Mercurial > vim
annotate src/regexp_nfa.c @ 4525:36ddcf4cecbc v7.3.1010
updated for version 7.3.1010
Problem: New regexp: adding \Z makes every character match.
Solution: Only apply ireg_icombine for composing characters.
Alsl add missing change from patch 1008. (Ken Takata)
author | Bram Moolenaar <bram@vim.org> |
---|---|
date | Fri, 24 May 2013 20:25:33 +0200 |
parents | 9a2183bd8295 |
children | 55bcaa1d2749 |
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 | |
8 #ifdef DEBUG | |
9 /* Comment this out to disable log files. They can get pretty big */ | |
10 # define ENABLE_LOG | |
11 # define LOG_NAME "log_nfarun.log" | |
4460 | 12 # define NFA_REGEXP_DEBUG_LOG |
13 # define NFA_REGEXP_DEBUG_LOG_NAME "nfa_regexp_debug.log" | |
4444 | 14 #endif |
15 | |
16 /* Upper limit allowed for {m,n} repetitions handled by NFA */ | |
17 #define NFA_BRACES_MAXLIMIT 10 | |
18 /* For allocating space for the postfix representation */ | |
19 #define NFA_POSTFIX_MULTIPLIER (NFA_BRACES_MAXLIMIT + 2)*2 | |
20 | |
21 enum | |
22 { | |
23 NFA_SPLIT = -1024, | |
24 NFA_MATCH, | |
25 NFA_SKIP_CHAR, /* matches a 0-length char */ | |
26 NFA_END_NEG_RANGE, /* Used when expanding [^ab] */ | |
27 | |
28 NFA_CONCAT, | |
29 NFA_OR, | |
30 NFA_STAR, | |
31 NFA_PLUS, | |
32 NFA_QUEST, | |
33 NFA_QUEST_NONGREEDY, /* Non-greedy version of \? */ | |
34 NFA_NOT, /* used for [^ab] negated char ranges */ | |
35 | |
36 NFA_BOL, /* ^ Begin line */ | |
37 NFA_EOL, /* $ End line */ | |
38 NFA_BOW, /* \< Begin word */ | |
39 NFA_EOW, /* \> End word */ | |
40 NFA_BOF, /* \%^ Begin file */ | |
41 NFA_EOF, /* \%$ End file */ | |
42 NFA_NEWL, | |
43 NFA_ZSTART, /* Used for \zs */ | |
44 NFA_ZEND, /* Used for \ze */ | |
45 NFA_NOPEN, /* Start of subexpression marked with \%( */ | |
46 NFA_NCLOSE, /* End of subexpr. marked with \%( ... \) */ | |
47 NFA_START_INVISIBLE, | |
48 NFA_END_INVISIBLE, | |
49 NFA_MULTIBYTE, /* Next nodes in NFA are part of the same | |
50 multibyte char */ | |
51 NFA_END_MULTIBYTE, /* End of multibyte char in the NFA */ | |
52 NFA_COMPOSING, /* Next nodes in NFA are part of the | |
53 composing multibyte char */ | |
54 NFA_END_COMPOSING, /* End of a composing char in the NFA */ | |
55 | |
56 /* The following are used only in the postfix form, not in the NFA */ | |
57 NFA_PREV_ATOM_NO_WIDTH, /* Used for \@= */ | |
58 NFA_PREV_ATOM_NO_WIDTH_NEG, /* Used for \@! */ | |
59 NFA_PREV_ATOM_JUST_BEFORE, /* Used for \@<= */ | |
60 NFA_PREV_ATOM_JUST_BEFORE_NEG, /* Used for \@<! */ | |
61 NFA_PREV_ATOM_LIKE_PATTERN, /* Used for \@> */ | |
62 | |
63 NFA_MOPEN, | |
64 NFA_MCLOSE = NFA_MOPEN + NSUBEXP, | |
65 | |
66 /* NFA_FIRST_NL */ | |
67 NFA_ANY = NFA_MCLOSE + NSUBEXP, /* Match any one character. */ | |
68 NFA_ANYOF, /* Match any character in this string. */ | |
69 NFA_ANYBUT, /* Match any character not in this string. */ | |
70 NFA_IDENT, /* Match identifier char */ | |
71 NFA_SIDENT, /* Match identifier char but no digit */ | |
72 NFA_KWORD, /* Match keyword char */ | |
73 NFA_SKWORD, /* Match word char but no digit */ | |
74 NFA_FNAME, /* Match file name char */ | |
75 NFA_SFNAME, /* Match file name char but no digit */ | |
76 NFA_PRINT, /* Match printable char */ | |
77 NFA_SPRINT, /* Match printable char but no digit */ | |
78 NFA_WHITE, /* Match whitespace char */ | |
79 NFA_NWHITE, /* Match non-whitespace char */ | |
80 NFA_DIGIT, /* Match digit char */ | |
81 NFA_NDIGIT, /* Match non-digit char */ | |
82 NFA_HEX, /* Match hex char */ | |
83 NFA_NHEX, /* Match non-hex char */ | |
84 NFA_OCTAL, /* Match octal char */ | |
85 NFA_NOCTAL, /* Match non-octal char */ | |
86 NFA_WORD, /* Match word char */ | |
87 NFA_NWORD, /* Match non-word char */ | |
88 NFA_HEAD, /* Match head char */ | |
89 NFA_NHEAD, /* Match non-head char */ | |
90 NFA_ALPHA, /* Match alpha char */ | |
91 NFA_NALPHA, /* Match non-alpha char */ | |
92 NFA_LOWER, /* Match lowercase char */ | |
93 NFA_NLOWER, /* Match non-lowercase char */ | |
94 NFA_UPPER, /* Match uppercase char */ | |
95 NFA_NUPPER, /* Match non-uppercase char */ | |
96 NFA_FIRST_NL = NFA_ANY + ADD_NL, | |
97 NFA_LAST_NL = NFA_NUPPER + ADD_NL, | |
98 | |
99 /* Character classes [:alnum:] etc */ | |
100 NFA_CLASS_ALNUM, | |
101 NFA_CLASS_ALPHA, | |
102 NFA_CLASS_BLANK, | |
103 NFA_CLASS_CNTRL, | |
104 NFA_CLASS_DIGIT, | |
105 NFA_CLASS_GRAPH, | |
106 NFA_CLASS_LOWER, | |
107 NFA_CLASS_PRINT, | |
108 NFA_CLASS_PUNCT, | |
109 NFA_CLASS_SPACE, | |
110 NFA_CLASS_UPPER, | |
111 NFA_CLASS_XDIGIT, | |
112 NFA_CLASS_TAB, | |
113 NFA_CLASS_RETURN, | |
114 NFA_CLASS_BACKSPACE, | |
115 NFA_CLASS_ESCAPE | |
116 }; | |
117 | |
118 /* Keep in sync with classchars. */ | |
119 static int nfa_classcodes[] = { | |
120 NFA_ANY, NFA_IDENT, NFA_SIDENT, NFA_KWORD,NFA_SKWORD, | |
121 NFA_FNAME, NFA_SFNAME, NFA_PRINT, NFA_SPRINT, | |
122 NFA_WHITE, NFA_NWHITE, NFA_DIGIT, NFA_NDIGIT, | |
123 NFA_HEX, NFA_NHEX, NFA_OCTAL, NFA_NOCTAL, | |
124 NFA_WORD, NFA_NWORD, NFA_HEAD, NFA_NHEAD, | |
125 NFA_ALPHA, NFA_NALPHA, NFA_LOWER, NFA_NLOWER, | |
126 NFA_UPPER, NFA_NUPPER | |
127 }; | |
128 | |
129 static char_u e_misplaced[] = N_("E866: (NFA regexp) Misplaced %c"); | |
130 | |
131 /* | |
132 * NFA errors can be of 3 types: | |
133 * *** NFA runtime errors, when something unknown goes wrong. The NFA fails | |
134 * silently and revert the to backtracking engine. | |
135 * syntax_error = FALSE; | |
136 * *** Regexp syntax errors, when the input regexp is not syntactically correct. | |
137 * The NFA engine displays an error message, and nothing else happens. | |
138 * syntax_error = TRUE | |
139 * *** Unsupported features, when the input regexp uses an operator that is not | |
140 * implemented in the NFA. The NFA engine fails silently, and reverts to the | |
141 * old backtracking engine. | |
142 * syntax_error = FALSE | |
143 * "The NFA fails" means that "compiling the regexp with the NFA fails": | |
144 * nfa_regcomp() returns FAIL. | |
145 */ | |
146 static int syntax_error = FALSE; | |
147 | |
148 /* NFA regexp \ze operator encountered. */ | |
149 static int nfa_has_zend = FALSE; | |
150 | |
151 static int *post_start; /* holds the postfix form of r.e. */ | |
152 static int *post_end; | |
153 static int *post_ptr; | |
154 | |
155 static int nstate; /* Number of states in the NFA. */ | |
156 static int istate; /* Index in the state vector, used in new_state() */ | |
157 static int nstate_max; /* Upper bound of estimated number of states. */ | |
158 | |
159 | |
160 static int nfa_regcomp_start __ARGS((char_u*expr, int re_flags)); | |
161 static int nfa_recognize_char_class __ARGS((char_u *start, char_u *end, int extra_newl)); | |
162 static int nfa_emit_equi_class __ARGS((int c, int neg)); | |
163 static void nfa_inc __ARGS((char_u **p)); | |
164 static void nfa_dec __ARGS((char_u **p)); | |
165 static int nfa_regatom __ARGS((void)); | |
166 static int nfa_regpiece __ARGS((void)); | |
167 static int nfa_regconcat __ARGS((void)); | |
168 static int nfa_regbranch __ARGS((void)); | |
169 static int nfa_reg __ARGS((int paren)); | |
170 #ifdef DEBUG | |
171 static void nfa_set_code __ARGS((int c)); | |
172 static void nfa_postfix_dump __ARGS((char_u *expr, int retval)); | |
173 static void nfa_print_state __ARGS((FILE *debugf, nfa_state_T *state, int ident)); | |
174 static void nfa_dump __ARGS((nfa_regprog_T *prog)); | |
175 #endif | |
176 static int *re2post __ARGS((void)); | |
177 static nfa_state_T *new_state __ARGS((int c, nfa_state_T *out, nfa_state_T *out1)); | |
178 static nfa_state_T *post2nfa __ARGS((int *postfix, int *end, int nfa_calc_size)); | |
179 static int check_char_class __ARGS((int class, int c)); | |
180 static void st_error __ARGS((int *postfix, int *end, int *p)); | |
181 static void nfa_save_listids __ARGS((nfa_state_T *start, int *list)); | |
182 static void nfa_restore_listids __ARGS((nfa_state_T *start, int *list)); | |
183 static void nfa_set_null_listids __ARGS((nfa_state_T *start)); | |
184 static void nfa_set_neg_listids __ARGS((nfa_state_T *start)); | |
185 static long nfa_regtry __ARGS((nfa_state_T *start, colnr_T col)); | |
186 static long nfa_regexec_both __ARGS((char_u *line, colnr_T col)); | |
187 static regprog_T *nfa_regcomp __ARGS((char_u *expr, int re_flags)); | |
188 static int nfa_regexec __ARGS((regmatch_T *rmp, char_u *line, colnr_T col)); | |
189 static long nfa_regexec_multi __ARGS((regmmatch_T *rmp, win_T *win, buf_T *buf, linenr_T lnum, colnr_T col, proftime_T *tm)); | |
190 | |
191 /* helper functions used when doing re2post() ... regatom() parsing */ | |
192 #define EMIT(c) do { \ | |
193 if (post_ptr >= post_end) \ | |
194 return FAIL; \ | |
195 *post_ptr++ = c; \ | |
196 } while (0) | |
197 | |
198 #define EMIT_MBYTE(c) \ | |
199 len = (*mb_char2bytes)(c, buf); \ | |
200 EMIT(buf[0]); \ | |
201 for (i = 1; i < len; i++) \ | |
202 { \ | |
203 EMIT(buf[i]); \ | |
204 EMIT(NFA_CONCAT); \ | |
205 } \ | |
206 EMIT(NFA_MULTIBYTE); | |
207 | |
208 #define EMIT_COMPOSING_UTF(input) \ | |
209 len = utfc_ptr2len(input); \ | |
210 EMIT(input[0]); \ | |
211 for (i = 1; i < len; i++) \ | |
212 { \ | |
213 EMIT(input[i]); \ | |
214 EMIT(NFA_CONCAT); \ | |
215 } \ | |
216 EMIT(NFA_COMPOSING); | |
217 | |
218 /* | |
219 * Initialize internal variables before NFA compilation. | |
220 * Return OK on success, FAIL otherwise. | |
221 */ | |
222 static int | |
223 nfa_regcomp_start(expr, re_flags) | |
224 char_u *expr; | |
225 int re_flags; /* see vim_regcomp() */ | |
226 { | |
4458 | 227 size_t postfix_size; |
4444 | 228 |
229 nstate = 0; | |
230 istate = 0; | |
231 /* A reasonable estimation for size */ | |
4458 | 232 nstate_max = (int)(STRLEN(expr) + 1) * NFA_POSTFIX_MULTIPLIER; |
4444 | 233 |
4454 | 234 /* Some items blow up in size, such as [A-z]. Add more space for that. |
235 * TODO: some patterns may still fail. */ | |
4458 | 236 nstate_max += 1000; |
4454 | 237 |
238 /* Size for postfix representation of expr. */ | |
4444 | 239 postfix_size = sizeof(*post_start) * nstate_max; |
4454 | 240 |
4444 | 241 post_start = (int *)lalloc(postfix_size, TRUE); |
242 if (post_start == NULL) | |
243 return FAIL; | |
244 vim_memset(post_start, 0, postfix_size); | |
245 post_ptr = post_start; | |
4454 | 246 post_end = post_start + nstate_max; |
4444 | 247 nfa_has_zend = FALSE; |
248 | |
249 regcomp_start(expr, re_flags); | |
250 | |
251 return OK; | |
252 } | |
253 | |
254 /* | |
255 * Search between "start" and "end" and try to recognize a | |
256 * character class in expanded form. For example [0-9]. | |
257 * On success, return the id the character class to be emitted. | |
258 * On failure, return 0 (=FAIL) | |
259 * Start points to the first char of the range, while end should point | |
260 * to the closing brace. | |
261 */ | |
262 static int | |
263 nfa_recognize_char_class(start, end, extra_newl) | |
264 char_u *start; | |
265 char_u *end; | |
266 int extra_newl; | |
267 { | |
268 int i; | |
269 /* Each of these variables takes up a char in "config[]", | |
270 * in the order they are here. */ | |
271 int not = FALSE, af = FALSE, AF = FALSE, az = FALSE, AZ = FALSE, | |
272 o7 = FALSE, o9 = FALSE, underscore = FALSE, newl = FALSE; | |
273 char_u *p; | |
274 #define NCONFIGS 16 | |
275 int classid[NCONFIGS] = { | |
276 NFA_DIGIT, NFA_NDIGIT, NFA_HEX, NFA_NHEX, | |
277 NFA_OCTAL, NFA_NOCTAL, NFA_WORD, NFA_NWORD, | |
278 NFA_HEAD, NFA_NHEAD, NFA_ALPHA, NFA_NALPHA, | |
279 NFA_LOWER, NFA_NLOWER, NFA_UPPER, NFA_NUPPER | |
280 }; | |
4450 | 281 char_u myconfig[10]; |
4444 | 282 char_u config[NCONFIGS][9] = { |
283 "000000100", /* digit */ | |
284 "100000100", /* non digit */ | |
285 "011000100", /* hex-digit */ | |
286 "111000100", /* non hex-digit */ | |
287 "000001000", /* octal-digit */ | |
288 "100001000", /* [^0-7] */ | |
289 "000110110", /* [0-9A-Za-z_] */ | |
290 "100110110", /* [^0-9A-Za-z_] */ | |
291 "000110010", /* head of word */ | |
292 "100110010", /* not head of word */ | |
293 "000110000", /* alphabetic char a-z */ | |
294 "100110000", /* non alphabetic char */ | |
295 "000100000", /* lowercase letter */ | |
296 "100100000", /* non lowercase */ | |
297 "000010000", /* uppercase */ | |
298 "100010000" /* non uppercase */ | |
299 }; | |
300 | |
301 if (extra_newl == TRUE) | |
302 newl = TRUE; | |
303 | |
304 if (*end != ']') | |
305 return FAIL; | |
306 p = start; | |
307 if (*p == '^') | |
308 { | |
309 not = TRUE; | |
310 p ++; | |
311 } | |
312 | |
313 while (p < end) | |
314 { | |
315 if (p + 2 < end && *(p + 1) == '-') | |
316 { | |
317 switch (*p) | |
318 { | |
319 case '0': | |
320 if (*(p + 2) == '9') | |
321 { | |
322 o9 = TRUE; | |
323 break; | |
324 } | |
325 else | |
326 if (*(p + 2) == '7') | |
327 { | |
328 o7 = TRUE; | |
329 break; | |
330 } | |
331 case 'a': | |
332 if (*(p + 2) == 'z') | |
333 { | |
334 az = TRUE; | |
335 break; | |
336 } | |
337 else | |
338 if (*(p + 2) == 'f') | |
339 { | |
340 af = TRUE; | |
341 break; | |
342 } | |
343 case 'A': | |
344 if (*(p + 2) == 'Z') | |
345 { | |
346 AZ = TRUE; | |
347 break; | |
348 } | |
349 else | |
350 if (*(p + 2) == 'F') | |
351 { | |
352 AF = TRUE; | |
353 break; | |
354 } | |
355 /* FALLTHROUGH */ | |
356 default: | |
357 return FAIL; | |
358 } | |
359 p += 3; | |
360 } | |
361 else if (p + 1 < end && *p == '\\' && *(p + 1) == 'n') | |
362 { | |
363 newl = TRUE; | |
364 p += 2; | |
365 } | |
366 else if (*p == '_') | |
367 { | |
368 underscore = TRUE; | |
369 p ++; | |
370 } | |
371 else if (*p == '\n') | |
372 { | |
373 newl = TRUE; | |
374 p ++; | |
375 } | |
376 else | |
377 return FAIL; | |
378 } /* while (p < end) */ | |
379 | |
380 if (p != end) | |
381 return FAIL; | |
382 | |
383 /* build the config that represents the ranges we gathered */ | |
384 STRCPY(myconfig, "000000000"); | |
385 if (not == TRUE) | |
386 myconfig[0] = '1'; | |
387 if (af == TRUE) | |
388 myconfig[1] = '1'; | |
389 if (AF == TRUE) | |
390 myconfig[2] = '1'; | |
391 if (az == TRUE) | |
392 myconfig[3] = '1'; | |
393 if (AZ == TRUE) | |
394 myconfig[4] = '1'; | |
395 if (o7 == TRUE) | |
396 myconfig[5] = '1'; | |
397 if (o9 == TRUE) | |
398 myconfig[6] = '1'; | |
399 if (underscore == TRUE) | |
400 myconfig[7] = '1'; | |
401 if (newl == TRUE) | |
402 { | |
403 myconfig[8] = '1'; | |
404 extra_newl = ADD_NL; | |
405 } | |
406 /* try to recognize character classes */ | |
407 for (i = 0; i < NCONFIGS; i++) | |
4450 | 408 if (STRNCMP(myconfig, config[i], 8) == 0) |
4444 | 409 return classid[i] + extra_newl; |
410 | |
411 /* fallthrough => no success so far */ | |
412 return FAIL; | |
413 | |
414 #undef NCONFIGS | |
415 } | |
416 | |
417 /* | |
418 * Produce the bytes for equivalence class "c". | |
419 * Currently only handles latin1, latin9 and utf-8. | |
