |
| children | 45f97c349537 |
| rev | line source |
|---|---|
| 4444 | 1 /* vi:set ts=8 sts=4 sw=4: |
| 2 * | |
| 3 * NFA regular expression implementation. | |
| 4 * | |
| 5 * This file is included in "regexp.c". | |
| 6 */ | |
| 7 | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
8 /* |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
9 * Logging of NFA engine. |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
10 * |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
11 * The NFA engine can write four log files: |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
12 * - Error log: Contains NFA engine's fatal errors. |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
13 * - Dump log: Contains compiled NFA state machine's information. |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
14 * - Run log: Contains information of matching procedure. |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
15 * - Debug log: Contains detailed information of matching procedure. Can be |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
16 * disabled by undefining NFA_REGEXP_DEBUG_LOG. |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
17 * The first one can also be used without debug mode. |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
18 * The last three are enabled when compiled as debug mode and individually |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
19 * disabled by commenting them out. |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
20 * The log files can get quite big! |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
21 * Do disable all of this when compiling Vim for debugging, undefine DEBUG in |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
22 * regexp.c |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
23 */ |
| 4444 | 24 #ifdef DEBUG |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
25 # define NFA_REGEXP_ERROR_LOG "nfa_regexp_error.log" |
| 4444 | 26 # define ENABLE_LOG |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
27 # define NFA_REGEXP_DUMP_LOG "nfa_regexp_dump.log" |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
28 # define NFA_REGEXP_RUN_LOG "nfa_regexp_run.log" |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
29 # define NFA_REGEXP_DEBUG_LOG "nfa_regexp_debug.log" |
| 4444 | 30 #endif |
| 31 | |
| 32 /* Upper limit allowed for {m,n} repetitions handled by NFA */ | |
| 33 #define NFA_BRACES_MAXLIMIT 10 | |
| 34 /* For allocating space for the postfix representation */ | |
| 35 #define NFA_POSTFIX_MULTIPLIER (NFA_BRACES_MAXLIMIT + 2)*2 | |
| 36 | |
| 37 enum | |
| 38 { | |
| 39 NFA_SPLIT = -1024, | |
| 40 NFA_MATCH, | |
| 41 NFA_SKIP_CHAR, /* matches a 0-length char */ | |
| 42 NFA_END_NEG_RANGE, /* Used when expanding [^ab] */ | |
| 43 | |
| 44 NFA_CONCAT, | |
| 45 NFA_OR, | |
| 46 NFA_STAR, | |
| 47 NFA_PLUS, | |
| 48 NFA_QUEST, | |
| 49 NFA_QUEST_NONGREEDY, /* Non-greedy version of \? */ | |
| 50 NFA_NOT, /* used for [^ab] negated char ranges */ | |
| 51 | |
| 52 NFA_BOL, /* ^ Begin line */ | |
| 53 NFA_EOL, /* $ End line */ | |
| 54 NFA_BOW, /* \< Begin word */ | |
| 55 NFA_EOW, /* \> End word */ | |
| 56 NFA_BOF, /* \%^ Begin file */ | |
| 57 NFA_EOF, /* \%$ End file */ | |
| 58 NFA_NEWL, | |
| 59 NFA_ZSTART, /* Used for \zs */ | |
| 60 NFA_ZEND, /* Used for \ze */ | |
| 61 NFA_NOPEN, /* Start of subexpression marked with \%( */ | |
| 62 NFA_NCLOSE, /* End of subexpr. marked with \%( ... \) */ | |
| 63 NFA_START_INVISIBLE, | |
| 64 NFA_END_INVISIBLE, | |
| 65 NFA_COMPOSING, /* Next nodes in NFA are part of the | |
| 66 composing multibyte char */ | |
| 67 NFA_END_COMPOSING, /* End of a composing char in the NFA */ | |
| 68 | |
| 69 /* The following are used only in the postfix form, not in the NFA */ | |
| 70 NFA_PREV_ATOM_NO_WIDTH, /* Used for \@= */ | |
| 71 NFA_PREV_ATOM_NO_WIDTH_NEG, /* Used for \@! */ | |
| 72 NFA_PREV_ATOM_JUST_BEFORE, /* Used for \@<= */ | |
| 73 NFA_PREV_ATOM_JUST_BEFORE_NEG, /* Used for \@<! */ | |
| 74 NFA_PREV_ATOM_LIKE_PATTERN, /* Used for \@> */ | |
| 75 | |
| 76 NFA_MOPEN, | |
| 77 NFA_MCLOSE = NFA_MOPEN + NSUBEXP, | |
| 78 | |
| 79 /* NFA_FIRST_NL */ | |
| 80 NFA_ANY = NFA_MCLOSE + NSUBEXP, /* Match any one character. */ | |
| 81 NFA_ANYOF, /* Match any character in this string. */ | |
| 82 NFA_ANYBUT, /* Match any character not in this string. */ | |
| 83 NFA_IDENT, /* Match identifier char */ | |
| 84 NFA_SIDENT, /* Match identifier char but no digit */ | |
| 85 NFA_KWORD, /* Match keyword char */ | |
| 86 NFA_SKWORD, /* Match word char but no digit */ | |
| 87 NFA_FNAME, /* Match file name char */ | |
| 88 NFA_SFNAME, /* Match file name char but no digit */ | |
| 89 NFA_PRINT, /* Match printable char */ | |
| 90 NFA_SPRINT, /* Match printable char but no digit */ | |
| 91 NFA_WHITE, /* Match whitespace char */ | |
| 92 NFA_NWHITE, /* Match non-whitespace char */ | |
| 93 NFA_DIGIT, /* Match digit char */ | |
| 94 NFA_NDIGIT, /* Match non-digit char */ | |
| 95 NFA_HEX, /* Match hex char */ | |
| 96 NFA_NHEX, /* Match non-hex char */ | |
| 97 NFA_OCTAL, /* Match octal char */ | |
| 98 NFA_NOCTAL, /* Match non-octal char */ | |
| 99 NFA_WORD, /* Match word char */ | |
| 100 NFA_NWORD, /* Match non-word char */ | |
| 101 NFA_HEAD, /* Match head char */ | |
| 102 NFA_NHEAD, /* Match non-head char */ | |
| 103 NFA_ALPHA, /* Match alpha char */ | |
| 104 NFA_NALPHA, /* Match non-alpha char */ | |
| 105 NFA_LOWER, /* Match lowercase char */ | |
| 106 NFA_NLOWER, /* Match non-lowercase char */ | |
| 107 NFA_UPPER, /* Match uppercase char */ | |
| 108 NFA_NUPPER, /* Match non-uppercase char */ | |
| 109 NFA_FIRST_NL = NFA_ANY + ADD_NL, | |
| 110 NFA_LAST_NL = NFA_NUPPER + ADD_NL, | |
| 111 | |
| 112 /* Character classes [:alnum:] etc */ | |
| 113 NFA_CLASS_ALNUM, | |
| 114 NFA_CLASS_ALPHA, | |
| 115 NFA_CLASS_BLANK, | |
| 116 NFA_CLASS_CNTRL, | |
| 117 NFA_CLASS_DIGIT, | |
| 118 NFA_CLASS_GRAPH, | |
| 119 NFA_CLASS_LOWER, | |
| 120 NFA_CLASS_PRINT, | |
| 121 NFA_CLASS_PUNCT, | |
| 122 NFA_CLASS_SPACE, | |
| 123 NFA_CLASS_UPPER, | |
| 124 NFA_CLASS_XDIGIT, | |
| 125 NFA_CLASS_TAB, | |
| 126 NFA_CLASS_RETURN, | |
| 127 NFA_CLASS_BACKSPACE, | |
| 128 NFA_CLASS_ESCAPE | |
| 129 }; | |
| 130 | |
| 131 /* Keep in sync with classchars. */ | |
| 132 static int nfa_classcodes[] = { | |
| 133 NFA_ANY, NFA_IDENT, NFA_SIDENT, NFA_KWORD,NFA_SKWORD, | |
| 134 NFA_FNAME, NFA_SFNAME, NFA_PRINT, NFA_SPRINT, | |
| 135 NFA_WHITE, NFA_NWHITE, NFA_DIGIT, NFA_NDIGIT, | |
| 136 NFA_HEX, NFA_NHEX, NFA_OCTAL, NFA_NOCTAL, | |
| 137 NFA_WORD, NFA_NWORD, NFA_HEAD, NFA_NHEAD, | |
| 138 NFA_ALPHA, NFA_NALPHA, NFA_LOWER, NFA_NLOWER, | |
| 139 NFA_UPPER, NFA_NUPPER | |
| 140 }; | |
| 141 | |
| 142 static char_u e_misplaced[] = N_("E866: (NFA regexp) Misplaced %c"); | |
| 143 | |
| 144 /* | |
| 145 * NFA errors can be of 3 types: | |
| 146 * *** NFA runtime errors, when something unknown goes wrong. The NFA fails | |
| 147 * silently and revert the to backtracking engine. | |
| 148 * syntax_error = FALSE; | |
| 149 * *** Regexp syntax errors, when the input regexp is not syntactically correct. | |
| 150 * The NFA engine displays an error message, and nothing else happens. | |
| 151 * syntax_error = TRUE | |
| 152 * *** Unsupported features, when the input regexp uses an operator that is not | |
| 153 * implemented in the NFA. The NFA engine fails silently, and reverts to the | |
| 154 * old backtracking engine. | |
| 155 * syntax_error = FALSE | |
| 156 * "The NFA fails" means that "compiling the regexp with the NFA fails": | |
| 157 * nfa_regcomp() returns FAIL. | |
| 158 */ | |
| 159 static int syntax_error = FALSE; | |
| 160 | |
| 161 /* NFA regexp \ze operator encountered. */ | |
| 162 static int nfa_has_zend = FALSE; | |
| 163 | |
| 164 static int *post_start; /* holds the postfix form of r.e. */ | |
| 165 static int *post_end; | |
| 166 static int *post_ptr; | |
| 167 | |
| 168 static int nstate; /* Number of states in the NFA. */ | |
| 169 static int istate; /* Index in the state vector, used in new_state() */ | |
| 170 static int nstate_max; /* Upper bound of estimated number of states. */ | |
| 171 | |
| 172 | |
| 173 static int nfa_regcomp_start __ARGS((char_u*expr, int re_flags)); | |
| 174 static int nfa_recognize_char_class __ARGS((char_u *start, char_u *end, int extra_newl)); | |
| 175 static int nfa_emit_equi_class __ARGS((int c, int neg)); | |
| 176 static void nfa_inc __ARGS((char_u **p)); | |
| 177 static void nfa_dec __ARGS((char_u **p)); | |
| 178 static int nfa_regatom __ARGS((void)); | |
| 179 static int nfa_regpiece __ARGS((void)); | |
| 180 static int nfa_regconcat __ARGS((void)); | |
| 181 static int nfa_regbranch __ARGS((void)); | |
| 182 static int nfa_reg __ARGS((int paren)); | |
| 183 #ifdef DEBUG | |
| 184 static void nfa_set_code __ARGS((int c)); | |
| 185 static void nfa_postfix_dump __ARGS((char_u *expr, int retval)); | |
|
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
186 static void nfa_print_state __ARGS((FILE *debugf, nfa_state_T *state)); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
187 static void nfa_print_state2 __ARGS((FILE *debugf, nfa_state_T *state, garray_T *indent)); |
| 4444 | 188 static void nfa_dump __ARGS((nfa_regprog_T *prog)); |
| 189 #endif | |
| 190 static int *re2post __ARGS((void)); | |
| 191 static nfa_state_T *new_state __ARGS((int c, nfa_state_T *out, nfa_state_T *out1)); | |
| 192 static nfa_state_T *post2nfa __ARGS((int *postfix, int *end, int nfa_calc_size)); | |
| 193 static int check_char_class __ARGS((int class, int c)); | |
| 194 static void st_error __ARGS((int *postfix, int *end, int *p)); | |
| 195 static void nfa_save_listids __ARGS((nfa_state_T *start, int *list)); | |
| 196 static void nfa_restore_listids __ARGS((nfa_state_T *start, int *list)); | |
| 197 static void nfa_set_null_listids __ARGS((nfa_state_T *start)); | |
| 198 static void nfa_set_neg_listids __ARGS((nfa_state_T *start)); | |
| 199 static long nfa_regtry __ARGS((nfa_state_T *start, colnr_T col)); | |
| 200 static long nfa_regexec_both __ARGS((char_u *line, colnr_T col)); | |
| 201 static regprog_T *nfa_regcomp __ARGS((char_u *expr, int re_flags)); | |
| 202 static int nfa_regexec __ARGS((regmatch_T *rmp, char_u *line, colnr_T col)); | |
| 203 static long nfa_regexec_multi __ARGS((regmmatch_T *rmp, win_T *win, buf_T *buf, linenr_T lnum, colnr_T col, proftime_T *tm)); | |
| 204 | |
| 205 /* helper functions used when doing re2post() ... regatom() parsing */ | |
| 206 #define EMIT(c) do { \ | |
| 207 if (post_ptr >= post_end) \ | |
| 208 return FAIL; \ | |
| 209 *post_ptr++ = c; \ | |
| 210 } while (0) | |
| 211 | |
| 212 /* | |
| 213 * Initialize internal variables before NFA compilation. | |
| 214 * Return OK on success, FAIL otherwise. | |
| 215 */ | |
| 216 static int | |
| 217 nfa_regcomp_start(expr, re_flags) | |
| 218 char_u *expr; | |
| 219 int re_flags; /* see vim_regcomp() */ | |
| 220 { | |
| 4458 | 221 size_t postfix_size; |
| 4444 | 222 |
| 223 nstate = 0; | |
| 224 istate = 0; | |
| 225 /* A reasonable estimation for size */ | |
| 4458 | 226 nstate_max = (int)(STRLEN(expr) + 1) * NFA_POSTFIX_MULTIPLIER; |
| 4444 | 227 |
| 4454 | 228 /* Some items blow up in size, such as [A-z]. Add more space for that. |
| 229 * TODO: some patterns may still fail. */ | |
| 4458 | 230 nstate_max += 1000; |
| 4454 | 231 |
| 232 /* Size for postfix representation of expr. */ | |
| 4444 | 233 postfix_size = sizeof(*post_start) * nstate_max; |
| 4454 | 234 |
| 4444 | 235 post_start = (int *)lalloc(postfix_size, TRUE); |
| 236 if (post_start == NULL) | |
| 237 return FAIL; | |
| 238 vim_memset(post_start, 0, postfix_size); | |
| 239 post_ptr = post_start; | |
| 4454 | 240 post_end = post_start + nstate_max; |
| 4444 | 241 nfa_has_zend = FALSE; |
| 242 | |
| 243 regcomp_start(expr, re_flags); | |
| 244 | |
| 245 return OK; | |
| 246 } | |
| 247 | |
| 248 /* | |
| 249 * Search between "start" and "end" and try to recognize a | |
| 250 * character class in expanded form. For example [0-9]. | |
| 251 * On success, return the id the character class to be emitted. | |
| 252 * On failure, return 0 (=FAIL) | |
| 253 * Start points to the first char of the range, while end should point | |
| 254 * to the closing brace. | |
| 255 */ | |
| 256 static int | |
| 257 nfa_recognize_char_class(start, end, extra_newl) | |
| 258 char_u *start; | |
| 259 char_u *end; | |
| 260 int extra_newl; | |
| 261 { | |
| 262 int i; | |
| 263 /* Each of these variables takes up a char in "config[]", | |
| 264 * in the order they are here. */ | |
| 265 int not = FALSE, af = FALSE, AF = FALSE, az = FALSE, AZ = FALSE, | |
| 266 o7 = FALSE, o9 = FALSE, underscore = FALSE, newl = FALSE; | |
| 267 char_u *p; | |
| 268 #define NCONFIGS 16 | |
| 269 int classid[NCONFIGS] = { | |
| 270 NFA_DIGIT, NFA_NDIGIT, NFA_HEX, NFA_NHEX, | |
| 271 NFA_OCTAL, NFA_NOCTAL, NFA_WORD, NFA_NWORD, | |
| 272 NFA_HEAD, NFA_NHEAD, NFA_ALPHA, NFA_NALPHA, | |
| 273 NFA_LOWER, NFA_NLOWER, NFA_UPPER, NFA_NUPPER | |
| 274 }; | |
| 4450 | 275 char_u myconfig[10]; |
| 4444 | 276 char_u config[NCONFIGS][9] = { |
| 277 "000000100", /* digit */ | |
| 278 "100000100", /* non digit */ | |
| 279 "011000100", /* hex-digit */ | |
| 280 "111000100", /* non hex-digit */ | |
| 281 "000001000", /* octal-digit */ | |
| 282 "100001000", /* [^0-7] */ | |
| 283 "000110110", /* [0-9A-Za-z_] */ | |
| 284 "100110110", /* [^0-9A-Za-z_] */ | |
| 285 "000110010", /* head of word */ | |
| 286 "100110010", /* not head of word */ | |
| 287 "000110000", /* alphabetic char a-z */ | |
| 288 "100110000", /* non alphabetic char */ | |
| 289 "000100000", /* lowercase letter */ | |
| 290 "100100000", /* non lowercase */ | |
| 291 "000010000", /* uppercase */ | |
| 292 "100010000" /* non uppercase */ | |
| 293 }; | |
| 294 | |
| 295 if (extra_newl == TRUE) | |
| 296 newl = TRUE; | |
| 297 | |
| 298 if (*end != ']') | |
| 299 return FAIL; | |
| 300 p = start; | |
| 301 if (*p == '^') | |
| 302 { | |
| 303 not = TRUE; | |
| 304 p ++; | |
| 305 } | |
| 306 | |
| 307 while (p < end) | |
| 308 { | |
| 309 if (p + 2 < end && *(p + 1) == '-') | |
| 310 { | |
| 311 switch (*p) | |
| 312 { | |
| 313 case '0': | |
| 314 if (*(p + 2) == '9') | |
| 315 { | |
| 316 o9 = TRUE; | |
| 317 break; | |
| 318 } | |
| 319 else | |
| 320 if (*(p + 2) == '7') | |
| 321 { | |
| 322 o7 = TRUE; | |
| 323 break; | |
| 324 } | |
| 325 case 'a': | |
| 326 if (*(p + 2) == 'z') | |
| 327 { | |
| 328 az = TRUE; | |
| 329 break; | |
| 330 } | |
| 331 else | |
| 332 if (*(p + 2) == 'f') | |
| 333 { | |
| 334 af = TRUE; | |
| 335 break; | |
| 336 } | |
| 337 case 'A': | |
| 338 if (*(p + 2) == 'Z') | |
| 339 { | |
| 340 AZ = TRUE; | |
| 341 break; | |
| 342 } | |
| 343 else | |
| 344 if (*(p + 2) == 'F') | |
| 345 { | |
| 346 AF = TRUE; | |
| 347 break; | |
| 348 } | |
| 349 /* FALLTHROUGH */ | |
| 350 default: | |
| 351 return FAIL; | |
| 352 } | |
| 353 p += 3; | |
| 354 } | |
| 355 else if (p + 1 < end && *p == '\\' && *(p + 1) == 'n') | |
| 356 { | |
| 357 newl = TRUE; | |
| 358 p += 2; | |
| 359 } | |
| 360 else if (*p == '_') | |
| 361 { | |
| 362 underscore = TRUE; | |
| 363 p ++; | |
| 364 } | |
| 365 else if (*p == '\n') | |
| 366 { | |
| 367 newl = TRUE; | |
| 368 p ++; | |
| 369 } | |
| 370 else | |
| 371 return FAIL; | |
| 372 } /* while (p < end) */ | |
| 373 | |
| 374 if (p != end) | |
| 375 return FAIL; | |
| 376 | |
| 377 /* build the config that represents the ranges we gathered */ | |
| 378 STRCPY(myconfig, "000000000"); | |
| 379 if (not == TRUE) | |
| 380 myconfig[0] = '1'; | |
| 381 if (af == TRUE) | |
| 382 myconfig[1] = '1'; | |
| 383 if (AF == TRUE) | |
| 384 myconfig[2] = '1'; | |
| 385 if (az == TRUE) | |
| 386 myconfig[3] = '1'; | |
| 387 if (AZ == TRUE) | |
| 388 myconfig[4] = '1'; | |
| 389 if (o7 == TRUE) | |
| 390 myconfig[5] = '1'; | |
| 391 if (o9 == TRUE) | |
| 392 myconfig[6] = '1'; | |
| 393 if (underscore == TRUE) | |
| 394 myconfig[7] = '1'; | |
| 395 if (newl == TRUE) | |
| 396 { | |
| 397 myconfig[8] = '1'; | |
| 398 extra_newl = ADD_NL; | |
| 399 } | |
| 400 /* try to recognize character classes */ | |
| 401 for (i = 0; i < NCONFIGS; i++) | |
| 4450 | 402 if (STRNCMP(myconfig, config[i], 8) == 0) |
| 4444 | 403 return classid[i] + extra_newl; |
| 404 | |
| 405 /* fallthrough => no success so far */ | |
| 406 return FAIL; | |
| 407 | |
| 408 #undef NCONFIGS | |
| 409 } | |
| 410 | |
| 411 /* | |
| 412 * Produce the bytes for equivalence class "c". | |
| 413 * Currently only handles latin1, latin9 and utf-8. | |
| 414 * Emits bytes in postfix notation: 'a,b,NFA_OR,c,NFA_OR' is | |
| 415 * equivalent to 'a OR b OR c' | |
| 416 * | |
| 417 * NOTE! When changing this function, also update reg_equi_class() | |
| 418 */ | |
| 419 static int | |
| 420 nfa_emit_equi_class(c, neg) | |
| 421 int c; | |
