Mercurial > vim
annotate src/regexp_nfa.c @ 4537:5cc98a5898cf v7.3.1016
updated for version 7.3.1016
Problem: Unused field in nfa_state.
Solution: Remove lastthread.
author | Bram Moolenaar <bram@vim.org> |
---|---|
date | Sat, 25 May 2013 15:31:05 +0200 |
parents | 45f97c349537 |
children | 532c2e850256 |
rev | line source |
---|---|
4444 | 1 /* vi:set ts=8 sts=4 sw=4: |
2 * | |
3 * NFA regular expression implementation. | |
4 * | |
5 * This file is included in "regexp.c". | |
6 */ | |
7 | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
8 /* |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
9 * Logging of NFA engine. |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
10 * |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
11 * The NFA engine can write four log files: |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
12 * - Error log: Contains NFA engine's fatal errors. |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
13 * - Dump log: Contains compiled NFA state machine's information. |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
14 * - Run log: Contains information of matching procedure. |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
15 * - Debug log: Contains detailed information of matching procedure. Can be |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
16 * disabled by undefining NFA_REGEXP_DEBUG_LOG. |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
17 * The first one can also be used without debug mode. |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
18 * The last three are enabled when compiled as debug mode and individually |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
19 * disabled by commenting them out. |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
20 * The log files can get quite big! |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
21 * Do disable all of this when compiling Vim for debugging, undefine DEBUG in |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
22 * regexp.c |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
23 */ |
4444 | 24 #ifdef DEBUG |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
25 # define NFA_REGEXP_ERROR_LOG "nfa_regexp_error.log" |
4444 | 26 # define ENABLE_LOG |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
27 # define NFA_REGEXP_DUMP_LOG "nfa_regexp_dump.log" |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
28 # define NFA_REGEXP_RUN_LOG "nfa_regexp_run.log" |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
29 # define NFA_REGEXP_DEBUG_LOG "nfa_regexp_debug.log" |
4444 | 30 #endif |
31 | |
32 /* Upper limit allowed for {m,n} repetitions handled by NFA */ | |
33 #define NFA_BRACES_MAXLIMIT 10 | |
34 /* For allocating space for the postfix representation */ | |
35 #define NFA_POSTFIX_MULTIPLIER (NFA_BRACES_MAXLIMIT + 2)*2 | |
36 | |
37 enum | |
38 { | |
39 NFA_SPLIT = -1024, | |
40 NFA_MATCH, | |
41 NFA_SKIP_CHAR, /* matches a 0-length char */ | |
42 NFA_END_NEG_RANGE, /* Used when expanding [^ab] */ | |
43 | |
44 NFA_CONCAT, | |
45 NFA_OR, | |
46 NFA_STAR, | |
47 NFA_PLUS, | |
48 NFA_QUEST, | |
49 NFA_QUEST_NONGREEDY, /* Non-greedy version of \? */ | |
50 NFA_NOT, /* used for [^ab] negated char ranges */ | |
51 | |
52 NFA_BOL, /* ^ Begin line */ | |
53 NFA_EOL, /* $ End line */ | |
54 NFA_BOW, /* \< Begin word */ | |
55 NFA_EOW, /* \> End word */ | |
56 NFA_BOF, /* \%^ Begin file */ | |
57 NFA_EOF, /* \%$ End file */ | |
58 NFA_NEWL, | |
59 NFA_ZSTART, /* Used for \zs */ | |
60 NFA_ZEND, /* Used for \ze */ | |
61 NFA_NOPEN, /* Start of subexpression marked with \%( */ | |
62 NFA_NCLOSE, /* End of subexpr. marked with \%( ... \) */ | |
63 NFA_START_INVISIBLE, | |
64 NFA_END_INVISIBLE, | |
65 NFA_COMPOSING, /* Next nodes in NFA are part of the | |
66 composing multibyte char */ | |
67 NFA_END_COMPOSING, /* End of a composing char in the NFA */ | |
68 | |
69 /* The following are used only in the postfix form, not in the NFA */ | |
70 NFA_PREV_ATOM_NO_WIDTH, /* Used for \@= */ | |
71 NFA_PREV_ATOM_NO_WIDTH_NEG, /* Used for \@! */ | |
72 NFA_PREV_ATOM_JUST_BEFORE, /* Used for \@<= */ | |