420 * Emits bytes in postfix notation: 'a,b,NFA_OR,c,NFA_OR' is | |
421 * equivalent to 'a OR b OR c' | |
422 * | |
423 * NOTE! When changing this function, also update reg_equi_class() | |
424 */ | |
425 static int | |
426 nfa_emit_equi_class(c, neg) | |
427 int c; | |
428 int neg; | |
429 { | |
430 int first = TRUE; | |
431 int glue = neg == TRUE ? NFA_CONCAT : NFA_OR; | |
432 #define EMIT2(c) \ | |
433 EMIT(c); \ | |
434 if (neg == TRUE) { \ | |
435 EMIT(NFA_NOT); \ | |
436 } \ | |
437 if (first == FALSE) \ | |
438 EMIT(glue); \ | |
439 else \ | |
440 first = FALSE; \ | |
441 | |
442 #ifdef FEAT_MBYTE | |
443 if (enc_utf8 || STRCMP(p_enc, "latin1") == 0 | |
444 || STRCMP(p_enc, "iso-8859-15") == 0) | |
445 #endif | |
446 { | |
447 switch (c) | |
448 { | |
449 case 'A': case '\300': case '\301': case '\302': | |
450 case '\303': case '\304': case '\305': | |
451 EMIT2('A'); EMIT2('\300'); EMIT2('\301'); | |
452 EMIT2('\302'); EMIT2('\303'); EMIT2('\304'); | |
453 EMIT2('\305'); | |
454 return OK; | |
455 | |
456 case 'C': case '\307': | |
457 EMIT2('C'); EMIT2('\307'); | |
458 return OK; | |
459 | |
460 case 'E': case '\310': case '\311': case '\312': case '\313': | |
461 EMIT2('E'); EMIT2('\310'); EMIT2('\311'); | |
462 EMIT2('\312'); EMIT2('\313'); | |
463 return OK; | |
464 | |
465 case 'I': case '\314': case '\315': case '\316': case '\317': | |
466 EMIT2('I'); EMIT2('\314'); EMIT2('\315'); | |
467 EMIT2('\316'); EMIT2('\317'); | |
468 return OK; | |
469 | |
470 case 'N': case '\321': | |
471 EMIT2('N'); EMIT2('\321'); | |
472 return OK; | |
473 | |
474 case 'O': case '\322': case '\323': case '\324': case '\325': | |
475 case '\326': | |
476 EMIT2('O'); EMIT2('\322'); EMIT2('\323'); | |
477 EMIT2('\324'); EMIT2('\325'); EMIT2('\326'); | |
478 return OK; | |
479 | |
480 case 'U': case '\331': case '\332': case '\333': case '\334': | |
481 EMIT2('U'); EMIT2('\331'); EMIT2('\332'); | |
482 EMIT2('\333'); EMIT2('\334'); | |
483 return OK; | |
484 | |
485 case 'Y': case '\335': | |
486 EMIT2('Y'); EMIT2('\335'); | |
487 return OK; | |
488 | |
489 case 'a': case '\340': case '\341': case '\342': | |
490 case '\343': case '\344': case '\345': | |
491 EMIT2('a'); EMIT2('\340'); EMIT2('\341'); | |
492 EMIT2('\342'); EMIT2('\343'); EMIT2('\344'); | |
493 EMIT2('\345'); | |
494 return OK; | |
495 | |
496 case 'c': case '\347': | |
497 EMIT2('c'); EMIT2('\347'); | |
498 return OK; | |
499 | |
500 case 'e': case '\350': case '\351': case '\352': case '\353': | |
501 EMIT2('e'); EMIT2('\350'); EMIT2('\351'); | |
502 EMIT2('\352'); EMIT2('\353'); | |
503 return OK; | |
504 | |
505 case 'i': case '\354': case '\355': case '\356': case '\357': | |
506 EMIT2('i'); EMIT2('\354'); EMIT2('\355'); | |
507 EMIT2('\356'); EMIT2('\357'); | |
508 return OK; | |
509 | |
510 case 'n': case '\361': | |
511 EMIT2('n'); EMIT2('\361'); | |
512 return OK; | |
513 | |
514 case 'o': case '\362': case '\363': case '\364': case '\365': | |
515 case '\366': | |
516 EMIT2('o'); EMIT2('\362'); EMIT2('\363'); | |
517 EMIT2('\364'); EMIT2('\365'); EMIT2('\366'); | |
518 return OK; | |
519 | |
520 case 'u': case '\371': case '\372': case '\373': case '\374': | |
521 EMIT2('u'); EMIT2('\371'); EMIT2('\372'); | |
522 EMIT2('\373'); EMIT2('\374'); | |
523 return OK; | |
524 | |
525 case 'y': case '\375': case '\377': | |
526 EMIT2('y'); EMIT2('\375'); EMIT2('\377'); | |
527 return OK; | |
528 | |
529 default: | |
530 return FAIL; | |
531 } | |
532 } | |
533 | |
534 EMIT(c); | |
535 return OK; | |
536 #undef EMIT2 | |
537 } | |
538 | |
539 /* | |
540 * Code to parse regular expression. | |
541 * | |
542 * We try to reuse parsing functions in regexp.c to | |
543 * minimize surprise and keep the syntax consistent. | |
544 */ | |
545 | |
546 /* | |
547 * Increments the pointer "p" by one (multi-byte) character. | |
548 */ | |
549 static void | |
550 nfa_inc(p) | |
551 char_u **p; | |
552 { | |
553 #ifdef FEAT_MBYTE | |
554 if (has_mbyte) | |
555 mb_ptr2char_adv(p); | |
556 else | |
557 #endif | |
558 *p = *p + 1; | |
559 } | |
560 | |
561 /* | |
562 * Decrements the pointer "p" by one (multi-byte) character. | |
563 */ | |
564 static void | |
565 nfa_dec(p) | |
566 char_u **p; | |
567 { | |
568 #ifdef FEAT_MBYTE | |
569 char_u *p2, *oldp; | |
570 | |
571 if (has_mbyte) | |
572 { | |
573 oldp = *p; | |
574 /* Try to find the multibyte char that advances to the current | |
575 * position. */ | |
576 do | |
577 { | |
578 *p = *p - 1; | |
579 p2 = *p; | |
580 mb_ptr2char_adv(&p2); | |
581 } while (p2 != oldp); | |
582 } | |
583 #else | |
584 *p = *p - 1; | |
585 #endif | |
586 } | |
587 | |
588 /* | |
589 * Parse the lowest level. | |
590 * | |
591 * An atom can be one of a long list of items. Many atoms match one character | |
592 * in the text. It is often an ordinary character or a character class. | |
593 * Braces can be used to make a pattern into an atom. The "\z(\)" construct | |
594 * is only for syntax highlighting. | |
595 * | |
596 * atom ::= ordinary-atom | |
597 * or \( pattern \) | |
598 * or \%( pattern \) | |
599 * or \z( pattern \) | |
600 */ | |
601 static int | |
602 nfa_regatom() | |
603 { | |
604 int c; | |
605 int charclass; | |
606 int equiclass; | |
607 int collclass; | |
608 int got_coll_char; | |
609 char_u *p; | |
610 char_u *endp; | |
611 #ifdef FEAT_MBYTE | |
612 char_u *old_regparse = regparse; | |
613 int clen; | |
614 int len; | |
615 static char_u buf[30]; | |
616 int i; | |
617 #endif | |
618 int extra = 0; | |
619 int first; | |
620 int emit_range; | |
621 int negated; | |
622 int result; | |
623 int startc = -1; | |
624 int endc = -1; | |
625 int oldstartc = -1; | |
626 int cpo_lit; /* 'cpoptions' contains 'l' flag */ | |
627 int cpo_bsl; /* 'cpoptions' contains '\' flag */ | |
628 int glue; /* ID that will "glue" nodes together */ | |
629 | |
630 cpo_lit = vim_strchr(p_cpo, CPO_LITERAL) != NULL; | |
631 cpo_bsl = vim_strchr(p_cpo, CPO_BACKSL) != NULL; | |
632 | |
633 c = getchr(); | |
634 | |
635 #ifdef FEAT_MBYTE | |
636 /* clen has the length of the current char, without composing chars */ | |
637 clen = (*mb_char2len)(c); | |
638 if (has_mbyte && clen > 1) | |
639 goto nfa_do_multibyte; | |
640 #endif | |
641 switch (c) | |
642 { | |
643 case Magic('^'): | |
644 EMIT(NFA_BOL); | |
645 break; | |
646 | |
647 case Magic('$'): | |
648 EMIT(NFA_EOL); | |
649 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
650 had_eol = TRUE; | |
651 #endif | |
652 break; | |
653 | |
654 case Magic('<'): | |
655 EMIT(NFA_BOW); | |
656 break; | |
657 | |
658 case Magic('>'): | |
659 EMIT(NFA_EOW); | |
660 break; | |
661 | |
662 case Magic('_'): | |
663 c = no_Magic(getchr()); | |
664 if (c == '^') /* "\_^" is start-of-line */ | |
665 { | |
666 EMIT(NFA_BOL); | |
667 break; | |
668 } | |
669 if (c == '$') /* "\_$" is end-of-line */ | |
670 { | |
671 EMIT(NFA_EOL); | |
672 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
673 had_eol = TRUE; | |
674 #endif | |
675 break; | |
676 } | |
677 | |
678 extra = ADD_NL; | |
679 | |
680 /* "\_[" is collection plus newline */ | |
681 if (c == '[') | |
4517
9a2183bd8295
updated for version 7.3.1006
Bram Moolenaar <bram@vim.org>
parents:
4515
diff
changeset
|
682 goto collection; |
4444 | 683 |
684 /* "\_x" is character class plus newline */ | |
685 /*FALLTHROUGH*/ | |
686 | |
687 /* | |
688 * Character classes. | |
689 */ | |
690 case Magic('.'): | |
691 case Magic('i'): | |
692 case Magic('I'): | |
693 case Magic('k'): | |
694 case Magic('K'): | |
695 case Magic('f'): | |
696 case Magic('F'): | |
697 case Magic('p'): | |
698 case Magic('P'): | |
699 case Magic('s'): | |
700 case Magic('S'): | |
701 case Magic('d'): | |
702 case Magic('D'): | |
703 case Magic('x'): | |
704 case Magic('X'): | |
705 case Magic('o'): | |
706 case Magic('O'): | |
707 case Magic('w'): | |
708 case Magic('W'): | |
709 case Magic('h'): | |
710 case Magic('H'): | |
711 case Magic('a'): | |
712 case Magic('A'): | |
713 case Magic('l'): | |
714 case Magic('L'): | |
715 case Magic('u'): | |
716 case Magic('U'): | |
717 p = vim_strchr(classchars, no_Magic(c)); | |
718 if (p == NULL) | |
719 { | |
720 return FAIL; /* runtime error */ | |
721 } | |
722 #ifdef FEAT_MBYTE | |
723 /* When '.' is followed by a composing char ignore the dot, so that | |
724 * the composing char is matched here. */ | |
725 if (enc_utf8 && c == Magic('.') && utf_iscomposing(peekchr())) | |
726 { | |
727 c = getchr(); | |
728 goto nfa_do_multibyte; | |
729 } | |
730 #endif | |
731 EMIT(nfa_classcodes[p - classchars]); | |
732 if (extra == ADD_NL) | |
733 { | |
734 EMIT(NFA_NEWL); | |
735 EMIT(NFA_OR); | |
736 regflags |= RF_HASNL; | |
737 } | |
738 break; | |
739 | |
740 case Magic('n'): | |
741 if (reg_string) | |
742 /* In a string "\n" matches a newline character. */ | |
743 EMIT(NL); | |
744 else | |
745 { | |
746 /* In buffer text "\n" matches the end of a line. */ | |
747 EMIT(NFA_NEWL); | |
748 regflags |= RF_HASNL; | |
749 } | |
750 break; | |
751 | |
752 case Magic('('): | |
753 if (nfa_reg(REG_PAREN) == FAIL) | |
754 return FAIL; /* cascaded error */ | |
755 break; | |
756 | |
757 case NUL: | |
758 syntax_error = TRUE; | |
759 EMSG_RET_FAIL(_("E865: (NFA) Regexp end encountered prematurely")); | |
760 | |
761 case Magic('|'): | |
762 case Magic('&'): | |
763 case Magic(')'): | |
764 syntax_error = TRUE; | |
4450 | 765 EMSGN(_(e_misplaced), no_Magic(c)); |
4444 | 766 return FAIL; |
767 | |
768 case Magic('='): | |
769 case Magic('?'): | |
770 case Magic('+'): | |
771 case Magic('@'): | |
772 case Magic('*'): | |
773 case Magic('{'): | |
774 /* these should follow an atom, not form an atom */ | |
775 syntax_error = TRUE; | |
4450 | 776 EMSGN(_(e_misplaced), no_Magic(c)); |
4444 | 777 return FAIL; |
778 | |
779 case Magic('~'): /* previous substitute pattern */ | |
780 /* Not supported yet */ | |
781 return FAIL; | |
782 | |
783 case Magic('1'): | |
784 case Magic('2'): | |
785 case Magic('3'): | |
786 case Magic('4'): | |
787 case Magic('5'): | |
788 case Magic('6'): | |
789 case Magic('7'): | |
790 case Magic('8'): | |
791 case Magic('9'): | |
792 /* not supported yet */ | |
793 return FAIL; | |
794 | |
795 case Magic('z'): | |
796 c = no_Magic(getchr()); | |
797 switch (c) | |
798 { | |
799 case 's': | |
800 EMIT(NFA_ZSTART); | |
801 break; | |
802 case 'e': | |
803 EMIT(NFA_ZEND); | |
804 nfa_has_zend = TRUE; | |
805 /* TODO: Currently \ze does not work properly. */ | |
806 return FAIL; | |
807 /* break; */ | |
808 case '1': | |
809 case '2': | |
810 case '3': | |
811 case '4': | |
812 case '5': | |
813 case '6': | |
814 case '7': | |
815 case '8': | |
816 case '9': | |
817 case '(': | |
818 /* \z1...\z9 and \z( not yet supported */ | |
819 return FAIL; | |
820 default: | |
821 syntax_error = TRUE; | |
4450 | 822 EMSGN(_("E867: (NFA) Unknown operator '\\z%c'"), |
4444 | 823 no_Magic(c)); |
824 return FAIL; | |
825 } | |
826 break; | |
827 | |
828 case Magic('%'): | |
829 c = no_Magic(getchr()); | |
830 switch (c) | |
831 { | |
832 /* () without a back reference */ | |
833 case '(': | |
834 if (nfa_reg(REG_NPAREN) == FAIL) | |
835 return FAIL; | |
836 EMIT(NFA_NOPEN); | |
837 break; | |
838 | |
839 case 'd': /* %d123 decimal */ | |
840 case 'o': /* %o123 octal */ | |
841 case 'x': /* %xab hex 2 */ | |
842 case 'u': /* %uabcd hex 4 */ | |
843 case 'U': /* %U1234abcd hex 8 */ | |
844 /* Not yet supported */ | |
845 return FAIL; | |
846 | |
847 c = coll_get_char(); | |
848 #ifdef FEAT_MBYTE | |
849 if ((*mb_char2len)(c) > 1) | |
850 { | |
851 EMIT_MBYTE(c); | |
852 } | |
853 else | |
854 #endif | |
855 EMIT(c); | |
856 break; | |
857 | |
858 /* Catch \%^ and \%$ regardless of where they appear in the | |
859 * pattern -- regardless of whether or not it makes sense. */ | |
860 case '^': | |
861 EMIT(NFA_BOF); | |
862 /* Not yet supported */ | |
863 return FAIL; | |
864 break; | |
865 | |
866 case '$': | |
867 EMIT(NFA_EOF); | |
868 /* Not yet supported */ | |
869 return FAIL; | |
870 break; | |
871 | |
872 case '#': | |
873 /* not supported yet */ | |
874 return FAIL; | |
875 break; | |
876 | |
877 case 'V': | |
878 /* not supported yet */ | |
879 return FAIL; | |
880 break; | |
881 | |
882 case '[': | |
883 /* \%[abc] not supported yet */ | |
884 return FAIL; | |
885 | |
886 default: | |
887 /* not supported yet */ | |
888 return FAIL; | |
889 } | |
890 break; | |
891 | |
892 case Magic('['): | |
4517
9a2183bd8295
updated for version 7.3.1006
Bram Moolenaar <bram@vim.org>
parents:
4515
diff
changeset
|
893 collection: |
4444 | 894 /* |
895 * Glue is emitted between several atoms from the []. | |
896 * It is either NFA_OR, or NFA_CONCAT. | |
897 * | |
898 * [abc] expands to 'a b NFA_OR c NFA_OR' (in postfix notation) | |
899 * [^abc] expands to 'a NFA_NOT b NFA_NOT NFA_CONCAT c NFA_NOT | |
900 * NFA_CONCAT NFA_END_NEG_RANGE NFA_CONCAT' (in postfix | |
901 * notation) | |
902 * | |
903 */ | |
904 | |
905 | |
906 /* Emit negation atoms, if needed. | |
907 * The CONCAT below merges the NOT with the previous node. */ | |
908 #define TRY_NEG() \ | |
909 if (negated == TRUE) \ | |
910 { \ | |
911 EMIT(NFA_NOT); \ | |
912 } | |
913 | |
914 /* Emit glue between important nodes : CONCAT or OR. */ | |
915 #define EMIT_GLUE() \ | |
916 if (first == FALSE) \ | |
917 EMIT(glue); \ | |
918 else \ | |
919 first = FALSE; | |
920 | |
921 p = regparse; | |
922 endp = skip_anyof(p); | |
923 if (*endp == ']') | |
924 { | |
925 /* | |
926 * Try to reverse engineer character classes. For example, | |
927 * recognize that [0-9] stands for \d and [A-Za-z_] with \h, | |
928 * and perform the necessary substitutions in the NFA. | |
929 */ | |
930 result = nfa_recognize_char_class(regparse, endp, | |
931 extra == ADD_NL); | |
932 if (result != FAIL) | |
933 { | |
934 if (result >= NFA_DIGIT && result <= NFA_NUPPER) | |
935 EMIT(result); | |
936 else /* must be char class + newline */ | |
937 { | |
938 EMIT(result - ADD_NL); | |
939 EMIT(NFA_NEWL); | |
940 EMIT(NFA_OR); | |
941 } | |
942 regparse = endp; | |
943 nfa_inc(®parse); | |
944 return OK; | |
945 } | |
946 /* | |
947 * Failed to recognize a character class. Use the simple | |
948 * version that turns [abc] into 'a' OR 'b' OR 'c' | |
949 */ | |
950 startc = endc = oldstartc = -1; | |
951 first = TRUE; /* Emitting first atom in this sequence? */ | |
952 negated = FALSE; | |
953 glue = NFA_OR; | |
954 if (*regparse == '^') /* negated range */ | |
955 { | |
956 negated = TRUE; | |
957 glue = NFA_CONCAT; | |
958 nfa_inc(®parse); | |
959 } | |
960 if (*regparse == '-') | |
961 { | |
962 startc = '-'; | |
963 EMIT(startc); | |
964 TRY_NEG(); | |
965 EMIT_GLUE(); | |
966 nfa_inc(®parse); | |
967 } | |