| 422 int neg; | |
| 423 { | |
| 424 int first = TRUE; | |
| 425 int glue = neg == TRUE ? NFA_CONCAT : NFA_OR; | |
| 426 #define EMIT2(c) \ | |
| 427 EMIT(c); \ | |
| 428 if (neg == TRUE) { \ | |
| 429 EMIT(NFA_NOT); \ | |
| 430 } \ | |
| 431 if (first == FALSE) \ | |
| 432 EMIT(glue); \ | |
| 433 else \ | |
| 434 first = FALSE; \ | |
| 435 | |
| 436 #ifdef FEAT_MBYTE | |
| 437 if (enc_utf8 || STRCMP(p_enc, "latin1") == 0 | |
| 438 || STRCMP(p_enc, "iso-8859-15") == 0) | |
| 439 #endif | |
| 440 { | |
| 441 switch (c) | |
| 442 { | |
| 443 case 'A': case '\300': case '\301': case '\302': | |
| 444 case '\303': case '\304': case '\305': | |
| 445 EMIT2('A'); EMIT2('\300'); EMIT2('\301'); | |
| 446 EMIT2('\302'); EMIT2('\303'); EMIT2('\304'); | |
| 447 EMIT2('\305'); | |
| 448 return OK; | |
| 449 | |
| 450 case 'C': case '\307': | |
| 451 EMIT2('C'); EMIT2('\307'); | |
| 452 return OK; | |
| 453 | |
| 454 case 'E': case '\310': case '\311': case '\312': case '\313': | |
| 455 EMIT2('E'); EMIT2('\310'); EMIT2('\311'); | |
| 456 EMIT2('\312'); EMIT2('\313'); | |
| 457 return OK; | |
| 458 | |
| 459 case 'I': case '\314': case '\315': case '\316': case '\317': | |
| 460 EMIT2('I'); EMIT2('\314'); EMIT2('\315'); | |
| 461 EMIT2('\316'); EMIT2('\317'); | |
| 462 return OK; | |
| 463 | |
| 464 case 'N': case '\321': | |
| 465 EMIT2('N'); EMIT2('\321'); | |
| 466 return OK; | |
| 467 | |
| 468 case 'O': case '\322': case '\323': case '\324': case '\325': | |
| 469 case '\326': | |
| 470 EMIT2('O'); EMIT2('\322'); EMIT2('\323'); | |
| 471 EMIT2('\324'); EMIT2('\325'); EMIT2('\326'); | |
| 472 return OK; | |
| 473 | |
| 474 case 'U': case '\331': case '\332': case '\333': case '\334': | |
| 475 EMIT2('U'); EMIT2('\331'); EMIT2('\332'); | |
| 476 EMIT2('\333'); EMIT2('\334'); | |
| 477 return OK; | |
| 478 | |
| 479 case 'Y': case '\335': | |
| 480 EMIT2('Y'); EMIT2('\335'); | |
| 481 return OK; | |
| 482 | |
| 483 case 'a': case '\340': case '\341': case '\342': | |
| 484 case '\343': case '\344': case '\345': | |
| 485 EMIT2('a'); EMIT2('\340'); EMIT2('\341'); | |
| 486 EMIT2('\342'); EMIT2('\343'); EMIT2('\344'); | |
| 487 EMIT2('\345'); | |
| 488 return OK; | |
| 489 | |
| 490 case 'c': case '\347': | |
| 491 EMIT2('c'); EMIT2('\347'); | |
| 492 return OK; | |
| 493 | |
| 494 case 'e': case '\350': case '\351': case '\352': case '\353': | |
| 495 EMIT2('e'); EMIT2('\350'); EMIT2('\351'); | |
| 496 EMIT2('\352'); EMIT2('\353'); | |
| 497 return OK; | |
| 498 | |
| 499 case 'i': case '\354': case '\355': case '\356': case '\357': | |
| 500 EMIT2('i'); EMIT2('\354'); EMIT2('\355'); | |
| 501 EMIT2('\356'); EMIT2('\357'); | |
| 502 return OK; | |
| 503 | |
| 504 case 'n': case '\361': | |
| 505 EMIT2('n'); EMIT2('\361'); | |
| 506 return OK; | |
| 507 | |
| 508 case 'o': case '\362': case '\363': case '\364': case '\365': | |
| 509 case '\366': | |
| 510 EMIT2('o'); EMIT2('\362'); EMIT2('\363'); | |
| 511 EMIT2('\364'); EMIT2('\365'); EMIT2('\366'); | |
| 512 return OK; | |
| 513 | |
| 514 case 'u': case '\371': case '\372': case '\373': case '\374': | |
| 515 EMIT2('u'); EMIT2('\371'); EMIT2('\372'); | |
| 516 EMIT2('\373'); EMIT2('\374'); | |
| 517 return OK; | |
| 518 | |
| 519 case 'y': case '\375': case '\377': | |
| 520 EMIT2('y'); EMIT2('\375'); EMIT2('\377'); | |
| 521 return OK; | |
| 522 | |
| 523 default: | |
| 524 return FAIL; | |
| 525 } | |
| 526 } | |
| 527 | |
| 528 EMIT(c); | |
| 529 return OK; | |
| 530 #undef EMIT2 | |
| 531 } | |
| 532 | |
| 533 /* | |
| 534 * Code to parse regular expression. | |
| 535 * | |
| 536 * We try to reuse parsing functions in regexp.c to | |
| 537 * minimize surprise and keep the syntax consistent. | |
| 538 */ | |
| 539 | |
| 540 /* | |
| 541 * Increments the pointer "p" by one (multi-byte) character. | |
| 542 */ | |
| 543 static void | |
| 544 nfa_inc(p) | |
| 545 char_u **p; | |
| 546 { | |
| 547 #ifdef FEAT_MBYTE | |
| 548 if (has_mbyte) | |
| 549 mb_ptr2char_adv(p); | |
| 550 else | |
| 551 #endif | |
| 552 *p = *p + 1; | |
| 553 } | |
| 554 | |
| 555 /* | |
| 556 * Decrements the pointer "p" by one (multi-byte) character. | |
| 557 */ | |
| 558 static void | |
| 559 nfa_dec(p) | |
| 560 char_u **p; | |
| 561 { | |
| 562 #ifdef FEAT_MBYTE | |
| 563 char_u *p2, *oldp; | |
| 564 | |
| 565 if (has_mbyte) | |
| 566 { | |
| 567 oldp = *p; | |
| 568 /* Try to find the multibyte char that advances to the current | |
| 569 * position. */ | |
| 570 do | |
| 571 { | |
| 572 *p = *p - 1; | |
| 573 p2 = *p; | |
| 574 mb_ptr2char_adv(&p2); | |
| 575 } while (p2 != oldp); | |
| 576 } | |
| 577 #else | |
| 578 *p = *p - 1; | |
| 579 #endif | |
| 580 } | |
| 581 | |
| 582 /* | |
| 583 * Parse the lowest level. | |
| 584 * | |
| 585 * An atom can be one of a long list of items. Many atoms match one character | |
| 586 * in the text. It is often an ordinary character or a character class. | |
| 587 * Braces can be used to make a pattern into an atom. The "\z(\)" construct | |
| 588 * is only for syntax highlighting. | |
| 589 * | |
| 590 * atom ::= ordinary-atom | |
| 591 * or \( pattern \) | |
| 592 * or \%( pattern \) | |
| 593 * or \z( pattern \) | |
| 594 */ | |
| 595 static int | |
| 596 nfa_regatom() | |
| 597 { | |
| 598 int c; | |
| 599 int charclass; | |
| 600 int equiclass; | |
| 601 int collclass; | |
| 602 int got_coll_char; | |
| 603 char_u *p; | |
| 604 char_u *endp; | |
| 605 #ifdef FEAT_MBYTE | |
| 606 char_u *old_regparse = regparse; | |
| 607 int clen; | |
| 608 int i; | |
| 609 #endif | |
| 610 int extra = 0; | |
| 611 int first; | |
| 612 int emit_range; | |
| 613 int negated; | |
| 614 int result; | |
| 615 int startc = -1; | |
| 616 int endc = -1; | |
| 617 int oldstartc = -1; | |
| 618 int cpo_lit; /* 'cpoptions' contains 'l' flag */ | |
| 619 int cpo_bsl; /* 'cpoptions' contains '\' flag */ | |
| 620 int glue; /* ID that will "glue" nodes together */ | |
| 621 | |
| 622 cpo_lit = vim_strchr(p_cpo, CPO_LITERAL) != NULL; | |
| 623 cpo_bsl = vim_strchr(p_cpo, CPO_BACKSL) != NULL; | |
| 624 | |
| 625 c = getchr(); | |
| 626 | |
| 627 #ifdef FEAT_MBYTE | |
| 628 /* clen has the length of the current char, without composing chars */ | |
| 629 clen = (*mb_char2len)(c); | |
| 630 if (has_mbyte && clen > 1) | |
| 631 goto nfa_do_multibyte; | |
| 632 #endif | |
| 633 switch (c) | |
| 634 { | |
| 635 case Magic('^'): | |
| 636 EMIT(NFA_BOL); | |
| 637 break; | |
| 638 | |
| 639 case Magic('$'): | |
| 640 EMIT(NFA_EOL); | |
| 641 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
| 642 had_eol = TRUE; | |
| 643 #endif | |
| 644 break; | |
| 645 | |
| 646 case Magic('<'): | |
| 647 EMIT(NFA_BOW); | |
| 648 break; | |
| 649 | |
| 650 case Magic('>'): | |
| 651 EMIT(NFA_EOW); | |
| 652 break; | |
| 653 | |
| 654 case Magic('_'): | |
| 655 c = no_Magic(getchr()); | |
| 656 if (c == '^') /* "\_^" is start-of-line */ | |
| 657 { | |
| 658 EMIT(NFA_BOL); | |
| 659 break; | |
| 660 } | |
| 661 if (c == '$') /* "\_$" is end-of-line */ | |
| 662 { | |
| 663 EMIT(NFA_EOL); | |
| 664 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
| 665 had_eol = TRUE; | |
| 666 #endif | |
| 667 break; | |
| 668 } | |
| 669 | |
| 670 extra = ADD_NL; | |
| 671 | |
| 672 /* "\_[" is collection plus newline */ | |
| 673 if (c == '[') | |
|
4517
9a2183bd8295
updated for version 7.3.1006
Bram Moolenaar <bram@vim.org>
parents:
4515
diff
changeset
|
674 goto collection; |
| 4444 | 675 |
| 676 /* "\_x" is character class plus newline */ | |
| 677 /*FALLTHROUGH*/ | |
| 678 | |
| 679 /* | |
| 680 * Character classes. | |
| 681 */ | |
| 682 case Magic('.'): | |
| 683 case Magic('i'): | |
| 684 case Magic('I'): | |
| 685 case Magic('k'): | |
| 686 case Magic('K'): | |
| 687 case Magic('f'): | |
| 688 case Magic('F'): | |
| 689 case Magic('p'): | |
| 690 case Magic('P'): | |
| 691 case Magic('s'): | |
| 692 case Magic('S'): | |
| 693 case Magic('d'): | |
| 694 case Magic('D'): | |
| 695 case Magic('x'): | |
| 696 case Magic('X'): | |
| 697 case Magic('o'): | |
| 698 case Magic('O'): | |
| 699 case Magic('w'): | |
| 700 case Magic('W'): | |
| 701 case Magic('h'): | |
| 702 case Magic('H'): | |
| 703 case Magic('a'): | |
| 704 case Magic('A'): | |
| 705 case Magic('l'): | |
| 706 case Magic('L'): | |
| 707 case Magic('u'): | |
| 708 case Magic('U'): | |
| 709 p = vim_strchr(classchars, no_Magic(c)); | |
| 710 if (p == NULL) | |
| 711 { | |
| 712 return FAIL; /* runtime error */ | |
| 713 } | |
| 714 #ifdef FEAT_MBYTE | |
| 715 /* When '.' is followed by a composing char ignore the dot, so that | |
| 716 * the composing char is matched here. */ | |
| 717 if (enc_utf8 && c == Magic('.') && utf_iscomposing(peekchr())) | |
| 718 { | |
| 719 c = getchr(); | |
| 720 goto nfa_do_multibyte; | |
| 721 } | |
| 722 #endif | |
| 723 EMIT(nfa_classcodes[p - classchars]); | |
| 724 if (extra == ADD_NL) | |
| 725 { | |
| 726 EMIT(NFA_NEWL); | |
| 727 EMIT(NFA_OR); | |
| 728 regflags |= RF_HASNL; | |
| 729 } | |
| 730 break; | |
| 731 | |
| 732 case Magic('n'): | |
| 733 if (reg_string) | |
| 734 /* In a string "\n" matches a newline character. */ | |
| 735 EMIT(NL); | |
| 736 else | |
| 737 { | |
| 738 /* In buffer text "\n" matches the end of a line. */ | |
| 739 EMIT(NFA_NEWL); | |
| 740 regflags |= RF_HASNL; | |
| 741 } | |
| 742 break; | |
| 743 | |
| 744 case Magic('('): | |
| 745 if (nfa_reg(REG_PAREN) == FAIL) | |
| 746 return FAIL; /* cascaded error */ | |
| 747 break; | |
| 748 | |
| 749 case NUL: | |
| 750 syntax_error = TRUE; | |
| 751 EMSG_RET_FAIL(_("E865: (NFA) Regexp end encountered prematurely")); | |
| 752 | |
| 753 case Magic('|'): | |
| 754 case Magic('&'): | |
| 755 case Magic(')'): | |
| 756 syntax_error = TRUE; | |
| 4450 | 757 EMSGN(_(e_misplaced), no_Magic(c)); |
| 4444 | 758 return FAIL; |
| 759 | |
| 760 case Magic('='): | |
| 761 case Magic('?'): | |
| 762 case Magic('+'): | |
| 763 case Magic('@'): | |
| 764 case Magic('*'): | |
| 765 case Magic('{'): | |
| 766 /* these should follow an atom, not form an atom */ | |
| 767 syntax_error = TRUE; | |
| 4450 | 768 EMSGN(_(e_misplaced), no_Magic(c)); |
| 4444 | 769 return FAIL; |
| 770 | |
| 771 case Magic('~'): /* previous substitute pattern */ | |
| 772 /* Not supported yet */ | |
| 773 return FAIL; | |
| 774 | |
| 775 case Magic('1'): | |
| 776 case Magic('2'): | |
| 777 case Magic('3'): | |
| 778 case Magic('4'): | |
| 779 case Magic('5'): | |
| 780 case Magic('6'): | |
| 781 case Magic('7'): | |
| 782 case Magic('8'): | |
| 783 case Magic('9'): | |
| 784 /* not supported yet */ | |
| 785 return FAIL; | |
| 786 | |
| 787 case Magic('z'): | |
| 788 c = no_Magic(getchr()); | |
| 789 switch (c) | |
| 790 { | |
| 791 case 's': | |
| 792 EMIT(NFA_ZSTART); | |
| 793 break; | |
| 794 case 'e': | |
| 795 EMIT(NFA_ZEND); | |
| 796 nfa_has_zend = TRUE; | |
| 797 /* TODO: Currently \ze does not work properly. */ | |
| 798 return FAIL; | |
| 799 /* break; */ | |
| 800 case '1': | |
| 801 case '2': | |
| 802 case '3': | |
| 803 case '4': | |
| 804 case '5': | |
| 805 case '6': | |
| 806 case '7': | |
| 807 case '8': | |
| 808 case '9': | |
| 809 case '(': | |
| 810 /* \z1...\z9 and \z( not yet supported */ | |
| 811 return FAIL; | |
| 812 default: | |
| 813 syntax_error = TRUE; | |
| 4450 | 814 EMSGN(_("E867: (NFA) Unknown operator '\\z%c'"), |
| 4444 | 815 no_Magic(c)); |
| 816 return FAIL; | |
| 817 } | |
| 818 break; | |
| 819 | |
| 820 case Magic('%'): | |
| 821 c = no_Magic(getchr()); | |
| 822 switch (c) | |
| 823 { | |
| 824 /* () without a back reference */ | |
| 825 case '(': | |
| 826 if (nfa_reg(REG_NPAREN) == FAIL) | |
| 827 return FAIL; | |
| 828 EMIT(NFA_NOPEN); | |
| 829 break; | |
| 830 | |
| 831 case 'd': /* %d123 decimal */ | |
| 832 case 'o': /* %o123 octal */ | |
| 833 case 'x': /* %xab hex 2 */ | |
| 834 case 'u': /* %uabcd hex 4 */ | |
| 835 case 'U': /* %U1234abcd hex 8 */ | |
| 836 /* Not yet supported */ | |
| 837 return FAIL; | |
| 838 | |
| 839 c = coll_get_char(); | |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
840 EMIT(c); |
| 4444 | 841 break; |
| 842 | |
| 843 /* Catch \%^ and \%$ regardless of where they appear in the | |
| 844 * pattern -- regardless of whether or not it makes sense. */ | |
| 845 case '^': | |
| 846 EMIT(NFA_BOF); | |
| 847 /* Not yet supported */ | |
| 848 return FAIL; | |
| 849 break; | |
| 850 | |
| 851 case '$': | |
| 852 EMIT(NFA_EOF); | |
| 853 /* Not yet supported */ | |
| 854 return FAIL; | |
| 855 break; | |
| 856 | |
| 857 case '#': | |
| 858 /* not supported yet */ | |
| 859 return FAIL; | |
| 860 break; | |
| 861 | |
| 862 case 'V': | |
| 863 /* not supported yet */ | |
| 864 return FAIL; | |
| 865 break; | |
| 866 | |
| 867 case '[': | |
| 868 /* \%[abc] not supported yet */ | |
| 869 return FAIL; | |
| 870 | |
| 871 default: | |
| 872 /* not supported yet */ | |
| 873 return FAIL; | |
| 874 } | |
| 875 break; | |
| 876 | |
| 877 case Magic('['): | |
|
4517
9a2183bd8295
updated for version 7.3.1006
Bram Moolenaar <bram@vim.org>
parents:
4515
diff
changeset
|
878 collection: |
| 4444 | 879 /* |
| 880 * Glue is emitted between several atoms from the []. | |
| 881 * It is either NFA_OR, or NFA_CONCAT. | |
| 882 * | |
| 883 * [abc] expands to 'a b NFA_OR c NFA_OR' (in postfix notation) | |
| 884 * [^abc] expands to 'a NFA_NOT b NFA_NOT NFA_CONCAT c NFA_NOT | |
| 885 * NFA_CONCAT NFA_END_NEG_RANGE NFA_CONCAT' (in postfix | |
| 886 * notation) | |
| 887 * | |
| 888 */ | |
| 889 | |
| 890 | |
| 891 /* Emit negation atoms, if needed. | |
| 892 * The CONCAT below merges the NOT with the previous node. */ | |
| 893 #define TRY_NEG() \ | |
| 894 if (negated == TRUE) \ | |
| 895 { \ | |
| 896 EMIT(NFA_NOT); \ | |
| 897 } | |
| 898 | |
| 899 /* Emit glue between important nodes : CONCAT or OR. */ | |
| 900 #define EMIT_GLUE() \ | |
| 901 if (first == FALSE) \ | |
| 902 EMIT(glue); \ | |
| 903 else \ | |
| 904 first = FALSE; | |
| 905 | |
| 906 p = regparse; | |
| 907 endp = skip_anyof(p); | |
| 908 if (*endp == ']') | |
| 909 { | |
| 910 /* | |
| 911 * Try to reverse engineer character classes. For example, | |
| 912 * recognize that [0-9] stands for \d and [A-Za-z_] with \h, | |
| 913 * and perform the necessary substitutions in the NFA. | |
| 914 */ | |
| 915 result = nfa_recognize_char_class(regparse, endp, | |
| 916 extra == ADD_NL); | |
| 917 if (result != FAIL) | |
| 918 { | |
| 919 if (result >= NFA_DIGIT && result <= NFA_NUPPER) | |
| 920 EMIT(result); | |
| 921 else /* must be char class + newline */ | |
| 922 { | |
| 923 EMIT(result - ADD_NL); | |
| 924 EMIT(NFA_NEWL); | |
| 925 EMIT(NFA_OR); | |
| 926 } | |
| 927 regparse = endp; | |
| 928 nfa_inc(®parse); | |
| 929 return OK; | |
| 930 } | |
| 931 /* | |
| 932 * Failed to recognize a character class. Use the simple | |
| 933 * version that turns [abc] into 'a' OR 'b' OR 'c' | |
| 934 */ | |
| 935 startc = endc = oldstartc = -1; | |
| 936 first = TRUE; /* Emitting first atom in this sequence? */ | |
| 937 negated = FALSE; | |
| 938 glue = NFA_OR; | |
| 939 if (*regparse == '^') /* negated range */ | |
| 940 { | |
| 941 negated = TRUE; | |
| 942 glue = NFA_CONCAT; | |
| 943 nfa_inc(®parse); | |
| 944 } | |
| 945 if (*regparse == '-') | |
| 946 { | |
| 947 startc = '-'; | |
| 948 EMIT(startc); | |
| 949 TRY_NEG(); | |
| 950 EMIT_GLUE(); | |
| 951 nfa_inc(®parse); | |
| 952 } | |
| 953 /* Emit the OR branches for each character in the [] */ | |
| 954 emit_range = FALSE; | |
| 955 while (regparse < endp) | |
| 956 { | |
| 957 oldstartc = startc; | |
| 958 startc = -1; | |
| 959 got_coll_char = FALSE; | |
| 960 if (*regparse == '[') | |
| 961 { | |
| 962 /* Check for [: :], [= =], [. .] */ | |
| 963 equiclass = collclass = 0; | |
| 964 charclass = get_char_class(®parse); | |
| 965 if (charclass == CLASS_NONE) | |
| 966 { | |
| 967 equiclass = get_equi_class(®parse); | |
| 968 if (equiclass == 0) | |
| 969 collclass = get_coll_element(®parse); | |
| 970 } | |
| 971 | |
| 972 /* Character class like [:alpha:] */ | |
| 973 if (charclass != CLASS_NONE) | |
| 974 { | |
| 975 switch (charclass) | |
| 976 { | |
| 977 case CLASS_ALNUM: | |
| 978 EMIT(NFA_CLASS_ALNUM); | |
| 979 break; | |
| 980 case CLASS_ALPHA: | |
| 981 EMIT(NFA_CLASS_ALPHA); | |
| 982 break; | |
| 983 case CLASS_BLANK: | |
| 984 EMIT(NFA_CLASS_BLANK); | |
| 985 break; | |
| 986 case CLASS_CNTRL: | |
| 987 EMIT(NFA_CLASS_CNTRL); | |
| 988 break; | |
| 989 case CLASS_DIGIT: | |
| 990 EMIT(NFA_CLASS_DIGIT); | |
| 991 break; | |
| 992 case CLASS_GRAPH: | |
| 993 EMIT(NFA_CLASS_GRAPH); | |
| 994 break; | |
| 995 case CLASS_LOWER: | |
| 996 EMIT(NFA_CLASS_LOWER); | |
| 997 break; | |
| 998 case CLASS_PRINT: | |
| 999 EMIT(NFA_CLASS_PRINT); | |
| 1000 break; | |
| 1001 case CLASS_PUNCT: | |
| 1002 EMIT(NFA_CLASS_PUNCT); | |
| 1003 break; | |
| 1004 case CLASS_SPACE: | |
| 1005 EMIT(NFA_CLASS_SPACE); | |