73 NFA_PREV_ATOM_JUST_BEFORE_NEG, /* Used for \@<! */ | |
74 NFA_PREV_ATOM_LIKE_PATTERN, /* Used for \@> */ | |
75 | |
76 NFA_MOPEN, | |
77 NFA_MCLOSE = NFA_MOPEN + NSUBEXP, | |
78 | |
79 /* NFA_FIRST_NL */ | |
80 NFA_ANY = NFA_MCLOSE + NSUBEXP, /* Match any one character. */ | |
81 NFA_ANYOF, /* Match any character in this string. */ | |
82 NFA_ANYBUT, /* Match any character not in this string. */ | |
83 NFA_IDENT, /* Match identifier char */ | |
84 NFA_SIDENT, /* Match identifier char but no digit */ | |
85 NFA_KWORD, /* Match keyword char */ | |
86 NFA_SKWORD, /* Match word char but no digit */ | |
87 NFA_FNAME, /* Match file name char */ | |
88 NFA_SFNAME, /* Match file name char but no digit */ | |
89 NFA_PRINT, /* Match printable char */ | |
90 NFA_SPRINT, /* Match printable char but no digit */ | |
91 NFA_WHITE, /* Match whitespace char */ | |
92 NFA_NWHITE, /* Match non-whitespace char */ | |
93 NFA_DIGIT, /* Match digit char */ | |
94 NFA_NDIGIT, /* Match non-digit char */ | |
95 NFA_HEX, /* Match hex char */ | |
96 NFA_NHEX, /* Match non-hex char */ | |
97 NFA_OCTAL, /* Match octal char */ | |
98 NFA_NOCTAL, /* Match non-octal char */ | |
99 NFA_WORD, /* Match word char */ | |
100 NFA_NWORD, /* Match non-word char */ | |
101 NFA_HEAD, /* Match head char */ | |
102 NFA_NHEAD, /* Match non-head char */ | |
103 NFA_ALPHA, /* Match alpha char */ | |
104 NFA_NALPHA, /* Match non-alpha char */ | |
105 NFA_LOWER, /* Match lowercase char */ | |
106 NFA_NLOWER, /* Match non-lowercase char */ | |
107 NFA_UPPER, /* Match uppercase char */ | |
108 NFA_NUPPER, /* Match non-uppercase char */ | |
109 NFA_FIRST_NL = NFA_ANY + ADD_NL, | |
110 NFA_LAST_NL = NFA_NUPPER + ADD_NL, | |
111 | |
112 /* Character classes [:alnum:] etc */ | |
113 NFA_CLASS_ALNUM, | |
114 NFA_CLASS_ALPHA, | |
115 NFA_CLASS_BLANK, | |
116 NFA_CLASS_CNTRL, | |
117 NFA_CLASS_DIGIT, | |
118 NFA_CLASS_GRAPH, | |
119 NFA_CLASS_LOWER, | |
120 NFA_CLASS_PRINT, | |
121 NFA_CLASS_PUNCT, | |
122 NFA_CLASS_SPACE, | |
123 NFA_CLASS_UPPER, | |
124 NFA_CLASS_XDIGIT, | |
125 NFA_CLASS_TAB, | |
126 NFA_CLASS_RETURN, | |
127 NFA_CLASS_BACKSPACE, | |
128 NFA_CLASS_ESCAPE | |
129 }; | |
130 | |
131 /* Keep in sync with classchars. */ | |
132 static int nfa_classcodes[] = { | |
133 NFA_ANY, NFA_IDENT, NFA_SIDENT, NFA_KWORD,NFA_SKWORD, | |
134 NFA_FNAME, NFA_SFNAME, NFA_PRINT, NFA_SPRINT, | |
135 NFA_WHITE, NFA_NWHITE, NFA_DIGIT, NFA_NDIGIT, | |
136 NFA_HEX, NFA_NHEX, NFA_OCTAL, NFA_NOCTAL, | |
137 NFA_WORD, NFA_NWORD, NFA_HEAD, NFA_NHEAD, | |
138 NFA_ALPHA, NFA_NALPHA, NFA_LOWER, NFA_NLOWER, | |
139 NFA_UPPER, NFA_NUPPER | |
140 }; | |
141 | |
142 static char_u e_misplaced[] = N_("E866: (NFA regexp) Misplaced %c"); | |
143 | |
144 /* | |
145 * NFA errors can be of 3 types: | |
146 * *** NFA runtime errors, when something unknown goes wrong. The NFA fails | |
147 * silently and revert the to backtracking engine. | |
148 * syntax_error = FALSE; | |
149 * *** Regexp syntax errors, when the input regexp is not syntactically correct. | |
150 * The NFA engine displays an error message, and nothing else happens. | |
151 * syntax_error = TRUE | |
152 * *** Unsupported features, when the input regexp uses an operator that is not | |
153 * implemented in the NFA. The NFA engine fails silently, and reverts to the | |
154 * old backtracking engine. | |
155 * syntax_error = FALSE | |
156 * "The NFA fails" means that "compiling the regexp with the NFA fails": | |
157 * nfa_regcomp() returns FAIL. | |
158 */ | |
159 static int syntax_error = FALSE; | |
160 | |
161 /* NFA regexp \ze operator encountered. */ | |
162 static int nfa_has_zend = FALSE; | |
163 | |
164 static int *post_start; /* holds the postfix form of r.e. */ | |
165 static int *post_end; | |
166 static int *post_ptr; | |
167 | |
168 static int nstate; /* Number of states in the NFA. */ | |
169 static int istate; /* Index in the state vector, used in new_state() */ | |
170 static int nstate_max; /* Upper bound of estimated number of states. */ | |
171 | |
172 | |
173 static int nfa_regcomp_start __ARGS((char_u*expr, int re_flags)); | |