968 /* Emit the OR branches for each character in the [] */ | |
969 emit_range = FALSE; | |
970 while (regparse < endp) | |
971 { | |
972 oldstartc = startc; | |
973 startc = -1; | |
974 got_coll_char = FALSE; | |
975 if (*regparse == '[') | |
976 { | |
977 /* Check for [: :], [= =], [. .] */ | |
978 equiclass = collclass = 0; | |
979 charclass = get_char_class(®parse); | |
980 if (charclass == CLASS_NONE) | |
981 { | |
982 equiclass = get_equi_class(®parse); | |
983 if (equiclass == 0) | |
984 collclass = get_coll_element(®parse); | |
985 } | |
986 | |
987 /* Character class like [:alpha:] */ | |
988 if (charclass != CLASS_NONE) | |
989 { | |
990 switch (charclass) | |
991 { | |
992 case CLASS_ALNUM: | |
993 EMIT(NFA_CLASS_ALNUM); | |
994 break; | |
995 case CLASS_ALPHA: | |
996 EMIT(NFA_CLASS_ALPHA); | |
997 break; | |
998 case CLASS_BLANK: | |
999 EMIT(NFA_CLASS_BLANK); | |
1000 break; | |
1001 case CLASS_CNTRL: | |
1002 EMIT(NFA_CLASS_CNTRL); | |
1003 break; | |
1004 case CLASS_DIGIT: | |
1005 EMIT(NFA_CLASS_DIGIT); | |
1006 break; | |
1007 case CLASS_GRAPH: | |
1008 EMIT(NFA_CLASS_GRAPH); | |
1009 break; | |
1010 case CLASS_LOWER: | |
1011 EMIT(NFA_CLASS_LOWER); | |
1012 break; | |
1013 case CLASS_PRINT: | |
1014 EMIT(NFA_CLASS_PRINT); | |
1015 break; | |
1016 case CLASS_PUNCT: | |
1017 EMIT(NFA_CLASS_PUNCT); | |
1018 break; | |
1019 case CLASS_SPACE: | |
1020 EMIT(NFA_CLASS_SPACE); | |
1021 break; | |
1022 case CLASS_UPPER: | |
1023 EMIT(NFA_CLASS_UPPER); | |
1024 break; | |
1025 case CLASS_XDIGIT: | |
1026 EMIT(NFA_CLASS_XDIGIT); | |
1027 break; | |
1028 case CLASS_TAB: | |
1029 EMIT(NFA_CLASS_TAB); | |
1030 break; | |
1031 case CLASS_RETURN: | |
1032 EMIT(NFA_CLASS_RETURN); | |
1033 break; | |
1034 case CLASS_BACKSPACE: | |
1035 EMIT(NFA_CLASS_BACKSPACE); | |
1036 break; | |
1037 case CLASS_ESCAPE: | |
1038 EMIT(NFA_CLASS_ESCAPE); | |
1039 break; | |
1040 } | |
1041 TRY_NEG(); | |
1042 EMIT_GLUE(); | |
1043 continue; | |
1044 } | |
1045 /* Try equivalence class [=a=] and the like */ | |
1046 if (equiclass != 0) | |
1047 { | |
1048 result = nfa_emit_equi_class(equiclass, negated); | |
1049 if (result == FAIL) | |
1050 { | |
1051 /* should never happen */ | |
1052 EMSG_RET_FAIL(_("E868: Error building NFA with equivalence class!")); | |
1053 } | |
1054 EMIT_GLUE(); | |
1055 continue; | |
1056 } | |
1057 /* Try collating class like [. .] */ | |
1058 if (collclass != 0) | |
1059 { | |
1060 startc = collclass; /* allow [.a.]-x as a range */ | |
1061 /* Will emit the proper atom at the end of the | |
1062 * while loop. */ | |
1063 } | |
1064 } | |
1065 /* Try a range like 'a-x' or '\t-z' */ | |
1066 if (*regparse == '-') | |
1067 { | |
1068 emit_range = TRUE; | |
1069 startc = oldstartc; | |
1070 nfa_inc(®parse); | |
1071 continue; /* reading the end of the range */ | |
1072 } | |
1073 | |
1074 /* Now handle simple and escaped characters. | |
1075 * Only "\]", "\^", "\]" and "\\" are special in Vi. Vim | |
1076 * accepts "\t", "\e", etc., but only when the 'l' flag in | |
1077 * 'cpoptions' is not included. | |
1078 * Posix doesn't recognize backslash at all. | |
1079 */ | |
1080 if (*regparse == '\\' | |
1081 && !cpo_bsl | |
1082 && regparse + 1 <= endp | |
1083 && (vim_strchr(REGEXP_INRANGE, regparse[1]) != NULL | |
1084 || (!cpo_lit | |
1085 && vim_strchr(REGEXP_ABBR, regparse[1]) | |
1086 != NULL) | |
1087 ) | |
1088 ) | |
1089 { | |
1090 nfa_inc(®parse); | |
1091 | |
4507
9dbbddb2ed10
updated for version 7.3.1001
Bram Moolenaar <bram@vim.org>
parents:
4503
diff
changeset
|
1092 if (*regparse == 'n') |
4444 | 1093 startc = reg_string ? NL : NFA_NEWL; |
1094 else | |
1095 if (*regparse == 'd' | |
1096 || *regparse == 'o' | |
1097 || *regparse == 'x' | |
1098 || *regparse == 'u' | |
1099 || *regparse == 'U' | |
1100 ) | |
1101 { | |
1102 /* TODO(RE) This needs more testing */ | |
1103 startc = coll_get_char(); | |
1104 got_coll_char = TRUE; | |
1105 nfa_dec(®parse); | |
1106 } | |
1107 else | |
1108 { | |
1109 /* \r,\t,\e,\b */ | |
1110 startc = backslash_trans(*regparse); | |
1111 } | |
1112 } | |
1113 | |
1114 /* Normal printable char */ | |
1115 if (startc == -1) | |
1116 #ifdef FEAT_MBYTE | |
1117 startc = (*mb_ptr2char)(regparse); | |
1118 #else | |
1119 startc = *regparse; | |
1120 #endif | |
1121 | |
1122 /* Previous char was '-', so this char is end of range. */ | |
1123 if (emit_range) | |
1124 { | |
1125 endc = startc; startc = oldstartc; | |
1126 if (startc > endc) | |
1127 EMSG_RET_FAIL(_(e_invrange)); | |
1128 #ifdef FEAT_MBYTE | |
1129 if (has_mbyte && ((*mb_char2len)(startc) > 1 | |
1130 || (*mb_char2len)(endc) > 1)) | |
1131 { | |
1132 if (endc > startc + 256) | |
1133 EMSG_RET_FAIL(_(e_invrange)); | |
1134 /* Emit the range. "startc" was already emitted, so | |
1135 * skip it. */ | |
1136 for (c = startc + 1; c <= endc; c++) | |
1137 { | |
1138 if ((*mb_char2len)(c) > 1) | |
1139 { | |
1140 EMIT_MBYTE(c); | |
1141 } | |
1142 else | |
1143 EMIT(c); | |
1144 TRY_NEG(); | |
1145 EMIT_GLUE(); | |
1146 } | |
1147 emit_range = FALSE; | |
1148 } | |
1149 else | |
1150 #endif | |
1151 { | |
1152 #ifdef EBCDIC | |
1153 int alpha_only = FALSE; | |
1154 | |
1155 /* for alphabetical range skip the gaps | |
1156 * 'i'-'j', 'r'-'s', 'I'-'J' and 'R'-'S'. */ | |
1157 if (isalpha(startc) && isalpha(endc)) | |
1158 alpha_only = TRUE; | |
1159 #endif | |
1160 /* Emit the range. "startc" was already emitted, so | |
1161 * skip it. */ | |
1162 for (c = startc + 1; c <= endc; c++) | |
1163 #ifdef EBCDIC | |
1164 if (!alpha_only || isalpha(startc)) | |
1165 #endif | |
1166 { | |
1167 EMIT(c); | |
1168 TRY_NEG(); | |
1169 EMIT_GLUE(); | |
1170 } | |
1171 emit_range = FALSE; | |
1172 } | |
1173 } | |
1174 else | |
1175 { | |
1176 /* | |
1177 * This char (startc) is not part of a range. Just | |
1178 * emit it. | |
1179 * | |
1180 * Normally, simply emit startc. But if we get char | |
1181 * code=0 from a collating char, then replace it with | |
1182 * 0x0a. | |
1183 * | |
1184 * This is needed to completely mimic the behaviour of | |
1185 * the backtracking engine. | |
1186 */ | |
1187 if (got_coll_char == TRUE && startc == 0) | |
1188 EMIT(0x0a); | |
1189 else | |
1190 #ifdef FEAT_MBYTE | |
1191 if ((*mb_char2len)(startc) > 1) | |
1192 { | |
1193 EMIT_MBYTE(startc); | |
1194 } | |
1195 else | |
1196 #endif | |
1197 EMIT(startc); | |
1198 TRY_NEG(); | |
1199 EMIT_GLUE(); | |
1200 } | |
1201 | |
1202 nfa_inc(®parse); | |
1203 } /* while (p < endp) */ | |
1204 | |
1205 nfa_dec(®parse); | |
1206 if (*regparse == '-') /* if last, '-' is just a char */ | |
1207 { | |
1208 EMIT('-'); | |
1209 TRY_NEG(); | |
1210 EMIT_GLUE(); | |
1211 } | |
1212 nfa_inc(®parse); | |
1213 | |
1214 if (extra == ADD_NL) /* \_[] also matches \n */ | |
1215 { | |
1216 EMIT(reg_string ? NL : NFA_NEWL); | |
1217 TRY_NEG(); | |
1218 EMIT_GLUE(); | |
1219 } | |
1220 | |
1221 /* skip the trailing ] */ | |
1222 regparse = endp; | |
1223 nfa_inc(®parse); | |
1224 if (negated == TRUE) | |
1225 { | |
1226 /* Mark end of negated char range */ | |
1227 EMIT(NFA_END_NEG_RANGE); | |
1228 EMIT(NFA_CONCAT); | |
1229 } | |
1230 return OK; | |
1231 } /* if exists closing ] */ | |
1232 else if (reg_strict) | |
1233 { | |
1234 syntax_error = TRUE; | |
1235 EMSG_RET_FAIL(_(e_missingbracket)); | |
1236 } | |
1237 | |
1238 /* FALLTHROUGH */ | |
1239 default: | |
1240 { | |
1241 #ifdef FEAT_MBYTE | |
1242 int plen; | |
1243 | |
1244 nfa_do_multibyte: | |
1245 /* length of current char, with composing chars, | |
1246 * from pointer */ | |
1247 plen = (*mb_ptr2len)(old_regparse); | |
1248 if (enc_utf8 && clen != plen) | |
1249 { | |
1250 /* A composing character is always handled as a | |
1251 * separate atom, surrounded by NFA_COMPOSING and | |
1252 * NFA_END_COMPOSING. Note that right now we are | |
1253 * building the postfix form, not the NFA itself; | |
1254 * a composing char could be: a, b, c, NFA_COMPOSING | |
1255 * where 'a', 'b', 'c' are chars with codes > 256. | |
1256 */ | |
1257 EMIT_COMPOSING_UTF(old_regparse); | |
1258 regparse = old_regparse + plen; | |
1259 } | |
1260 else | |
1261 /* A multi-byte character is always handled as a | |
1262 * separate atom, surrounded by NFA_MULTIBYTE and | |
1263 * NFA_END_MULTIBYTE */ | |
1264 if (plen > 1) | |
1265 { | |
1266 EMIT_MBYTE(c); | |
1267 } | |
1268 else | |
1269 #endif | |
1270 { | |
1271 c = no_Magic(c); | |
1272 EMIT(c); | |
1273 } | |
1274 return OK; | |
1275 } | |
1276 } | |
1277 | |
1278 #undef TRY_NEG | |
1279 #undef EMIT_GLUE | |
1280 | |
1281 return OK; | |
1282 } | |
1283 | |
1284 /* | |
1285 * Parse something followed by possible [*+=]. | |
1286 * | |
1287 * A piece is an atom, possibly followed by a multi, an indication of how many | |
1288 * times the atom can be matched. Example: "a*" matches any sequence of "a" | |
1289 * characters: "", "a", "aa", etc. | |
1290 * | |
1291 * piece ::= atom | |
1292 * or atom multi | |
1293 */ | |
1294 static int | |
1295 nfa_regpiece() | |
1296 { | |
1297 int i; | |
1298 int op; | |
1299 int ret; | |
1300 long minval, maxval; | |
1301 int greedy = TRUE; /* Braces are prefixed with '-' ? */ | |
1302 char_u *old_regparse, *new_regparse; | |
1303 int c2; | |
1304 int *old_post_ptr, *my_post_start; | |
1305 int old_regnpar; | |
1306 int quest; | |
1307 | |
1308 /* Save the current position in the regexp, so that we can use it if | |
1309 * <atom>{m,n} is next. */ | |
1310 old_regparse = regparse; | |
1311 /* Save current number of open parenthesis, so we can use it if | |
1312 * <atom>{m,n} is next */ | |
1313 old_regnpar = regnpar; | |
1314 /* store current pos in the postfix form, for \{m,n} involving 0s */ | |
1315 my_post_start = post_ptr; | |
1316 | |
1317 ret = nfa_regatom(); | |
1318 if (ret == FAIL) | |
1319 return FAIL; /* cascaded error */ | |
1320 | |
1321 op = peekchr(); | |
1322 if (re_multi_type(op) == NOT_MULTI) | |
1323 return OK; | |
1324 | |
1325 skipchr(); | |
1326 switch (op) | |
1327 { | |
1328 case Magic('*'): | |
1329 EMIT(NFA_STAR); | |
1330 break; | |
1331 | |
1332 case Magic('+'): | |
1333 /* | |
1334 * Trick: Normally, (a*)\+ would match the whole input "aaa". The | |
1335 * first and only submatch would be "aaa". But the backtracking | |
1336 * engine interprets the plus as "try matching one more time", and | |
1337 * a* matches a second time at the end of the input, the empty | |
1338 * string. | |
1339 * The submatch will the empty string. | |
1340 * | |
1341 * In order to be consistent with the old engine, we disable | |
1342 * NFA_PLUS, and replace <atom>+ with <atom><atom>* | |
1343 */ | |
1344 /* EMIT(NFA_PLUS); */ | |
1345 regnpar = old_regnpar; | |
1346 regparse = old_regparse; | |
1347 curchr = -1; | |
1348 if (nfa_regatom() == FAIL) | |
1349 return FAIL; | |
1350 EMIT(NFA_STAR); | |
1351 EMIT(NFA_CONCAT); | |
1352 skipchr(); /* skip the \+ */ | |
1353 break; | |
1354 | |
1355 case Magic('@'): | |
1356 op = no_Magic(getchr()); | |
1357 switch(op) | |
1358 { | |
1359 case '=': | |
1360 EMIT(NFA_PREV_ATOM_NO_WIDTH); | |
1361 break; | |
1362 case '!': | |
1363 case '<': | |
1364 case '>': | |
1365 /* Not supported yet */ | |
1366 return FAIL; | |
1367 default: | |
1368 syntax_error = TRUE; | |
4450 | 1369 EMSGN(_("E869: (NFA) Unknown operator '\\@%c'"), op); |
4444 | 1370 return FAIL; |
1371 } | |
1372 break; | |
1373 | |
1374 case Magic('?'): | |
1375 case Magic('='): | |
1376 EMIT(NFA_QUEST); | |
1377 break; | |
1378 | |
1379 case Magic('{'): | |
1380 /* a{2,5} will expand to 'aaa?a?a?' | |
1381 * a{-1,3} will expand to 'aa??a??', where ?? is the nongreedy | |
1382 * version of '?' | |
1383 * \v(ab){2,3} will expand to '(ab)(ab)(ab)?', where all the | |
1384 * parenthesis have the same id | |
1385 */ | |
1386 | |
1387 greedy = TRUE; | |
1388 c2 = peekchr(); | |
1389 if (c2 == '-' || c2 == Magic('-')) | |
1390 { | |
1391 skipchr(); | |
1392 greedy = FALSE; | |
1393 } | |
1394 if (!read_limits(&minval, &maxval)) | |
1395 { | |
1396 syntax_error = TRUE; | |
1397 EMSG_RET_FAIL(_("E870: (NFA regexp) Error reading repetition limits")); | |
1398 } | |
1399 /* <atom>{0,inf}, <atom>{0,} and <atom>{} are equivalent to | |
1400 * <atom>* */ | |
1401 if (minval == 0 && maxval == MAX_LIMIT && greedy) | |
1402 { | |
1403 EMIT(NFA_STAR); | |
1404 break; | |
1405 } | |
1406 | |
1407 if (maxval > NFA_BRACES_MAXLIMIT) | |
1408 { | |
1409 /* This would yield a huge automaton and use too much memory. | |
1410 * Revert to old engine */ | |
1411 return FAIL; | |
1412 } | |
1413 | |
1414 /* Special case: x{0} or x{-0} */ | |
1415 if (maxval == 0) | |
1416 { | |
1417 /* Ignore result of previous call to nfa_regatom() */ | |
1418 post_ptr = my_post_start; | |
1419 /* NFA_SKIP_CHAR has 0-length and works everywhere */ | |
1420 EMIT(NFA_SKIP_CHAR); | |
1421 return OK; | |
1422 } | |
1423 | |
1424 /* Ignore previous call to nfa_regatom() */ | |
1425 post_ptr = my_post_start; | |
1426 /* Save pos after the repeated atom and the \{} */ | |
1427 new_regparse = regparse; | |
1428 | |
1429 new_regparse = regparse; | |
1430 quest = (greedy == TRUE? NFA_QUEST : NFA_QUEST_NONGREEDY); | |
1431 for (i = 0; i < maxval; i++) | |
1432 { | |
1433 /* Goto beginning of the repeated atom */ | |
1434 regparse = old_regparse; | |
1435 curchr = -1; | |
1436 /* Restore count of parenthesis */ | |
1437 regnpar = old_regnpar; | |
1438 old_post_ptr = post_ptr; | |
1439 if (nfa_regatom() == FAIL) | |
1440 return FAIL; | |
1441 /* after "minval" times, atoms are optional */ | |
1442 if (i + 1 > minval) | |
1443 EMIT(quest); | |
1444 if (old_post_ptr != my_post_start) | |
1445 EMIT(NFA_CONCAT); | |
1446 } | |
1447 | |
1448 /* Go to just after the repeated atom and the \{} */ | |
1449 regparse = new_regparse; | |
1450 curchr = -1; | |
1451 | |
1452 break; | |
1453 | |
1454 | |
1455 default: | |
1456 break; | |
1457 } /* end switch */ | |
1458 | |
1459 if (re_multi_type(peekchr()) != NOT_MULTI) | |
1460 { | |
1461 /* Can't have a multi follow a multi. */ | |
1462 syntax_error = TRUE; | |
1463 EMSG_RET_FAIL(_("E871: (NFA regexp) Can't have a multi follow a multi !")); | |
1464 } | |
1465 | |
1466 return OK; | |
1467 } | |
1468 | |
1469 /* | |
1470 * Parse one or more pieces, concatenated. It matches a match for the | |
1471 * first piece, followed by a match for the second piece, etc. Example: | |
1472 * "f[0-9]b", first matches "f", then a digit and then "b". | |
1473 * | |
1474 * concat ::= piece | |
1475 * or piece piece | |
1476 * or piece piece piece | |
1477 * etc. | |
1478 */ | |
1479 static int | |
1480 nfa_regconcat() | |
1481 { | |
1482 int cont = TRUE; | |
1483 int first = TRUE; | |
1484 | |
1485 while (cont) | |
1486 { | |
1487 switch (peekchr()) | |
1488 { | |