| 1006 break; | |
| 1007 case CLASS_UPPER: | |
| 1008 EMIT(NFA_CLASS_UPPER); | |
| 1009 break; | |
| 1010 case CLASS_XDIGIT: | |
| 1011 EMIT(NFA_CLASS_XDIGIT); | |
| 1012 break; | |
| 1013 case CLASS_TAB: | |
| 1014 EMIT(NFA_CLASS_TAB); | |
| 1015 break; | |
| 1016 case CLASS_RETURN: | |
| 1017 EMIT(NFA_CLASS_RETURN); | |
| 1018 break; | |
| 1019 case CLASS_BACKSPACE: | |
| 1020 EMIT(NFA_CLASS_BACKSPACE); | |
| 1021 break; | |
| 1022 case CLASS_ESCAPE: | |
| 1023 EMIT(NFA_CLASS_ESCAPE); | |
| 1024 break; | |
| 1025 } | |
| 1026 TRY_NEG(); | |
| 1027 EMIT_GLUE(); | |
| 1028 continue; | |
| 1029 } | |
| 1030 /* Try equivalence class [=a=] and the like */ | |
| 1031 if (equiclass != 0) | |
| 1032 { | |
| 1033 result = nfa_emit_equi_class(equiclass, negated); | |
| 1034 if (result == FAIL) | |
| 1035 { | |
| 1036 /* should never happen */ | |
| 1037 EMSG_RET_FAIL(_("E868: Error building NFA with equivalence class!")); | |
| 1038 } | |
| 1039 EMIT_GLUE(); | |
| 1040 continue; | |
| 1041 } | |
| 1042 /* Try collating class like [. .] */ | |
| 1043 if (collclass != 0) | |
| 1044 { | |
| 1045 startc = collclass; /* allow [.a.]-x as a range */ | |
| 1046 /* Will emit the proper atom at the end of the | |
| 1047 * while loop. */ | |
| 1048 } | |
| 1049 } | |
| 1050 /* Try a range like 'a-x' or '\t-z' */ | |
| 1051 if (*regparse == '-') | |
| 1052 { | |
| 1053 emit_range = TRUE; | |
| 1054 startc = oldstartc; | |
| 1055 nfa_inc(®parse); | |
| 1056 continue; /* reading the end of the range */ | |
| 1057 } | |
| 1058 | |
| 1059 /* Now handle simple and escaped characters. | |
| 1060 * Only "\]", "\^", "\]" and "\\" are special in Vi. Vim | |
| 1061 * accepts "\t", "\e", etc., but only when the 'l' flag in | |
| 1062 * 'cpoptions' is not included. | |
| 1063 * Posix doesn't recognize backslash at all. | |
| 1064 */ | |
| 1065 if (*regparse == '\\' | |
| 1066 && !cpo_bsl | |
| 1067 && regparse + 1 <= endp | |
| 1068 && (vim_strchr(REGEXP_INRANGE, regparse[1]) != NULL | |
| 1069 || (!cpo_lit | |
| 1070 && vim_strchr(REGEXP_ABBR, regparse[1]) | |
| 1071 != NULL) | |
| 1072 ) | |
| 1073 ) | |
| 1074 { | |
| 1075 nfa_inc(®parse); | |
| 1076 | |
|
4507
9dbbddb2ed10
updated for version 7.3.1001
Bram Moolenaar <bram@vim.org>
parents:
4503
diff
changeset
|
1077 if (*regparse == 'n') |
| 4444 | 1078 startc = reg_string ? NL : NFA_NEWL; |
| 1079 else | |
| 1080 if (*regparse == 'd' | |
| 1081 || *regparse == 'o' | |
| 1082 || *regparse == 'x' | |
| 1083 || *regparse == 'u' | |
| 1084 || *regparse == 'U' | |
| 1085 ) | |
| 1086 { | |
| 1087 /* TODO(RE) This needs more testing */ | |
| 1088 startc = coll_get_char(); | |
| 1089 got_coll_char = TRUE; | |
| 1090 nfa_dec(®parse); | |
| 1091 } | |
| 1092 else | |
| 1093 { | |
| 1094 /* \r,\t,\e,\b */ | |
| 1095 startc = backslash_trans(*regparse); | |
| 1096 } | |
| 1097 } | |
| 1098 | |
| 1099 /* Normal printable char */ | |
| 1100 if (startc == -1) | |
| 1101 #ifdef FEAT_MBYTE | |
| 1102 startc = (*mb_ptr2char)(regparse); | |
| 1103 #else | |
| 1104 startc = *regparse; | |
| 1105 #endif | |
| 1106 | |
| 1107 /* Previous char was '-', so this char is end of range. */ | |
| 1108 if (emit_range) | |
| 1109 { | |
| 1110 endc = startc; startc = oldstartc; | |
| 1111 if (startc > endc) | |
| 1112 EMSG_RET_FAIL(_(e_invrange)); | |
| 1113 #ifdef FEAT_MBYTE | |
| 1114 if (has_mbyte && ((*mb_char2len)(startc) > 1 | |
| 1115 || (*mb_char2len)(endc) > 1)) | |
| 1116 { | |
| 1117 if (endc > startc + 256) | |
| 1118 EMSG_RET_FAIL(_(e_invrange)); | |
| 1119 /* Emit the range. "startc" was already emitted, so | |
| 1120 * skip it. */ | |
| 1121 for (c = startc + 1; c <= endc; c++) | |
| 1122 { | |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1123 EMIT(c); |
| 4444 | 1124 TRY_NEG(); |
| 1125 EMIT_GLUE(); | |
| 1126 } | |
| 1127 emit_range = FALSE; | |
| 1128 } | |
| 1129 else | |
| 1130 #endif | |
| 1131 { | |
| 1132 #ifdef EBCDIC | |
| 1133 int alpha_only = FALSE; | |
| 1134 | |
| 1135 /* for alphabetical range skip the gaps | |
| 1136 * 'i'-'j', 'r'-'s', 'I'-'J' and 'R'-'S'. */ | |
| 1137 if (isalpha(startc) && isalpha(endc)) | |
| 1138 alpha_only = TRUE; | |
| 1139 #endif | |
| 1140 /* Emit the range. "startc" was already emitted, so | |
| 1141 * skip it. */ | |
| 1142 for (c = startc + 1; c <= endc; c++) | |
| 1143 #ifdef EBCDIC | |
| 1144 if (!alpha_only || isalpha(startc)) | |
| 1145 #endif | |
| 1146 { | |
| 1147 EMIT(c); | |
| 1148 TRY_NEG(); | |
| 1149 EMIT_GLUE(); | |
| 1150 } | |
| 1151 emit_range = FALSE; | |
| 1152 } | |
| 1153 } | |
| 1154 else | |
| 1155 { | |
| 1156 /* | |
| 1157 * This char (startc) is not part of a range. Just | |
| 1158 * emit it. | |
| 1159 * | |
| 1160 * Normally, simply emit startc. But if we get char | |
| 1161 * code=0 from a collating char, then replace it with | |
| 1162 * 0x0a. | |
| 1163 * | |
| 1164 * This is needed to completely mimic the behaviour of | |
| 1165 * the backtracking engine. | |
| 1166 */ | |
| 1167 if (got_coll_char == TRUE && startc == 0) | |
| 1168 EMIT(0x0a); | |
| 1169 else | |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1170 EMIT(startc); |
| 4444 | 1171 TRY_NEG(); |
| 1172 EMIT_GLUE(); | |
| 1173 } | |
| 1174 | |
| 1175 nfa_inc(®parse); | |
| 1176 } /* while (p < endp) */ | |
| 1177 | |
| 1178 nfa_dec(®parse); | |
| 1179 if (*regparse == '-') /* if last, '-' is just a char */ | |
| 1180 { | |
| 1181 EMIT('-'); | |
| 1182 TRY_NEG(); | |
| 1183 EMIT_GLUE(); | |
| 1184 } | |
| 1185 nfa_inc(®parse); | |
| 1186 | |
| 1187 if (extra == ADD_NL) /* \_[] also matches \n */ | |
| 1188 { | |
| 1189 EMIT(reg_string ? NL : NFA_NEWL); | |
| 1190 TRY_NEG(); | |
| 1191 EMIT_GLUE(); | |
| 1192 } | |
| 1193 | |
| 1194 /* skip the trailing ] */ | |
| 1195 regparse = endp; | |
| 1196 nfa_inc(®parse); | |
| 1197 if (negated == TRUE) | |
| 1198 { | |
| 1199 /* Mark end of negated char range */ | |
| 1200 EMIT(NFA_END_NEG_RANGE); | |
| 1201 EMIT(NFA_CONCAT); | |
| 1202 } | |
| 1203 return OK; | |
|
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
1204 } /* if exists closing ] */ |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
1205 |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
1206 if (reg_strict) |
| 4444 | 1207 { |
| 1208 syntax_error = TRUE; | |
| 1209 EMSG_RET_FAIL(_(e_missingbracket)); | |
| 1210 } | |
|
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
1211 /* FALLTHROUGH */ |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
1212 |
| 4444 | 1213 default: |
| 1214 { | |
| 1215 #ifdef FEAT_MBYTE | |
| 1216 int plen; | |
| 1217 | |
| 1218 nfa_do_multibyte: | |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1219 /* Length of current char with composing chars. */ |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1220 if (enc_utf8 && clen != (plen = (*mb_ptr2len)(old_regparse))) |
| 4444 | 1221 { |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1222 /* A base character plus composing characters. |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1223 * This requires creating a separate atom as if enclosing |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1224 * the characters in (), where NFA_COMPOSING is the ( and |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1225 * NFA_END_COMPOSING is the ). Note that right now we are |
| 4444 | 1226 * building the postfix form, not the NFA itself; |
| 1227 * a composing char could be: a, b, c, NFA_COMPOSING | |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1228 * where 'b' and 'c' are chars with codes > 256. */ |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1229 i = 0; |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1230 for (;;) |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1231 { |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1232 EMIT(c); |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1233 if (i > 0) |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1234 EMIT(NFA_CONCAT); |
|
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
1235 if ((i += utf_char2len(c)) >= plen) |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1236 break; |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1237 c = utf_ptr2char(old_regparse + i); |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1238 } |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1239 EMIT(NFA_COMPOSING); |
| 4444 | 1240 regparse = old_regparse + plen; |
| 1241 } | |
| 1242 else | |
| 1243 #endif | |
| 1244 { | |
| 1245 c = no_Magic(c); | |
| 1246 EMIT(c); | |
| 1247 } | |
| 1248 return OK; | |
| 1249 } | |
| 1250 } | |
| 1251 | |
| 1252 #undef TRY_NEG | |
| 1253 #undef EMIT_GLUE | |
| 1254 | |
| 1255 return OK; | |
| 1256 } | |
| 1257 | |
| 1258 /* | |
| 1259 * Parse something followed by possible [*+=]. | |
| 1260 * | |
| 1261 * A piece is an atom, possibly followed by a multi, an indication of how many | |
| 1262 * times the atom can be matched. Example: "a*" matches any sequence of "a" | |
| 1263 * characters: "", "a", "aa", etc. | |
| 1264 * | |
| 1265 * piece ::= atom | |
| 1266 * or atom multi | |
| 1267 */ | |
| 1268 static int | |
| 1269 nfa_regpiece() | |
| 1270 { | |
| 1271 int i; | |
| 1272 int op; | |
| 1273 int ret; | |
| 1274 long minval, maxval; | |
| 1275 int greedy = TRUE; /* Braces are prefixed with '-' ? */ | |
| 1276 char_u *old_regparse, *new_regparse; | |
| 1277 int c2; | |
| 1278 int *old_post_ptr, *my_post_start; | |
| 1279 int old_regnpar; | |
| 1280 int quest; | |
| 1281 | |
| 1282 /* Save the current position in the regexp, so that we can use it if | |
| 1283 * <atom>{m,n} is next. */ | |
| 1284 old_regparse = regparse; | |
| 1285 /* Save current number of open parenthesis, so we can use it if | |
| 1286 * <atom>{m,n} is next */ | |
| 1287 old_regnpar = regnpar; | |
| 1288 /* store current pos in the postfix form, for \{m,n} involving 0s */ | |
| 1289 my_post_start = post_ptr; | |
| 1290 | |
| 1291 ret = nfa_regatom(); | |
| 1292 if (ret == FAIL) | |
| 1293 return FAIL; /* cascaded error */ | |
| 1294 | |
| 1295 op = peekchr(); | |
| 1296 if (re_multi_type(op) == NOT_MULTI) | |
| 1297 return OK; | |
| 1298 | |
| 1299 skipchr(); | |
| 1300 switch (op) | |
| 1301 { | |
| 1302 case Magic('*'): | |
| 1303 EMIT(NFA_STAR); | |
| 1304 break; | |
| 1305 | |
| 1306 case Magic('+'): | |
| 1307 /* | |
| 1308 * Trick: Normally, (a*)\+ would match the whole input "aaa". The | |
| 1309 * first and only submatch would be "aaa". But the backtracking | |
| 1310 * engine interprets the plus as "try matching one more time", and | |
| 1311 * a* matches a second time at the end of the input, the empty | |
| 1312 * string. | |
| 1313 * The submatch will the empty string. | |
| 1314 * | |
| 1315 * In order to be consistent with the old engine, we disable | |
| 1316 * NFA_PLUS, and replace <atom>+ with <atom><atom>* | |
| 1317 */ | |
| 1318 /* EMIT(NFA_PLUS); */ | |
| 1319 regnpar = old_regnpar; | |
| 1320 regparse = old_regparse; | |
| 1321 curchr = -1; | |
| 1322 if (nfa_regatom() == FAIL) | |
| 1323 return FAIL; | |
| 1324 EMIT(NFA_STAR); | |
| 1325 EMIT(NFA_CONCAT); | |
| 1326 skipchr(); /* skip the \+ */ | |
| 1327 break; | |
| 1328 | |
| 1329 case Magic('@'): | |
| 1330 op = no_Magic(getchr()); | |
| 1331 switch(op) | |
| 1332 { | |
| 1333 case '=': | |
| 1334 EMIT(NFA_PREV_ATOM_NO_WIDTH); | |
| 1335 break; | |
| 1336 case '!': | |
| 1337 case '<': | |
| 1338 case '>': | |
| 1339 /* Not supported yet */ | |
| 1340 return FAIL; | |
| 1341 default: | |
| 1342 syntax_error = TRUE; | |
| 4450 | 1343 EMSGN(_("E869: (NFA) Unknown operator '\\@%c'"), op); |
| 4444 | 1344 return FAIL; |
| 1345 } | |
| 1346 break; | |
| 1347 | |
| 1348 case Magic('?'): | |
| 1349 case Magic('='): | |
| 1350 EMIT(NFA_QUEST); | |
| 1351 break; | |
| 1352 | |
| 1353 case Magic('{'): | |
| 1354 /* a{2,5} will expand to 'aaa?a?a?' | |
| 1355 * a{-1,3} will expand to 'aa??a??', where ?? is the nongreedy | |
| 1356 * version of '?' | |
| 1357 * \v(ab){2,3} will expand to '(ab)(ab)(ab)?', where all the | |
| 1358 * parenthesis have the same id | |
| 1359 */ | |
| 1360 | |
| 1361 greedy = TRUE; | |
| 1362 c2 = peekchr(); | |
| 1363 if (c2 == '-' || c2 == Magic('-')) | |
| 1364 { | |
| 1365 skipchr(); | |
| 1366 greedy = FALSE; | |
| 1367 } | |
| 1368 if (!read_limits(&minval, &maxval)) | |
| 1369 { | |
| 1370 syntax_error = TRUE; | |
| 1371 EMSG_RET_FAIL(_("E870: (NFA regexp) Error reading repetition limits")); | |
| 1372 } | |
| 1373 /* <atom>{0,inf}, <atom>{0,} and <atom>{} are equivalent to | |
| 1374 * <atom>* */ | |
| 1375 if (minval == 0 && maxval == MAX_LIMIT && greedy) | |
| 1376 { | |
| 1377 EMIT(NFA_STAR); | |
| 1378 break; | |
| 1379 } | |
| 1380 | |
| 1381 if (maxval > NFA_BRACES_MAXLIMIT) | |
| 1382 { | |
| 1383 /* This would yield a huge automaton and use too much memory. | |
| 1384 * Revert to old engine */ | |
| 1385 return FAIL; | |
| 1386 } | |
| 1387 | |
| 1388 /* Special case: x{0} or x{-0} */ | |
| 1389 if (maxval == 0) | |
| 1390 { | |
| 1391 /* Ignore result of previous call to nfa_regatom() */ | |
| 1392 post_ptr = my_post_start; | |
| 1393 /* NFA_SKIP_CHAR has 0-length and works everywhere */ | |
| 1394 EMIT(NFA_SKIP_CHAR); | |
| 1395 return OK; | |
| 1396 } | |
| 1397 | |
| 1398 /* Ignore previous call to nfa_regatom() */ | |
| 1399 post_ptr = my_post_start; | |
| 1400 /* Save pos after the repeated atom and the \{} */ | |
| 1401 new_regparse = regparse; | |
| 1402 | |
| 1403 new_regparse = regparse; | |
| 1404 quest = (greedy == TRUE? NFA_QUEST : NFA_QUEST_NONGREEDY); | |
| 1405 for (i = 0; i < maxval; i++) | |
| 1406 { | |
| 1407 /* Goto beginning of the repeated atom */ | |
| 1408 regparse = old_regparse; | |
| 1409 curchr = -1; | |
| 1410 /* Restore count of parenthesis */ | |
| 1411 regnpar = old_regnpar; | |
| 1412 old_post_ptr = post_ptr; | |
| 1413 if (nfa_regatom() == FAIL) | |
| 1414 return FAIL; | |
| 1415 /* after "minval" times, atoms are optional */ | |
| 1416 if (i + 1 > minval) | |
| 1417 EMIT(quest); | |
| 1418 if (old_post_ptr != my_post_start) | |
| 1419 EMIT(NFA_CONCAT); | |
| 1420 } | |
| 1421 | |
| 1422 /* Go to just after the repeated atom and the \{} */ | |
| 1423 regparse = new_regparse; | |
| 1424 curchr = -1; | |
| 1425 | |
| 1426 break; | |
| 1427 | |
| 1428 | |
| 1429 default: | |
| 1430 break; | |
| 1431 } /* end switch */ | |
| 1432 | |
| 1433 if (re_multi_type(peekchr()) != NOT_MULTI) | |
| 1434 { | |
| 1435 /* Can't have a multi follow a multi. */ | |
| 1436 syntax_error = TRUE; | |
| 1437 EMSG_RET_FAIL(_("E871: (NFA regexp) Can't have a multi follow a multi !")); | |
| 1438 } | |
| 1439 | |
| 1440 return OK; | |
| 1441 } | |
| 1442 | |
| 1443 /* | |
| 1444 * Parse one or more pieces, concatenated. It matches a match for the | |
| 1445 * first piece, followed by a match for the second piece, etc. Example: | |
| 1446 * "f[0-9]b", first matches "f", then a digit and then "b". | |
| 1447 * | |
| 1448 * concat ::= piece | |
| 1449 * or piece piece | |
| 1450 * or piece piece piece | |
| 1451 * etc. | |
| 1452 */ | |
| 1453 static int | |
| 1454 nfa_regconcat() | |
| 1455 { | |
| 1456 int cont = TRUE; | |
| 1457 int first = TRUE; | |
| 1458 | |
| 1459 while (cont) | |
| 1460 { | |
| 1461 switch (peekchr()) | |
| 1462 { | |
| 1463 case NUL: | |
| 1464 case Magic('|'): | |
| 1465 case Magic('&'): | |
| 1466 case Magic(')'): | |
| 1467 cont = FALSE; | |
| 1468 break; | |
| 1469 | |
| 1470 case Magic('Z'): | |
| 1471 #ifdef FEAT_MBYTE | |
| 1472 regflags |= RF_ICOMBINE; | |
| 1473 #endif | |
| 1474 skipchr_keepstart(); | |
| 1475 break; | |
| 1476 case Magic('c'): | |
| 1477 regflags |= RF_ICASE; | |
| 1478 skipchr_keepstart(); | |
| 1479 break; | |
| 1480 case Magic('C'): | |
| 1481 regflags |= RF_NOICASE; | |
| 1482 skipchr_keepstart(); | |
| 1483 break; | |
| 1484 case Magic('v'): | |
| 1485 reg_magic = MAGIC_ALL; | |
| 1486 skipchr_keepstart(); | |
| 1487 curchr = -1; | |
| 1488 break; | |
| 1489 case Magic('m'): | |
| 1490 reg_magic = MAGIC_ON; | |
| 1491 skipchr_keepstart(); | |
| 1492 curchr = -1; | |
| 1493 break; | |
| 1494 case Magic('M'): | |
| 1495 reg_magic = MAGIC_OFF; | |
| 1496 skipchr_keepstart(); | |
| 1497 curchr = -1; | |
| 1498 break; | |
| 1499 case Magic('V'): | |
| 1500 reg_magic = MAGIC_NONE; | |
| 1501 skipchr_keepstart(); | |
| 1502 curchr = -1; | |
| 1503 break; | |
| 1504 | |
| 1505 default: | |
| 1506 if (nfa_regpiece() == FAIL) | |
| 1507 return FAIL; | |
| 1508 if (first == FALSE) | |
| 1509 EMIT(NFA_CONCAT); | |
| 1510 else | |
| 1511 first = FALSE; | |
| 1512 break; | |
| 1513 } | |
| 1514 } | |
| 1515 | |
| 1516 return OK; | |
| 1517 } | |
| 1518 | |
| 1519 /* | |
| 1520 * Parse a branch, one or more concats, separated by "\&". It matches the | |
| 1521 * last concat, but only if all the preceding concats also match at the same | |
| 1522 * position. Examples: | |
| 1523 * "foobeep\&..." matches "foo" in "foobeep". | |