174 static int nfa_recognize_char_class __ARGS((char_u *start, char_u *end, int extra_newl)); | |
175 static int nfa_emit_equi_class __ARGS((int c, int neg)); | |
176 static void nfa_inc __ARGS((char_u **p)); | |
177 static void nfa_dec __ARGS((char_u **p)); | |
178 static int nfa_regatom __ARGS((void)); | |
179 static int nfa_regpiece __ARGS((void)); | |
180 static int nfa_regconcat __ARGS((void)); | |
181 static int nfa_regbranch __ARGS((void)); | |
182 static int nfa_reg __ARGS((int paren)); | |
183 #ifdef DEBUG | |
184 static void nfa_set_code __ARGS((int c)); | |
185 static void nfa_postfix_dump __ARGS((char_u *expr, int retval)); | |
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
186 static void nfa_print_state __ARGS((FILE *debugf, nfa_state_T *state)); |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
187 static void nfa_print_state2 __ARGS((FILE *debugf, nfa_state_T *state, garray_T *indent)); |
4444 | 188 static void nfa_dump __ARGS((nfa_regprog_T *prog)); |
189 #endif | |
190 static int *re2post __ARGS((void)); | |
191 static nfa_state_T *new_state __ARGS((int c, nfa_state_T *out, nfa_state_T *out1)); | |
192 static nfa_state_T *post2nfa __ARGS((int *postfix, int *end, int nfa_calc_size)); | |
193 static int check_char_class __ARGS((int class, int c)); | |
194 static void st_error __ARGS((int *postfix, int *end, int *p)); | |
195 static void nfa_save_listids __ARGS((nfa_state_T *start, int *list)); | |
196 static void nfa_restore_listids __ARGS((nfa_state_T *start, int *list)); | |
197 static void nfa_set_null_listids __ARGS((nfa_state_T *start)); | |
198 static void nfa_set_neg_listids __ARGS((nfa_state_T *start)); | |
199 static long nfa_regtry __ARGS((nfa_state_T *start, colnr_T col)); | |
200 static long nfa_regexec_both __ARGS((char_u *line, colnr_T col)); | |
201 static regprog_T *nfa_regcomp __ARGS((char_u *expr, int re_flags)); | |
202 static int nfa_regexec __ARGS((regmatch_T *rmp, char_u *line, colnr_T col)); | |
203 static long nfa_regexec_multi __ARGS((regmmatch_T *rmp, win_T *win, buf_T *buf, linenr_T lnum, colnr_T col, proftime_T *tm)); | |
204 | |
205 /* helper functions used when doing re2post() ... regatom() parsing */ | |
206 #define EMIT(c) do { \ | |
207 if (post_ptr >= post_end) \ | |
208 return FAIL; \ | |
209 *post_ptr++ = c; \ | |
210 } while (0) | |
211 | |
212 /* | |
213 * Initialize internal variables before NFA compilation. | |
214 * Return OK on success, FAIL otherwise. | |
215 */ | |
216 static int | |
217 nfa_regcomp_start(expr, re_flags) | |
218 char_u *expr; | |
219 int re_flags; /* see vim_regcomp() */ | |
220 { | |
4458 | 221 size_t postfix_size; |
4444 | 222 |
223 nstate = 0; | |
224 istate = 0; | |
225 /* A reasonable estimation for size */ | |
4458 | 226 nstate_max = (int)(STRLEN(expr) + 1) * NFA_POSTFIX_MULTIPLIER; |
4444 | 227 |
4454 | 228 /* Some items blow up in size, such as [A-z]. Add more space for that. |
229 * TODO: some patterns may still fail. */ | |
4458 | 230 nstate_max += 1000; |
4454 | 231 |
232 /* Size for postfix representation of expr. */ | |
4444 | 233 postfix_size = sizeof(*post_start) * nstate_max; |
4454 | 234 |
4444 | 235 post_start = (int *)lalloc(postfix_size, TRUE); |
236 if (post_start == NULL) | |
237 return FAIL; | |
238 vim_memset(post_start, 0, postfix_size); | |
239 post_ptr = post_start; | |
4454 | 240 post_end = post_start + nstate_max; |
4444 | 241 nfa_has_zend = FALSE; |
242 | |
243 regcomp_start(expr, re_flags); | |
244 | |
245 return OK; | |
246 } | |
247 | |
248 /* | |
249 * Search between "start" and "end" and try to recognize a | |
250 * character class in expanded form. For example [0-9]. | |
251 * On success, return the id the character class to be emitted. | |
252 * On failure, return 0 (=FAIL) | |
253 * Start points to the first char of the range, while end should point | |
254 * to the closing brace. | |
255 */ | |
256 static int | |
257 nfa_recognize_char_class(start, end, extra_newl) | |
258 char_u *start; | |
259 char_u *end; | |
260 int extra_newl; | |
261 { | |
262 int i; | |
263 /* Each of these variables takes up a char in "config[]", | |