1489 case NUL: | |
1490 case Magic('|'): | |
1491 case Magic('&'): | |
1492 case Magic(')'): | |
1493 cont = FALSE; | |
1494 break; | |
1495 | |
1496 case Magic('Z'): | |
1497 #ifdef FEAT_MBYTE | |
1498 regflags |= RF_ICOMBINE; | |
1499 #endif | |
1500 skipchr_keepstart(); | |
1501 break; | |
1502 case Magic('c'): | |
1503 regflags |= RF_ICASE; | |
1504 skipchr_keepstart(); | |
1505 break; | |
1506 case Magic('C'): | |
1507 regflags |= RF_NOICASE; | |
1508 skipchr_keepstart(); | |
1509 break; | |
1510 case Magic('v'): | |
1511 reg_magic = MAGIC_ALL; | |
1512 skipchr_keepstart(); | |
1513 curchr = -1; | |
1514 break; | |
1515 case Magic('m'): | |
1516 reg_magic = MAGIC_ON; | |
1517 skipchr_keepstart(); | |
1518 curchr = -1; | |
1519 break; | |
1520 case Magic('M'): | |
1521 reg_magic = MAGIC_OFF; | |
1522 skipchr_keepstart(); | |
1523 curchr = -1; | |
1524 break; | |
1525 case Magic('V'): | |
1526 reg_magic = MAGIC_NONE; | |
1527 skipchr_keepstart(); | |
1528 curchr = -1; | |
1529 break; | |
1530 | |
1531 default: | |
1532 if (nfa_regpiece() == FAIL) | |
1533 return FAIL; | |
1534 if (first == FALSE) | |
1535 EMIT(NFA_CONCAT); | |
1536 else | |
1537 first = FALSE; | |
1538 break; | |
1539 } | |
1540 } | |
1541 | |
1542 return OK; | |
1543 } | |
1544 | |
1545 /* | |
1546 * Parse a branch, one or more concats, separated by "\&". It matches the | |
1547 * last concat, but only if all the preceding concats also match at the same | |
1548 * position. Examples: | |
1549 * "foobeep\&..." matches "foo" in "foobeep". | |
1550 * ".*Peter\&.*Bob" matches in a line containing both "Peter" and "Bob" | |
1551 * | |
1552 * branch ::= concat | |
1553 * or concat \& concat | |
1554 * or concat \& concat \& concat | |
1555 * etc. | |
1556 */ | |
1557 static int | |
1558 nfa_regbranch() | |
1559 { | |
1560 int ch; | |
1561 int *old_post_ptr; | |
1562 | |
1563 old_post_ptr = post_ptr; | |
1564 | |
1565 /* First branch, possibly the only one */ | |
1566 if (nfa_regconcat() == FAIL) | |
1567 return FAIL; | |
1568 | |
1569 ch = peekchr(); | |
1570 /* Try next concats */ | |
1571 while (ch == Magic('&')) | |
1572 { | |
1573 skipchr(); | |
1574 EMIT(NFA_NOPEN); | |
1575 EMIT(NFA_PREV_ATOM_NO_WIDTH); | |
1576 old_post_ptr = post_ptr; | |
1577 if (nfa_regconcat() == FAIL) | |
1578 return FAIL; | |
1579 /* if concat is empty, skip a input char. But do emit a node */ | |
1580 if (old_post_ptr == post_ptr) | |
1581 EMIT(NFA_SKIP_CHAR); | |
1582 EMIT(NFA_CONCAT); | |
1583 ch = peekchr(); | |
1584 } | |
1585 | |
1586 /* Even if a branch is empty, emit one node for it */ | |
1587 if (old_post_ptr == post_ptr) | |
1588 EMIT(NFA_SKIP_CHAR); | |
1589 | |
1590 return OK; | |
1591 } | |
1592 | |
1593 /* | |
1594 * Parse a pattern, one or more branches, separated by "\|". It matches | |
1595 * anything that matches one of the branches. Example: "foo\|beep" matches | |
1596 * "foo" and matches "beep". If more than one branch matches, the first one | |
1597 * is used. | |
1598 * | |
1599 * pattern ::= branch | |
1600 * or branch \| branch | |
1601 * or branch \| branch \| branch | |
1602 * etc. | |
1603 */ | |
1604 static int | |
1605 nfa_reg(paren) | |
1606 int paren; /* REG_NOPAREN, REG_PAREN, REG_NPAREN or REG_ZPAREN */ | |
1607 { | |
1608 int parno = 0; | |
1609 | |
1610 #ifdef FEAT_SYN_HL | |
1611 #endif | |
1612 if (paren == REG_PAREN) | |
1613 { | |
1614 if (regnpar >= NSUBEXP) /* Too many `(' */ | |
1615 { | |
1616 syntax_error = TRUE; | |
1617 EMSG_RET_FAIL(_("E872: (NFA regexp) Too many '('")); | |
1618 } | |
1619 parno = regnpar++; | |
1620 } | |
1621 | |
1622 if (nfa_regbranch() == FAIL) | |
1623 return FAIL; /* cascaded error */ | |
1624 | |
1625 while (peekchr() == Magic('|')) | |
1626 { | |
1627 skipchr(); | |
1628 if (nfa_regbranch() == FAIL) | |
1629 return FAIL; /* cascaded error */ | |
1630 EMIT(NFA_OR); | |
1631 } | |
1632 | |
1633 /* Check for proper termination. */ | |
1634 if (paren != REG_NOPAREN && getchr() != Magic(')')) | |
1635 { | |
1636 syntax_error = TRUE; | |
1637 if (paren == REG_NPAREN) | |
1638 EMSG2_RET_FAIL(_(e_unmatchedpp), reg_magic == MAGIC_ALL); | |
1639 else | |
1640 EMSG2_RET_FAIL(_(e_unmatchedp), reg_magic == MAGIC_ALL); | |
1641 } | |
1642 else if (paren == REG_NOPAREN && peekchr() != NUL) | |
1643 { | |
1644 syntax_error = TRUE; | |
1645 if (peekchr() == Magic(')')) | |
1646 EMSG2_RET_FAIL(_(e_unmatchedpar), reg_magic == MAGIC_ALL); | |
1647 else | |
1648 EMSG_RET_FAIL(_("E873: (NFA regexp) proper termination error")); | |
1649 } | |
1650 /* | |
1651 * Here we set the flag allowing back references to this set of | |
1652 * parentheses. | |
1653 */ | |
1654 if (paren == REG_PAREN) | |
1655 { | |
1656 had_endbrace[parno] = TRUE; /* have seen the close paren */ | |
1657 EMIT(NFA_MOPEN + parno); | |
1658 } | |
1659 | |
1660 return OK; | |
1661 } | |
1662 | |
1663 typedef struct | |
1664 { | |
1665 char_u *start[NSUBEXP]; | |
1666 char_u *end[NSUBEXP]; | |
1667 lpos_T startpos[NSUBEXP]; | |
1668 lpos_T endpos[NSUBEXP]; | |
1669 } regsub_T; | |
1670 | |
1671 static int nfa_regmatch __ARGS((nfa_state_T *start, regsub_T *submatch, regsub_T *m)); | |
1672 | |
1673 #ifdef DEBUG | |
1674 static char_u code[50]; | |
1675 | |
1676 static void | |
1677 nfa_set_code(c) | |
1678 int c; | |
1679 { | |
1680 int addnl = FALSE; | |
1681 | |
1682 if (c >= NFA_FIRST_NL && c <= NFA_LAST_NL) | |
1683 { | |
1684 addnl = TRUE; | |
1685 c -= ADD_NL; | |
1686 } | |
1687 | |
1688 STRCPY(code, ""); | |
1689 switch (c) | |
1690 { | |
1691 case NFA_MATCH: STRCPY(code, "NFA_MATCH "); break; | |
1692 case NFA_SPLIT: STRCPY(code, "NFA_SPLIT "); break; | |
1693 case NFA_CONCAT: STRCPY(code, "NFA_CONCAT "); break; | |
1694 case NFA_NEWL: STRCPY(code, "NFA_NEWL "); break; | |
1695 case NFA_ZSTART: STRCPY(code, "NFA_ZSTART"); break; | |
1696 case NFA_ZEND: STRCPY(code, "NFA_ZEND"); break; | |
1697 | |
1698 case NFA_PREV_ATOM_NO_WIDTH: | |
1699 STRCPY(code, "NFA_PREV_ATOM_NO_WIDTH"); break; | |
1700 case NFA_NOPEN: STRCPY(code, "NFA_MOPEN_INVISIBLE"); break; | |
1701 case NFA_NCLOSE: STRCPY(code, "NFA_MCLOSE_INVISIBLE"); break; | |
1702 case NFA_START_INVISIBLE: STRCPY(code, "NFA_START_INVISIBLE"); break; | |
1703 case NFA_END_INVISIBLE: STRCPY(code, "NFA_END_INVISIBLE"); break; | |
1704 | |
1705 case NFA_MULTIBYTE: STRCPY(code, "NFA_MULTIBYTE"); break; | |
1706 case NFA_END_MULTIBYTE: STRCPY(code, "NFA_END_MULTIBYTE"); break; | |
1707 | |
1708 case NFA_COMPOSING: STRCPY(code, "NFA_COMPOSING"); break; | |
1709 case NFA_END_COMPOSING: STRCPY(code, "NFA_END_COMPOSING"); break; | |
1710 | |
1711 case NFA_MOPEN + 0: | |
1712 case NFA_MOPEN + 1: | |
1713 case NFA_MOPEN + 2: | |
1714 case NFA_MOPEN + 3: | |
1715 case NFA_MOPEN + 4: | |
1716 case NFA_MOPEN + 5: | |
1717 case NFA_MOPEN + 6: | |
1718 case NFA_MOPEN + 7: | |
1719 case NFA_MOPEN + 8: | |
1720 case NFA_MOPEN + 9: | |
1721 STRCPY(code, "NFA_MOPEN(x)"); | |
1722 code[10] = c - NFA_MOPEN + '0'; | |
1723 break; | |
1724 case NFA_MCLOSE + 0: | |
1725 case NFA_MCLOSE + 1: | |
1726 case NFA_MCLOSE + 2: | |
1727 case NFA_MCLOSE + 3: | |
1728 case NFA_MCLOSE + 4: | |
1729 case NFA_MCLOSE + 5: | |
1730 case NFA_MCLOSE + 6: | |
1731 case NFA_MCLOSE + 7: | |
1732 case NFA_MCLOSE + 8: | |
1733 case NFA_MCLOSE + 9: | |
1734 STRCPY(code, "NFA_MCLOSE(x)"); | |
1735 code[11] = c - NFA_MCLOSE + '0'; | |
1736 break; | |
1737 case NFA_EOL: STRCPY(code, "NFA_EOL "); break; | |
1738 case NFA_BOL: STRCPY(code, "NFA_BOL "); break; | |
1739 case NFA_EOW: STRCPY(code, "NFA_EOW "); break; | |
1740 case NFA_BOW: STRCPY(code, "NFA_BOW "); break; | |
1741 case NFA_STAR: STRCPY(code, "NFA_STAR "); break; | |
1742 case NFA_PLUS: STRCPY(code, "NFA_PLUS "); break; | |
1743 case NFA_NOT: STRCPY(code, "NFA_NOT "); break; | |
1744 case NFA_SKIP_CHAR: STRCPY(code, "NFA_SKIP_CHAR"); break; | |
1745 case NFA_OR: STRCPY(code, "NFA_OR"); break; | |
1746 case NFA_QUEST: STRCPY(code, "NFA_QUEST"); break; | |
1747 case NFA_QUEST_NONGREEDY: STRCPY(code, "NFA_QUEST_NON_GREEDY"); break; | |
1748 case NFA_END_NEG_RANGE: STRCPY(code, "NFA_END_NEG_RANGE"); break; | |
1749 case NFA_CLASS_ALNUM: STRCPY(code, "NFA_CLASS_ALNUM"); break; | |
1750 case NFA_CLASS_ALPHA: STRCPY(code, "NFA_CLASS_ALPHA"); break; | |
1751 case NFA_CLASS_BLANK: STRCPY(code, "NFA_CLASS_BLANK"); break; | |
1752 case NFA_CLASS_CNTRL: STRCPY(code, "NFA_CLASS_CNTRL"); break; | |
1753 case NFA_CLASS_DIGIT: STRCPY(code, "NFA_CLASS_DIGIT"); break; | |
1754 case NFA_CLASS_GRAPH: STRCPY(code, "NFA_CLASS_GRAPH"); break; | |
1755 case NFA_CLASS_LOWER: STRCPY(code, "NFA_CLASS_LOWER"); break; | |
1756 case NFA_CLASS_PRINT: STRCPY(code, "NFA_CLASS_PRINT"); break; | |
1757 case NFA_CLASS_PUNCT: STRCPY(code, "NFA_CLASS_PUNCT"); break; | |
1758 case NFA_CLASS_SPACE: STRCPY(code, "NFA_CLASS_SPACE"); break; | |
1759 case NFA_CLASS_UPPER: STRCPY(code, "NFA_CLASS_UPPER"); break; | |
1760 case NFA_CLASS_XDIGIT: STRCPY(code, "NFA_CLASS_XDIGIT"); break; | |
1761 case NFA_CLASS_TAB: STRCPY(code, "NFA_CLASS_TAB"); break; | |
1762 case NFA_CLASS_RETURN: STRCPY(code, "NFA_CLASS_RETURN"); break; | |
1763 case NFA_CLASS_BACKSPACE: STRCPY(code, "NFA_CLASS_BACKSPACE"); break; | |
1764 case NFA_CLASS_ESCAPE: STRCPY(code, "NFA_CLASS_ESCAPE"); break; | |
1765 | |
1766 case NFA_ANY: STRCPY(code, "NFA_ANY"); break; | |
1767 case NFA_IDENT: STRCPY(code, "NFA_IDENT"); break; | |
1768 case NFA_SIDENT:STRCPY(code, "NFA_SIDENT"); break; | |
1769 case NFA_KWORD: STRCPY(code, "NFA_KWORD"); break; | |
1770 case NFA_SKWORD:STRCPY(code, "NFA_SKWORD"); break; | |
1771 case NFA_FNAME: STRCPY(code, "NFA_FNAME"); break; | |
1772 case NFA_SFNAME:STRCPY(code, "NFA_SFNAME"); break; | |
1773 case NFA_PRINT: STRCPY(code, "NFA_PRINT"); break; | |
1774 case NFA_SPRINT:STRCPY(code, "NFA_SPRINT"); break; | |
1775 case NFA_WHITE: STRCPY(code, "NFA_WHITE"); break; | |
1776 case NFA_NWHITE:STRCPY(code, "NFA_NWHITE"); break; | |
1777 case NFA_DIGIT: STRCPY(code, "NFA_DIGIT"); break; | |
1778 case NFA_NDIGIT:STRCPY(code, "NFA_NDIGIT"); break; | |
1779 case NFA_HEX: STRCPY(code, "NFA_HEX"); break; | |
1780 case NFA_NHEX: STRCPY(code, "NFA_NHEX"); break; | |
1781 case NFA_OCTAL: STRCPY(code, "NFA_OCTAL"); break; | |
1782 case NFA_NOCTAL:STRCPY(code, "NFA_NOCTAL"); break; | |
1783 case NFA_WORD: STRCPY(code, "NFA_WORD"); break; | |
1784 case NFA_NWORD: STRCPY(code, "NFA_NWORD"); break; | |
1785 case NFA_HEAD: STRCPY(code, "NFA_HEAD"); break; | |
1786 case NFA_NHEAD: STRCPY(code, "NFA_NHEAD"); break; | |
1787 case NFA_ALPHA: STRCPY(code, "NFA_ALPHA"); break; | |
1788 case NFA_NALPHA:STRCPY(code, "NFA_NALPHA"); break; | |
1789 case NFA_LOWER: STRCPY(code, "NFA_LOWER"); break; | |
1790 case NFA_NLOWER:STRCPY(code, "NFA_NLOWER"); break; | |
1791 case NFA_UPPER: STRCPY(code, "NFA_UPPER"); break; | |
1792 case NFA_NUPPER:STRCPY(code, "NFA_NUPPER"); break; | |
1793 | |
1794 default: | |
1795 STRCPY(code, "CHAR(x)"); | |
1796 code[5] = c; | |
1797 } | |
1798 | |
1799 if (addnl == TRUE) | |
1800 STRCAT(code, " + NEWLINE "); | |
1801 | |
1802 } | |
1803 | |
1804 #ifdef ENABLE_LOG | |
1805 static FILE *log_fd; | |
1806 | |
1807 /* | |
1808 * Print the postfix notation of the current regexp. | |
1809 */ | |
1810 static void | |
1811 nfa_postfix_dump(expr, retval) | |
1812 char_u *expr; | |
1813 int retval; | |
1814 { | |
1815 int *p; | |
1816 FILE *f; | |
1817 | |
1818 f = fopen("LOG.log", "a"); | |
1819 if (f != NULL) | |
1820 { | |
1821 fprintf(f, "\n-------------------------\n"); | |
1822 if (retval == FAIL) | |
1823 fprintf(f, ">>> NFA engine failed ... \n"); | |
1824 else if (retval == OK) | |
1825 fprintf(f, ">>> NFA engine succeeded !\n"); | |
1826 fprintf(f, "Regexp: \"%s\"\nPostfix notation (char): \"", expr); | |
4464 | 1827 for (p = post_start; *p && p < post_end; p++) |
4444 | 1828 { |
1829 nfa_set_code(*p); | |
1830 fprintf(f, "%s, ", code); | |
1831 } | |
1832 fprintf(f, "\"\nPostfix notation (int): "); | |
4464 | 1833 for (p = post_start; *p && p < post_end; p++) |
4444 | 1834 fprintf(f, "%d ", *p); |
1835 fprintf(f, "\n\n"); | |
1836 fclose(f); | |
1837 } | |
1838 } | |
1839 | |
1840 /* | |
1841 * Print the NFA starting with a root node "state". | |
1842 */ | |
1843 static void | |
1844 nfa_print_state(debugf, state, ident) | |
1845 FILE *debugf; | |
1846 nfa_state_T *state; | |
1847 int ident; | |
1848 { | |
1849 int i; | |
1850 | |
1851 if (state == NULL) | |
1852 return; | |
1853 | |
1854 fprintf(debugf, "(%2d)", abs(state->id)); | |
1855 for (i = 0; i < ident; i++) | |
1856 fprintf(debugf, "%c", ' '); | |
1857 | |
1858 nfa_set_code(state->c); | |
1859 fprintf(debugf, "%s %s (%d) (id=%d)\n", | |
1860 state->negated ? "NOT" : "", code, state->c, abs(state->id)); | |
1861 if (state->id < 0) | |
1862 return; | |
1863 | |
1864 state->id = abs(state->id) * -1; | |
1865 nfa_print_state(debugf, state->out, ident + 4); | |
1866 nfa_print_state(debugf, state->out1, ident + 4); | |
1867 } | |
1868 | |
1869 /* | |
1870 * Print the NFA state machine. | |
1871 */ | |
1872 static void | |
1873 nfa_dump(prog) | |
1874 nfa_regprog_T *prog; | |
1875 { | |
1876 FILE *debugf = fopen("LOG.log", "a"); | |
1877 | |
1878 if (debugf != NULL) | |
1879 { | |
1880 nfa_print_state(debugf, prog->start, 0); | |
1881 fclose(debugf); | |
1882 } | |
1883 } | |
1884 #endif /* ENABLE_LOG */ | |
1885 #endif /* DEBUG */ | |
1886 | |
1887 /* | |
1888 * Parse r.e. @expr and convert it into postfix form. | |
1889 * Return the postfix string on success, NULL otherwise. | |
1890 */ | |
1891 static int * | |
1892 re2post() | |
1893 { | |
1894 if (nfa_reg(REG_NOPAREN) == FAIL) | |
1895 return NULL; | |
1896 EMIT(NFA_MOPEN); | |
1897 return post_start; | |
1898 } | |
1899 | |
1900 /* NB. Some of the code below is inspired by Russ's. */ | |
1901 | |
1902 /* | |
1903 * Represents an NFA state plus zero or one or two arrows exiting. | |
1904 * if c == MATCH, no arrows out; matching state. | |
1905 * If c == SPLIT, unlabeled arrows to out and out1 (if != NULL). | |
1906 * If c < 256, labeled arrow with character c to out. | |
1907 */ | |
1908 | |
1909 static nfa_state_T *state_ptr; /* points to nfa_prog->state */ | |
1910 | |
1911 /* | |
1912 * Allocate and initialize nfa_state_T. | |
1913 */ | |
1914 static nfa_state_T * | |
1915 new_state(c, out, out1) | |
1916 int c; | |
1917 nfa_state_T *out; | |
1918 nfa_state_T *out1; | |
1919 { | |
1920 nfa_state_T *s; | |
1921 | |
1922 if (istate >= nstate) | |
1923 return NULL; | |
1924 | |
1925 s = &state_ptr[istate++]; | |
1926 | |
1927 s->c = c; | |
1928 s->out = out; | |
1929 s->out1 = out1; | |
1930 | |
1931 s->id = istate; | |
1932 s->lastlist = 0; | |
1933 s->lastthread = NULL; | |
1934 s->visits = 0; | |
1935 s->negated = FALSE; | |