| 1524 * ".*Peter\&.*Bob" matches in a line containing both "Peter" and "Bob" | |
| 1525 * | |
| 1526 * branch ::= concat | |
| 1527 * or concat \& concat | |
| 1528 * or concat \& concat \& concat | |
| 1529 * etc. | |
| 1530 */ | |
| 1531 static int | |
| 1532 nfa_regbranch() | |
| 1533 { | |
| 1534 int ch; | |
| 1535 int *old_post_ptr; | |
| 1536 | |
| 1537 old_post_ptr = post_ptr; | |
| 1538 | |
| 1539 /* First branch, possibly the only one */ | |
| 1540 if (nfa_regconcat() == FAIL) | |
| 1541 return FAIL; | |
| 1542 | |
| 1543 ch = peekchr(); | |
| 1544 /* Try next concats */ | |
| 1545 while (ch == Magic('&')) | |
| 1546 { | |
| 1547 skipchr(); | |
| 1548 EMIT(NFA_NOPEN); | |
| 1549 EMIT(NFA_PREV_ATOM_NO_WIDTH); | |
| 1550 old_post_ptr = post_ptr; | |
| 1551 if (nfa_regconcat() == FAIL) | |
| 1552 return FAIL; | |
| 1553 /* if concat is empty, skip a input char. But do emit a node */ | |
| 1554 if (old_post_ptr == post_ptr) | |
| 1555 EMIT(NFA_SKIP_CHAR); | |
| 1556 EMIT(NFA_CONCAT); | |
| 1557 ch = peekchr(); | |
| 1558 } | |
| 1559 | |
| 1560 /* Even if a branch is empty, emit one node for it */ | |
| 1561 if (old_post_ptr == post_ptr) | |
| 1562 EMIT(NFA_SKIP_CHAR); | |
| 1563 | |
| 1564 return OK; | |
| 1565 } | |
| 1566 | |
| 1567 /* | |
| 1568 * Parse a pattern, one or more branches, separated by "\|". It matches | |
| 1569 * anything that matches one of the branches. Example: "foo\|beep" matches | |
| 1570 * "foo" and matches "beep". If more than one branch matches, the first one | |
| 1571 * is used. | |
| 1572 * | |
| 1573 * pattern ::= branch | |
| 1574 * or branch \| branch | |
| 1575 * or branch \| branch \| branch | |
| 1576 * etc. | |
| 1577 */ | |
| 1578 static int | |
| 1579 nfa_reg(paren) | |
| 1580 int paren; /* REG_NOPAREN, REG_PAREN, REG_NPAREN or REG_ZPAREN */ | |
| 1581 { | |
| 1582 int parno = 0; | |
| 1583 | |
| 1584 #ifdef FEAT_SYN_HL | |
| 1585 #endif | |
| 1586 if (paren == REG_PAREN) | |
| 1587 { | |
| 1588 if (regnpar >= NSUBEXP) /* Too many `(' */ | |
| 1589 { | |
| 1590 syntax_error = TRUE; | |
| 1591 EMSG_RET_FAIL(_("E872: (NFA regexp) Too many '('")); | |
| 1592 } | |
| 1593 parno = regnpar++; | |
| 1594 } | |
| 1595 | |
| 1596 if (nfa_regbranch() == FAIL) | |
| 1597 return FAIL; /* cascaded error */ | |
| 1598 | |
| 1599 while (peekchr() == Magic('|')) | |
| 1600 { | |
| 1601 skipchr(); | |
| 1602 if (nfa_regbranch() == FAIL) | |
| 1603 return FAIL; /* cascaded error */ | |
| 1604 EMIT(NFA_OR); | |
| 1605 } | |
| 1606 | |
| 1607 /* Check for proper termination. */ | |
| 1608 if (paren != REG_NOPAREN && getchr() != Magic(')')) | |
| 1609 { | |
| 1610 syntax_error = TRUE; | |
| 1611 if (paren == REG_NPAREN) | |
| 1612 EMSG2_RET_FAIL(_(e_unmatchedpp), reg_magic == MAGIC_ALL); | |
| 1613 else | |
| 1614 EMSG2_RET_FAIL(_(e_unmatchedp), reg_magic == MAGIC_ALL); | |
| 1615 } | |
| 1616 else if (paren == REG_NOPAREN && peekchr() != NUL) | |
| 1617 { | |
| 1618 syntax_error = TRUE; | |
| 1619 if (peekchr() == Magic(')')) | |
| 1620 EMSG2_RET_FAIL(_(e_unmatchedpar), reg_magic == MAGIC_ALL); | |
| 1621 else | |
| 1622 EMSG_RET_FAIL(_("E873: (NFA regexp) proper termination error")); | |
| 1623 } | |
| 1624 /* | |
| 1625 * Here we set the flag allowing back references to this set of | |
| 1626 * parentheses. | |
| 1627 */ | |
| 1628 if (paren == REG_PAREN) | |
| 1629 { | |
| 1630 had_endbrace[parno] = TRUE; /* have seen the close paren */ | |
| 1631 EMIT(NFA_MOPEN + parno); | |
| 1632 } | |
| 1633 | |
| 1634 return OK; | |
| 1635 } | |
| 1636 | |
| 1637 typedef struct | |
| 1638 { | |
| 1639 char_u *start[NSUBEXP]; | |
| 1640 char_u *end[NSUBEXP]; | |
| 1641 lpos_T startpos[NSUBEXP]; | |
| 1642 lpos_T endpos[NSUBEXP]; | |
| 1643 } regsub_T; | |
| 1644 | |
| 1645 static int nfa_regmatch __ARGS((nfa_state_T *start, regsub_T *submatch, regsub_T *m)); | |
| 1646 | |
| 1647 #ifdef DEBUG | |
| 1648 static char_u code[50]; | |
| 1649 | |
| 1650 static void | |
| 1651 nfa_set_code(c) | |
| 1652 int c; | |
| 1653 { | |
| 1654 int addnl = FALSE; | |
| 1655 | |
| 1656 if (c >= NFA_FIRST_NL && c <= NFA_LAST_NL) | |
| 1657 { | |
| 1658 addnl = TRUE; | |
| 1659 c -= ADD_NL; | |
| 1660 } | |
| 1661 | |
| 1662 STRCPY(code, ""); | |
| 1663 switch (c) | |
| 1664 { | |
| 1665 case NFA_MATCH: STRCPY(code, "NFA_MATCH "); break; | |
| 1666 case NFA_SPLIT: STRCPY(code, "NFA_SPLIT "); break; | |
| 1667 case NFA_CONCAT: STRCPY(code, "NFA_CONCAT "); break; | |
| 1668 case NFA_NEWL: STRCPY(code, "NFA_NEWL "); break; | |
| 1669 case NFA_ZSTART: STRCPY(code, "NFA_ZSTART"); break; | |
| 1670 case NFA_ZEND: STRCPY(code, "NFA_ZEND"); break; | |
| 1671 | |
| 1672 case NFA_PREV_ATOM_NO_WIDTH: | |
| 1673 STRCPY(code, "NFA_PREV_ATOM_NO_WIDTH"); break; | |
| 1674 case NFA_NOPEN: STRCPY(code, "NFA_MOPEN_INVISIBLE"); break; | |
| 1675 case NFA_NCLOSE: STRCPY(code, "NFA_MCLOSE_INVISIBLE"); break; | |
| 1676 case NFA_START_INVISIBLE: STRCPY(code, "NFA_START_INVISIBLE"); break; | |
| 1677 case NFA_END_INVISIBLE: STRCPY(code, "NFA_END_INVISIBLE"); break; | |
| 1678 | |
| 1679 case NFA_COMPOSING: STRCPY(code, "NFA_COMPOSING"); break; | |
| 1680 case NFA_END_COMPOSING: STRCPY(code, "NFA_END_COMPOSING"); break; | |
| 1681 | |
| 1682 case NFA_MOPEN + 0: | |
| 1683 case NFA_MOPEN + 1: | |
| 1684 case NFA_MOPEN + 2: | |
| 1685 case NFA_MOPEN + 3: | |
| 1686 case NFA_MOPEN + 4: | |
| 1687 case NFA_MOPEN + 5: | |
| 1688 case NFA_MOPEN + 6: | |
| 1689 case NFA_MOPEN + 7: | |
| 1690 case NFA_MOPEN + 8: | |
| 1691 case NFA_MOPEN + 9: | |
| 1692 STRCPY(code, "NFA_MOPEN(x)"); | |
| 1693 code[10] = c - NFA_MOPEN + '0'; | |
| 1694 break; | |
| 1695 case NFA_MCLOSE + 0: | |
| 1696 case NFA_MCLOSE + 1: | |
| 1697 case NFA_MCLOSE + 2: | |
| 1698 case NFA_MCLOSE + 3: | |
| 1699 case NFA_MCLOSE + 4: | |
| 1700 case NFA_MCLOSE + 5: | |
| 1701 case NFA_MCLOSE + 6: | |
| 1702 case NFA_MCLOSE + 7: | |
| 1703 case NFA_MCLOSE + 8: | |
| 1704 case NFA_MCLOSE + 9: | |
| 1705 STRCPY(code, "NFA_MCLOSE(x)"); | |
| 1706 code[11] = c - NFA_MCLOSE + '0'; | |
| 1707 break; | |
| 1708 case NFA_EOL: STRCPY(code, "NFA_EOL "); break; | |
| 1709 case NFA_BOL: STRCPY(code, "NFA_BOL "); break; | |
| 1710 case NFA_EOW: STRCPY(code, "NFA_EOW "); break; | |
| 1711 case NFA_BOW: STRCPY(code, "NFA_BOW "); break; | |
| 1712 case NFA_STAR: STRCPY(code, "NFA_STAR "); break; | |
| 1713 case NFA_PLUS: STRCPY(code, "NFA_PLUS "); break; | |
| 1714 case NFA_NOT: STRCPY(code, "NFA_NOT "); break; | |
| 1715 case NFA_SKIP_CHAR: STRCPY(code, "NFA_SKIP_CHAR"); break; | |
| 1716 case NFA_OR: STRCPY(code, "NFA_OR"); break; | |
| 1717 case NFA_QUEST: STRCPY(code, "NFA_QUEST"); break; | |
| 1718 case NFA_QUEST_NONGREEDY: STRCPY(code, "NFA_QUEST_NON_GREEDY"); break; | |
| 1719 case NFA_END_NEG_RANGE: STRCPY(code, "NFA_END_NEG_RANGE"); break; | |
| 1720 case NFA_CLASS_ALNUM: STRCPY(code, "NFA_CLASS_ALNUM"); break; | |
| 1721 case NFA_CLASS_ALPHA: STRCPY(code, "NFA_CLASS_ALPHA"); break; | |
| 1722 case NFA_CLASS_BLANK: STRCPY(code, "NFA_CLASS_BLANK"); break; | |
| 1723 case NFA_CLASS_CNTRL: STRCPY(code, "NFA_CLASS_CNTRL"); break; | |
| 1724 case NFA_CLASS_DIGIT: STRCPY(code, "NFA_CLASS_DIGIT"); break; | |
| 1725 case NFA_CLASS_GRAPH: STRCPY(code, "NFA_CLASS_GRAPH"); break; | |
| 1726 case NFA_CLASS_LOWER: STRCPY(code, "NFA_CLASS_LOWER"); break; | |
| 1727 case NFA_CLASS_PRINT: STRCPY(code, "NFA_CLASS_PRINT"); break; | |
| 1728 case NFA_CLASS_PUNCT: STRCPY(code, "NFA_CLASS_PUNCT"); break; | |
| 1729 case NFA_CLASS_SPACE: STRCPY(code, "NFA_CLASS_SPACE"); break; | |
| 1730 case NFA_CLASS_UPPER: STRCPY(code, "NFA_CLASS_UPPER"); break; | |
| 1731 case NFA_CLASS_XDIGIT: STRCPY(code, "NFA_CLASS_XDIGIT"); break; | |
| 1732 case NFA_CLASS_TAB: STRCPY(code, "NFA_CLASS_TAB"); break; | |
| 1733 case NFA_CLASS_RETURN: STRCPY(code, "NFA_CLASS_RETURN"); break; | |
| 1734 case NFA_CLASS_BACKSPACE: STRCPY(code, "NFA_CLASS_BACKSPACE"); break; | |
| 1735 case NFA_CLASS_ESCAPE: STRCPY(code, "NFA_CLASS_ESCAPE"); break; | |
| 1736 | |
| 1737 case NFA_ANY: STRCPY(code, "NFA_ANY"); break; | |
| 1738 case NFA_IDENT: STRCPY(code, "NFA_IDENT"); break; | |
| 1739 case NFA_SIDENT:STRCPY(code, "NFA_SIDENT"); break; | |
| 1740 case NFA_KWORD: STRCPY(code, "NFA_KWORD"); break; | |
| 1741 case NFA_SKWORD:STRCPY(code, "NFA_SKWORD"); break; | |
| 1742 case NFA_FNAME: STRCPY(code, "NFA_FNAME"); break; | |
| 1743 case NFA_SFNAME:STRCPY(code, "NFA_SFNAME"); break; | |
| 1744 case NFA_PRINT: STRCPY(code, "NFA_PRINT"); break; | |
| 1745 case NFA_SPRINT:STRCPY(code, "NFA_SPRINT"); break; | |
| 1746 case NFA_WHITE: STRCPY(code, "NFA_WHITE"); break; | |
| 1747 case NFA_NWHITE:STRCPY(code, "NFA_NWHITE"); break; | |
| 1748 case NFA_DIGIT: STRCPY(code, "NFA_DIGIT"); break; | |
| 1749 case NFA_NDIGIT:STRCPY(code, "NFA_NDIGIT"); break; | |
| 1750 case NFA_HEX: STRCPY(code, "NFA_HEX"); break; | |
| 1751 case NFA_NHEX: STRCPY(code, "NFA_NHEX"); break; | |
| 1752 case NFA_OCTAL: STRCPY(code, "NFA_OCTAL"); break; | |
| 1753 case NFA_NOCTAL:STRCPY(code, "NFA_NOCTAL"); break; | |
| 1754 case NFA_WORD: STRCPY(code, "NFA_WORD"); break; | |
| 1755 case NFA_NWORD: STRCPY(code, "NFA_NWORD"); break; | |
| 1756 case NFA_HEAD: STRCPY(code, "NFA_HEAD"); break; | |
| 1757 case NFA_NHEAD: STRCPY(code, "NFA_NHEAD"); break; | |
| 1758 case NFA_ALPHA: STRCPY(code, "NFA_ALPHA"); break; | |
| 1759 case NFA_NALPHA:STRCPY(code, "NFA_NALPHA"); break; | |
| 1760 case NFA_LOWER: STRCPY(code, "NFA_LOWER"); break; | |
| 1761 case NFA_NLOWER:STRCPY(code, "NFA_NLOWER"); break; | |
| 1762 case NFA_UPPER: STRCPY(code, "NFA_UPPER"); break; | |
| 1763 case NFA_NUPPER:STRCPY(code, "NFA_NUPPER"); break; | |
| 1764 | |
| 1765 default: | |
| 1766 STRCPY(code, "CHAR(x)"); | |
| 1767 code[5] = c; | |
| 1768 } | |
| 1769 | |
| 1770 if (addnl == TRUE) | |
| 1771 STRCAT(code, " + NEWLINE "); | |
| 1772 | |
| 1773 } | |
| 1774 | |
| 1775 #ifdef ENABLE_LOG | |
| 1776 static FILE *log_fd; | |
| 1777 | |
| 1778 /* | |
| 1779 * Print the postfix notation of the current regexp. | |
| 1780 */ | |
| 1781 static void | |
| 1782 nfa_postfix_dump(expr, retval) | |
| 1783 char_u *expr; | |
| 1784 int retval; | |
| 1785 { | |
| 1786 int *p; | |
| 1787 FILE *f; | |
| 1788 | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
1789 f = fopen(NFA_REGEXP_DUMP_LOG, "a"); |
| 4444 | 1790 if (f != NULL) |
| 1791 { | |
| 1792 fprintf(f, "\n-------------------------\n"); | |
| 1793 if (retval == FAIL) | |
| 1794 fprintf(f, ">>> NFA engine failed ... \n"); | |
| 1795 else if (retval == OK) | |
| 1796 fprintf(f, ">>> NFA engine succeeded !\n"); | |
| 1797 fprintf(f, "Regexp: \"%s\"\nPostfix notation (char): \"", expr); | |
| 4464 | 1798 for (p = post_start; *p && p < post_end; p++) |
| 4444 | 1799 { |
| 1800 nfa_set_code(*p); | |
| 1801 fprintf(f, "%s, ", code); | |
| 1802 } | |
| 1803 fprintf(f, "\"\nPostfix notation (int): "); | |
| 4464 | 1804 for (p = post_start; *p && p < post_end; p++) |
| 4444 | 1805 fprintf(f, "%d ", *p); |
| 1806 fprintf(f, "\n\n"); | |
| 1807 fclose(f); | |
| 1808 } | |
| 1809 } | |
| 1810 | |
| 1811 /* | |
| 1812 * Print the NFA starting with a root node "state". | |
| 1813 */ | |
| 1814 static void | |
|
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1815 nfa_print_state(debugf, state) |
| 4444 | 1816 FILE *debugf; |
| 1817 nfa_state_T *state; | |
| 1818 { | |
|
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1819 garray_T indent; |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1820 |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1821 ga_init2(&indent, 1, 64); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1822 ga_append(&indent, '\0'); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1823 nfa_print_state2(debugf, state, &indent); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1824 ga_clear(&indent); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1825 } |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1826 |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1827 static void |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1828 nfa_print_state2(debugf, state, indent) |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1829 FILE *debugf; |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1830 nfa_state_T *state; |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1831 garray_T *indent; |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1832 { |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1833 char_u *p; |
| 4444 | 1834 |
| 1835 if (state == NULL) | |
| 1836 return; | |
| 1837 | |
| 1838 fprintf(debugf, "(%2d)", abs(state->id)); | |
|
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1839 |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1840 /* Output indent */ |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1841 p = (char_u *)indent->ga_data; |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1842 if (indent->ga_len >= 3) |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1843 { |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1844 int last = indent->ga_len - 3; |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1845 char_u save[2]; |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1846 |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1847 STRNCPY(save, &p[last], 2); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1848 STRNCPY(&p[last], "+-", 2); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1849 fprintf(debugf, " %s", p); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1850 STRNCPY(&p[last], save, 2); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1851 } |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1852 else |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1853 fprintf(debugf, " %s", p); |
| 4444 | 1854 |
| 1855 nfa_set_code(state->c); | |
|
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1856 fprintf(debugf, "%s%s (%d) (id=%d)\n", |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1857 state->negated ? "NOT " : "", code, state->c, abs(state->id)); |
| 4444 | 1858 if (state->id < 0) |
| 1859 return; | |
| 1860 | |
| 1861 state->id = abs(state->id) * -1; | |
|
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1862 |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1863 /* grow indent for state->out */ |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1864 indent->ga_len -= 1; |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1865 if (state->out1) |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1866 ga_concat(indent, "| "); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1867 else |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1868 ga_concat(indent, " "); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1869 ga_append(indent, '\0'); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1870 |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1871 nfa_print_state2(debugf, state->out, indent); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1872 |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1873 /* replace last part of indent for state->out1 */ |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1874 indent->ga_len -= 3; |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1875 ga_concat(indent, " "); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1876 ga_append(indent, '\0'); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1877 |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1878 nfa_print_state2(debugf, state->out1, indent); |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1879 |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1880 /* shrink indent */ |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1881 indent->ga_len -= 3; |
|
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1882 ga_append(indent, '\0'); |
| 4444 | 1883 } |
| 1884 | |
| 1885 /* | |
| 1886 * Print the NFA state machine. | |
| 1887 */ | |
| 1888 static void | |
| 1889 nfa_dump(prog) | |
| 1890 nfa_regprog_T *prog; | |
| 1891 { | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
1892 FILE *debugf = fopen(NFA_REGEXP_DUMP_LOG, "a"); |
| 4444 | 1893 |
| 1894 if (debugf != NULL) | |
| 1895 { | |
|
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1896 nfa_print_state(debugf, prog->start); |
| 4444 | 1897 fclose(debugf); |
| 1898 } | |
| 1899 } | |
| 1900 #endif /* ENABLE_LOG */ | |
| 1901 #endif /* DEBUG */ | |
| 1902 | |
| 1903 /* | |
| 1904 * Parse r.e. @expr and convert it into postfix form. | |
| 1905 * Return the postfix string on success, NULL otherwise. | |
| 1906 */ | |
| 1907 static int * | |
| 1908 re2post() | |
| 1909 { | |
| 1910 if (nfa_reg(REG_NOPAREN) == FAIL) | |
| 1911 return NULL; | |
| 1912 EMIT(NFA_MOPEN); | |
| 1913 return post_start; | |
| 1914 } | |
| 1915 | |
| 1916 /* NB. Some of the code below is inspired by Russ's. */ | |
| 1917 | |
| 1918 /* | |
| 1919 * Represents an NFA state plus zero or one or two arrows exiting. | |