264 * in the order they are here. */ | |
265 int not = FALSE, af = FALSE, AF = FALSE, az = FALSE, AZ = FALSE, | |
266 o7 = FALSE, o9 = FALSE, underscore = FALSE, newl = FALSE; | |
267 char_u *p; | |
268 #define NCONFIGS 16 | |
269 int classid[NCONFIGS] = { | |
270 NFA_DIGIT, NFA_NDIGIT, NFA_HEX, NFA_NHEX, | |
271 NFA_OCTAL, NFA_NOCTAL, NFA_WORD, NFA_NWORD, | |
272 NFA_HEAD, NFA_NHEAD, NFA_ALPHA, NFA_NALPHA, | |
273 NFA_LOWER, NFA_NLOWER, NFA_UPPER, NFA_NUPPER | |
274 }; | |
4450 | 275 char_u myconfig[10]; |
4444 | 276 char_u config[NCONFIGS][9] = { |
277 "000000100", /* digit */ | |
278 "100000100", /* non digit */ | |
279 "011000100", /* hex-digit */ | |
280 "111000100", /* non hex-digit */ | |
281 "000001000", /* octal-digit */ | |
282 "100001000", /* [^0-7] */ | |
283 "000110110", /* [0-9A-Za-z_] */ | |
284 "100110110", /* [^0-9A-Za-z_] */ | |
285 "000110010", /* head of word */ | |
286 "100110010", /* not head of word */ | |
287 "000110000", /* alphabetic char a-z */ | |
288 "100110000", /* non alphabetic char */ | |
289 "000100000", /* lowercase letter */ | |
290 "100100000", /* non lowercase */ | |
291 "000010000", /* uppercase */ | |
292 "100010000" /* non uppercase */ | |
293 }; | |
294 | |
295 if (extra_newl == TRUE) | |
296 newl = TRUE; | |
297 | |
298 if (*end != ']') | |
299 return FAIL; | |
300 p = start; | |
301 if (*p == '^') | |
302 { | |
303 not = TRUE; | |
304 p ++; | |
305 } | |
306 | |
307 while (p < end) | |
308 { | |
309 if (p + 2 < end && *(p + 1) == '-') | |
310 { | |
311 switch (*p) | |
312 { | |
313 case '0': | |
314 if (*(p + 2) == '9') | |
315 { | |
316 o9 = TRUE; | |
317 break; | |
318 } | |
319 else | |
320 if (*(p + 2) == '7') | |
321 { | |
322 o7 = TRUE; | |
323 break; | |
324 } | |
325 case 'a': | |
326 if (*(p + 2) == 'z') | |
327 { | |
328 az = TRUE; | |
329 break; | |
330 } | |
331 else | |
332 if (*(p + 2) == 'f') | |
333 { | |
334 af = TRUE; | |
335 break; | |
336 } | |
337 case 'A': | |
338 if (*(p + 2) == 'Z') | |
339 { | |
340 AZ = TRUE; | |
341 break; | |
342 } | |
343 else | |
344 if (*(p + 2) == 'F') | |
345 { | |
346 AF = TRUE; | |
347 break; | |
348 } | |
349 /* FALLTHROUGH */ | |
350 default: | |
351 return FAIL; | |
352 } | |
353 p += 3; | |
354 } | |
355 else if (p + 1 < end && *p == '\\' && *(p + 1) == 'n') | |
356 { | |
357 newl = TRUE; | |
358 p += 2; | |
359 } | |
360 else if (*p == '_') | |
361 { | |
362 underscore = TRUE; | |
363 p ++; | |
364 } | |
365 else if (*p == '\n') | |
366 { | |
367 newl = TRUE; | |
368 p ++; | |
369 } | |
370 else | |
371 return FAIL; | |
372 } /* while (p < end) */ | |
373 | |
374 if (p != end) | |
375 return FAIL; | |
376 | |
377 /* build the config that represents the ranges we gathered */ | |
378 STRCPY(myconfig, "000000000"); | |
379 if (not == TRUE) | |
380 myconfig[0] = '1'; | |
381 if (af == TRUE) | |
382 myconfig[1] = '1'; | |
383 if (AF == TRUE) | |
384 myconfig[2] = '1'; | |
385 if (az == TRUE) | |
386 myconfig[3] = '1'; | |
387 if (AZ == TRUE) | |
388 myconfig[4] = '1'; | |
389 if (o7 == TRUE) | |
390 myconfig[5] = '1'; | |
391 if (o9 == TRUE) | |
392 myconfig[6] = '1'; | |
393 if (underscore == TRUE) | |
394 myconfig[7] = '1'; | |
395 if (newl == TRUE) | |
396 { | |
397 myconfig[8] = '1'; | |
398 extra_newl = ADD_NL; | |
399 } | |
400 /* try to recognize character classes */ | |
401 for (i = 0; i < NCONFIGS; i++) | |
4450 | 402 if (STRNCMP(myconfig, config[i], 8) == 0) |
4444 | 403 return classid[i] + extra_newl; |
404 | |
405 /* fallthrough => no success so far */ | |
406 return FAIL; | |
407 | |
408 #undef NCONFIGS | |
409 } | |
410 | |
411 /* | |
412 * Produce the bytes for equivalence class "c". | |
413 * Currently only handles latin1, latin9 and utf-8. | |
414 * Emits bytes in postfix notation: 'a,b,NFA_OR,c,NFA_OR' is | |
415 * equivalent to 'a OR b OR c' | |
416 * | |
417 * NOTE! When changing this function, also update reg_equi_class() | |
418 */ | |
419 static int | |
420 nfa_emit_equi_class(c, neg) | |
421 int c; | |
422 int neg; | |
423 { | |
424 int first = TRUE; | |
425 int glue = neg == TRUE ? NFA_CONCAT : NFA_OR; | |
426 #define EMIT2(c) \ | |
427 EMIT(c); \ | |