1936 | |
1937 return s; | |
1938 } | |
1939 | |
1940 /* | |
1941 * A partially built NFA without the matching state filled in. | |
1942 * Frag_T.start points at the start state. | |
1943 * Frag_T.out is a list of places that need to be set to the | |
1944 * next state for this fragment. | |
1945 */ | |
1946 typedef union Ptrlist Ptrlist; | |
1947 struct Frag | |
1948 { | |
1949 nfa_state_T *start; | |
1950 Ptrlist *out; | |
1951 }; | |
1952 typedef struct Frag Frag_T; | |
1953 | |
1954 static Frag_T frag __ARGS((nfa_state_T *start, Ptrlist *out)); | |
1955 static Ptrlist *list1 __ARGS((nfa_state_T **outp)); | |
1956 static void patch __ARGS((Ptrlist *l, nfa_state_T *s)); | |
1957 static Ptrlist *append __ARGS((Ptrlist *l1, Ptrlist *l2)); | |
1958 static void st_push __ARGS((Frag_T s, Frag_T **p, Frag_T *stack_end)); | |
1959 static Frag_T st_pop __ARGS((Frag_T **p, Frag_T *stack)); | |
1960 | |
1961 /* | |
4456 | 1962 * Initialize a Frag_T struct and return it. |
4444 | 1963 */ |
1964 static Frag_T | |
1965 frag(start, out) | |
1966 nfa_state_T *start; | |
1967 Ptrlist *out; | |
1968 { | |
4456 | 1969 Frag_T n; |
1970 | |
1971 n.start = start; | |
1972 n.out = out; | |
4444 | 1973 return n; |
1974 } | |
1975 | |
1976 /* | |
1977 * Since the out pointers in the list are always | |
1978 * uninitialized, we use the pointers themselves | |
1979 * as storage for the Ptrlists. | |
1980 */ | |
1981 union Ptrlist | |
1982 { | |
1983 Ptrlist *next; | |
1984 nfa_state_T *s; | |
1985 }; | |
1986 | |
1987 /* | |
1988 * Create singleton list containing just outp. | |
1989 */ | |
1990 static Ptrlist * | |
1991 list1(outp) | |
1992 nfa_state_T **outp; | |
1993 { | |
1994 Ptrlist *l; | |
1995 | |
1996 l = (Ptrlist *)outp; | |
1997 l->next = NULL; | |
1998 return l; | |
1999 } | |
2000 | |
2001 /* | |
2002 * Patch the list of states at out to point to start. | |
2003 */ | |
2004 static void | |
2005 patch(l, s) | |
2006 Ptrlist *l; | |
2007 nfa_state_T *s; | |
2008 { | |
2009 Ptrlist *next; | |
2010 | |
2011 for (; l; l = next) | |
2012 { | |
2013 next = l->next; | |
2014 l->s = s; | |
2015 } | |
2016 } | |
2017 | |
2018 | |
2019 /* | |
2020 * Join the two lists l1 and l2, returning the combination. | |
2021 */ | |
2022 static Ptrlist * | |
2023 append(l1, l2) | |
2024 Ptrlist *l1; | |
2025 Ptrlist *l2; | |
2026 { | |
2027 Ptrlist *oldl1; | |
2028 | |
2029 oldl1 = l1; | |
2030 while (l1->next) | |
2031 l1 = l1->next; | |
2032 l1->next = l2; | |
2033 return oldl1; | |
2034 } | |
2035 | |
2036 /* | |
2037 * Stack used for transforming postfix form into NFA. | |
2038 */ | |
2039 static Frag_T empty; | |
2040 | |
2041 static void | |
2042 st_error(postfix, end, p) | |
2043 int *postfix; | |
2044 int *end; | |
2045 int *p; | |
2046 { | |
2047 FILE *df; | |
2048 int *p2; | |
2049 | |
2050 df = fopen("stack.err", "a"); | |
2051 if (df) | |
2052 { | |
2053 fprintf(df, "Error popping the stack!\n"); | |
2054 #ifdef DEBUG | |
2055 fprintf(df, "Current regexp is \"%s\"\n", nfa_regengine.expr); | |
2056 #endif | |
2057 fprintf(df, "Postfix form is: "); | |
2058 #ifdef DEBUG | |
2059 for (p2 = postfix; p2 < end; p2++) | |
2060 { | |
2061 nfa_set_code(*p2); | |
2062 fprintf(df, "%s, ", code); | |
2063 } | |
2064 nfa_set_code(*p); | |
2065 fprintf(df, "\nCurrent position is: "); | |
2066 for (p2 = postfix; p2 <= p; p2 ++) | |
2067 { | |
2068 nfa_set_code(*p2); | |
2069 fprintf(df, "%s, ", code); | |
2070 } | |
2071 #else | |
2072 for (p2 = postfix; p2 < end; p2++) | |
2073 { | |
2074 fprintf(df, "%d, ", *p2); | |
2075 } | |
2076 fprintf(df, "\nCurrent position is: "); | |
2077 for (p2 = postfix; p2 <= p; p2 ++) | |
2078 { | |
2079 fprintf(df, "%d, ", *p2); | |
2080 } | |
2081 #endif | |
2082 fprintf(df, "\n--------------------------\n"); | |
2083 fclose(df); | |
2084 } | |
2085 EMSG(_("E874: (NFA) Could not pop the stack !")); | |
2086 } | |
2087 | |
2088 /* | |
2089 * Push an item onto the stack. | |
2090 */ | |
2091 static void | |
2092 st_push(s, p, stack_end) | |
2093 Frag_T s; | |
2094 Frag_T **p; | |
2095 Frag_T *stack_end; | |
2096 { | |
2097 Frag_T *stackp = *p; | |
2098 | |
2099 if (stackp >= stack_end) | |
2100 return; | |
2101 *stackp = s; | |
2102 *p = *p + 1; | |
2103 } | |
2104 | |
2105 /* | |
2106 * Pop an item from the stack. | |
2107 */ | |
2108 static Frag_T | |
2109 st_pop(p, stack) | |
2110 Frag_T **p; | |
2111 Frag_T *stack; | |
2112 { | |
2113 Frag_T *stackp; | |
2114 | |
2115 *p = *p - 1; | |
2116 stackp = *p; | |
2117 if (stackp < stack) | |
2118 return empty; | |
2119 return **p; | |
2120 } | |
2121 | |
2122 /* | |
2123 * Convert a postfix form into its equivalent NFA. | |
2124 * Return the NFA start state on success, NULL otherwise. | |
2125 */ | |
2126 static nfa_state_T * | |
2127 post2nfa(postfix, end, nfa_calc_size) | |
2128 int *postfix; | |
2129 int *end; | |
2130 int nfa_calc_size; | |
2131 { | |
2132 int *p; | |
2133 int mopen; | |
2134 int mclose; | |
2135 Frag_T *stack = NULL; | |
2136 Frag_T *stackp = NULL; | |
2137 Frag_T *stack_end = NULL; | |
2138 Frag_T e1; | |
2139 Frag_T e2; | |
2140 Frag_T e; | |
2141 nfa_state_T *s; | |
2142 nfa_state_T *s1; | |
2143 nfa_state_T *matchstate; | |
4484 | 2144 nfa_state_T *ret = NULL; |
4444 | 2145 |
2146 if (postfix == NULL) | |
2147 return NULL; | |
2148 | |
4456 | 2149 #define PUSH(s) st_push((s), &stackp, stack_end) |
4444 | 2150 #define POP() st_pop(&stackp, stack); \ |
2151 if (stackp < stack) \ | |
2152 { \ | |
2153 st_error(postfix, end, p); \ | |
2154 return NULL; \ | |
2155 } | |
2156 | |
2157 if (nfa_calc_size == FALSE) | |
2158 { | |
2159 /* Allocate space for the stack. Max states on the stack : nstate */ | |
4462 | 2160 stack = (Frag_T *) lalloc((nstate + 1) * sizeof(Frag_T), TRUE); |
4444 | 2161 stackp = stack; |
4462 | 2162 stack_end = stack + (nstate + 1); |
4444 | 2163 } |
2164 | |
2165 for (p = postfix; p < end; ++p) | |
2166 { | |
2167 switch (*p) | |
2168 { | |
2169 case NFA_CONCAT: | |
2170 /* Catenation. | |
2171 * Pay attention: this operator does not exist | |
2172 * in the r.e. itself (it is implicit, really). | |
2173 * It is added when r.e. is translated to postfix | |
2174 * form in re2post(). | |
2175 * | |
2176 * No new state added here. */ | |
2177 if (nfa_calc_size == TRUE) | |
2178 { | |
4458 | 2179 /* nstate += 0; */ |
4444 | 2180 break; |
2181 } | |
2182 e2 = POP(); | |
2183 e1 = POP(); | |
2184 patch(e1.out, e2.start); | |
2185 PUSH(frag(e1.start, e2.out)); | |
2186 break; | |
2187 | |
2188 case NFA_NOT: | |
2189 /* Negation of a character */ | |
2190 if (nfa_calc_size == TRUE) | |
2191 { | |
4458 | 2192 /* nstate += 0; */ |
4444 | 2193 break; |
2194 } | |
2195 e1 = POP(); | |
2196 e1.start->negated = TRUE; | |
2197 if (e1.start->c == NFA_MULTIBYTE || e1.start->c == NFA_COMPOSING) | |
2198 e1.start->out1->negated = TRUE; | |
2199 PUSH(e1); | |
2200 break; | |
2201 | |
2202 case NFA_OR: | |
2203 /* Alternation */ | |
2204 if (nfa_calc_size == TRUE) | |
2205 { | |
4458 | 2206 nstate++; |
4444 | 2207 break; |
2208 } | |
2209 e2 = POP(); | |
2210 e1 = POP(); | |
2211 s = new_state(NFA_SPLIT, e1.start, e2.start); | |
2212 if (s == NULL) | |
4484 | 2213 goto theend; |
4444 | 2214 PUSH(frag(s, append(e1.out, e2.out))); |
2215 break; | |
2216 | |
2217 case NFA_STAR: | |
2218 /* Zero or more */ | |
2219 if (nfa_calc_size == TRUE) | |
2220 { | |
4458 | 2221 nstate++; |
4444 | 2222 break; |
2223 } | |
2224 e = POP(); | |
2225 s = new_state(NFA_SPLIT, e.start, NULL); | |
2226 if (s == NULL) | |
4484 | 2227 goto theend; |
4444 | 2228 patch(e.out, s); |
2229 PUSH(frag(s, list1(&s->out1))); | |
2230 break; | |
2231 | |
2232 case NFA_QUEST: | |
2233 /* one or zero atoms=> greedy match */ | |
2234 if (nfa_calc_size == TRUE) | |
2235 { | |
4458 | 2236 nstate++; |
4444 | 2237 break; |
2238 } | |
2239 e = POP(); | |
2240 s = new_state(NFA_SPLIT, e.start, NULL); | |
2241 if (s == NULL) | |
4484 | 2242 goto theend; |
4444 | 2243 PUSH(frag(s, append(e.out, list1(&s->out1)))); |
2244 break; | |
2245 | |
2246 case NFA_QUEST_NONGREEDY: | |
2247 /* zero or one atoms => non-greedy match */ | |
2248 if (nfa_calc_size == TRUE) | |
2249 { | |
4458 | 2250 nstate++; |
4444 | 2251 break; |
2252 } | |
2253 e = POP(); | |
2254 s = new_state(NFA_SPLIT, NULL, e.start); | |
2255 if (s == NULL) | |
4484 | 2256 goto theend; |
4444 | 2257 PUSH(frag(s, append(e.out, list1(&s->out)))); |
2258 break; | |
2259 | |
2260 case NFA_PLUS: | |
2261 /* One or more */ | |
2262 if (nfa_calc_size == TRUE) | |
2263 { | |
4458 | 2264 nstate++; |
4444 | 2265 break; |
2266 } | |
2267 e = POP(); | |
2268 s = new_state(NFA_SPLIT, e.start, NULL); | |
2269 if (s == NULL) | |
4484 | 2270 goto theend; |
4444 | 2271 patch(e.out, s); |
2272 PUSH(frag(e.start, list1(&s->out1))); | |
2273 break; | |
2274 | |
2275 case NFA_SKIP_CHAR: | |
2276 /* Symbol of 0-length, Used in a repetition | |
2277 * with max/min count of 0 */ | |
2278 if (nfa_calc_size == TRUE) | |
2279 { | |
4458 | 2280 nstate++; |
4444 | 2281 break; |
2282 } | |
2283 s = new_state(NFA_SKIP_CHAR, NULL, NULL); | |
2284 if (s == NULL) | |
4484 | 2285 goto theend; |
4444 | 2286 PUSH(frag(s, list1(&s->out))); |
2287 break; | |
2288 | |
2289 case NFA_PREV_ATOM_NO_WIDTH: | |
2290 /* The \@= operator: match the preceding atom with 0 width. | |
2291 * Surrounds the preceding atom with START_INVISIBLE and | |
2292 * END_INVISIBLE, similarly to MOPEN. | |
2293 */ | |
2294 /* TODO: Maybe this drops the speed? */ | |
4484 | 2295 goto theend; |
4444 | 2296 |
2297 if (nfa_calc_size == TRUE) | |
2298 { | |
2299 nstate += 2; | |
2300 break; | |
2301 } | |
2302 e = POP(); | |
2303 s1 = new_state(NFA_END_INVISIBLE, NULL, NULL); | |
2304 if (s1 == NULL) | |
4484 | 2305 goto theend; |
4444 | 2306 patch(e.out, s1); |
2307 | |
2308 s = new_state(NFA_START_INVISIBLE, e.start, s1); | |
2309 if (s == NULL) | |
4484 | 2310 goto theend; |
4444 | 2311 PUSH(frag(s, list1(&s1->out))); |
2312 break; | |
2313 | |
2314 case NFA_MOPEN + 0: /* Submatch */ | |
2315 case NFA_MOPEN + 1: | |
2316 case NFA_MOPEN + 2: | |
2317 case NFA_MOPEN + 3: | |
2318 case NFA_MOPEN + 4: | |
2319 case NFA_MOPEN + 5: | |
2320 case NFA_MOPEN + 6: | |
2321 case NFA_MOPEN + 7: | |
2322 case NFA_MOPEN + 8: | |
2323 case NFA_MOPEN + 9: | |
2324 case NFA_NOPEN: /* \%( "Invisible Submatch" */ | |
2325 case NFA_MULTIBYTE: /* mbyte char */ | |
2326 case NFA_COMPOSING: /* composing char */ | |
2327 if (nfa_calc_size == TRUE) | |
2328 { | |
2329 nstate += 2; | |
2330 break; | |
2331 } | |
2332 | |
2333 mopen = *p; | |
2334 switch (*p) | |
2335 { | |
2336 case NFA_NOPEN: | |
2337 mclose = NFA_NCLOSE; | |
2338 break; | |
2339 case NFA_MULTIBYTE: | |
2340 mclose = NFA_END_MULTIBYTE; | |
2341 break; | |
2342 case NFA_COMPOSING: | |
2343 mclose = NFA_END_COMPOSING; | |
2344 break; | |
2345 default: | |
2346 /* NFA_MOPEN(0) ... NFA_MOPEN(9) */ | |
2347 mclose = *p + NSUBEXP; | |
2348 break; | |
2349 } | |
2350 | |
2351 /* Allow "NFA_MOPEN" as a valid postfix representation for | |
2352 * the empty regexp "". In this case, the NFA will be | |
2353 * NFA_MOPEN -> NFA_MCLOSE. Note that this also allows | |
2354 * empty groups of parenthesis, and empty mbyte chars */ | |
2355 if (stackp == stack) | |
2356 { | |
2357 s = new_state(mopen, NULL, NULL); | |
2358 if (s == NULL) | |
4484 | 2359 goto theend; |
4444 | 2360 s1 = new_state(mclose, NULL, NULL); |
2361 if (s1 == NULL) | |
4484 | 2362 goto theend; |
4444 | 2363 patch(list1(&s->out), s1); |
2364 PUSH(frag(s, list1(&s1->out))); | |
2365 break; | |
2366 } | |
2367 | |
2368 /* At least one node was emitted before NFA_MOPEN, so | |
2369 * at least one node will be between NFA_MOPEN and NFA_MCLOSE */ | |
2370 e = POP(); | |
2371 s = new_state(mopen, e.start, NULL); /* `(' */ | |
2372 if (s == NULL) | |
4484 | 2373 goto theend; |
4444 | 2374 |
2375 s1 = new_state(mclose, NULL, NULL); /* `)' */ | |
2376 if (s1 == NULL) | |
4484 | 2377 goto theend; |
4444 | 2378 patch(e.out, s1); |
2379 | |
2380 if (mopen == NFA_MULTIBYTE || mopen == NFA_COMPOSING) | |
2381 /* MULTIBYTE->out1 = END_MULTIBYTE | |
2382 * COMPOSING->out1 = END_COMPOSING */ | |
2383 patch(list1(&s->out1), s1); | |
2384 | |
2385 PUSH(frag(s, list1(&s1->out))); | |
2386 break; | |
2387 | |
2388 case NFA_ZSTART: | |
2389 case NFA_ZEND: | |
2390 default: | |
2391 /* Operands */ | |
2392 if (nfa_calc_size == TRUE) | |
2393 { | |
4458 | 2394 nstate++; |
4444 | 2395 break; |
2396 } | |
2397 s = new_state(*p, NULL, NULL); | |
2398 if (s == NULL) | |
4484 | 2399 goto theend; |
4444 | 2400 PUSH(frag(s, list1(&s->out))); |
2401 break; | |
2402 | |
2403 } /* switch(*p) */ | |
2404 | |
2405 } /* for(p = postfix; *p; ++p) */ | |
2406 | |
2407 if (nfa_calc_size == TRUE) | |
2408 { | |
4458 | 2409 nstate++; |
4484 | 2410 goto theend; /* Return value when counting size is ignored anyway */ |
4444 | 2411 } |
2412 | |
2413 e = POP(); | |
2414 if (stackp != stack) | |
2415 EMSG_RET_NULL(_("E875: (NFA regexp) (While converting from postfix to NFA), too many states left on stack")); | |
2416 | |
2417 if (istate >= nstate) | |
2418 EMSG_RET_NULL(_("E876: (NFA regexp) Not enough space to store the whole NFA ")); | |
2419 | |
2420 matchstate = &state_ptr[istate++]; /* the match state */ | |
2421 matchstate->c = NFA_MATCH; | |
2422 matchstate->out = matchstate->out1 = NULL; | |
2423 | |
2424 patch(e.out, matchstate); | |
4484 | 2425 ret = e.start; |
2426 | |
2427 theend: | |
2428 vim_free(stack); | |
2429 return ret; | |
4444 | 2430 |
2431 #undef POP1 | |
2432 #undef PUSH1 | |
2433 #undef POP2 | |
2434 #undef PUSH2 | |
2435 #undef POP | |
2436 #undef PUSH | |
2437 } | |
2438 | |
2439 /**************************************************************** | |
2440 * NFA execution code. | |
2441 ****************************************************************/ | |
2442 | |
2443 /* thread_T contains runtime information of a NFA state */ | |
2444 struct thread | |
2445 { | |
2446 nfa_state_T *state; | |
2447 regsub_T sub; /* submatch info */ | |
2448 }; | |
2449 | |
2450 typedef struct | |
2451 { | |
2452 thread_T *t; | |
2453 int n; | |
2454 } List; | |
2455 | |
2456 static void addstate __ARGS((List *l, nfa_state_T *state, regsub_T *m, int off, int lid, int *match)); | |
2457 | |
2458 static void | |
2459 addstate(l, state, m, off, lid, match) | |
2460 List *l; /* runtime state list */ | |
2461 nfa_state_T *state; /* state to update */ | |
2462 regsub_T *m; /* pointers to subexpressions */ | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2463 int off; /* byte offset, when -1 go to next line */ |
4444 | 2464 int lid; |
2465 int *match; /* found match? */ | |
2466 { | |
2467 regsub_T save; | |
2468 int subidx = 0; | |