| 1920 * if c == MATCH, no arrows out; matching state. | |
| 1921 * If c == SPLIT, unlabeled arrows to out and out1 (if != NULL). | |
| 1922 * If c < 256, labeled arrow with character c to out. | |
| 1923 */ | |
| 1924 | |
| 1925 static nfa_state_T *state_ptr; /* points to nfa_prog->state */ | |
| 1926 | |
| 1927 /* | |
| 1928 * Allocate and initialize nfa_state_T. | |
| 1929 */ | |
| 1930 static nfa_state_T * | |
| 1931 new_state(c, out, out1) | |
| 1932 int c; | |
| 1933 nfa_state_T *out; | |
| 1934 nfa_state_T *out1; | |
| 1935 { | |
| 1936 nfa_state_T *s; | |
| 1937 | |
| 1938 if (istate >= nstate) | |
| 1939 return NULL; | |
| 1940 | |
| 1941 s = &state_ptr[istate++]; | |
| 1942 | |
| 1943 s->c = c; | |
| 1944 s->out = out; | |
| 1945 s->out1 = out1; | |
| 1946 | |
| 1947 s->id = istate; | |
| 1948 s->lastlist = 0; | |
| 1949 s->lastthread = NULL; | |
| 1950 s->visits = 0; | |
| 1951 s->negated = FALSE; | |
| 1952 | |
| 1953 return s; | |
| 1954 } | |
| 1955 | |
| 1956 /* | |
| 1957 * A partially built NFA without the matching state filled in. | |
| 1958 * Frag_T.start points at the start state. | |
| 1959 * Frag_T.out is a list of places that need to be set to the | |
| 1960 * next state for this fragment. | |
| 1961 */ | |
| 1962 typedef union Ptrlist Ptrlist; | |
| 1963 struct Frag | |
| 1964 { | |
| 1965 nfa_state_T *start; | |
| 1966 Ptrlist *out; | |
| 1967 }; | |
| 1968 typedef struct Frag Frag_T; | |
| 1969 | |
| 1970 static Frag_T frag __ARGS((nfa_state_T *start, Ptrlist *out)); | |
| 1971 static Ptrlist *list1 __ARGS((nfa_state_T **outp)); | |
| 1972 static void patch __ARGS((Ptrlist *l, nfa_state_T *s)); | |
| 1973 static Ptrlist *append __ARGS((Ptrlist *l1, Ptrlist *l2)); | |
| 1974 static void st_push __ARGS((Frag_T s, Frag_T **p, Frag_T *stack_end)); | |
| 1975 static Frag_T st_pop __ARGS((Frag_T **p, Frag_T *stack)); | |
| 1976 | |
| 1977 /* | |
| 4456 | 1978 * Initialize a Frag_T struct and return it. |
| 4444 | 1979 */ |
| 1980 static Frag_T | |
| 1981 frag(start, out) | |
| 1982 nfa_state_T *start; | |
| 1983 Ptrlist *out; | |
| 1984 { | |
| 4456 | 1985 Frag_T n; |
| 1986 | |
| 1987 n.start = start; | |
| 1988 n.out = out; | |
| 4444 | 1989 return n; |
| 1990 } | |
| 1991 | |
| 1992 /* | |
| 1993 * Since the out pointers in the list are always | |
| 1994 * uninitialized, we use the pointers themselves | |
| 1995 * as storage for the Ptrlists. | |
| 1996 */ | |
| 1997 union Ptrlist | |
| 1998 { | |
| 1999 Ptrlist *next; | |
| 2000 nfa_state_T *s; | |
| 2001 }; | |
| 2002 | |
| 2003 /* | |
| 2004 * Create singleton list containing just outp. | |
| 2005 */ | |
| 2006 static Ptrlist * | |
| 2007 list1(outp) | |
| 2008 nfa_state_T **outp; | |
| 2009 { | |
| 2010 Ptrlist *l; | |
| 2011 | |
| 2012 l = (Ptrlist *)outp; | |
| 2013 l->next = NULL; | |
| 2014 return l; | |
| 2015 } | |
| 2016 | |
| 2017 /* | |
| 2018 * Patch the list of states at out to point to start. | |
| 2019 */ | |
| 2020 static void | |
| 2021 patch(l, s) | |
| 2022 Ptrlist *l; | |
| 2023 nfa_state_T *s; | |
| 2024 { | |
| 2025 Ptrlist *next; | |
| 2026 | |
| 2027 for (; l; l = next) | |
| 2028 { | |
| 2029 next = l->next; | |
| 2030 l->s = s; | |
| 2031 } | |
| 2032 } | |
| 2033 | |
| 2034 | |
| 2035 /* | |
| 2036 * Join the two lists l1 and l2, returning the combination. | |
| 2037 */ | |
| 2038 static Ptrlist * | |
| 2039 append(l1, l2) | |
| 2040 Ptrlist *l1; | |
| 2041 Ptrlist *l2; | |
| 2042 { | |
| 2043 Ptrlist *oldl1; | |
| 2044 | |
| 2045 oldl1 = l1; | |
| 2046 while (l1->next) | |
| 2047 l1 = l1->next; | |
| 2048 l1->next = l2; | |
| 2049 return oldl1; | |
| 2050 } | |
| 2051 | |
| 2052 /* | |
| 2053 * Stack used for transforming postfix form into NFA. | |
| 2054 */ | |
| 2055 static Frag_T empty; | |
| 2056 | |
| 2057 static void | |
| 2058 st_error(postfix, end, p) | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2059 int *postfix UNUSED; |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2060 int *end UNUSED; |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2061 int *p UNUSED; |
| 4444 | 2062 { |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2063 #ifdef NFA_REGEXP_ERROR_LOG |
| 4444 | 2064 FILE *df; |
| 2065 int *p2; | |
| 2066 | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2067 df = fopen(NFA_REGEXP_ERROR_LOG, "a"); |
| 4444 | 2068 if (df) |
| 2069 { | |
| 2070 fprintf(df, "Error popping the stack!\n"); | |
| 2071 #ifdef DEBUG | |
| 2072 fprintf(df, "Current regexp is \"%s\"\n", nfa_regengine.expr); | |
| 2073 #endif | |
| 2074 fprintf(df, "Postfix form is: "); | |
| 2075 #ifdef DEBUG | |
| 2076 for (p2 = postfix; p2 < end; p2++) | |
| 2077 { | |
| 2078 nfa_set_code(*p2); | |
| 2079 fprintf(df, "%s, ", code); | |
| 2080 } | |
| 2081 nfa_set_code(*p); | |
| 2082 fprintf(df, "\nCurrent position is: "); | |
| 2083 for (p2 = postfix; p2 <= p; p2 ++) | |
| 2084 { | |
| 2085 nfa_set_code(*p2); | |
| 2086 fprintf(df, "%s, ", code); | |
| 2087 } | |
| 2088 #else | |
| 2089 for (p2 = postfix; p2 < end; p2++) | |
| 2090 { | |
| 2091 fprintf(df, "%d, ", *p2); | |
| 2092 } | |
| 2093 fprintf(df, "\nCurrent position is: "); | |
| 2094 for (p2 = postfix; p2 <= p; p2 ++) | |
| 2095 { | |
| 2096 fprintf(df, "%d, ", *p2); | |
| 2097 } | |
| 2098 #endif | |
| 2099 fprintf(df, "\n--------------------------\n"); | |
| 2100 fclose(df); | |
| 2101 } | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2102 #endif |
| 4444 | 2103 EMSG(_("E874: (NFA) Could not pop the stack !")); |
| 2104 } | |
| 2105 | |
| 2106 /* | |
| 2107 * Push an item onto the stack. | |
| 2108 */ | |
| 2109 static void | |
| 2110 st_push(s, p, stack_end) | |
| 2111 Frag_T s; | |
| 2112 Frag_T **p; | |
| 2113 Frag_T *stack_end; | |
| 2114 { | |
| 2115 Frag_T *stackp = *p; | |
| 2116 | |
| 2117 if (stackp >= stack_end) | |
| 2118 return; | |
| 2119 *stackp = s; | |
| 2120 *p = *p + 1; | |
| 2121 } | |
| 2122 | |
| 2123 /* | |
| 2124 * Pop an item from the stack. | |
| 2125 */ | |
| 2126 static Frag_T | |
| 2127 st_pop(p, stack) | |
| 2128 Frag_T **p; | |
| 2129 Frag_T *stack; | |
| 2130 { | |
| 2131 Frag_T *stackp; | |
| 2132 | |
| 2133 *p = *p - 1; | |
| 2134 stackp = *p; | |
| 2135 if (stackp < stack) | |
| 2136 return empty; | |
| 2137 return **p; | |
| 2138 } | |
| 2139 | |
| 2140 /* | |
| 2141 * Convert a postfix form into its equivalent NFA. | |
| 2142 * Return the NFA start state on success, NULL otherwise. | |
| 2143 */ | |
| 2144 static nfa_state_T * | |
| 2145 post2nfa(postfix, end, nfa_calc_size) | |
| 2146 int *postfix; | |
| 2147 int *end; | |
| 2148 int nfa_calc_size; | |
| 2149 { | |
| 2150 int *p; | |
| 2151 int mopen; | |
| 2152 int mclose; | |
| 2153 Frag_T *stack = NULL; | |
| 2154 Frag_T *stackp = NULL; | |
| 2155 Frag_T *stack_end = NULL; | |
| 2156 Frag_T e1; | |
| 2157 Frag_T e2; | |
| 2158 Frag_T e; | |
| 2159 nfa_state_T *s; | |
| 2160 nfa_state_T *s1; | |
| 2161 nfa_state_T *matchstate; | |
| 4484 | 2162 nfa_state_T *ret = NULL; |
| 4444 | 2163 |
| 2164 if (postfix == NULL) | |
| 2165 return NULL; | |
| 2166 | |
| 4456 | 2167 #define PUSH(s) st_push((s), &stackp, stack_end) |
| 4444 | 2168 #define POP() st_pop(&stackp, stack); \ |
| 2169 if (stackp < stack) \ | |
| 2170 { \ | |
| 2171 st_error(postfix, end, p); \ | |
| 2172 return NULL; \ | |
| 2173 } | |
| 2174 | |
| 2175 if (nfa_calc_size == FALSE) | |
| 2176 { | |
| 2177 /* Allocate space for the stack. Max states on the stack : nstate */ | |
| 4462 | 2178 stack = (Frag_T *) lalloc((nstate + 1) * sizeof(Frag_T), TRUE); |
| 4444 | 2179 stackp = stack; |
| 4462 | 2180 stack_end = stack + (nstate + 1); |
| 4444 | 2181 } |
| 2182 | |
| 2183 for (p = postfix; p < end; ++p) | |
| 2184 { | |
| 2185 switch (*p) | |
| 2186 { | |
| 2187 case NFA_CONCAT: | |
| 2188 /* Catenation. | |
| 2189 * Pay attention: this operator does not exist | |
| 2190 * in the r.e. itself (it is implicit, really). | |
| 2191 * It is added when r.e. is translated to postfix | |
| 2192 * form in re2post(). | |
| 2193 * | |
| 2194 * No new state added here. */ | |
| 2195 if (nfa_calc_size == TRUE) | |
| 2196 { | |
| 4458 | 2197 /* nstate += 0; */ |
| 4444 | 2198 break; |
| 2199 } | |
| 2200 e2 = POP(); | |
| 2201 e1 = POP(); | |
| 2202 patch(e1.out, e2.start); | |
| 2203 PUSH(frag(e1.start, e2.out)); | |
| 2204 break; | |
| 2205 | |
| 2206 case NFA_NOT: | |
| 2207 /* Negation of a character */ | |
| 2208 if (nfa_calc_size == TRUE) | |
| 2209 { | |
| 4458 | 2210 /* nstate += 0; */ |
| 4444 | 2211 break; |
| 2212 } | |
| 2213 e1 = POP(); | |
| 2214 e1.start->negated = TRUE; | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2215 #ifdef FEAT_MBYTE |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2216 if (e1.start->c == NFA_COMPOSING) |
| 4444 | 2217 e1.start->out1->negated = TRUE; |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2218 #endif |
| 4444 | 2219 PUSH(e1); |
| 2220 break; | |
| 2221 | |
| 2222 case NFA_OR: | |
| 2223 /* Alternation */ | |
| 2224 if (nfa_calc_size == TRUE) | |
| 2225 { | |
| 4458 | 2226 nstate++; |
| 4444 | 2227 break; |
| 2228 } | |
| 2229 e2 = POP(); | |
| 2230 e1 = POP(); | |
| 2231 s = new_state(NFA_SPLIT, e1.start, e2.start); | |
| 2232 if (s == NULL) | |
| 4484 | 2233 goto theend; |
| 4444 | 2234 PUSH(frag(s, append(e1.out, e2.out))); |
| 2235 break; | |
| 2236 | |
| 2237 case NFA_STAR: | |
| 2238 /* Zero or more */ | |
| 2239 if (nfa_calc_size == TRUE) | |
| 2240 { | |
| 4458 | 2241 nstate++; |
| 4444 | 2242 break; |
| 2243 } | |
| 2244 e = POP(); | |
| 2245 s = new_state(NFA_SPLIT, e.start, NULL); | |
| 2246 if (s == NULL) | |
| 4484 | 2247 goto theend; |
| 4444 | 2248 patch(e.out, s); |
| 2249 PUSH(frag(s, list1(&s->out1))); | |
| 2250 break; | |
| 2251 | |
| 2252 case NFA_QUEST: | |
| 2253 /* one or zero atoms=> greedy match */ | |
| 2254 if (nfa_calc_size == TRUE) | |
| 2255 { | |
| 4458 | 2256 nstate++; |
| 4444 | 2257 break; |
| 2258 } | |
| 2259 e = POP(); | |
| 2260 s = new_state(NFA_SPLIT, e.start, NULL); | |
| 2261 if (s == NULL) | |
| 4484 | 2262 goto theend; |
| 4444 | 2263 PUSH(frag(s, append(e.out, list1(&s->out1)))); |
| 2264 break; | |
| 2265 | |
| 2266 case NFA_QUEST_NONGREEDY: | |
| 2267 /* zero or one atoms => non-greedy match */ | |
| 2268 if (nfa_calc_size == TRUE) | |
| 2269 { | |
| 4458 | 2270 nstate++; |
| 4444 | 2271 break; |
| 2272 } | |
| 2273 e = POP(); | |
| 2274 s = new_state(NFA_SPLIT, NULL, e.start); | |
| 2275 if (s == NULL) | |
| 4484 | 2276 goto theend; |
| 4444 | 2277 PUSH(frag(s, append(e.out, list1(&s->out)))); |
| 2278 break; | |
| 2279 | |
| 2280 case NFA_PLUS: | |
| 2281 /* One or more */ | |
| 2282 if (nfa_calc_size == TRUE) | |
| 2283 { | |
| 4458 | 2284 nstate++; |
| 4444 | 2285 break; |
| 2286 } | |
| 2287 e = POP(); | |
| 2288 s = new_state(NFA_SPLIT, e.start, NULL); | |
| 2289 if (s == NULL) | |
| 4484 | 2290 goto theend; |
| 4444 | 2291 patch(e.out, s); |
| 2292 PUSH(frag(e.start, list1(&s->out1))); | |
| 2293 break; | |
| 2294 | |
| 2295 case NFA_SKIP_CHAR: | |
| 2296 /* Symbol of 0-length, Used in a repetition | |
| 2297 * with max/min count of 0 */ | |
| 2298 if (nfa_calc_size == TRUE) | |
| 2299 { | |
| 4458 | 2300 nstate++; |
| 4444 | 2301 break; |
| 2302 } | |
| 2303 s = new_state(NFA_SKIP_CHAR, NULL, NULL); | |
| 2304 if (s == NULL) | |
| 4484 | 2305 goto theend; |
| 4444 | 2306 PUSH(frag(s, list1(&s->out))); |
| 2307 break; | |
| 2308 | |
| 2309 case NFA_PREV_ATOM_NO_WIDTH: | |
| 2310 /* The \@= operator: match the preceding atom with 0 width. | |
| 2311 * Surrounds the preceding atom with START_INVISIBLE and | |
| 2312 * END_INVISIBLE, similarly to MOPEN. | |
| 2313 */ | |
| 2314 /* TODO: Maybe this drops the speed? */ | |
| 4484 | 2315 goto theend; |
| 4444 | 2316 |
| 2317 if (nfa_calc_size == TRUE) | |
| 2318 { | |
| 2319 nstate += 2; | |
| 2320 break; | |
| 2321 } | |
| 2322 e = POP(); | |
| 2323 s1 = new_state(NFA_END_INVISIBLE, NULL, NULL); | |
| 2324 if (s1 == NULL) | |
| 4484 | 2325 goto theend; |
| 4444 | 2326 patch(e.out, s1); |
| 2327 | |
| 2328 s = new_state(NFA_START_INVISIBLE, e.start, s1); | |
| 2329 if (s == NULL) | |
| 4484 | 2330 goto theend; |
| 4444 | 2331 PUSH(frag(s, list1(&s1->out))); |
| 2332 break; | |
| 2333 | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2334 #ifdef FEAT_MBYTE |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2335 case NFA_COMPOSING: /* char with composing char */ |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2336 #if 0 |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2337 /* TODO */ |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2338 if (regflags & RF_ICOMBINE) |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2339 { |
|
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
2340 /* use the base character only */ |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2341 } |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2342 #endif |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2343 /* FALLTHROUGH */ |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2344 #endif |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2345 |
| 4444 | 2346 case NFA_MOPEN + 0: /* Submatch */ |
| 2347 case NFA_MOPEN + 1: | |
| 2348 case NFA_MOPEN + 2: | |
| 2349 case NFA_MOPEN + 3: | |
| 2350 case NFA_MOPEN + 4: | |
| 2351 case NFA_MOPEN + 5: | |
| 2352 case NFA_MOPEN + 6: | |
| 2353 case NFA_MOPEN + 7: | |
| 2354 case NFA_MOPEN + 8: | |
| 2355 case NFA_MOPEN + 9: | |
| 2356 case NFA_NOPEN: /* \%( "Invisible Submatch" */ | |
| 2357 if (nfa_calc_size == TRUE) | |
| 2358 { | |
| 2359 nstate += 2; | |
| 2360 break; | |
| 2361 } | |
| 2362 | |
| 2363 mopen = *p; | |
| 2364 switch (*p) | |
| 2365 { | |
| 2366 case NFA_NOPEN: | |
| 2367 mclose = NFA_NCLOSE; | |
| 2368 break; | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2369 #ifdef FEAT_MBYTE |
| 4444 | 2370 case NFA_COMPOSING: |
| 2371 mclose = NFA_END_COMPOSING; | |
| 2372 break; | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2373 #endif |
| 4444 | 2374 default: |
| 2375 /* NFA_MOPEN(0) ... NFA_MOPEN(9) */ | |
| 2376 mclose = *p + NSUBEXP; | |
| 2377 break; | |
| 2378 } | |
| 2379 | |
| 2380 /* Allow "NFA_MOPEN" as a valid postfix representation for | |
| 2381 * the empty regexp "". In this case, the NFA will be | |
| 2382 * NFA_MOPEN -> NFA_MCLOSE. Note that this also allows | |
| 2383 * empty groups of parenthesis, and empty mbyte chars */ | |
| 2384 if (stackp == stack) | |
| 2385 { | |
| 2386 s = new_state(mopen, NULL, NULL); | |
| 2387 if (s == NULL) | |
| 4484 | 2388 goto theend; |
| 4444 | 2389 s1 = new_state(mclose, NULL, NULL); |
| 2390 if (s1 == NULL) | |
| 4484 | 2391 goto theend; |
| 4444 | 2392 patch(list1(&s->out), s1); |
| 2393 PUSH(frag(s, list1(&s1->out))); | |
| 2394 break; | |
| 2395 } | |
| 2396 | |
| 2397 /* At least one node was emitted before NFA_MOPEN, so | |
| 2398 * at least one node will be between NFA_MOPEN and NFA_MCLOSE */ | |
| 2399 e = POP(); | |
| 2400 s = new_state(mopen, e.start, NULL); /* `(' */ | |
| 2401 if (s == NULL) | |
| 4484 | 2402 goto theend; |
| 4444 | 2403 |
| 2404 s1 = new_state(mclose, NULL, NULL); /* `)' */ | |
| 2405 if (s1 == NULL) | |
| 4484 | 2406 goto theend; |
| 4444 | 2407 patch(e.out, s1); |
| 2408 | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2409 #ifdef FEAT_MBYTE |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2410 if (mopen == NFA_COMPOSING) |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2411 /* COMPOSING->out1 = END_COMPOSING */ |
| 4444 | 2412 patch(list1(&s->out1), s1); |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2413 #endif |
| 4444 | 2414 |
| 2415 PUSH(frag(s, list1(&s1->out))); | |
| 2416 break; | |
| 2417 | |
| 2418 case NFA_ZSTART: | |
| 2419 case NFA_ZEND: | |
| 2420 default: | |
| 2421 /* Operands */ | |
| 2422 if (nfa_calc_size == TRUE) | |
| 2423 { | |
| 4458 | 2424 nstate++; |
| 4444 | 2425 break; |
| 2426 } | |
| 2427 s = new_state(*p, NULL, NULL); | |
| 2428 if (s == NULL) | |
| 4484 | 2429 goto theend; |
| 4444 | 2430 PUSH(frag(s, list1(&s->out))); |
| 2431 break; | |
| 2432 | |
| 2433 } /* switch(*p) */ | |
| 2434 | |
| 2435 } /* for(p = postfix; *p; ++p) */ | |
| 2436 | |
| 2437 if (nfa_calc_size == TRUE) | |
| 2438 { | |
| 4458 | 2439 nstate++; |
| 4484 | 2440 goto theend; /* Return value when counting size is ignored anyway */ |
| 4444 | 2441 } |
| 2442 | |
| 2443 e = POP(); | |
| 2444 if (stackp != stack) | |
| 2445 EMSG_RET_NULL(_("E875: (NFA regexp) (While converting from postfix to NFA), too many states left on stack")); | |
| 2446 | |
| 2447 if (istate >= nstate) | |
| 2448 EMSG_RET_NULL(_("E876: (NFA regexp) Not enough space to store the whole NFA ")); | |
| 2449 | |
| 2450 matchstate = &state_ptr[istate++]; /* the match state */ | |
| 2451 matchstate->c = NFA_MATCH; | |
| 2452 matchstate->out = matchstate->out1 = NULL; | |
| 2453 | |