428 if (neg == TRUE) { \ | |
429 EMIT(NFA_NOT); \ | |
430 } \ | |
431 if (first == FALSE) \ | |
432 EMIT(glue); \ | |
433 else \ | |
434 first = FALSE; \ | |
435 | |
436 #ifdef FEAT_MBYTE | |
437 if (enc_utf8 || STRCMP(p_enc, "latin1") == 0 | |
438 || STRCMP(p_enc, "iso-8859-15") == 0) | |
439 #endif | |
440 { | |
441 switch (c) | |
442 { | |
443 case 'A': case '\300': case '\301': case '\302': | |
444 case '\303': case '\304': case '\305': | |
445 EMIT2('A'); EMIT2('\300'); EMIT2('\301'); | |
446 EMIT2('\302'); EMIT2('\303'); EMIT2('\304'); | |
447 EMIT2('\305'); | |
448 return OK; | |
449 | |
450 case 'C': case '\307': | |
451 EMIT2('C'); EMIT2('\307'); | |
452 return OK; | |
453 | |
454 case 'E': case '\310': case '\311': case '\312': case '\313': | |
455 EMIT2('E'); EMIT2('\310'); EMIT2('\311'); | |
456 EMIT2('\312'); EMIT2('\313'); | |
457 return OK; | |
458 | |
459 case 'I': case '\314': case '\315': case '\316': case '\317': | |
460 EMIT2('I'); EMIT2('\314'); EMIT2('\315'); | |
461 EMIT2('\316'); EMIT2('\317'); | |
462 return OK; | |
463 | |
464 case 'N': case '\321': | |
465 EMIT2('N'); EMIT2('\321'); | |
466 return OK; | |
467 | |
468 case 'O': case '\322': case '\323': case '\324': case '\325': | |
469 case '\326': | |
470 EMIT2('O'); EMIT2('\322'); EMIT2('\323'); | |
471 EMIT2('\324'); EMIT2('\325'); EMIT2('\326'); | |
472 return OK; | |
473 | |
474 case 'U': case '\331': case '\332': case '\333': case '\334': | |
475 EMIT2('U'); EMIT2('\331'); EMIT2('\332'); | |
476 EMIT2('\333'); EMIT2('\334'); | |
477 return OK; | |
478 | |
479 case 'Y': case '\335': | |
480 EMIT2('Y'); EMIT2('\335'); | |
481 return OK; | |
482 | |
483 case 'a': case '\340': case '\341': case '\342': | |
484 case '\343': case '\344': case '\345': | |
485 EMIT2('a'); EMIT2('\340'); EMIT2('\341'); | |
486 EMIT2('\342'); EMIT2('\343'); EMIT2('\344'); | |
487 EMIT2('\345'); | |
488 return OK; | |
489 | |
490 case 'c': case '\347': | |
491 EMIT2('c'); EMIT2('\347'); | |
492 return OK; | |
493 | |
494 case 'e': case '\350': case '\351': case '\352': case '\353': | |
495 EMIT2('e'); EMIT2('\350'); EMIT2('\351'); | |
496 EMIT2('\352'); EMIT2('\353'); | |
497 return OK; | |
498 | |
499 case 'i': case '\354': case '\355': case '\356': case '\357': | |
500 EMIT2('i'); EMIT2('\354'); EMIT2('\355'); | |
501 EMIT2('\356'); EMIT2('\357'); | |
502 return OK; | |
503 | |
504 case 'n': case '\361': | |
505 EMIT2('n'); EMIT2('\361'); | |
506 return OK; | |
507 | |
508 case 'o': case '\362': case '\363': case '\364': case '\365': | |
509 case '\366': | |
510 EMIT2('o'); EMIT2('\362'); EMIT2('\363'); | |
511 EMIT2('\364'); EMIT2('\365'); EMIT2('\366'); | |
512 return OK; | |
513 | |
514 case 'u': case '\371': case '\372': case '\373': case '\374': | |
515 EMIT2('u'); EMIT2('\371'); EMIT2('\372'); | |
516 EMIT2('\373'); EMIT2('\374'); | |
517 return OK; | |
518 | |
519 case 'y': case '\375': case '\377': | |
520 EMIT2('y'); EMIT2('\375'); EMIT2('\377'); | |
521 return OK; | |
522 | |
523 default: | |
524 return FAIL; | |
525 } | |
526 } | |
527 | |
528 EMIT(c); | |
529 return OK; | |
530 #undef EMIT2 | |
531 } | |
532 | |
533 /* | |
534 * Code to parse regular expression. | |
535 * | |
536 * We try to reuse parsing functions in regexp.c to | |
537 * minimize surprise and keep the syntax consistent. | |
538 */ | |
539 | |
540 /* | |
541 * Increments the pointer "p" by one (multi-byte) character. | |
542 */ | |
543 static void | |
544 nfa_inc(p) | |
545 char_u **p; | |
546 { | |
547 #ifdef FEAT_MBYTE | |
548 if (has_mbyte) | |
549 mb_ptr2char_adv(p); | |
550 else | |
551 #endif | |
552 *p = *p + 1; | |
553 } | |
554 | |
555 /* | |
556 * Decrements the pointer "p" by one (multi-byte) character. | |
557 */ | |
558 static void | |
559 nfa_dec(p) | |
560 char_u **p; | |
561 { | |
562 #ifdef FEAT_MBYTE | |
563 char_u *p2, *oldp; | |
564 | |
565 if (has_mbyte) | |
566 { | |
567 oldp = *p; | |
568 /* Try to find the multibyte char that advances to the current | |
569 * position. */ | |
570 do | |
571 { | |
572 *p = *p - 1; | |
573 p2 = *p; | |
574 mb_ptr2char_adv(&p2); | |
575 } while (p2 != oldp); | |