2469 | |
2470 if (l == NULL || state == NULL) | |
2471 return; | |
2472 | |
2473 switch (state->c) | |
2474 { | |
2475 case NFA_SPLIT: | |
2476 case NFA_NOT: | |
2477 case NFA_NOPEN: | |
2478 case NFA_NCLOSE: | |
2479 case NFA_MCLOSE: | |
2480 case NFA_MCLOSE + 1: | |
2481 case NFA_MCLOSE + 2: | |
2482 case NFA_MCLOSE + 3: | |
2483 case NFA_MCLOSE + 4: | |
2484 case NFA_MCLOSE + 5: | |
2485 case NFA_MCLOSE + 6: | |
2486 case NFA_MCLOSE + 7: | |
2487 case NFA_MCLOSE + 8: | |
2488 case NFA_MCLOSE + 9: | |
2489 /* Do not remember these nodes in list "thislist" or "nextlist" */ | |
2490 break; | |
2491 | |
2492 default: | |
2493 if (state->lastlist == lid) | |
2494 { | |
2495 if (++state->visits > 2) | |
2496 return; | |
2497 } | |
2498 else | |
2499 { | |
2500 /* add the state to the list */ | |
2501 state->lastlist = lid; | |
2502 state->lastthread = &l->t[l->n++]; | |
2503 state->lastthread->state = state; | |
2504 state->lastthread->sub = *m; | |
2505 } | |
2506 } | |
2507 | |
2508 #ifdef ENABLE_LOG | |
2509 nfa_set_code(state->c); | |
2510 fprintf(log_fd, "> Adding state %d to list. Character %s, code %d\n", | |
2511 abs(state->id), code, state->c); | |
2512 #endif | |
2513 switch (state->c) | |
2514 { | |
2515 case NFA_MATCH: | |
2516 *match = TRUE; | |
2517 break; | |
2518 | |
2519 case NFA_SPLIT: | |
2520 addstate(l, state->out, m, off, lid, match); | |
2521 addstate(l, state->out1, m, off, lid, match); | |
2522 break; | |
2523 | |
2524 case NFA_SKIP_CHAR: | |
2525 addstate(l, state->out, m, off, lid, match); | |
2526 break; | |
2527 | |
2528 #if 0 | |
2529 case NFA_END_NEG_RANGE: | |
2530 /* Nothing to handle here. nfa_regmatch() will take care of it */ | |
2531 break; | |
2532 | |
2533 case NFA_NOT: | |
2534 EMSG(_("E999: (NFA regexp internal error) Should not process NOT node !")); | |
2535 #ifdef ENABLE_LOG | |
2536 fprintf(f, "\n\n>>> E999: Added state NFA_NOT to a list ... Something went wrong ! Why wasn't it processed already? \n\n"); | |
2537 #endif | |
2538 break; | |
2539 | |
2540 case NFA_COMPOSING: | |
2541 /* nfa_regmatch() will match all the bytes of this composing char. */ | |
2542 break; | |
2543 | |
2544 case NFA_MULTIBYTE: | |
2545 /* nfa_regmatch() will match all the bytes of this multibyte char. */ | |
2546 break; | |
2547 #endif | |
2548 | |
2549 case NFA_END_MULTIBYTE: | |
2550 /* Successfully matched this mbyte char */ | |
2551 addstate(l, state->out, m, off, lid, match); | |
2552 break; | |
2553 | |
2554 case NFA_NOPEN: | |
2555 case NFA_NCLOSE: | |
2556 addstate(l, state->out, m, off, lid, match); | |
2557 break; | |
2558 | |
2559 /* If this state is reached, then a recursive call of nfa_regmatch() | |
2560 * succeeded. the next call saves the found submatches in the | |
2561 * first state after the "invisible" branch. */ | |
2562 #if 0 | |
2563 case NFA_END_INVISIBLE: | |
2564 break; | |
2565 #endif | |
2566 | |
2567 case NFA_MOPEN + 0: | |
2568 case NFA_MOPEN + 1: | |
2569 case NFA_MOPEN + 2: | |
2570 case NFA_MOPEN + 3: | |
2571 case NFA_MOPEN + 4: | |
2572 case NFA_MOPEN + 5: | |
2573 case NFA_MOPEN + 6: | |
2574 case NFA_MOPEN + 7: | |
2575 case NFA_MOPEN + 8: | |
2576 case NFA_MOPEN + 9: | |
2577 case NFA_ZSTART: | |
2578 subidx = state->c - NFA_MOPEN; | |
2579 if (state->c == NFA_ZSTART) | |
2580 subidx = 0; | |
2581 | |
2582 if (REG_MULTI) | |
2583 { | |
2584 save.startpos[subidx] = m->startpos[subidx]; | |
2585 save.endpos[subidx] = m->endpos[subidx]; | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2586 if (off == -1) |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2587 { |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2588 m->startpos[subidx].lnum = reglnum + 1; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2589 m->startpos[subidx].col = 0; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2590 } |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2591 else |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2592 { |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2593 m->startpos[subidx].lnum = reglnum; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2594 m->startpos[subidx].col = |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2595 (colnr_T)(reginput - regline + off); |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2596 } |
4444 | 2597 } |
2598 else | |
2599 { | |
2600 save.start[subidx] = m->start[subidx]; | |
2601 save.end[subidx] = m->end[subidx]; | |
2602 m->start[subidx] = reginput + off; | |
2603 } | |
2604 | |
2605 addstate(l, state->out, m, off, lid, match); | |
2606 | |
2607 if (REG_MULTI) | |
2608 { | |
2609 m->startpos[subidx] = save.startpos[subidx]; | |
2610 m->endpos[subidx] = save.endpos[subidx]; | |
2611 } | |
2612 else | |
2613 { | |
2614 m->start[subidx] = save.start[subidx]; | |
2615 m->end[subidx] = save.end[subidx]; | |
2616 } | |
2617 break; | |
2618 | |
2619 case NFA_MCLOSE + 0: | |
2620 if (nfa_has_zend == TRUE) | |
2621 { | |
2622 addstate(l, state->out, m, off, lid, match); | |
2623 break; | |
2624 } | |
2625 case NFA_MCLOSE + 1: | |
2626 case NFA_MCLOSE + 2: | |
2627 case NFA_MCLOSE + 3: | |
2628 case NFA_MCLOSE + 4: | |
2629 case NFA_MCLOSE + 5: | |
2630 case NFA_MCLOSE + 6: | |
2631 case NFA_MCLOSE + 7: | |
2632 case NFA_MCLOSE + 8: | |
2633 case NFA_MCLOSE + 9: | |
2634 case NFA_ZEND: | |
2635 subidx = state->c - NFA_MCLOSE; | |
2636 if (state->c == NFA_ZEND) | |
2637 subidx = 0; | |
2638 | |
2639 if (REG_MULTI) | |
2640 { | |
2641 save.startpos[subidx] = m->startpos[subidx]; | |
2642 save.endpos[subidx] = m->endpos[subidx]; | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2643 if (off == -1) |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2644 { |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2645 m->endpos[subidx].lnum = reglnum + 1; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2646 m->endpos[subidx].col = 0; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2647 } |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2648 else |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2649 { |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2650 m->endpos[subidx].lnum = reglnum; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2651 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
|
2652 } |
4444 | 2653 } |
2654 else | |
2655 { | |
2656 save.start[subidx] = m->start[subidx]; | |
2657 save.end[subidx] = m->end[subidx]; | |
2658 m->end[subidx] = reginput + off; | |
2659 } | |
2660 | |
2661 addstate(l, state->out, m, off, lid, match); | |
2662 | |
2663 if (REG_MULTI) | |
2664 { | |
2665 m->startpos[subidx] = save.startpos[subidx]; | |
2666 m->endpos[subidx] = save.endpos[subidx]; | |
2667 } | |
2668 else | |
2669 { | |
2670 m->start[subidx] = save.start[subidx]; | |
2671 m->end[subidx] = save.end[subidx]; | |
2672 } | |
2673 break; | |
2674 } | |
2675 } | |
2676 | |
2677 /* | |
2678 * Check character class "class" against current character c. | |
2679 */ | |
2680 static int | |
2681 check_char_class(class, c) | |
2682 int class; | |
2683 int c; | |
2684 { | |
2685 switch (class) | |
2686 { | |
2687 case NFA_CLASS_ALNUM: | |
4464 | 2688 if (c >= 1 && c <= 255 && isalnum(c)) |
4444 | 2689 return OK; |
2690 break; | |
2691 case NFA_CLASS_ALPHA: | |
4464 | 2692 if (c >= 1 && c <= 255 && isalpha(c)) |
4444 | 2693 return OK; |
2694 break; | |
2695 case NFA_CLASS_BLANK: | |
2696 if (c == ' ' || c == '\t') | |
2697 return OK; | |
2698 break; | |
2699 case NFA_CLASS_CNTRL: | |
4464 | 2700 if (c >= 1 && c <= 255 && iscntrl(c)) |
4444 | 2701 return OK; |
2702 break; | |
2703 case NFA_CLASS_DIGIT: | |
2704 if (VIM_ISDIGIT(c)) | |
2705 return OK; | |
2706 break; | |
2707 case NFA_CLASS_GRAPH: | |
4464 | 2708 if (c >= 1 && c <= 255 && isgraph(c)) |
4444 | 2709 return OK; |
2710 break; | |
2711 case NFA_CLASS_LOWER: | |
2712 if (MB_ISLOWER(c)) | |
2713 return OK; | |
2714 break; | |
2715 case NFA_CLASS_PRINT: | |
2716 if (vim_isprintc(c)) | |
2717 return OK; | |
2718 break; | |
2719 case NFA_CLASS_PUNCT: | |
4464 | 2720 if (c >= 1 && c <= 255 && ispunct(c)) |
4444 | 2721 return OK; |
2722 break; | |
2723 case NFA_CLASS_SPACE: | |
2724 if ((c >=9 && c <= 13) || (c == ' ')) | |
2725 return OK; | |
2726 break; | |
2727 case NFA_CLASS_UPPER: | |
2728 if (MB_ISUPPER(c)) | |
2729 return OK; | |
2730 break; | |
2731 case NFA_CLASS_XDIGIT: | |
2732 if (vim_isxdigit(c)) | |
2733 return OK; | |
2734 break; | |
2735 case NFA_CLASS_TAB: | |
2736 if (c == '\t') | |
2737 return OK; | |
2738 break; | |
2739 case NFA_CLASS_RETURN: | |
2740 if (c == '\r') | |
2741 return OK; | |
2742 break; | |
2743 case NFA_CLASS_BACKSPACE: | |
2744 if (c == '\b') | |
2745 return OK; | |
2746 break; | |
2747 case NFA_CLASS_ESCAPE: | |
2748 if (c == '\033') | |
2749 return OK; | |
2750 break; | |
2751 | |
2752 default: | |
2753 /* should not be here :P */ | |
2754 EMSG_RET_FAIL(_("E877: (NFA regexp) Invalid character class ")); | |
2755 } | |
2756 return FAIL; | |
2757 } | |
2758 | |
2759 /* | |
2760 * Set all NFA nodes' list ID equal to -1. | |
2761 */ | |
2762 static void | |
2763 nfa_set_neg_listids(start) | |
2764 nfa_state_T *start; | |
2765 { | |
2766 if (start == NULL) | |
2767 return; | |
2768 if (start->lastlist >= 0) | |
2769 { | |
2770 start->lastlist = -1; | |
2771 nfa_set_neg_listids(start->out); | |
2772 nfa_set_neg_listids(start->out1); | |
2773 } | |
2774 } | |
2775 | |
2776 /* | |
2777 * Set all NFA nodes' list ID equal to 0. | |
2778 */ | |
2779 static void | |
2780 nfa_set_null_listids(start) | |
2781 nfa_state_T *start; | |
2782 { | |
2783 if (start == NULL) | |
2784 return; | |
2785 if (start->lastlist == -1) | |
2786 { | |
2787 start->lastlist = 0; | |
2788 nfa_set_null_listids(start->out); | |
2789 nfa_set_null_listids(start->out1); | |
2790 } | |
2791 } | |
2792 | |
2793 /* | |
2794 * Save list IDs for all NFA states in "list". | |
2795 */ | |
2796 static void | |
2797 nfa_save_listids(start, list) | |
2798 nfa_state_T *start; | |
2799 int *list; | |
2800 { | |
2801 if (start == NULL) | |
2802 return; | |
2803 if (start->lastlist != -1) | |
2804 { | |
2805 list[abs(start->id)] = start->lastlist; | |
2806 start->lastlist = -1; | |
2807 nfa_save_listids(start->out, list); | |
2808 nfa_save_listids(start->out1, list); | |
2809 } | |
2810 } | |
2811 | |
2812 /* | |
2813 * Restore list IDs from "list" to all NFA states. | |
2814 */ | |
2815 static void | |
2816 nfa_restore_listids(start, list) | |
2817 nfa_state_T *start; | |
2818 int *list; | |
2819 { | |
2820 if (start == NULL) | |
2821 return; | |
2822 if (start->lastlist == -1) | |
2823 { | |
2824 start->lastlist = list[abs(start->id)]; | |
2825 nfa_restore_listids(start->out, list); | |
2826 nfa_restore_listids(start->out1, list); | |
2827 } | |
2828 } | |
2829 | |
2830 /* | |
2831 * Main matching routine. | |
2832 * | |
2833 * Run NFA to determine whether it matches reginput. | |
2834 * | |
2835 * Return TRUE if there is a match, FALSE otherwise. | |
2836 * Note: Caller must ensure that: start != NULL. | |
2837 */ | |
2838 static int | |
2839 nfa_regmatch(start, submatch, m) | |
2840 nfa_state_T *start; | |
2841 regsub_T *submatch; | |
2842 regsub_T *m; | |
2843 { | |
2844 int c = -1; | |
2845 int n; | |
2846 int i = 0; | |
2847 int result; | |
2848 int size = 0; | |
2849 int match = FALSE; | |
2850 int flag = 0; | |
2851 int old_reglnum = -1; | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2852 int go_to_nextline; |
4444 | 2853 thread_T *t; |
2854 char_u *old_reginput = NULL; | |
2855 char_u *old_regline = NULL; | |
2856 nfa_state_T *sta; | |
2857 nfa_state_T *end; | |
2858 List list[3]; | |
2859 List *listtbl[2][2]; | |
2860 List *ll; | |
2861 int listid = 1; | |
4525
36ddcf4cecbc
updated for version 7.3.1010
Bram Moolenaar <bram@vim.org>
parents:
4517
diff
changeset
|
2862 int endnode; |
4444 | 2863 List *thislist; |
2864 List *nextlist; | |
2865 List *neglist; | |
2866 int *listids = NULL; | |
2867 int j = 0; | |
2868 int len = 0; | |
4460 | 2869 #ifdef NFA_REGEXP_DEBUG_LOG |
2870 FILE *debug = fopen(NFA_REGEXP_DEBUG_LOG_NAME, "a"); | |
4444 | 2871 |
2872 if (debug == NULL) | |
2873 { | |
4460 | 2874 EMSG2(_("(NFA) COULD NOT OPEN %s !"), NFA_REGEXP_DEBUG_LOG_NAME); |
4444 | 2875 return FALSE; |
2876 } | |
2877 #endif | |
2878 | |
2879 /* Allocate memory for the lists of nodes */ | |
2880 size = (nstate + 1) * sizeof(thread_T); | |
2881 list[0].t = (thread_T *)lalloc(size, TRUE); | |
2882 list[1].t = (thread_T *)lalloc(size, TRUE); | |
2883 list[2].t = (thread_T *)lalloc(size, TRUE); | |
2884 if (list[0].t == NULL || list[1].t == NULL || list[2].t == NULL) | |
2885 goto theend; | |
2886 vim_memset(list[0].t, 0, size); | |
2887 vim_memset(list[1].t, 0, size); | |
2888 vim_memset(list[2].t, 0, size); | |
2889 | |
2890 #ifdef ENABLE_LOG | |
2891 log_fd = fopen(LOG_NAME, "a"); | |
2892 if (log_fd != NULL) | |
2893 { | |
2894 fprintf(log_fd, "**********************************\n"); | |
2895 nfa_set_code(start->c); | |
2896 fprintf(log_fd, " RUNNING nfa_regmatch() starting with state %d, code %s\n", | |
2897 abs(start->id), code); | |
2898 fprintf(log_fd, "**********************************\n"); | |
2899 } | |
2900 else | |
2901 { | |
2902 EMSG(_("Could not open temporary log file for writing, displaying on stderr ... ")); | |
2903 log_fd = stderr; | |
2904 } | |
2905 #endif | |
2906 | |
2907 thislist = &list[0]; | |
2908 thislist->n = 0; | |
2909 nextlist = &list[1]; | |
2910 nextlist->n = 0; | |
2911 neglist = &list[2]; | |
2912 neglist->n = 0; | |
2913 #ifdef ENABLE_LOG | |
2914 fprintf(log_fd, "(---) STARTSTATE\n"); | |
2915 #endif | |
2916 addstate(thislist, start, m, 0, listid, &match); | |
2917 | |
2918 /* There are two cases when the NFA advances: 1. input char matches the | |
2919 * NFA node and 2. input char does not match the NFA node, but the next | |
2920 * node is NFA_NOT. The following macro calls addstate() according to | |
2921 * these rules. It is used A LOT, so use the "listtbl" table for speed */ | |
2922 listtbl[0][0] = NULL; | |
2923 listtbl[0][1] = neglist; | |
2924 listtbl[1][0] = nextlist; | |
2925 listtbl[1][1] = NULL; | |
2926 #define ADD_POS_NEG_STATE(node) \ | |
2927 ll = listtbl[result ? 1 : 0][node->negated]; \ | |
2928 if (ll != NULL) \ | |
2929 addstate(ll, node->out , &t->sub, n, listid + 1, &match); | |
2930 | |
2931 | |
2932 /* | |
2933 * Run for each character. | |
2934 */ | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2935 for (;;) |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2936 { |
4444 | 2937 #ifdef FEAT_MBYTE |
2938 if (has_mbyte) | |
2939 { | |
2940 c = (*mb_ptr2char)(reginput); | |
2941 n = (*mb_ptr2len)(reginput); | |
2942 } | |
2943 else | |
2944 #endif | |
2945 { | |