| 2454 patch(e.out, matchstate); | |
| 4484 | 2455 ret = e.start; |
| 2456 | |
| 2457 theend: | |
| 2458 vim_free(stack); | |
| 2459 return ret; | |
| 4444 | 2460 |
| 2461 #undef POP1 | |
| 2462 #undef PUSH1 | |
| 2463 #undef POP2 | |
| 2464 #undef PUSH2 | |
| 2465 #undef POP | |
| 2466 #undef PUSH | |
| 2467 } | |
| 2468 | |
| 2469 /**************************************************************** | |
| 2470 * NFA execution code. | |
| 2471 ****************************************************************/ | |
| 2472 | |
| 2473 /* thread_T contains runtime information of a NFA state */ | |
| 2474 struct thread | |
| 2475 { | |
| 2476 nfa_state_T *state; | |
| 2477 regsub_T sub; /* submatch info */ | |
| 2478 }; | |
| 2479 | |
| 2480 typedef struct | |
| 2481 { | |
| 2482 thread_T *t; | |
| 2483 int n; | |
| 2484 } List; | |
| 2485 | |
| 2486 static void addstate __ARGS((List *l, nfa_state_T *state, regsub_T *m, int off, int lid, int *match)); | |
| 2487 | |
| 2488 static void | |
| 2489 addstate(l, state, m, off, lid, match) | |
| 2490 List *l; /* runtime state list */ | |
| 2491 nfa_state_T *state; /* state to update */ | |
| 2492 regsub_T *m; /* pointers to subexpressions */ | |
|
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2493 int off; /* byte offset, when -1 go to next line */ |
| 4444 | 2494 int lid; |
| 2495 int *match; /* found match? */ | |
| 2496 { | |
| 2497 regsub_T save; | |
| 2498 int subidx = 0; | |
| 2499 | |
| 2500 if (l == NULL || state == NULL) | |
| 2501 return; | |
| 2502 | |
| 2503 switch (state->c) | |
| 2504 { | |
| 2505 case NFA_SPLIT: | |
| 2506 case NFA_NOT: | |
| 2507 case NFA_NOPEN: | |
| 2508 case NFA_NCLOSE: | |
| 2509 case NFA_MCLOSE: | |
| 2510 case NFA_MCLOSE + 1: | |
| 2511 case NFA_MCLOSE + 2: | |
| 2512 case NFA_MCLOSE + 3: | |
| 2513 case NFA_MCLOSE + 4: | |
| 2514 case NFA_MCLOSE + 5: | |
| 2515 case NFA_MCLOSE + 6: | |
| 2516 case NFA_MCLOSE + 7: | |
| 2517 case NFA_MCLOSE + 8: | |
| 2518 case NFA_MCLOSE + 9: | |
| 2519 /* Do not remember these nodes in list "thislist" or "nextlist" */ | |
| 2520 break; | |
| 2521 | |
| 2522 default: | |
| 2523 if (state->lastlist == lid) | |
| 2524 { | |
| 2525 if (++state->visits > 2) | |
| 2526 return; | |
| 2527 } | |
| 2528 else | |
| 2529 { | |
| 2530 /* add the state to the list */ | |
| 2531 state->lastlist = lid; | |
| 2532 state->lastthread = &l->t[l->n++]; | |
| 2533 state->lastthread->state = state; | |
| 2534 state->lastthread->sub = *m; | |
| 2535 } | |
| 2536 } | |
| 2537 | |
| 2538 #ifdef ENABLE_LOG | |
| 2539 nfa_set_code(state->c); | |
| 2540 fprintf(log_fd, "> Adding state %d to list. Character %s, code %d\n", | |
| 2541 abs(state->id), code, state->c); | |
| 2542 #endif | |
| 2543 switch (state->c) | |
| 2544 { | |
| 2545 case NFA_MATCH: | |
| 2546 *match = TRUE; | |
| 2547 break; | |
| 2548 | |
| 2549 case NFA_SPLIT: | |
| 2550 addstate(l, state->out, m, off, lid, match); | |
| 2551 addstate(l, state->out1, m, off, lid, match); | |
| 2552 break; | |
| 2553 | |
| 2554 case NFA_SKIP_CHAR: | |
| 2555 addstate(l, state->out, m, off, lid, match); | |
| 2556 break; | |
| 2557 | |
| 2558 #if 0 | |
| 2559 case NFA_END_NEG_RANGE: | |
| 2560 /* Nothing to handle here. nfa_regmatch() will take care of it */ | |
| 2561 break; | |
| 2562 | |
| 2563 case NFA_NOT: | |
| 2564 EMSG(_("E999: (NFA regexp internal error) Should not process NOT node !")); | |
| 2565 #ifdef ENABLE_LOG | |
| 2566 fprintf(f, "\n\n>>> E999: Added state NFA_NOT to a list ... Something went wrong ! Why wasn't it processed already? \n\n"); | |
| 2567 #endif | |
| 2568 break; | |
| 2569 | |
| 2570 case NFA_COMPOSING: | |
| 2571 /* nfa_regmatch() will match all the bytes of this composing char. */ | |
| 2572 break; | |
| 2573 #endif | |
| 2574 | |
| 2575 case NFA_NOPEN: | |
| 2576 case NFA_NCLOSE: | |
| 2577 addstate(l, state->out, m, off, lid, match); | |
| 2578 break; | |
| 2579 | |
| 2580 /* If this state is reached, then a recursive call of nfa_regmatch() | |
| 2581 * succeeded. the next call saves the found submatches in the | |
| 2582 * first state after the "invisible" branch. */ | |
| 2583 #if 0 | |
| 2584 case NFA_END_INVISIBLE: | |
| 2585 break; | |
| 2586 #endif | |
| 2587 | |
| 2588 case NFA_MOPEN + 0: | |
| 2589 case NFA_MOPEN + 1: | |
| 2590 case NFA_MOPEN + 2: | |
| 2591 case NFA_MOPEN + 3: | |
| 2592 case NFA_MOPEN + 4: | |
| 2593 case NFA_MOPEN + 5: | |
| 2594 case NFA_MOPEN + 6: | |
| 2595 case NFA_MOPEN + 7: | |
| 2596 case NFA_MOPEN + 8: | |
| 2597 case NFA_MOPEN + 9: | |
| 2598 case NFA_ZSTART: | |
| 2599 subidx = state->c - NFA_MOPEN; | |
| 2600 if (state->c == NFA_ZSTART) | |
| 2601 subidx = 0; | |
| 2602 | |
| 2603 if (REG_MULTI) | |
| 2604 { | |
| 2605 save.startpos[subidx] = m->startpos[subidx]; | |
| 2606 save.endpos[subidx] = m->endpos[subidx]; | |
|
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2607 if (off == -1) |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2608 { |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2609 m->startpos[subidx].lnum = reglnum + 1; |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2610 m->startpos[subidx].col = 0; |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2611 } |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2612 else |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2613 { |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2614 m->startpos[subidx].lnum = reglnum; |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2615 m->startpos[subidx].col = |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2616 (colnr_T)(reginput - regline + off); |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2617 } |
| 4444 | 2618 } |
| 2619 else | |
| 2620 { | |
| 2621 save.start[subidx] = m->start[subidx]; | |
| 2622 save.end[subidx] = m->end[subidx]; | |
| 2623 m->start[subidx] = reginput + off; | |
| 2624 } | |
| 2625 | |
| 2626 addstate(l, state->out, m, off, lid, match); | |
| 2627 | |
| 2628 if (REG_MULTI) | |
| 2629 { | |
| 2630 m->startpos[subidx] = save.startpos[subidx]; | |
| 2631 m->endpos[subidx] = save.endpos[subidx]; | |
| 2632 } | |
| 2633 else | |
| 2634 { | |
| 2635 m->start[subidx] = save.start[subidx]; | |
| 2636 m->end[subidx] = save.end[subidx]; | |
| 2637 } | |
| 2638 break; | |
| 2639 | |
| 2640 case NFA_MCLOSE + 0: | |
| 2641 if (nfa_has_zend == TRUE) | |
| 2642 { | |
| 2643 addstate(l, state->out, m, off, lid, match); | |
| 2644 break; | |
| 2645 } | |
| 2646 case NFA_MCLOSE + 1: | |
| 2647 case NFA_MCLOSE + 2: | |
| 2648 case NFA_MCLOSE + 3: | |
| 2649 case NFA_MCLOSE + 4: | |
| 2650 case NFA_MCLOSE + 5: | |
| 2651 case NFA_MCLOSE + 6: | |
| 2652 case NFA_MCLOSE + 7: | |
| 2653 case NFA_MCLOSE + 8: | |
| 2654 case NFA_MCLOSE + 9: | |
| 2655 case NFA_ZEND: | |
| 2656 subidx = state->c - NFA_MCLOSE; | |
| 2657 if (state->c == NFA_ZEND) | |
| 2658 subidx = 0; | |
| 2659 | |
| 2660 if (REG_MULTI) | |
| 2661 { | |
| 2662 save.startpos[subidx] = m->startpos[subidx]; | |
| 2663 save.endpos[subidx] = m->endpos[subidx]; | |
|
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2664 if (off == -1) |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2665 { |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2666 m->endpos[subidx].lnum = reglnum + 1; |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2667 m->endpos[subidx].col = 0; |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2668 } |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2669 else |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2670 { |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2671 m->endpos[subidx].lnum = reglnum; |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2672 m->endpos[subidx].col = (colnr_T)(reginput - regline + off); |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2673 } |
| 4444 | 2674 } |
| 2675 else | |
| 2676 { | |
| 2677 save.start[subidx] = m->start[subidx]; | |
| 2678 save.end[subidx] = m->end[subidx]; | |
| 2679 m->end[subidx] = reginput + off; | |
| 2680 } | |
| 2681 | |
| 2682 addstate(l, state->out, m, off, lid, match); | |
| 2683 | |
| 2684 if (REG_MULTI) | |
| 2685 { | |
| 2686 m->startpos[subidx] = save.startpos[subidx]; | |
| 2687 m->endpos[subidx] = save.endpos[subidx]; | |
| 2688 } | |
| 2689 else | |
| 2690 { | |
| 2691 m->start[subidx] = save.start[subidx]; | |
| 2692 m->end[subidx] = save.end[subidx]; | |
| 2693 } | |
| 2694 break; | |
| 2695 } | |
| 2696 } | |
| 2697 | |
| 2698 /* | |
| 2699 * Check character class "class" against current character c. | |
| 2700 */ | |
| 2701 static int | |
| 2702 check_char_class(class, c) | |
| 2703 int class; | |
| 2704 int c; | |
| 2705 { | |
| 2706 switch (class) | |
| 2707 { | |
| 2708 case NFA_CLASS_ALNUM: | |
| 4464 | 2709 if (c >= 1 && c <= 255 && isalnum(c)) |
| 4444 | 2710 return OK; |
| 2711 break; | |
| 2712 case NFA_CLASS_ALPHA: | |
| 4464 | 2713 if (c >= 1 && c <= 255 && isalpha(c)) |
| 4444 | 2714 return OK; |
| 2715 break; | |
| 2716 case NFA_CLASS_BLANK: | |
| 2717 if (c == ' ' || c == '\t') | |
| 2718 return OK; | |
| 2719 break; | |
| 2720 case NFA_CLASS_CNTRL: | |
| 4464 | 2721 if (c >= 1 && c <= 255 && iscntrl(c)) |
| 4444 | 2722 return OK; |
| 2723 break; | |
| 2724 case NFA_CLASS_DIGIT: | |
| 2725 if (VIM_ISDIGIT(c)) | |
| 2726 return OK; | |
| 2727 break; | |
| 2728 case NFA_CLASS_GRAPH: | |
| 4464 | 2729 if (c >= 1 && c <= 255 && isgraph(c)) |
| 4444 | 2730 return OK; |
| 2731 break; | |
| 2732 case NFA_CLASS_LOWER: | |
| 2733 if (MB_ISLOWER(c)) | |
| 2734 return OK; | |
| 2735 break; | |
| 2736 case NFA_CLASS_PRINT: | |
| 2737 if (vim_isprintc(c)) | |
| 2738 return OK; | |
| 2739 break; | |
| 2740 case NFA_CLASS_PUNCT: | |
| 4464 | 2741 if (c >= 1 && c <= 255 && ispunct(c)) |
| 4444 | 2742 return OK; |
| 2743 break; | |
| 2744 case NFA_CLASS_SPACE: | |
| 2745 if ((c >=9 && c <= 13) || (c == ' ')) | |
| 2746 return OK; | |
| 2747 break; | |
| 2748 case NFA_CLASS_UPPER: | |
| 2749 if (MB_ISUPPER(c)) | |
| 2750 return OK; | |
| 2751 break; | |
| 2752 case NFA_CLASS_XDIGIT: | |
| 2753 if (vim_isxdigit(c)) | |
| 2754 return OK; | |
| 2755 break; | |
| 2756 case NFA_CLASS_TAB: | |
| 2757 if (c == '\t') | |
| 2758 return OK; | |
| 2759 break; | |
| 2760 case NFA_CLASS_RETURN: | |
| 2761 if (c == '\r') | |
| 2762 return OK; | |
| 2763 break; | |
| 2764 case NFA_CLASS_BACKSPACE: | |
| 2765 if (c == '\b') | |
| 2766 return OK; | |
| 2767 break; | |
| 2768 case NFA_CLASS_ESCAPE: | |
| 2769 if (c == '\033') | |
| 2770 return OK; | |
| 2771 break; | |
| 2772 | |
| 2773 default: | |
| 2774 /* should not be here :P */ | |
| 2775 EMSG_RET_FAIL(_("E877: (NFA regexp) Invalid character class ")); | |
| 2776 } | |
| 2777 return FAIL; | |
| 2778 } | |
| 2779 | |
| 2780 /* | |
| 2781 * Set all NFA nodes' list ID equal to -1. | |
| 2782 */ | |
| 2783 static void | |
| 2784 nfa_set_neg_listids(start) | |
| 2785 nfa_state_T *start; | |
| 2786 { | |
| 2787 if (start == NULL) | |
| 2788 return; | |
| 2789 if (start->lastlist >= 0) | |
| 2790 { | |
| 2791 start->lastlist = -1; | |
| 2792 nfa_set_neg_listids(start->out); | |
| 2793 nfa_set_neg_listids(start->out1); | |
| 2794 } | |
| 2795 } | |
| 2796 | |
| 2797 /* | |
| 2798 * Set all NFA nodes' list ID equal to 0. | |
| 2799 */ | |
| 2800 static void | |
| 2801 nfa_set_null_listids(start) | |
| 2802 nfa_state_T *start; | |
| 2803 { | |
| 2804 if (start == NULL) | |
| 2805 return; | |
| 2806 if (start->lastlist == -1) | |
| 2807 { | |
| 2808 start->lastlist = 0; | |
| 2809 nfa_set_null_listids(start->out); | |
| 2810 nfa_set_null_listids(start->out1); | |
| 2811 } | |
| 2812 } | |
| 2813 | |
| 2814 /* | |
| 2815 * Save list IDs for all NFA states in "list". | |
| 2816 */ | |
| 2817 static void | |
| 2818 nfa_save_listids(start, list) | |
| 2819 nfa_state_T *start; | |
| 2820 int *list; | |
| 2821 { | |
| 2822 if (start == NULL) | |
| 2823 return; | |
| 2824 if (start->lastlist != -1) | |
| 2825 { | |
| 2826 list[abs(start->id)] = start->lastlist; | |
| 2827 start->lastlist = -1; | |
| 2828 nfa_save_listids(start->out, list); | |
| 2829 nfa_save_listids(start->out1, list); | |
| 2830 } | |
| 2831 } | |
| 2832 | |
| 2833 /* | |
| 2834 * Restore list IDs from "list" to all NFA states. | |
| 2835 */ | |
| 2836 static void | |
| 2837 nfa_restore_listids(start, list) | |
| 2838 nfa_state_T *start; | |
| 2839 int *list; | |
| 2840 { | |
| 2841 if (start == NULL) | |
| 2842 return; | |
| 2843 if (start->lastlist == -1) | |
| 2844 { | |
| 2845 start->lastlist = list[abs(start->id)]; | |
| 2846 nfa_restore_listids(start->out, list); | |
| 2847 nfa_restore_listids(start->out1, list); | |
| 2848 } | |
| 2849 } | |
| 2850 | |
| 2851 /* | |
| 2852 * Main matching routine. | |
| 2853 * | |
| 2854 * Run NFA to determine whether it matches reginput. | |
| 2855 * | |
| 2856 * Return TRUE if there is a match, FALSE otherwise. | |
| 2857 * Note: Caller must ensure that: start != NULL. | |
| 2858 */ | |
| 2859 static int | |
| 2860 nfa_regmatch(start, submatch, m) | |
| 2861 nfa_state_T *start; | |
| 2862 regsub_T *submatch; | |
| 2863 regsub_T *m; | |
| 2864 { | |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2865 int c; |
| 4444 | 2866 int n; |
| 2867 int i = 0; | |
| 2868 int result; | |
| 2869 int size = 0; | |
| 2870 int match = FALSE; | |
| 2871 int flag = 0; | |
| 2872 int old_reglnum = -1; | |
|
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2873 int go_to_nextline; |
| 4444 | 2874 thread_T *t; |
| 2875 char_u *old_reginput = NULL; | |
| 2876 char_u *old_regline = NULL; | |
| 2877 List list[3]; | |
| 2878 List *listtbl[2][2]; | |
| 2879 List *ll; | |
| 2880 int listid = 1; | |
| 2881 List *thislist; | |
| 2882 List *nextlist; | |
| 2883 List *neglist; | |
| 2884 int *listids = NULL; | |
| 2885 int j = 0; | |
| 4460 | 2886 #ifdef NFA_REGEXP_DEBUG_LOG |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2887 FILE *debug = fopen(NFA_REGEXP_DEBUG_LOG, "a"); |
| 4444 | 2888 |
| 2889 if (debug == NULL) | |
| 2890 { | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2891 EMSG2(_("(NFA) COULD NOT OPEN %s !"), NFA_REGEXP_DEBUG_LOG); |
| 4444 | 2892 return FALSE; |
| 2893 } | |
| 2894 #endif | |
| 2895 | |
| 2896 /* Allocate memory for the lists of nodes */ | |
| 2897 size = (nstate + 1) * sizeof(thread_T); | |
| 2898 list[0].t = (thread_T *)lalloc(size, TRUE); | |
| 2899 list[1].t = (thread_T *)lalloc(size, TRUE); | |
| 2900 list[2].t = (thread_T *)lalloc(size, TRUE); | |
| 2901 if (list[0].t == NULL || list[1].t == NULL || list[2].t == NULL) | |
| 2902 goto theend; | |
| 2903 vim_memset(list[0].t, 0, size); | |
| 2904 vim_memset(list[1].t, 0, size); | |
| 2905 vim_memset(list[2].t, 0, size); | |
| 2906 | |
| 2907 #ifdef ENABLE_LOG | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2908 log_fd = fopen(NFA_REGEXP_RUN_LOG, "a"); |
| 4444 | 2909 if (log_fd != NULL) |
| 2910 { | |
| 2911 fprintf(log_fd, "**********************************\n"); | |
| 2912 nfa_set_code(start->c); | |
| 2913 fprintf(log_fd, " RUNNING nfa_regmatch() starting with state %d, code %s\n", | |
| 2914 abs(start->id), code); | |
| 2915 fprintf(log_fd, "**********************************\n"); | |
| 2916 } | |
| 2917 else | |
| 2918 { | |
| 2919 EMSG(_("Could not open temporary log file for writing, displaying on stderr ... ")); | |
| 2920 log_fd = stderr; | |
| 2921 } | |
| 2922 #endif | |
| 2923 | |
| 2924 thislist = &list[0]; | |
| 2925 thislist->n = 0; | |
| 2926 nextlist = &list[1]; | |
| 2927 nextlist->n = 0; | |
| 2928 neglist = &list[2]; | |
| 2929 neglist->n = 0; | |
| 2930 #ifdef ENABLE_LOG | |
| 2931 fprintf(log_fd, "(---) STARTSTATE\n"); | |
| 2932 #endif | |
| 2933 addstate(thislist, start, m, 0, listid, &match); | |
| 2934 | |
| 2935 /* There are two cases when the NFA advances: 1. input char matches the | |
| 2936 * NFA node and 2. input char does not match the NFA node, but the next | |
| 2937 * node is NFA_NOT. The following macro calls addstate() according to | |
| 2938 * these rules. It is used A LOT, so use the "listtbl" table for speed */ | |
| 2939 listtbl[0][0] = NULL; | |
| 2940 listtbl[0][1] = neglist; | |
| 2941 listtbl[1][0] = nextlist; | |
| 2942 listtbl[1][1] = NULL; | |
| 2943 #define ADD_POS_NEG_STATE(node) \ | |
| 2944 ll = listtbl[result ? 1 : 0][node->negated]; \ | |
| 2945 if (ll != NULL) \ | |
| 2946 addstate(ll, node->out , &t->sub, n, listid + 1, &match); | |
| 2947 | |
| 2948 | |
| 2949 /* | |
| 2950 * Run for each character. | |
| 2951 */ | |
|
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2952 for (;;) |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2953 { |
| 4444 | 2954 #ifdef FEAT_MBYTE |
| 2955 if (has_mbyte) | |
| 2956 { | |
| 2957 c = (*mb_ptr2char)(reginput); | |
| 2958 n = (*mb_ptr2len)(reginput); | |
| 2959 } | |
| 2960 else | |
| 2961 #endif | |
| 2962 { | |
| 2963 c = *reginput; | |
| 2964 n = 1; | |
| 2965 } | |
| 2966 if (c == NUL) | |