576 } | |
577 #else | |
578 *p = *p - 1; | |
579 #endif | |
580 } | |
581 | |
582 /* | |
583 * Parse the lowest level. | |
584 * | |
585 * An atom can be one of a long list of items. Many atoms match one character | |
586 * in the text. It is often an ordinary character or a character class. | |
587 * Braces can be used to make a pattern into an atom. The "\z(\)" construct | |
588 * is only for syntax highlighting. | |
589 * | |
590 * atom ::= ordinary-atom | |
591 * or \( pattern \) | |
592 * or \%( pattern \) | |
593 * or \z( pattern \) | |
594 */ | |
595 static int | |
596 nfa_regatom() | |
597 { | |
598 int c; | |
599 int charclass; | |
600 int equiclass; | |
601 int collclass; | |
602 int got_coll_char; | |
603 char_u *p; | |
604 char_u *endp; | |
605 #ifdef FEAT_MBYTE | |
606 char_u *old_regparse = regparse; | |
607 int clen; | |
608 int i; | |
609 #endif | |
610 int extra = 0; | |
611 int first; | |
612 int emit_range; | |
613 int negated; | |
614 int result; | |
615 int startc = -1; | |
616 int endc = -1; | |
617 int oldstartc = -1; | |
618 int cpo_lit; /* 'cpoptions' contains 'l' flag */ | |
619 int cpo_bsl; /* 'cpoptions' contains '\' flag */ | |
620 int glue; /* ID that will "glue" nodes together */ | |
621 | |
622 cpo_lit = vim_strchr(p_cpo, CPO_LITERAL) != NULL; | |
623 cpo_bsl = vim_strchr(p_cpo, CPO_BACKSL) != NULL; | |
624 | |
625 c = getchr(); | |
626 | |
627 #ifdef FEAT_MBYTE | |
628 /* clen has the length of the current char, without composing chars */ | |
629 clen = (*mb_char2len)(c); | |
630 if (has_mbyte && clen > 1) | |
631 goto nfa_do_multibyte; | |
632 #endif | |
633 switch (c) | |
634 { | |
635 case Magic('^'): | |
636 EMIT(NFA_BOL); | |
637 break; | |
638 | |
639 case Magic('$'): | |
640 EMIT(NFA_EOL); | |
641 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
642 had_eol = TRUE; | |
643 #endif | |
644 break; | |
645 | |
646 case Magic('<'): | |
647 EMIT(NFA_BOW); | |
648 break; | |
649 | |
650 case Magic('>'): | |
651 EMIT(NFA_EOW); | |
652 break; | |
653 | |
654 case Magic('_'): | |
655 c = no_Magic(getchr()); | |
656 if (c == '^') /* "\_^" is start-of-line */ | |
657 { | |
658 EMIT(NFA_BOL); | |
659 break; | |
660 } | |
661 if (c == '$') /* "\_$" is end-of-line */ | |
662 { | |
663 EMIT(NFA_EOL); | |
664 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
665 had_eol = TRUE; | |
666 #endif | |
667 break; | |
668 } | |
669 | |
670 extra = ADD_NL; | |
671 | |
672 /* "\_[" is collection plus newline */ | |
673 if (c == '[') | |
4517
9a2183bd8295
updated for version 7.3.1006
Bram Moolenaar <bram@vim.org>
parents:
4515
diff
changeset
|
674 goto collection; |
4444 | 675 |
676 /* "\_x" is character class plus newline */ | |
677 /*FALLTHROUGH*/ | |
678 | |
679 /* | |
680 * Character classes. | |
681 */ | |
682 case Magic('.'): | |
683 case Magic('i'): | |
684 case Magic('I'): | |
685 case Magic('k'): | |
686 case Magic('K'): | |
687 case Magic('f'): | |
688 case Magic('F'): | |
689 case Magic('p'): | |
690 case Magic('P'): | |
691 case Magic('s'): | |
692 case Magic('S'): | |
693 case Magic('d'): | |
694 case Magic('D'): | |
695 case Magic('x'): | |
696 case Magic('X'): | |
697 case Magic('o'): | |
698 case Magic('O'): | |
699 case Magic('w'): | |
700 case Magic('W'): | |
701 case Magic('h'): | |
702 case Magic('H'): | |
703 case Magic('a'): | |
704 case Magic('A'): | |
705 case Magic('l'): | |
706 case Magic('L'): | |
707 case Magic('u'): | |
708 case Magic('U'): | |
709 p = vim_strchr(classchars, no_Magic(c)); | |
710 if (p == NULL) | |
711 { | |
712 return FAIL; /* runtime error */ | |
713 } | |
714 #ifdef FEAT_MBYTE | |
715 /* When '.' is followed by a composing char ignore the dot, so that | |
716 * the composing char is matched here. */ | |
717 if (enc_utf8 && c == Magic('.') && utf_iscomposing(peekchr())) | |
718 { | |
4535
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
719 old_regparse = regparse; |
4444 | 720 c = getchr(); |
721 goto nfa_do_multibyte; | |
722 } | |
723 #endif | |
724 EMIT(nfa_classcodes[p - classchars]); | |
725 if (extra == ADD_NL) | |
726 { | |
727 EMIT(NFA_NEWL); | |
728 EMIT(NFA_OR); | |