2946 c = *reginput; | |
2947 n = 1; | |
2948 } | |
2949 if (c == NUL) | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2950 { |
4444 | 2951 n = 0; |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2952 go_to_nextline = FALSE; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2953 } |
4444 | 2954 |
2955 /* swap lists */ | |
2956 thislist = &list[flag]; | |
2957 nextlist = &list[flag ^= 1]; | |
2958 nextlist->n = 0; /* `clear' nextlist */ | |
2959 listtbl[1][0] = nextlist; | |
2960 ++listid; | |
2961 | |
2962 #ifdef ENABLE_LOG | |
2963 fprintf(log_fd, "------------------------------------------\n"); | |
2964 fprintf(log_fd, ">>> Reginput is \"%s\"\n", reginput); | |
2965 fprintf(log_fd, ">>> Advanced one character ... Current char is %c (code %d) \n", c, (int)c); | |
2966 fprintf(log_fd, ">>> Thislist has %d states available: ", thislist->n); | |
2967 for (i = 0; i< thislist->n; i++) | |
2968 fprintf(log_fd, "%d ", abs(thislist->t[i].state->id)); | |
2969 fprintf(log_fd, "\n"); | |
2970 #endif | |
2971 | |
4460 | 2972 #ifdef NFA_REGEXP_DEBUG_LOG |
4444 | 2973 fprintf(debug, "\n-------------------\n"); |
2974 #endif | |
4480 | 2975 /* |
2976 * If the state lists are empty we can stop. | |
2977 */ | |
2978 if (thislist->n == 0 && neglist->n == 0) | |
2979 break; | |
4444 | 2980 |
2981 /* compute nextlist */ | |
2982 for (i = 0; i < thislist->n || neglist->n > 0; ++i) | |
2983 { | |
2984 if (neglist->n > 0) | |
2985 { | |
2986 t = &neglist->t[0]; | |
4468 | 2987 neglist->n--; |
4444 | 2988 i--; |
2989 } | |
2990 else | |
2991 t = &thislist->t[i]; | |
2992 | |
4460 | 2993 #ifdef NFA_REGEXP_DEBUG_LOG |
4444 | 2994 nfa_set_code(t->state->c); |
2995 fprintf(debug, "%s, ", code); | |
2996 #endif | |
2997 #ifdef ENABLE_LOG | |
2998 nfa_set_code(t->state->c); | |
2999 fprintf(log_fd, "(%d) %s, code %d ... \n", abs(t->state->id), | |
3000 code, (int)t->state->c); | |
3001 #endif | |
3002 | |
3003 /* | |
3004 * Handle the possible codes of the current state. | |
3005 * The most important is NFA_MATCH. | |
3006 */ | |
3007 switch (t->state->c) | |
3008 { | |
3009 case NFA_MATCH: | |
3010 match = TRUE; | |
3011 *submatch = t->sub; | |
3012 #ifdef ENABLE_LOG | |
3013 for (j = 0; j < 4; j++) | |
3014 if (REG_MULTI) | |
3015 fprintf(log_fd, "\n *** group %d, start: c=%d, l=%d, end: c=%d, l=%d", | |
3016 j, | |
3017 t->sub.startpos[j].col, | |
3018 (int)t->sub.startpos[j].lnum, | |
3019 t->sub.endpos[j].col, | |
3020 (int)t->sub.endpos[j].lnum); | |
3021 else | |
3022 fprintf(log_fd, "\n *** group %d, start: \"%s\", end: \"%s\"", | |
3023 j, | |
3024 (char *)t->sub.start[j], | |
3025 (char *)t->sub.end[j]); | |
3026 fprintf(log_fd, "\n"); | |
3027 #endif | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3028 /* 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
|
3029 * states at this position. */ |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3030 goto nextchar; |
4444 | 3031 |
3032 case NFA_END_INVISIBLE: | |
3033 /* This is only encountered after a NFA_START_INVISIBLE node. | |
3034 * They surround a zero-width group, used with "\@=" and "\&". | |
3035 * If we got here, it means that the current "invisible" group | |
3036 * finished successfully, so return control to the parent | |
3037 * nfa_regmatch(). Submatches are stored in *m, and used in | |
3038 * the parent call. */ | |
3039 if (start->c == NFA_MOPEN + 0) | |
3040 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
3041 &match); | |
3042 else | |
3043 { | |
3044 *m = t->sub; | |
3045 match = TRUE; | |
3046 } | |
3047 break; | |
3048 | |
3049 case NFA_START_INVISIBLE: | |
3050 /* Save global variables, and call nfa_regmatch() to check if | |
3051 * the current concat matches at this position. The concat | |
3052 * ends with the node NFA_END_INVISIBLE */ | |
3053 old_reginput = reginput; | |
3054 old_regline = regline; | |
3055 old_reglnum = reglnum; | |
3056 if (listids == NULL) | |
3057 { | |
3058 listids = (int *) lalloc(sizeof(int) * nstate, TRUE); | |
3059 if (listids == NULL) | |
3060 { | |
3061 EMSG(_("E878: (NFA) Could not allocate memory for branch traversal!")); | |
3062 return 0; | |
3063 } | |
3064 } | |
3065 #ifdef ENABLE_LOG | |
3066 if (log_fd != stderr) | |
3067 fclose(log_fd); | |
3068 log_fd = NULL; | |
3069 #endif | |
3070 /* Have to clear the listid field of the NFA nodes, so that | |
3071 * nfa_regmatch() and addstate() can run properly after | |
3072 * recursion. */ | |
3073 nfa_save_listids(start, listids); | |
3074 nfa_set_null_listids(start); | |
3075 result = nfa_regmatch(t->state->out, submatch, m); | |
3076 nfa_set_neg_listids(start); | |
3077 nfa_restore_listids(start, listids); | |
3078 | |
3079 #ifdef ENABLE_LOG | |
3080 log_fd = fopen(LOG_NAME, "a"); | |
3081 if (log_fd != NULL) | |
3082 { | |
3083 fprintf(log_fd, "****************************\n"); | |
3084 fprintf(log_fd, "FINISHED RUNNING nfa_regmatch() recursively\n"); | |
3085 fprintf(log_fd, "MATCH = %s\n", result == TRUE ? "OK" : "FALSE"); | |
3086 fprintf(log_fd, "****************************\n"); | |
3087 } | |
3088 else | |
3089 { | |
3090 EMSG(_("Could not open temporary log file for writing, displaying on stderr ... ")); | |
3091 log_fd = stderr; | |
3092 } | |
3093 #endif | |
3094 if (result == TRUE) | |
3095 { | |
3096 /* Restore position in input text */ | |
3097 reginput = old_reginput; | |
3098 regline = old_regline; | |
3099 reglnum = old_reglnum; | |
3100 /* Copy submatch info from the recursive call */ | |
3101 if (REG_MULTI) | |
3102 for (j = 1; j < NSUBEXP; j++) | |
3103 { | |
3104 t->sub.startpos[j] = m->startpos[j]; | |
3105 t->sub.endpos[j] = m->endpos[j]; | |
3106 } | |
3107 else | |
3108 for (j = 1; j < NSUBEXP; j++) | |
3109 { | |
3110 t->sub.start[j] = m->start[j]; | |
3111 t->sub.end[j] = m->end[j]; | |
3112 } | |
3113 /* t->state->out1 is the corresponding END_INVISIBLE node */ | |
3114 addstate(thislist, t->state->out1->out, &t->sub, 0, listid, | |
3115 &match); | |
3116 } | |
3117 else | |
3118 { | |
3119 /* continue with next input char */ | |
3120 reginput = old_reginput; | |
3121 } | |
3122 break; | |
3123 | |
3124 case NFA_BOL: | |
3125 if (reginput == regline) | |
3126 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
3127 &match); | |
3128 break; | |
3129 | |
3130 case NFA_EOL: | |
3131 if (c == NUL) | |
3132 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
3133 &match); | |
3134 break; | |
3135 | |
3136 case NFA_BOW: | |
3137 { | |
3138 int bow = TRUE; | |
3139 | |
3140 if (c == NUL) | |
3141 bow = FALSE; | |
3142 #ifdef FEAT_MBYTE | |
3143 else if (has_mbyte) | |
3144 { | |
3145 int this_class; | |
3146 | |
3147 /* Get class of current and previous char (if it exists). */ | |
4503 | 3148 this_class = mb_get_class_buf(reginput, reg_buf); |
4444 | 3149 if (this_class <= 1) |
3150 bow = FALSE; | |
3151 else if (reg_prev_class() == this_class) | |
3152 bow = FALSE; | |
3153 } | |
3154 #endif | |
4503 | 3155 else if (!vim_iswordc_buf(c, reg_buf) |
3156 || (reginput > regline | |
3157 && vim_iswordc_buf(reginput[-1], reg_buf))) | |
4444 | 3158 bow = FALSE; |
3159 if (bow) | |
3160 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
3161 &match); | |
3162 break; | |
3163 } | |
3164 | |
3165 case NFA_EOW: | |
3166 { | |
3167 int eow = TRUE; | |
3168 | |
3169 if (reginput == regline) | |
3170 eow = FALSE; | |
3171 #ifdef FEAT_MBYTE | |
3172 else if (has_mbyte) | |
3173 { | |
3174 int this_class, prev_class; | |
3175 | |
3176 /* Get class of current and previous char (if it exists). */ | |
4503 | 3177 this_class = mb_get_class_buf(reginput, reg_buf); |
4444 | 3178 prev_class = reg_prev_class(); |
3179 if (this_class == prev_class | |
3180 || prev_class == 0 || prev_class == 1) | |
3181 eow = FALSE; | |
3182 } | |
3183 #endif | |
4503 | 3184 else if (!vim_iswordc_buf(reginput[-1], reg_buf) |
3185 || (reginput[0] != NUL && vim_iswordc_buf(c, reg_buf))) | |
4444 | 3186 eow = FALSE; |
3187 if (eow) | |
3188 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
3189 &match); | |
3190 break; | |
3191 } | |
3192 | |
3193 case NFA_MULTIBYTE: | |
3194 case NFA_COMPOSING: | |
4525
36ddcf4cecbc
updated for version 7.3.1010
Bram Moolenaar <bram@vim.org>
parents:
4517
diff
changeset
|
3195 endnode = t->state->c + 1; |
4444 | 3196 result = OK; |
3197 sta = t->state->out; | |
3198 len = 1; | |
3199 while (sta->c != endnode && len <= n) | |
3200 { | |
3201 if (reginput[len-1] != sta->c) | |
3202 { | |
4525
36ddcf4cecbc
updated for version 7.3.1010
Bram Moolenaar <bram@vim.org>
parents:
4517
diff
changeset
|
3203 result = FAIL; |
4444 | 3204 break; |
3205 } | |
3206 len++; | |
3207 sta = sta->out; | |
3208 } | |
3209 | |
3210 /* if input char length doesn't match regexp char length */ | |
3211 if (len -1 < n || sta->c != endnode) | |
4525
36ddcf4cecbc
updated for version 7.3.1010
Bram Moolenaar <bram@vim.org>
parents:
4517
diff
changeset
|
3212 result = FAIL; |
4444 | 3213 end = t->state->out1; /* NFA_END_MULTIBYTE or |
3214 NFA_END_COMPOSING */ | |
3215 /* If \Z was present, then ignore composing characters */ | |
4525
36ddcf4cecbc
updated for version 7.3.1010
Bram Moolenaar <bram@vim.org>
parents:
4517
diff
changeset
|
3216 if (ireg_icombine && endnode == NFA_END_COMPOSING) |
4444 | 3217 result = 1 ^ sta->negated; |
3218 ADD_POS_NEG_STATE(end); | |
3219 break; | |
3220 | |
3221 case NFA_NEWL: | |
3222 if (!reg_line_lbr && REG_MULTI | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3223 && c == NUL && reglnum <= reg_maxline) |
4444 | 3224 { |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3225 go_to_nextline = TRUE; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3226 /* 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
|
3227 * at the start of the next line. */ |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3228 addstate(nextlist, t->state->out, &t->sub, -1, |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3229 listid + 1, &match); |
4444 | 3230 } |
3231 break; | |
3232 | |
3233 case NFA_CLASS_ALNUM: | |
3234 case NFA_CLASS_ALPHA: | |
3235 case NFA_CLASS_BLANK: | |
3236 case NFA_CLASS_CNTRL: | |
3237 case NFA_CLASS_DIGIT: | |
3238 case NFA_CLASS_GRAPH: | |
3239 case NFA_CLASS_LOWER: | |
3240 case NFA_CLASS_PRINT: | |
3241 case NFA_CLASS_PUNCT: | |
3242 case NFA_CLASS_SPACE: | |
3243 case NFA_CLASS_UPPER: | |
3244 case NFA_CLASS_XDIGIT: | |
3245 case NFA_CLASS_TAB: | |
3246 case NFA_CLASS_RETURN: | |
3247 case NFA_CLASS_BACKSPACE: | |
3248 case NFA_CLASS_ESCAPE: | |
3249 result = check_char_class(t->state->c, c); | |
3250 ADD_POS_NEG_STATE(t->state); | |
3251 break; | |
3252 | |
3253 case NFA_END_NEG_RANGE: | |
3254 /* This follows a series of negated nodes, like: | |
3255 * CHAR(x), NFA_NOT, CHAR(y), NFA_NOT etc. */ | |
3256 if (c > 0) | |
3257 addstate(nextlist, t->state->out, &t->sub, n, listid + 1, | |
3258 &match); | |
3259 break; | |
3260 | |
3261 case NFA_ANY: | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3262 /* Any char except '\0', (end of input) does not match. */ |
4444 | 3263 if (c > 0) |
3264 addstate(nextlist, t->state->out, &t->sub, n, listid + 1, | |
3265 &match); | |
3266 break; | |
3267 | |
3268 /* | |
3269 * Character classes like \a for alpha, \d for digit etc. | |
3270 */ | |
3271 case NFA_IDENT: /* \i */ | |
3272 result = vim_isIDc(c); | |
3273 ADD_POS_NEG_STATE(t->state); | |
3274 break; | |
3275 | |
3276 case NFA_SIDENT: /* \I */ | |
3277 result = !VIM_ISDIGIT(c) && vim_isIDc(c); | |
3278 ADD_POS_NEG_STATE(t->state); | |
3279 break; | |
3280 | |
3281 case NFA_KWORD: /* \k */ | |
4503 | 3282 result = vim_iswordp_buf(reginput, reg_buf); |
4444 | 3283 ADD_POS_NEG_STATE(t->state); |
3284 break; | |
3285 | |
3286 case NFA_SKWORD: /* \K */ | |
4503 | 3287 result = !VIM_ISDIGIT(c) && vim_iswordp_buf(reginput, reg_buf); |
4444 | 3288 ADD_POS_NEG_STATE(t->state); |
3289 break; | |
3290 | |
3291 case NFA_FNAME: /* \f */ | |
3292 result = vim_isfilec(c); | |
3293 ADD_POS_NEG_STATE(t->state); | |
3294 break; | |
3295 | |
3296 case NFA_SFNAME: /* \F */ | |
3297 result = !VIM_ISDIGIT(c) && vim_isfilec(c); | |
3298 ADD_POS_NEG_STATE(t->state); | |
3299 break; | |
3300 | |
3301 case NFA_PRINT: /* \p */ | |
4470 | 3302 result = ptr2cells(reginput) == 1; |
4444 | 3303 ADD_POS_NEG_STATE(t->state); |
3304 break; | |
3305 | |
3306 case NFA_SPRINT: /* \P */ | |
4470 | 3307 result = !VIM_ISDIGIT(c) && ptr2cells(reginput) == 1; |
4444 | 3308 ADD_POS_NEG_STATE(t->state); |
3309 break; | |
3310 | |
3311 case NFA_WHITE: /* \s */ | |
3312 result = vim_iswhite(c); | |
3313 ADD_POS_NEG_STATE(t->state); | |
3314 break; | |
3315 | |
3316 case NFA_NWHITE: /* \S */ | |
3317 result = c != NUL && !vim_iswhite(c); | |
3318 ADD_POS_NEG_STATE(t->state); | |
3319 break; | |
3320 | |
3321 case NFA_DIGIT: /* \d */ | |
3322 result = ri_digit(c); | |
3323 ADD_POS_NEG_STATE(t->state); | |
3324 break; | |
3325 | |
3326 case NFA_NDIGIT: /* \D */ | |
3327 result = c != NUL && !ri_digit(c); | |
3328 ADD_POS_NEG_STATE(t->state); | |
3329 break; | |
3330 | |
3331 case NFA_HEX: /* \x */ | |
3332 result = ri_hex(c); | |
3333 ADD_POS_NEG_STATE(t->state); | |
3334 break; | |
3335 | |
3336 case NFA_NHEX: /* \X */ | |
3337 result = c != NUL && !ri_hex(c); | |
3338 ADD_POS_NEG_STATE(t->state); | |
3339 break; | |
3340 | |
3341 case NFA_OCTAL: /* \o */ | |
3342 result = ri_octal(c); | |
3343 ADD_POS_NEG_STATE(t->state); | |
3344 break; | |
3345 | |
3346 case NFA_NOCTAL: /* \O */ | |
3347 result = c != NUL && !ri_octal(c); | |
3348 ADD_POS_NEG_STATE(t->state); | |
3349 break; | |
3350 | |
3351 case NFA_WORD: /* \w */ | |
3352 result = ri_word(c); | |
3353 ADD_POS_NEG_STATE(t->state); | |
3354 break; | |
3355 | |
3356 case NFA_NWORD: /* \W */ | |
3357 result = c != NUL && !ri_word(c); | |
3358 ADD_POS_NEG_STATE(t->state); | |
3359 break; | |
3360 | |
3361 case NFA_HEAD: /* \h */ | |
3362 result = ri_head(c); | |
3363 ADD_POS_NEG_STATE(t->state); | |
3364 break; | |
3365 | |
3366 case NFA_NHEAD: /* \H */ | |
3367 result = c != NUL && !ri_head(c); | |
3368 ADD_POS_NEG_STATE(t->state); | |
3369 break; | |
3370 | |
3371 case NFA_ALPHA: /* \a */ | |
3372 result = ri_alpha(c); | |
3373 ADD_POS_NEG_STATE(t->state); | |
3374 break; | |
3375 | |
3376 case NFA_NALPHA: /* \A */ | |
3377 result = c != NUL && !ri_alpha(c); | |
3378 ADD_POS_NEG_STATE(t->state); | |
3379 break; | |
3380 | |
3381 case NFA_LOWER: /* \l */ | |
3382 result = ri_lower(c); | |
3383 ADD_POS_NEG_STATE(t->state); | |
3384 break; | |
3385 | |
3386 case NFA_NLOWER: /* \L */ | |
3387 result = c != NUL && !ri_lower(c); | |
3388 ADD_POS_NEG_STATE(t->state); | |
3389 break; | |
3390 | |
3391 case NFA_UPPER: /* \u */ | |
3392 result = ri_upper(c); | |
3393 ADD_POS_NEG_STATE(t->state); | |