|
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2967 { |
| 4444 | 2968 n = 0; |
|
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2969 go_to_nextline = FALSE; |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2970 } |
| 4444 | 2971 |
| 2972 /* swap lists */ | |
| 2973 thislist = &list[flag]; | |
| 2974 nextlist = &list[flag ^= 1]; | |
| 2975 nextlist->n = 0; /* `clear' nextlist */ | |
| 2976 listtbl[1][0] = nextlist; | |
| 2977 ++listid; | |
| 2978 | |
| 2979 #ifdef ENABLE_LOG | |
| 2980 fprintf(log_fd, "------------------------------------------\n"); | |
| 2981 fprintf(log_fd, ">>> Reginput is \"%s\"\n", reginput); | |
| 2982 fprintf(log_fd, ">>> Advanced one character ... Current char is %c (code %d) \n", c, (int)c); | |
| 2983 fprintf(log_fd, ">>> Thislist has %d states available: ", thislist->n); | |
| 2984 for (i = 0; i< thislist->n; i++) | |
| 2985 fprintf(log_fd, "%d ", abs(thislist->t[i].state->id)); | |
| 2986 fprintf(log_fd, "\n"); | |
| 2987 #endif | |
| 2988 | |
| 4460 | 2989 #ifdef NFA_REGEXP_DEBUG_LOG |
| 4444 | 2990 fprintf(debug, "\n-------------------\n"); |
| 2991 #endif | |
| 4480 | 2992 /* |
| 2993 * If the state lists are empty we can stop. | |
| 2994 */ | |
| 2995 if (thislist->n == 0 && neglist->n == 0) | |
| 2996 break; | |
| 4444 | 2997 |
| 2998 /* compute nextlist */ | |
| 2999 for (i = 0; i < thislist->n || neglist->n > 0; ++i) | |
| 3000 { | |
| 3001 if (neglist->n > 0) | |
| 3002 { | |
| 3003 t = &neglist->t[0]; | |
| 4468 | 3004 neglist->n--; |
| 4444 | 3005 i--; |
| 3006 } | |
| 3007 else | |
| 3008 t = &thislist->t[i]; | |
| 3009 | |
| 4460 | 3010 #ifdef NFA_REGEXP_DEBUG_LOG |
| 4444 | 3011 nfa_set_code(t->state->c); |
| 3012 fprintf(debug, "%s, ", code); | |
| 3013 #endif | |
| 3014 #ifdef ENABLE_LOG | |
| 3015 nfa_set_code(t->state->c); | |
| 3016 fprintf(log_fd, "(%d) %s, code %d ... \n", abs(t->state->id), | |
| 3017 code, (int)t->state->c); | |
| 3018 #endif | |
| 3019 | |
| 3020 /* | |
| 3021 * Handle the possible codes of the current state. | |
| 3022 * The most important is NFA_MATCH. | |
| 3023 */ | |
| 3024 switch (t->state->c) | |
| 3025 { | |
| 3026 case NFA_MATCH: | |
| 3027 match = TRUE; | |
| 3028 *submatch = t->sub; | |
| 3029 #ifdef ENABLE_LOG | |
| 3030 for (j = 0; j < 4; j++) | |
| 3031 if (REG_MULTI) | |
| 3032 fprintf(log_fd, "\n *** group %d, start: c=%d, l=%d, end: c=%d, l=%d", | |
| 3033 j, | |
| 3034 t->sub.startpos[j].col, | |
| 3035 (int)t->sub.startpos[j].lnum, | |
| 3036 t->sub.endpos[j].col, | |
| 3037 (int)t->sub.endpos[j].lnum); | |
| 3038 else | |
| 3039 fprintf(log_fd, "\n *** group %d, start: \"%s\", end: \"%s\"", | |
| 3040 j, | |
| 3041 (char *)t->sub.start[j], | |
| 3042 (char *)t->sub.end[j]); | |
| 3043 fprintf(log_fd, "\n"); | |
| 3044 #endif | |
|
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3045 /* Found the left-most longest match, do not look at any other |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3046 * states at this position. */ |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3047 goto nextchar; |
| 4444 | 3048 |
| 3049 case NFA_END_INVISIBLE: | |
| 3050 /* This is only encountered after a NFA_START_INVISIBLE node. | |
| 3051 * They surround a zero-width group, used with "\@=" and "\&". | |
| 3052 * If we got here, it means that the current "invisible" group | |
| 3053 * finished successfully, so return control to the parent | |
| 3054 * nfa_regmatch(). Submatches are stored in *m, and used in | |
| 3055 * the parent call. */ | |
| 3056 if (start->c == NFA_MOPEN + 0) | |
| 3057 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
| 3058 &match); | |
| 3059 else | |
| 3060 { | |
| 3061 *m = t->sub; | |
| 3062 match = TRUE; | |
| 3063 } | |
| 3064 break; | |
| 3065 | |
| 3066 case NFA_START_INVISIBLE: | |
| 3067 /* Save global variables, and call nfa_regmatch() to check if | |
| 3068 * the current concat matches at this position. The concat | |
| 3069 * ends with the node NFA_END_INVISIBLE */ | |
| 3070 old_reginput = reginput; | |
| 3071 old_regline = regline; | |
| 3072 old_reglnum = reglnum; | |
| 3073 if (listids == NULL) | |
| 3074 { | |
| 3075 listids = (int *) lalloc(sizeof(int) * nstate, TRUE); | |
| 3076 if (listids == NULL) | |
| 3077 { | |
| 3078 EMSG(_("E878: (NFA) Could not allocate memory for branch traversal!")); | |
| 3079 return 0; | |
| 3080 } | |
| 3081 } | |
| 3082 #ifdef ENABLE_LOG | |
| 3083 if (log_fd != stderr) | |
| 3084 fclose(log_fd); | |
| 3085 log_fd = NULL; | |
| 3086 #endif | |
| 3087 /* Have to clear the listid field of the NFA nodes, so that | |
| 3088 * nfa_regmatch() and addstate() can run properly after | |
| 3089 * recursion. */ | |
| 3090 nfa_save_listids(start, listids); | |
| 3091 nfa_set_null_listids(start); | |
| 3092 result = nfa_regmatch(t->state->out, submatch, m); | |
| 3093 nfa_set_neg_listids(start); | |
| 3094 nfa_restore_listids(start, listids); | |
| 3095 | |
| 3096 #ifdef ENABLE_LOG | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
3097 log_fd = fopen(NFA_REGEXP_RUN_LOG, "a"); |
| 4444 | 3098 if (log_fd != NULL) |
| 3099 { | |
| 3100 fprintf(log_fd, "****************************\n"); | |
| 3101 fprintf(log_fd, "FINISHED RUNNING nfa_regmatch() recursively\n"); | |
| 3102 fprintf(log_fd, "MATCH = %s\n", result == TRUE ? "OK" : "FALSE"); | |
| 3103 fprintf(log_fd, "****************************\n"); | |
| 3104 } | |
| 3105 else | |
| 3106 { | |
| 3107 EMSG(_("Could not open temporary log file for writing, displaying on stderr ... ")); | |
| 3108 log_fd = stderr; | |
| 3109 } | |
| 3110 #endif | |
| 3111 if (result == TRUE) | |
| 3112 { | |
| 3113 /* Restore position in input text */ | |
| 3114 reginput = old_reginput; | |
| 3115 regline = old_regline; | |
| 3116 reglnum = old_reglnum; | |
| 3117 /* Copy submatch info from the recursive call */ | |
| 3118 if (REG_MULTI) | |
| 3119 for (j = 1; j < NSUBEXP; j++) | |
| 3120 { | |
| 3121 t->sub.startpos[j] = m->startpos[j]; | |
| 3122 t->sub.endpos[j] = m->endpos[j]; | |
| 3123 } | |
| 3124 else | |
| 3125 for (j = 1; j < NSUBEXP; j++) | |
| 3126 { | |
| 3127 t->sub.start[j] = m->start[j]; | |
| 3128 t->sub.end[j] = m->end[j]; | |
| 3129 } | |
| 3130 /* t->state->out1 is the corresponding END_INVISIBLE node */ | |
| 3131 addstate(thislist, t->state->out1->out, &t->sub, 0, listid, | |
| 3132 &match); | |
| 3133 } | |
| 3134 else | |
| 3135 { | |
| 3136 /* continue with next input char */ | |
| 3137 reginput = old_reginput; | |
| 3138 } | |
| 3139 break; | |
| 3140 | |
| 3141 case NFA_BOL: | |
| 3142 if (reginput == regline) | |
| 3143 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
| 3144 &match); | |
| 3145 break; | |
| 3146 | |
| 3147 case NFA_EOL: | |
| 3148 if (c == NUL) | |
| 3149 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
| 3150 &match); | |
| 3151 break; | |
| 3152 | |
| 3153 case NFA_BOW: | |
| 3154 { | |
| 3155 int bow = TRUE; | |
| 3156 | |
| 3157 if (c == NUL) | |
| 3158 bow = FALSE; | |
| 3159 #ifdef FEAT_MBYTE | |
| 3160 else if (has_mbyte) | |
| 3161 { | |
| 3162 int this_class; | |
| 3163 | |
| 3164 /* Get class of current and previous char (if it exists). */ | |
| 4503 | 3165 this_class = mb_get_class_buf(reginput, reg_buf); |
| 4444 | 3166 if (this_class <= 1) |
| 3167 bow = FALSE; | |
| 3168 else if (reg_prev_class() == this_class) | |
| 3169 bow = FALSE; | |
| 3170 } | |
| 3171 #endif | |
| 4503 | 3172 else if (!vim_iswordc_buf(c, reg_buf) |
| 3173 || (reginput > regline | |
| 3174 && vim_iswordc_buf(reginput[-1], reg_buf))) | |
| 4444 | 3175 bow = FALSE; |
| 3176 if (bow) | |
| 3177 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
| 3178 &match); | |
| 3179 break; | |
| 3180 } | |
| 3181 | |
| 3182 case NFA_EOW: | |
| 3183 { | |
| 3184 int eow = TRUE; | |
| 3185 | |
| 3186 if (reginput == regline) | |
| 3187 eow = FALSE; | |
| 3188 #ifdef FEAT_MBYTE | |
| 3189 else if (has_mbyte) | |
| 3190 { | |
| 3191 int this_class, prev_class; | |
| 3192 | |
| 3193 /* Get class of current and previous char (if it exists). */ | |
| 4503 | 3194 this_class = mb_get_class_buf(reginput, reg_buf); |
| 4444 | 3195 prev_class = reg_prev_class(); |
| 3196 if (this_class == prev_class | |
| 3197 || prev_class == 0 || prev_class == 1) | |
| 3198 eow = FALSE; | |
| 3199 } | |
| 3200 #endif | |
| 4503 | 3201 else if (!vim_iswordc_buf(reginput[-1], reg_buf) |
| 3202 || (reginput[0] != NUL && vim_iswordc_buf(c, reg_buf))) | |
| 4444 | 3203 eow = FALSE; |
| 3204 if (eow) | |
| 3205 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
| 3206 &match); | |
| 3207 break; | |
| 3208 } | |
| 3209 | |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3210 #ifdef FEAT_MBYTE |
| 4444 | 3211 case NFA_COMPOSING: |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3212 { |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
3213 int mc = c; |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
3214 int len = 0; |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
3215 nfa_state_T *end; |
|
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
3216 nfa_state_T *sta; |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3217 |
| 4444 | 3218 result = OK; |
| 3219 sta = t->state->out; | |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3220 len = 0; |
|
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3221 if (ireg_icombine) |
| 4444 | 3222 { |
|
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3223 /* If \Z was present, then ignore composing characters. */ |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3224 /* TODO: How about negated? */ |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3225 if (sta->c != c) |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3226 result = FAIL; |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3227 len = n; |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3228 while (sta->c != NFA_END_COMPOSING) |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3229 sta = sta->out; |
| 4444 | 3230 } |
|
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3231 else |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3232 while (sta->c != NFA_END_COMPOSING && len < n) |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3233 { |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3234 if (len > 0) |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3235 mc = mb_ptr2char(reginput + len); |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3236 if (mc != sta->c) |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3237 break; |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3238 len += mb_char2len(mc); |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3239 sta = sta->out; |
|
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3240 } |
| 4444 | 3241 |
| 3242 /* if input char length doesn't match regexp char length */ | |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3243 if (len < n || sta->c != NFA_END_COMPOSING) |
|
4525
36ddcf4cecbc
updated for version 7.3.1010
Bram Moolenaar <bram@vim.org>
parents:
4517
diff
changeset
|
3244 result = FAIL; |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3245 end = t->state->out1; /* NFA_END_COMPOSING */ |
| 4444 | 3246 ADD_POS_NEG_STATE(end); |
| 3247 break; | |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3248 } |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3249 #endif |
| 4444 | 3250 |
| 3251 case NFA_NEWL: | |
| 3252 if (!reg_line_lbr && REG_MULTI | |
|
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3253 && c == NUL && reglnum <= reg_maxline) |
| 4444 | 3254 { |
|
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3255 go_to_nextline = TRUE; |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3256 /* Pass -1 for the offset, which means taking the position |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3257 * at the start of the next line. */ |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3258 addstate(nextlist, t->state->out, &t->sub, -1, |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3259 listid + 1, &match); |
| 4444 | 3260 } |
| 3261 break; | |
| 3262 | |
| 3263 case NFA_CLASS_ALNUM: | |
| 3264 case NFA_CLASS_ALPHA: | |
| 3265 case NFA_CLASS_BLANK: | |
| 3266 case NFA_CLASS_CNTRL: | |
| 3267 case NFA_CLASS_DIGIT: | |
| 3268 case NFA_CLASS_GRAPH: | |
| 3269 case NFA_CLASS_LOWER: | |
| 3270 case NFA_CLASS_PRINT: | |
| 3271 case NFA_CLASS_PUNCT: | |
| 3272 case NFA_CLASS_SPACE: | |
| 3273 case NFA_CLASS_UPPER: | |
| 3274 case NFA_CLASS_XDIGIT: | |
| 3275 case NFA_CLASS_TAB: | |
| 3276 case NFA_CLASS_RETURN: | |
| 3277 case NFA_CLASS_BACKSPACE: | |
| 3278 case NFA_CLASS_ESCAPE: | |
| 3279 result = check_char_class(t->state->c, c); | |
| 3280 ADD_POS_NEG_STATE(t->state); | |
| 3281 break; | |
| 3282 | |
| 3283 case NFA_END_NEG_RANGE: | |
| 3284 /* This follows a series of negated nodes, like: | |
| 3285 * CHAR(x), NFA_NOT, CHAR(y), NFA_NOT etc. */ | |
| 3286 if (c > 0) | |
| 3287 addstate(nextlist, t->state->out, &t->sub, n, listid + 1, | |
| 3288 &match); | |
| 3289 break; | |
| 3290 | |
| 3291 case NFA_ANY: | |
|
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3292 /* Any char except '\0', (end of input) does not match. */ |
| 4444 | 3293 if (c > 0) |
| 3294 addstate(nextlist, t->state->out, &t->sub, n, listid + 1, | |
| 3295 &match); | |
| 3296 break; | |
| 3297 | |
| 3298 /* | |
| 3299 * Character classes like \a for alpha, \d for digit etc. | |
| 3300 */ | |
| 3301 case NFA_IDENT: /* \i */ | |
| 3302 result = vim_isIDc(c); | |
| 3303 ADD_POS_NEG_STATE(t->state); | |
| 3304 break; | |
| 3305 | |
| 3306 case NFA_SIDENT: /* \I */ | |
| 3307 result = !VIM_ISDIGIT(c) && vim_isIDc(c); | |
| 3308 ADD_POS_NEG_STATE(t->state); | |
| 3309 break; | |
| 3310 | |
| 3311 case NFA_KWORD: /* \k */ | |
| 4503 | 3312 result = vim_iswordp_buf(reginput, reg_buf); |
| 4444 | 3313 ADD_POS_NEG_STATE(t->state); |
| 3314 break; | |
| 3315 | |
| 3316 case NFA_SKWORD: /* \K */ | |
| 4503 | 3317 result = !VIM_ISDIGIT(c) && vim_iswordp_buf(reginput, reg_buf); |
| 4444 | 3318 ADD_POS_NEG_STATE(t->state); |
| 3319 break; | |
| 3320 | |
| 3321 case NFA_FNAME: /* \f */ | |
| 3322 result = vim_isfilec(c); | |
| 3323 ADD_POS_NEG_STATE(t->state); | |
| 3324 break; | |
| 3325 | |
| 3326 case NFA_SFNAME: /* \F */ | |
| 3327 result = !VIM_ISDIGIT(c) && vim_isfilec(c); | |
| 3328 ADD_POS_NEG_STATE(t->state); | |
| 3329 break; | |
| 3330 | |
| 3331 case NFA_PRINT: /* \p */ | |
| 4470 | 3332 result = ptr2cells(reginput) == 1; |
| 4444 | 3333 ADD_POS_NEG_STATE(t->state); |
| 3334 break; | |
| 3335 | |
| 3336 case NFA_SPRINT: /* \P */ | |
| 4470 | 3337 result = !VIM_ISDIGIT(c) && ptr2cells(reginput) == 1; |
| 4444 | 3338 ADD_POS_NEG_STATE(t->state); |
| 3339 break; | |
| 3340 | |
| 3341 case NFA_WHITE: /* \s */ | |
| 3342 result = vim_iswhite(c); | |
| 3343 ADD_POS_NEG_STATE(t->state); | |
| 3344 break; | |
| 3345 | |
| 3346 case NFA_NWHITE: /* \S */ | |
| 3347 result = c != NUL && !vim_iswhite(c); | |
| 3348 ADD_POS_NEG_STATE(t->state); | |
| 3349 break; | |
| 3350 | |
| 3351 case NFA_DIGIT: /* \d */ | |
| 3352 result = ri_digit(c); | |
| 3353 ADD_POS_NEG_STATE(t->state); | |
| 3354 break; | |
| 3355 | |
| 3356 case NFA_NDIGIT: /* \D */ | |
| 3357 result = c != NUL && !ri_digit(c); | |
| 3358 ADD_POS_NEG_STATE(t->state); | |
| 3359 break; | |
| 3360 | |
| 3361 case NFA_HEX: /* \x */ | |
| 3362 result = ri_hex(c); | |
| 3363 ADD_POS_NEG_STATE(t->state); | |
| 3364 break; | |
| 3365 | |
| 3366 case NFA_NHEX: /* \X */ | |
| 3367 result = c != NUL && !ri_hex(c); | |
| 3368 ADD_POS_NEG_STATE(t->state); | |
| 3369 break; | |
| 3370 | |
| 3371 case NFA_OCTAL: /* \o */ | |
| 3372 result = ri_octal(c); | |
| 3373 ADD_POS_NEG_STATE(t->state); | |
| 3374 break; | |
| 3375 | |
| 3376 case NFA_NOCTAL: /* \O */ | |
| 3377 result = c != NUL && !ri_octal(c); | |
| 3378 ADD_POS_NEG_STATE(t->state); | |
| 3379 break; | |
| 3380 | |
| 3381 case NFA_WORD: /* \w */ | |
| 3382 result = ri_word(c); | |
| 3383 ADD_POS_NEG_STATE(t->state); | |
| 3384 break; | |
| 3385 | |
| 3386 case NFA_NWORD: /* \W */ | |
| 3387 result = c != NUL && !ri_word(c); | |
| 3388 ADD_POS_NEG_STATE(t->state); | |
| 3389 break; | |
| 3390 | |
| 3391 case NFA_HEAD: /* \h */ | |
| 3392 result = ri_head(c); | |
| 3393 ADD_POS_NEG_STATE(t->state); | |
| 3394 break; | |
| 3395 | |
| 3396 case NFA_NHEAD: /* \H */ | |
| 3397 result = c != NUL && !ri_head(c); | |
| 3398 ADD_POS_NEG_STATE(t->state); | |
| 3399 break; | |
| 3400 | |
| 3401 case NFA_ALPHA: /* \a */ | |
| 3402 result = ri_alpha(c); | |
| 3403 ADD_POS_NEG_STATE(t->state); | |
| 3404 break; | |
| 3405 | |
| 3406 case NFA_NALPHA: /* \A */ | |
| 3407 result = c != NUL && !ri_alpha(c); | |
| 3408 ADD_POS_NEG_STATE(t->state); | |
| 3409 break; | |
| 3410 | |
| 3411 case NFA_LOWER: /* \l */ | |
| 3412 result = ri_lower(c); | |
| 3413 ADD_POS_NEG_STATE(t->state); | |
| 3414 break; | |