729 regflags |= RF_HASNL; | |
730 } | |
731 break; | |
732 | |
733 case Magic('n'): | |
734 if (reg_string) | |
735 /* In a string "\n" matches a newline character. */ | |
736 EMIT(NL); | |
737 else | |
738 { | |
739 /* In buffer text "\n" matches the end of a line. */ | |
740 EMIT(NFA_NEWL); | |
741 regflags |= RF_HASNL; | |
742 } | |
743 break; | |
744 | |
745 case Magic('('): | |
746 if (nfa_reg(REG_PAREN) == FAIL) | |
747 return FAIL; /* cascaded error */ | |
748 break; | |
749 | |
750 case NUL: | |
751 syntax_error = TRUE; | |
752 EMSG_RET_FAIL(_("E865: (NFA) Regexp end encountered prematurely")); | |
753 | |
754 case Magic('|'): | |
755 case Magic('&'): | |
756 case Magic(')'): | |
757 syntax_error = TRUE; | |
4450 | 758 EMSGN(_(e_misplaced), no_Magic(c)); |
4444 | 759 return FAIL; |
760 | |
761 case Magic('='): | |
762 case Magic('?'): | |
763 case Magic('+'): | |
764 case Magic('@'): | |
765 case Magic('*'): | |
766 case Magic('{'): | |
767 /* these should follow an atom, not form an atom */ | |
768 syntax_error = TRUE; | |
4450 | 769 EMSGN(_(e_misplaced), no_Magic(c)); |
4444 | 770 return FAIL; |
771 | |
772 case Magic('~'): /* previous substitute pattern */ | |
773 /* Not supported yet */ | |
774 return FAIL; | |
775 | |
776 case Magic('1'): | |
777 case Magic('2'): | |
778 case Magic('3'): | |
779 case Magic('4'): | |
780 case Magic('5'): | |
781 case Magic('6'): | |
782 case Magic('7'): | |
783 case Magic('8'): | |
784 case Magic('9'): | |
785 /* not supported yet */ | |
786 return FAIL; | |
787 | |
788 case Magic('z'): | |
789 c = no_Magic(getchr()); | |
790 switch (c) | |
791 { | |
792 case 's': | |
793 EMIT(NFA_ZSTART); | |
794 break; | |
795 case 'e': | |
796 EMIT(NFA_ZEND); | |
797 nfa_has_zend = TRUE; | |
798 /* TODO: Currently \ze does not work properly. */ | |
799 return FAIL; | |
800 /* break; */ | |
801 case '1': | |
802 case '2': | |
803 case '3': | |
804 case '4': | |
805 case '5': | |
806 case '6': | |
807 case '7': | |
808 case '8': | |
809 case '9': | |
810 case '(': | |
811 /* \z1...\z9 and \z( not yet supported */ | |
812 return FAIL; | |
813 default: | |
814 syntax_error = TRUE; | |
4450 | 815 EMSGN(_("E867: (NFA) Unknown operator '\\z%c'"), |
4444 | 816 no_Magic(c)); |
817 return FAIL; | |
818 } | |
819 break; | |
820 | |
821 case Magic('%'): | |
822 c = no_Magic(getchr()); | |
823 switch (c) | |
824 { | |
825 /* () without a back reference */ | |
826 case '(': | |
827 if (nfa_reg(REG_NPAREN) == FAIL) | |
828 return FAIL; | |
829 EMIT(NFA_NOPEN); | |
830 break; | |
831 | |
832 case 'd': /* %d123 decimal */ | |
833 case 'o': /* %o123 octal */ | |
834 case 'x': /* %xab hex 2 */ | |
835 case 'u': /* %uabcd hex 4 */ | |
836 case 'U': /* %U1234abcd hex 8 */ | |
837 /* Not yet supported */ | |
838 return FAIL; | |
839 | |
840 c = coll_get_char(); | |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
841 EMIT(c); |
4444 | 842 break; |
843 | |
844 /* Catch \%^ and \%$ regardless of where they appear in the | |
845 * pattern -- regardless of whether or not it makes sense. */ | |
846 case '^': | |
847 EMIT(NFA_BOF); | |
848 /* Not yet supported */ | |
849 return FAIL; | |
850 break; | |
851 | |
852 case '$': | |
853 EMIT(NFA_EOF); | |
854 /* Not yet supported */ | |
855 return FAIL; | |
856 break; | |
857 | |
858 case '#': | |
859 /* not supported yet */ | |
860 return FAIL; | |
861 break; | |
862 | |
863 case 'V': | |
864 /* not supported yet */ | |
865 return FAIL; | |
866 break; | |
867 | |
868 case '[': | |
869 /* \%[abc] not supported yet */ | |
870 return FAIL; | |
871 | |
872 default: | |
873 /* not supported yet */ | |
874 return FAIL; | |
875 } | |
876 break; | |
877 | |
878 case Magic('['): | |
4517
9a2183bd8295
updated for version 7.3.1006
Bram Moolenaar <bram@vim.org>
parents:
4515
diff
changeset
|
879 collection: |
4444 | 880 /* |
881 * Glue is emitted between several atoms from the []. | |
882 * It is either NFA_OR, or NFA_CONCAT. | |
883 * | |
884 * [abc] expands to 'a b NFA_OR c NFA_OR' (in postfix notation) | |
885 * [^abc] expands to 'a NFA_NOT b NFA_NOT NFA_CONCAT c NFA_NOT | |