3394 break; | |
3395 | |
3396 case NFA_NUPPER: /* \U */ | |
3397 result = c != NUL && !ri_upper(c); | |
3398 ADD_POS_NEG_STATE(t->state); | |
3399 break; | |
3400 | |
4482 | 3401 case NFA_MOPEN + 0: |
3402 case NFA_MOPEN + 1: | |
3403 case NFA_MOPEN + 2: | |
3404 case NFA_MOPEN + 3: | |
3405 case NFA_MOPEN + 4: | |
3406 case NFA_MOPEN + 5: | |
3407 case NFA_MOPEN + 6: | |
3408 case NFA_MOPEN + 7: | |
3409 case NFA_MOPEN + 8: | |
3410 case NFA_MOPEN + 9: | |
3411 /* handled below */ | |
3412 break; | |
3413 | |
3414 case NFA_SKIP_CHAR: | |
3415 case NFA_ZSTART: | |
3416 /* TODO: should not happen? */ | |
3417 break; | |
3418 | |
4444 | 3419 default: /* regular character */ |
4482 | 3420 /* TODO: put this in #ifdef later */ |
3421 if (t->state->c < -256) | |
3422 EMSGN("INTERNAL: Negative state char: %ld", t->state->c); | |
4444 | 3423 result = (no_Magic(t->state->c) == c); |
4482 | 3424 |
4444 | 3425 if (!result) |
3426 result = ireg_ic == TRUE | |
3427 && MB_TOLOWER(t->state->c) == MB_TOLOWER(c); | |
3428 ADD_POS_NEG_STATE(t->state); | |
3429 break; | |
3430 } | |
3431 | |
3432 } /* for (thislist = thislist; thislist->state; thislist++) */ | |
3433 | |
3434 /* The first found match is the leftmost one, but there may be a | |
3435 * longer one. Keep running the NFA, but don't start from the | |
3436 * beginning. Also, do not add the start state in recursive calls of | |
3437 * nfa_regmatch(), because recursive calls should only start in the | |
3438 * first position. */ | |
3439 if (match == FALSE && start->c == NFA_MOPEN + 0) | |
3440 { | |
3441 #ifdef ENABLE_LOG | |
3442 fprintf(log_fd, "(---) STARTSTATE\n"); | |
3443 #endif | |
3444 addstate(nextlist, start, m, n, listid + 1, &match); | |
3445 } | |
3446 | |
3447 #ifdef ENABLE_LOG | |
3448 fprintf(log_fd, ">>> Thislist had %d states available: ", thislist->n); | |
3449 for (i = 0; i< thislist->n; i++) | |
3450 fprintf(log_fd, "%d ", abs(thislist->t[i].state->id)); | |
3451 fprintf(log_fd, "\n"); | |
3452 #endif | |
3453 | |
3454 nextchar: | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3455 /* 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
|
3456 * finish. */ |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3457 if (n != 0) |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3458 reginput += n; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3459 else if (go_to_nextline) |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3460 reg_nextline(); |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3461 else |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3462 break; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3463 } |
4444 | 3464 |
3465 #ifdef ENABLE_LOG | |
3466 if (log_fd != stderr) | |
3467 fclose(log_fd); | |
3468 log_fd = NULL; | |
3469 #endif | |
3470 | |
3471 theend: | |
3472 /* Free memory */ | |
3473 vim_free(list[0].t); | |
3474 vim_free(list[1].t); | |
3475 vim_free(list[2].t); | |
3476 list[0].t = list[1].t = list[2].t = NULL; | |
3477 if (listids != NULL) | |
3478 vim_free(listids); | |
3479 #undef ADD_POS_NEG_STATE | |
4460 | 3480 #ifdef NFA_REGEXP_DEBUG_LOG |
4444 | 3481 fclose(debug); |
3482 #endif | |
3483 | |
3484 return match; | |
3485 } | |
3486 | |
3487 /* | |
3488 * Try match of "prog" with at regline["col"]. | |
3489 * Returns 0 for failure, number of lines contained in the match otherwise. | |
3490 */ | |
3491 static long | |
3492 nfa_regtry(start, col) | |
3493 nfa_state_T *start; | |
3494 colnr_T col; | |
3495 { | |
3496 int i; | |
3497 regsub_T sub, m; | |
3498 #ifdef ENABLE_LOG | |
3499 FILE *f; | |
3500 #endif | |
3501 | |
3502 reginput = regline + col; | |
3503 need_clear_subexpr = TRUE; | |
3504 | |
3505 #ifdef ENABLE_LOG | |
3506 f = fopen(LOG_NAME, "a"); | |
3507 if (f != NULL) | |
3508 { | |
3509 fprintf(f, "\n\n\n\n\n\n\t\t=======================================================\n"); | |
3510 fprintf(f, " =======================================================\n"); | |
3511 #ifdef DEBUG | |
3512 fprintf(f, "\tRegexp is \"%s\"\n", nfa_regengine.expr); | |
3513 #endif | |
3514 fprintf(f, "\tInput text is \"%s\" \n", reginput); | |
3515 fprintf(f, " =======================================================\n\n\n\n\n\n\n"); | |
3516 nfa_print_state(f, start, 0); | |
3517 fprintf(f, "\n\n"); | |
3518 fclose(f); | |
3519 } | |
3520 else | |
3521 EMSG(_("Could not open temporary log file for writing ")); | |
3522 #endif | |
3523 | |
3524 if (REG_MULTI) | |
3525 { | |
3526 /* Use 0xff to set lnum to -1 */ | |
3527 vim_memset(sub.startpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
3528 vim_memset(sub.endpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
3529 vim_memset(m.startpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
3530 vim_memset(m.endpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
3531 } | |
3532 else | |
3533 { | |
3534 vim_memset(sub.start, 0, sizeof(char_u *) * NSUBEXP); | |
3535 vim_memset(sub.end, 0, sizeof(char_u *) * NSUBEXP); | |
3536 vim_memset(m.start, 0, sizeof(char_u *) * NSUBEXP); | |
3537 vim_memset(m.end, 0, sizeof(char_u *) * NSUBEXP); | |
3538 } | |
3539 | |
3540 if (nfa_regmatch(start, &sub, &m) == FALSE) | |
3541 return 0; | |
3542 | |
3543 cleanup_subexpr(); | |
3544 if (REG_MULTI) | |
3545 { | |
3546 for (i = 0; i < NSUBEXP; i++) | |
3547 { | |
3548 reg_startpos[i] = sub.startpos[i]; | |
3549 reg_endpos[i] = sub.endpos[i]; | |
3550 } | |
3551 | |
3552 if (reg_startpos[0].lnum < 0) | |
3553 { | |
3554 reg_startpos[0].lnum = 0; | |
3555 reg_startpos[0].col = col; | |
3556 } | |
3557 if (reg_endpos[0].lnum < 0) | |
3558 { | |
3559 reg_endpos[0].lnum = reglnum; | |
3560 reg_endpos[0].col = (int)(reginput - regline); | |
3561 } | |
3562 else | |
3563 /* Use line number of "\ze". */ | |
3564 reglnum = reg_endpos[0].lnum; | |
3565 } | |
3566 else | |
3567 { | |
3568 for (i = 0; i < NSUBEXP; i++) | |
3569 { | |
3570 reg_startp[i] = sub.start[i]; | |
3571 reg_endp[i] = sub.end[i]; | |
3572 } | |
3573 | |
3574 if (reg_startp[0] == NULL) | |
3575 reg_startp[0] = regline + col; | |
3576 if (reg_endp[0] == NULL) | |
3577 reg_endp[0] = reginput; | |
3578 } | |
3579 | |
3580 return 1 + reglnum; | |
3581 } | |
3582 | |
3583 /* | |
3584 * Match a regexp against a string ("line" points to the string) or multiple | |
3585 * lines ("line" is NULL, use reg_getline()). | |
3586 * | |
3587 * Returns 0 for failure, number of lines contained in the match otherwise. | |
3588 */ | |
3589 static long | |
3590 nfa_regexec_both(line, col) | |
3591 char_u *line; | |
3592 colnr_T col; /* column to start looking for match */ | |
3593 { | |
3594 nfa_regprog_T *prog; | |
3595 long retval = 0L; | |
3596 int i; | |
3597 | |
3598 if (REG_MULTI) | |
3599 { | |
3600 prog = (nfa_regprog_T *)reg_mmatch->regprog; | |
3601 line = reg_getline((linenr_T)0); /* relative to the cursor */ | |
3602 reg_startpos = reg_mmatch->startpos; | |
3603 reg_endpos = reg_mmatch->endpos; | |
3604 } | |
3605 else | |
3606 { | |
3607 prog = (nfa_regprog_T *)reg_match->regprog; | |
3608 reg_startp = reg_match->startp; | |
3609 reg_endp = reg_match->endp; | |
3610 } | |
3611 | |
3612 /* Be paranoid... */ | |
3613 if (prog == NULL || line == NULL) | |
3614 { | |
3615 EMSG(_(e_null)); | |
3616 goto theend; | |
3617 } | |
3618 | |
3619 /* If the start column is past the maximum column: no need to try. */ | |
3620 if (ireg_maxcol > 0 && col >= ireg_maxcol) | |
3621 goto theend; | |
3622 | |
3623 /* If pattern contains "\c" or "\C": overrule value of ireg_ic */ | |
3624 if (prog->regflags & RF_ICASE) | |
3625 ireg_ic = TRUE; | |
3626 else if (prog->regflags & RF_NOICASE) | |
3627 ireg_ic = FALSE; | |
3628 | |
3629 #ifdef FEAT_MBYTE | |
3630 /* If pattern contains "\Z" overrule value of ireg_icombine */ | |
3631 if (prog->regflags & RF_ICOMBINE) | |
3632 ireg_icombine = TRUE; | |
3633 #endif | |
3634 | |
3635 regline = line; | |
3636 reglnum = 0; /* relative to line */ | |
3637 | |
3638 nstate = prog->nstate; | |
3639 | |
3640 for (i = 0; i < nstate; ++i) | |
3641 { | |
3642 prog->state[i].id = i; | |
3643 prog->state[i].lastlist = 0; | |
3644 prog->state[i].visits = 0; | |
3645 prog->state[i].lastthread = NULL; | |
3646 } | |
3647 | |
3648 retval = nfa_regtry(prog->start, col); | |
3649 | |
3650 theend: | |
3651 return retval; | |
3652 } | |
3653 | |
3654 /* | |
3655 * Compile a regular expression into internal code for the NFA matcher. | |
3656 * Returns the program in allocated space. Returns NULL for an error. | |
3657 */ | |
3658 static regprog_T * | |
3659 nfa_regcomp(expr, re_flags) | |
3660 char_u *expr; | |
3661 int re_flags; | |
3662 { | |
3663 nfa_regprog_T *prog; | |
4458 | 3664 size_t prog_size; |
4444 | 3665 int *postfix; |
3666 | |
3667 if (expr == NULL) | |
3668 return NULL; | |
3669 | |
3670 #ifdef DEBUG | |
3671 nfa_regengine.expr = expr; | |
3672 #endif | |
3673 | |
3674 init_class_tab(); | |
3675 | |
3676 if (nfa_regcomp_start(expr, re_flags) == FAIL) | |
3677 return NULL; | |
3678 | |
3679 /* Space for compiled regexp */ | |
3680 prog_size = sizeof(nfa_regprog_T) + sizeof(nfa_state_T) * nstate_max; | |
3681 prog = (nfa_regprog_T *)lalloc(prog_size, TRUE); | |
3682 if (prog == NULL) | |
3683 goto fail; | |
3684 vim_memset(prog, 0, prog_size); | |
3685 | |
3686 /* Build postfix form of the regexp. Needed to build the NFA | |
3687 * (and count its size) */ | |
3688 postfix = re2post(); | |
3689 if (postfix == NULL) | |
3690 goto fail; /* Cascaded (syntax?) error */ | |
3691 | |
3692 /* | |
3693 * In order to build the NFA, we parse the input regexp twice: | |
3694 * 1. first pass to count size (so we can allocate space) | |
3695 * 2. second to emit code | |
3696 */ | |
3697 #ifdef ENABLE_LOG | |
3698 { | |
3699 FILE *f = fopen(LOG_NAME, "a"); | |
3700 | |
3701 if (f != NULL) | |
3702 { | |
3703 fprintf(f, "\n*****************************\n\n\n\n\tCompiling regexp \"%s\" ... hold on !\n", expr); | |
3704 fclose(f); | |
3705 } | |
3706 } | |
3707 #endif | |
3708 | |
3709 /* | |
3710 * PASS 1 | |
3711 * Count number of NFA states in "nstate". Do not build the NFA. | |
3712 */ | |
3713 post2nfa(postfix, post_ptr, TRUE); | |
3714 state_ptr = prog->state; | |
3715 | |
3716 /* | |
3717 * PASS 2 | |
3718 * Build the NFA | |
3719 */ | |
3720 prog->start = post2nfa(postfix, post_ptr, FALSE); | |
3721 if (prog->start == NULL) | |
3722 goto fail; | |
3723 | |
3724 prog->regflags = regflags; | |
3725 prog->engine = &nfa_regengine; | |
3726 prog->nstate = nstate; | |
3727 #ifdef ENABLE_LOG | |
3728 nfa_postfix_dump(expr, OK); | |
3729 nfa_dump(prog); | |
3730 #endif | |
3731 | |
3732 out: | |
3733 vim_free(post_start); | |
3734 post_start = post_ptr = post_end = NULL; | |
3735 state_ptr = NULL; | |
3736 return (regprog_T *)prog; | |
3737 | |
3738 fail: | |
3739 vim_free(prog); | |
3740 prog = NULL; | |
3741 #ifdef ENABLE_LOG | |
3742 nfa_postfix_dump(expr, FAIL); | |
3743 #endif | |
3744 #ifdef DEBUG | |
3745 nfa_regengine.expr = NULL; | |
3746 #endif | |
3747 goto out; | |
3748 } | |
3749 | |
3750 | |
3751 /* | |
3752 * Match a regexp against a string. | |
3753 * "rmp->regprog" is a compiled regexp as returned by nfa_regcomp(). | |
3754 * Uses curbuf for line count and 'iskeyword'. | |
3755 * | |
3756 * Return TRUE if there is a match, FALSE if not. | |
3757 */ | |
3758 static int | |
3759 nfa_regexec(rmp, line, col) | |
3760 regmatch_T *rmp; | |
3761 char_u *line; /* string to match against */ | |
3762 colnr_T col; /* column to start looking for match */ | |
3763 { | |
3764 reg_match = rmp; | |
3765 reg_mmatch = NULL; | |
3766 reg_maxline = 0; | |
3767 reg_line_lbr = FALSE; | |
3768 reg_buf = curbuf; | |
3769 reg_win = NULL; | |
3770 ireg_ic = rmp->rm_ic; | |
3771 #ifdef FEAT_MBYTE | |
3772 ireg_icombine = FALSE; | |
3773 #endif | |
3774 ireg_maxcol = 0; | |
3775 return (nfa_regexec_both(line, col) != 0); | |
3776 } | |
3777 | |
3778 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ | |
3779 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
3780 | |
3781 static int nfa_regexec_nl __ARGS((regmatch_T *rmp, char_u *line, colnr_T col)); | |
3782 | |
3783 /* | |
3784 * Like nfa_regexec(), but consider a "\n" in "line" to be a line break. | |
3785 */ | |
3786 static int | |
3787 nfa_regexec_nl(rmp, line, col) | |
3788 regmatch_T *rmp; | |
3789 char_u *line; /* string to match against */ | |
3790 colnr_T col; /* column to start looking for match */ | |
3791 { | |
3792 reg_match = rmp; | |
3793 reg_mmatch = NULL; | |
3794 reg_maxline = 0; | |
3795 reg_line_lbr = TRUE; | |
3796 reg_buf = curbuf; | |
3797 reg_win = NULL; | |
3798 ireg_ic = rmp->rm_ic; | |
3799 #ifdef FEAT_MBYTE | |
3800 ireg_icombine = FALSE; | |
3801 #endif | |
3802 ireg_maxcol = 0; | |
3803 return (nfa_regexec_both(line, col) != 0); | |
3804 } | |
3805 #endif | |
3806 | |
3807 | |
3808 /* | |
3809 * Match a regexp against multiple lines. | |
3810 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
3811 * Uses curbuf for line count and 'iskeyword'. | |
3812 * | |
3813 * Return zero if there is no match. Return number of lines contained in the | |
3814 * match otherwise. | |
3815 * | |
3816 * Note: the body is the same as bt_regexec() except for nfa_regexec_both() | |
3817 * | |
3818 * ! Also NOTE : match may actually be in another line. e.g.: | |
3819 * when r.e. is \nc, cursor is at 'a' and the text buffer looks like | |
3820 * | |
3821 * +-------------------------+ | |
3822 * |a | | |
3823 * |b | | |
3824 * |c | | |
3825 * | | | |
3826 * +-------------------------+ | |
3827 * | |
3828 * then nfa_regexec_multi() returns 3. while the original | |
3829 * vim_regexec_multi() returns 0 and a second call at line 2 will return 2. | |
3830 * | |
3831 * FIXME if this behavior is not compatible. | |
3832 */ | |
3833 static long | |
3834 nfa_regexec_multi(rmp, win, buf, lnum, col, tm) | |
3835 regmmatch_T *rmp; | |
3836 win_T *win; /* window in which to search or NULL */ | |
3837 buf_T *buf; /* buffer in which to search */ | |
3838 linenr_T lnum; /* nr of line to start looking for match */ | |
3839 colnr_T col; /* column to start looking for match */ | |
3840 proftime_T *tm UNUSED; /* timeout limit or NULL */ | |
3841 { | |
3842 reg_match = NULL; | |
3843 reg_mmatch = rmp; | |
3844 reg_buf = buf; | |
3845 reg_win = win; | |
3846 reg_firstlnum = lnum; | |
3847 reg_maxline = reg_buf->b_ml.ml_line_count - lnum; | |
3848 reg_line_lbr = FALSE; | |
3849 ireg_ic = rmp->rmm_ic; | |
3850 #ifdef FEAT_MBYTE | |
3851 ireg_icombine = FALSE; | |
3852 #endif | |
3853 ireg_maxcol = rmp->rmm_maxcol; | |
3854 | |
4503 | 3855 return nfa_regexec_both(NULL, col); |
4444 | 3856 } |
3857 | |
3858 #ifdef DEBUG | |
3859 # undef ENABLE_LOG | |
3860 #endif |