| 3415 | |
| 3416 case NFA_NLOWER: /* \L */ | |
| 3417 result = c != NUL && !ri_lower(c); | |
| 3418 ADD_POS_NEG_STATE(t->state); | |
| 3419 break; | |
| 3420 | |
| 3421 case NFA_UPPER: /* \u */ | |
| 3422 result = ri_upper(c); | |
| 3423 ADD_POS_NEG_STATE(t->state); | |
| 3424 break; | |
| 3425 | |
| 3426 case NFA_NUPPER: /* \U */ | |
| 3427 result = c != NUL && !ri_upper(c); | |
| 3428 ADD_POS_NEG_STATE(t->state); | |
| 3429 break; | |
| 3430 | |
| 4482 | 3431 case NFA_MOPEN + 0: |
| 3432 case NFA_MOPEN + 1: | |
| 3433 case NFA_MOPEN + 2: | |
| 3434 case NFA_MOPEN + 3: | |
| 3435 case NFA_MOPEN + 4: | |
| 3436 case NFA_MOPEN + 5: | |
| 3437 case NFA_MOPEN + 6: | |
| 3438 case NFA_MOPEN + 7: | |
| 3439 case NFA_MOPEN + 8: | |
| 3440 case NFA_MOPEN + 9: | |
| 3441 /* handled below */ | |
| 3442 break; | |
| 3443 | |
| 3444 case NFA_SKIP_CHAR: | |
| 3445 case NFA_ZSTART: | |
| 3446 /* TODO: should not happen? */ | |
| 3447 break; | |
| 3448 | |
| 4444 | 3449 default: /* regular character */ |
| 4482 | 3450 /* TODO: put this in #ifdef later */ |
| 3451 if (t->state->c < -256) | |
| 3452 EMSGN("INTERNAL: Negative state char: %ld", t->state->c); | |
| 4444 | 3453 result = (no_Magic(t->state->c) == c); |
| 4482 | 3454 |
| 4444 | 3455 if (!result) |
| 3456 result = ireg_ic == TRUE | |
| 3457 && MB_TOLOWER(t->state->c) == MB_TOLOWER(c); | |
|
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3458 #ifdef FEAT_MBYTE |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3459 /* If there is a composing character which is not being |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3460 * ignored there can be no match. Match with composing |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3461 * character uses NFA_COMPOSING above. */ |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3462 if (result && enc_utf8 && !ireg_icombine |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3463 && n != utf_char2len(c)) |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3464 result = FALSE; |
|
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3465 #endif |
| 4444 | 3466 ADD_POS_NEG_STATE(t->state); |
| 3467 break; | |
| 3468 } | |
| 3469 | |
| 3470 } /* for (thislist = thislist; thislist->state; thislist++) */ | |
| 3471 | |
| 3472 /* The first found match is the leftmost one, but there may be a | |
| 3473 * longer one. Keep running the NFA, but don't start from the | |
| 3474 * beginning. Also, do not add the start state in recursive calls of | |
| 3475 * nfa_regmatch(), because recursive calls should only start in the | |
| 3476 * first position. */ | |
| 3477 if (match == FALSE && start->c == NFA_MOPEN + 0) | |
| 3478 { | |
| 3479 #ifdef ENABLE_LOG | |
| 3480 fprintf(log_fd, "(---) STARTSTATE\n"); | |
| 3481 #endif | |
| 3482 addstate(nextlist, start, m, n, listid + 1, &match); | |
| 3483 } | |
| 3484 | |
| 3485 #ifdef ENABLE_LOG | |
| 3486 fprintf(log_fd, ">>> Thislist had %d states available: ", thislist->n); | |
| 3487 for (i = 0; i< thislist->n; i++) | |
| 3488 fprintf(log_fd, "%d ", abs(thislist->t[i].state->id)); | |
| 3489 fprintf(log_fd, "\n"); | |
| 3490 #endif | |
| 3491 | |
| 3492 nextchar: | |
|
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3493 /* Advance to the next character, or advance to the next line, or |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3494 * finish. */ |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3495 if (n != 0) |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3496 reginput += n; |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3497 else if (go_to_nextline) |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3498 reg_nextline(); |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3499 else |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3500 break; |
|
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3501 } |
| 4444 | 3502 |
| 3503 #ifdef ENABLE_LOG | |
| 3504 if (log_fd != stderr) | |
| 3505 fclose(log_fd); | |
| 3506 log_fd = NULL; | |
| 3507 #endif | |
| 3508 | |
| 3509 theend: | |
| 3510 /* Free memory */ | |
| 3511 vim_free(list[0].t); | |
| 3512 vim_free(list[1].t); | |
| 3513 vim_free(list[2].t); | |
| 3514 list[0].t = list[1].t = list[2].t = NULL; | |
| 3515 if (listids != NULL) | |
| 3516 vim_free(listids); | |
| 3517 #undef ADD_POS_NEG_STATE | |
| 4460 | 3518 #ifdef NFA_REGEXP_DEBUG_LOG |
| 4444 | 3519 fclose(debug); |
| 3520 #endif | |
| 3521 | |
| 3522 return match; | |
| 3523 } | |
| 3524 | |
| 3525 /* | |
| 3526 * Try match of "prog" with at regline["col"]. | |
| 3527 * Returns 0 for failure, number of lines contained in the match otherwise. | |
| 3528 */ | |
| 3529 static long | |
| 3530 nfa_regtry(start, col) | |
| 3531 nfa_state_T *start; | |
| 3532 colnr_T col; | |
| 3533 { | |
| 3534 int i; | |
| 3535 regsub_T sub, m; | |
| 3536 #ifdef ENABLE_LOG | |
| 3537 FILE *f; | |
| 3538 #endif | |
| 3539 | |
| 3540 reginput = regline + col; | |
| 3541 need_clear_subexpr = TRUE; | |
| 3542 | |
| 3543 #ifdef ENABLE_LOG | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
3544 f = fopen(NFA_REGEXP_RUN_LOG, "a"); |
| 4444 | 3545 if (f != NULL) |
| 3546 { | |
| 3547 fprintf(f, "\n\n\n\n\n\n\t\t=======================================================\n"); | |
| 3548 fprintf(f, " =======================================================\n"); | |
| 3549 #ifdef DEBUG | |
| 3550 fprintf(f, "\tRegexp is \"%s\"\n", nfa_regengine.expr); | |
| 3551 #endif | |
| 3552 fprintf(f, "\tInput text is \"%s\" \n", reginput); | |
| 3553 fprintf(f, " =======================================================\n\n\n\n\n\n\n"); | |
|
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
3554 nfa_print_state(f, start); |
| 4444 | 3555 fprintf(f, "\n\n"); |
| 3556 fclose(f); | |
| 3557 } | |
| 3558 else | |
| 3559 EMSG(_("Could not open temporary log file for writing ")); | |
| 3560 #endif | |
| 3561 | |
| 3562 if (REG_MULTI) | |
| 3563 { | |
| 3564 /* Use 0xff to set lnum to -1 */ | |
| 3565 vim_memset(sub.startpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
| 3566 vim_memset(sub.endpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
| 3567 vim_memset(m.startpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
| 3568 vim_memset(m.endpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
| 3569 } | |
| 3570 else | |
| 3571 { | |
| 3572 vim_memset(sub.start, 0, sizeof(char_u *) * NSUBEXP); | |
| 3573 vim_memset(sub.end, 0, sizeof(char_u *) * NSUBEXP); | |
| 3574 vim_memset(m.start, 0, sizeof(char_u *) * NSUBEXP); | |
| 3575 vim_memset(m.end, 0, sizeof(char_u *) * NSUBEXP); | |
| 3576 } | |
| 3577 | |
| 3578 if (nfa_regmatch(start, &sub, &m) == FALSE) | |
| 3579 return 0; | |
| 3580 | |
| 3581 cleanup_subexpr(); | |
| 3582 if (REG_MULTI) | |
| 3583 { | |
| 3584 for (i = 0; i < NSUBEXP; i++) | |
| 3585 { | |
| 3586 reg_startpos[i] = sub.startpos[i]; | |
| 3587 reg_endpos[i] = sub.endpos[i]; | |
| 3588 } | |
| 3589 | |
| 3590 if (reg_startpos[0].lnum < 0) | |
| 3591 { | |
| 3592 reg_startpos[0].lnum = 0; | |
| 3593 reg_startpos[0].col = col; | |
| 3594 } | |
| 3595 if (reg_endpos[0].lnum < 0) | |
| 3596 { | |
| 3597 reg_endpos[0].lnum = reglnum; | |
| 3598 reg_endpos[0].col = (int)(reginput - regline); | |
| 3599 } | |
| 3600 else | |
| 3601 /* Use line number of "\ze". */ | |
| 3602 reglnum = reg_endpos[0].lnum; | |
| 3603 } | |
| 3604 else | |
| 3605 { | |
| 3606 for (i = 0; i < NSUBEXP; i++) | |
| 3607 { | |
| 3608 reg_startp[i] = sub.start[i]; | |
| 3609 reg_endp[i] = sub.end[i]; | |
| 3610 } | |
| 3611 | |
| 3612 if (reg_startp[0] == NULL) | |
| 3613 reg_startp[0] = regline + col; | |
| 3614 if (reg_endp[0] == NULL) | |
| 3615 reg_endp[0] = reginput; | |
| 3616 } | |
| 3617 | |
| 3618 return 1 + reglnum; | |
| 3619 } | |
| 3620 | |
| 3621 /* | |
| 3622 * Match a regexp against a string ("line" points to the string) or multiple | |
| 3623 * lines ("line" is NULL, use reg_getline()). | |
| 3624 * | |
| 3625 * Returns 0 for failure, number of lines contained in the match otherwise. | |
| 3626 */ | |
| 3627 static long | |
| 3628 nfa_regexec_both(line, col) | |
| 3629 char_u *line; | |
| 3630 colnr_T col; /* column to start looking for match */ | |
| 3631 { | |
| 3632 nfa_regprog_T *prog; | |
| 3633 long retval = 0L; | |
| 3634 int i; | |
| 3635 | |
| 3636 if (REG_MULTI) | |
| 3637 { | |
| 3638 prog = (nfa_regprog_T *)reg_mmatch->regprog; | |
| 3639 line = reg_getline((linenr_T)0); /* relative to the cursor */ | |
| 3640 reg_startpos = reg_mmatch->startpos; | |
| 3641 reg_endpos = reg_mmatch->endpos; | |
| 3642 } | |
| 3643 else | |
| 3644 { | |
| 3645 prog = (nfa_regprog_T *)reg_match->regprog; | |
| 3646 reg_startp = reg_match->startp; | |
| 3647 reg_endp = reg_match->endp; | |
| 3648 } | |
| 3649 | |
| 3650 /* Be paranoid... */ | |
| 3651 if (prog == NULL || line == NULL) | |
| 3652 { | |
| 3653 EMSG(_(e_null)); | |
| 3654 goto theend; | |
| 3655 } | |
| 3656 | |
| 3657 /* If the start column is past the maximum column: no need to try. */ | |
| 3658 if (ireg_maxcol > 0 && col >= ireg_maxcol) | |
| 3659 goto theend; | |
| 3660 | |
| 3661 /* If pattern contains "\c" or "\C": overrule value of ireg_ic */ | |
| 3662 if (prog->regflags & RF_ICASE) | |
| 3663 ireg_ic = TRUE; | |
| 3664 else if (prog->regflags & RF_NOICASE) | |
| 3665 ireg_ic = FALSE; | |
| 3666 | |
| 3667 #ifdef FEAT_MBYTE | |
| 3668 /* If pattern contains "\Z" overrule value of ireg_icombine */ | |
| 3669 if (prog->regflags & RF_ICOMBINE) | |
| 3670 ireg_icombine = TRUE; | |
| 3671 #endif | |
| 3672 | |
| 3673 regline = line; | |
| 3674 reglnum = 0; /* relative to line */ | |
| 3675 | |
| 3676 nstate = prog->nstate; | |
| 3677 | |
| 3678 for (i = 0; i < nstate; ++i) | |
| 3679 { | |
| 3680 prog->state[i].id = i; | |
| 3681 prog->state[i].lastlist = 0; | |
| 3682 prog->state[i].visits = 0; | |
| 3683 prog->state[i].lastthread = NULL; | |
| 3684 } | |
| 3685 | |
| 3686 retval = nfa_regtry(prog->start, col); | |
| 3687 | |
| 3688 theend: | |
| 3689 return retval; | |
| 3690 } | |
| 3691 | |
| 3692 /* | |
| 3693 * Compile a regular expression into internal code for the NFA matcher. | |
| 3694 * Returns the program in allocated space. Returns NULL for an error. | |
| 3695 */ | |
| 3696 static regprog_T * | |
| 3697 nfa_regcomp(expr, re_flags) | |
| 3698 char_u *expr; | |
| 3699 int re_flags; | |
| 3700 { | |
| 3701 nfa_regprog_T *prog; | |
| 4458 | 3702 size_t prog_size; |
| 4444 | 3703 int *postfix; |
| 3704 | |
| 3705 if (expr == NULL) | |
| 3706 return NULL; | |
| 3707 | |
| 3708 #ifdef DEBUG | |
| 3709 nfa_regengine.expr = expr; | |
| 3710 #endif | |
| 3711 | |
| 3712 init_class_tab(); | |
| 3713 | |
| 3714 if (nfa_regcomp_start(expr, re_flags) == FAIL) | |
| 3715 return NULL; | |
| 3716 | |
| 3717 /* Space for compiled regexp */ | |
| 3718 prog_size = sizeof(nfa_regprog_T) + sizeof(nfa_state_T) * nstate_max; | |
| 3719 prog = (nfa_regprog_T *)lalloc(prog_size, TRUE); | |
| 3720 if (prog == NULL) | |
| 3721 goto fail; | |
| 3722 vim_memset(prog, 0, prog_size); | |
| 3723 | |
| 3724 /* Build postfix form of the regexp. Needed to build the NFA | |
| 3725 * (and count its size) */ | |
| 3726 postfix = re2post(); | |
| 3727 if (postfix == NULL) | |
| 3728 goto fail; /* Cascaded (syntax?) error */ | |
| 3729 | |
| 3730 /* | |
| 3731 * In order to build the NFA, we parse the input regexp twice: | |
| 3732 * 1. first pass to count size (so we can allocate space) | |
| 3733 * 2. second to emit code | |
| 3734 */ | |
| 3735 #ifdef ENABLE_LOG | |
| 3736 { | |
|
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
3737 FILE *f = fopen(NFA_REGEXP_RUN_LOG, "a"); |
| 4444 | 3738 |
| 3739 if (f != NULL) | |
| 3740 { | |
| 3741 fprintf(f, "\n*****************************\n\n\n\n\tCompiling regexp \"%s\" ... hold on !\n", expr); | |
| 3742 fclose(f); | |
| 3743 } | |
| 3744 } | |
| 3745 #endif | |
| 3746 | |
| 3747 /* | |
| 3748 * PASS 1 | |
| 3749 * Count number of NFA states in "nstate". Do not build the NFA. | |
| 3750 */ | |
| 3751 post2nfa(postfix, post_ptr, TRUE); | |
| 3752 state_ptr = prog->state; | |
| 3753 | |
| 3754 /* | |
| 3755 * PASS 2 | |
| 3756 * Build the NFA | |
| 3757 */ | |
| 3758 prog->start = post2nfa(postfix, post_ptr, FALSE); | |
| 3759 if (prog->start == NULL) | |
| 3760 goto fail; | |
| 3761 | |
| 3762 prog->regflags = regflags; | |
| 3763 prog->engine = &nfa_regengine; | |
| 3764 prog->nstate = nstate; | |
| 3765 #ifdef ENABLE_LOG | |
| 3766 nfa_postfix_dump(expr, OK); | |
| 3767 nfa_dump(prog); | |
| 3768 #endif | |
| 3769 | |
| 3770 out: | |
| 3771 vim_free(post_start); | |
| 3772 post_start = post_ptr = post_end = NULL; | |
| 3773 state_ptr = NULL; | |
| 3774 return (regprog_T *)prog; | |
| 3775 | |
| 3776 fail: | |
| 3777 vim_free(prog); | |
| 3778 prog = NULL; | |
| 3779 #ifdef ENABLE_LOG | |
| 3780 nfa_postfix_dump(expr, FAIL); | |
| 3781 #endif | |
| 3782 #ifdef DEBUG | |
| 3783 nfa_regengine.expr = NULL; | |
| 3784 #endif | |
| 3785 goto out; | |
| 3786 } | |
| 3787 | |
| 3788 | |
| 3789 /* | |
| 3790 * Match a regexp against a string. | |
| 3791 * "rmp->regprog" is a compiled regexp as returned by nfa_regcomp(). | |
| 3792 * Uses curbuf for line count and 'iskeyword'. | |
| 3793 * | |
| 3794 * Return TRUE if there is a match, FALSE if not. | |
| 3795 */ | |
| 3796 static int | |
| 3797 nfa_regexec(rmp, line, col) | |
| 3798 regmatch_T *rmp; | |
| 3799 char_u *line; /* string to match against */ | |
| 3800 colnr_T col; /* column to start looking for match */ | |
| 3801 { | |
| 3802 reg_match = rmp; | |
| 3803 reg_mmatch = NULL; | |
| 3804 reg_maxline = 0; | |
| 3805 reg_line_lbr = FALSE; | |
| 3806 reg_buf = curbuf; | |
| 3807 reg_win = NULL; | |
| 3808 ireg_ic = rmp->rm_ic; | |
| 3809 #ifdef FEAT_MBYTE | |
| 3810 ireg_icombine = FALSE; | |
| 3811 #endif | |
| 3812 ireg_maxcol = 0; | |
| 3813 return (nfa_regexec_both(line, col) != 0); | |
| 3814 } | |
| 3815 | |
| 3816 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ | |
| 3817 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
| 3818 | |
| 3819 static int nfa_regexec_nl __ARGS((regmatch_T *rmp, char_u *line, colnr_T col)); | |
| 3820 | |
| 3821 /* | |
| 3822 * Like nfa_regexec(), but consider a "\n" in "line" to be a line break. | |
| 3823 */ | |
| 3824 static int | |
| 3825 nfa_regexec_nl(rmp, line, col) | |
| 3826 regmatch_T *rmp; | |
| 3827 char_u *line; /* string to match against */ | |
| 3828 colnr_T col; /* column to start looking for match */ | |
| 3829 { | |
| 3830 reg_match = rmp; | |
| 3831 reg_mmatch = NULL; | |
| 3832 reg_maxline = 0; | |
| 3833 reg_line_lbr = TRUE; | |
| 3834 reg_buf = curbuf; | |
| 3835 reg_win = NULL; | |
| 3836 ireg_ic = rmp->rm_ic; | |
| 3837 #ifdef FEAT_MBYTE | |
| 3838 ireg_icombine = FALSE; | |
| 3839 #endif | |
| 3840 ireg_maxcol = 0; | |
| 3841 return (nfa_regexec_both(line, col) != 0); | |
| 3842 } | |
| 3843 #endif | |
| 3844 | |
| 3845 | |
| 3846 /* | |
| 3847 * Match a regexp against multiple lines. | |
| 3848 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
| 3849 * Uses curbuf for line count and 'iskeyword'. | |
| 3850 * | |
| 3851 * Return zero if there is no match. Return number of lines contained in the | |
| 3852 * match otherwise. | |
| 3853 * | |
| 3854 * Note: the body is the same as bt_regexec() except for nfa_regexec_both() | |
| 3855 * | |
| 3856 * ! Also NOTE : match may actually be in another line. e.g.: | |
| 3857 * when r.e. is \nc, cursor is at 'a' and the text buffer looks like | |
| 3858 * | |
| 3859 * +-------------------------+ | |
| 3860 * |a | | |
| 3861 * |b | | |
| 3862 * |c | | |
| 3863 * | | | |
| 3864 * +-------------------------+ | |
| 3865 * | |
| 3866 * then nfa_regexec_multi() returns 3. while the original | |
| 3867 * vim_regexec_multi() returns 0 and a second call at line 2 will return 2. | |
| 3868 * | |
| 3869 * FIXME if this behavior is not compatible. | |
| 3870 */ | |
| 3871 static long | |
| 3872 nfa_regexec_multi(rmp, win, buf, lnum, col, tm) | |
| 3873 regmmatch_T *rmp; | |
| 3874 win_T *win; /* window in which to search or NULL */ | |
| 3875 buf_T *buf; /* buffer in which to search */ | |
| 3876 linenr_T lnum; /* nr of line to start looking for match */ | |
| 3877 colnr_T col; /* column to start looking for match */ | |
| 3878 proftime_T *tm UNUSED; /* timeout limit or NULL */ | |
| 3879 { | |
| 3880 reg_match = NULL; | |
| 3881 reg_mmatch = rmp; | |
| 3882 reg_buf = buf; | |
| 3883 reg_win = win; | |
| 3884 reg_firstlnum = lnum; | |
| 3885 reg_maxline = reg_buf->b_ml.ml_line_count - lnum; | |
| 3886 reg_line_lbr = FALSE; | |
| 3887 ireg_ic = rmp->rmm_ic; | |
| 3888 #ifdef FEAT_MBYTE | |
| 3889 ireg_icombine = FALSE; | |
| 3890 #endif | |
| 3891 ireg_maxcol = rmp->rmm_maxcol; | |
| 3892 | |
| 4503 | 3893 return nfa_regexec_both(NULL, col); |
| 4444 | 3894 } |
| 3895 | |
| 3896 #ifdef DEBUG | |
| 3897 # undef ENABLE_LOG | |
| 3898 #endif |