886 * NFA_CONCAT NFA_END_NEG_RANGE NFA_CONCAT' (in postfix | |
887 * notation) | |
888 * | |
889 */ | |
890 | |
891 | |
892 /* Emit negation atoms, if needed. | |
893 * The CONCAT below merges the NOT with the previous node. */ | |
894 #define TRY_NEG() \ | |
895 if (negated == TRUE) \ | |
896 { \ | |
897 EMIT(NFA_NOT); \ | |
898 } | |
899 | |
900 /* Emit glue between important nodes : CONCAT or OR. */ | |
901 #define EMIT_GLUE() \ | |
902 if (first == FALSE) \ | |
903 EMIT(glue); \ | |
904 else \ | |
905 first = FALSE; | |
906 | |
907 p = regparse; | |
908 endp = skip_anyof(p); | |
909 if (*endp == ']') | |
910 { | |
911 /* | |
912 * Try to reverse engineer character classes. For example, | |
913 * recognize that [0-9] stands for \d and [A-Za-z_] with \h, | |
914 * and perform the necessary substitutions in the NFA. | |
915 */ | |
916 result = nfa_recognize_char_class(regparse, endp, | |
917 extra == ADD_NL); | |
918 if (result != FAIL) | |
919 { | |
920 if (result >= NFA_DIGIT && result <= NFA_NUPPER) | |
921 EMIT(result); | |
922 else /* must be char class + newline */ | |
923 { | |
924 EMIT(result - ADD_NL); | |
925 EMIT(NFA_NEWL); | |
926 EMIT(NFA_OR); | |
927 } | |
928 regparse = endp; | |
929 nfa_inc(®parse); | |
930 return OK; | |
931 } | |
932 /* | |
933 * Failed to recognize a character class. Use the simple | |
934 * version that turns [abc] into 'a' OR 'b' OR 'c' | |
935 */ | |
936 startc = endc = oldstartc = -1; | |
937 first = TRUE; /* Emitting first atom in this sequence? */ | |
938 negated = FALSE; | |
939 glue = NFA_OR; | |
940 if (*regparse == '^') /* negated range */ | |
941 { | |
942 negated = TRUE; | |
943 glue = NFA_CONCAT; | |
944 nfa_inc(®parse); | |
945 } | |
946 if (*regparse == '-') | |
947 { | |
948 startc = '-'; | |
949 EMIT(startc); | |
950 TRY_NEG(); | |
951 EMIT_GLUE(); | |
952 nfa_inc(®parse); | |
953 } | |
954 /* Emit the OR branches for each character in the [] */ | |
955 emit_range = FALSE; | |
956 while (regparse < endp) | |
957 { | |
958 oldstartc = startc; | |
959 startc = -1; | |
960 got_coll_char = FALSE; | |
961 if (*regparse == '[') | |
962 { | |
963 /* Check for [: :], [= =], [. .] */ | |
964 equiclass = collclass = 0; | |
965 charclass = get_char_class(®parse); | |
966 if (charclass == CLASS_NONE) | |
967 { | |
968 equiclass = get_equi_class(®parse); | |
969 if (equiclass == 0) | |
970 collclass = get_coll_element(®parse); | |
971 } | |
972 | |
973 /* Character class like [:alpha:] */ | |
974 if (charclass != CLASS_NONE) | |
975 { | |
976 switch (charclass) | |
977 { | |
978 case CLASS_ALNUM: | |
979 EMIT(NFA_CLASS_ALNUM); | |
980 break; | |
981 case CLASS_ALPHA: | |
982 EMIT(NFA_CLASS_ALPHA); | |
983 break; | |
984 case CLASS_BLANK: | |
985 EMIT(NFA_CLASS_BLANK); | |
986 break; | |
987 case CLASS_CNTRL: | |
988 EMIT(NFA_CLASS_CNTRL); | |
989 break; | |
990 case CLASS_DIGIT: | |
991 EMIT(NFA_CLASS_DIGIT); | |
992 break; | |
993 case CLASS_GRAPH: | |
994 EMIT(NFA_CLASS_GRAPH); | |
995 break; | |
996 case CLASS_LOWER: | |
997 EMIT(NFA_CLASS_LOWER); | |
998 break; | |
999 case CLASS_PRINT: | |
1000 EMIT(NFA_CLASS_PRINT); | |
1001 break; | |
1002 case CLASS_PUNCT: | |
1003 EMIT(NFA_CLASS_PUNCT); | |
1004 break; | |
1005 case CLASS_SPACE: | |
1006 EMIT(NFA_CLASS_SPACE); | |
1007 break; | |
1008 case CLASS_UPPER: | |
1009 EMIT(NFA_CLASS_UPPER); | |
1010 break; | |
1011 case CLASS_XDIGIT: | |
1012 EMIT(NFA_CLASS_XDIGIT); | |
1013 break; | |
1014 case CLASS_TAB: | |
1015 EMIT(NFA_CLASS_TAB); | |
1016 break; | |
1017 case CLASS_RETURN: | |
1018 EMIT(NFA_CLASS_RETURN); | |
1019 break; | |
1020 case CLASS_BACKSPACE: | |
1021 EMIT(NFA_CLASS_BACKSPACE); | |
1022 break; | |
1023 case CLASS_ESCAPE: | |
1024 EMIT(NFA_CLASS_ESCAPE); | |
1025 break; | |
1026 } | |
1027 TRY_NEG(); | |
1028 EMIT_GLUE(); | |
1029 continue; | |
1030 } | |
1031 /* Try equivalence class [=a=] and the like */ | |
1032 if (equiclass != 0) | |
1033 { | |
1034 result = nfa_emit_equi_class(equiclass, negated); | |
1035 if (result == FAIL) | |
1036 { | |
1037 /* should never happen */ | |
1038 EMSG_RET_FAIL(_("E868: Error building NFA with equivalence class!")); | |
1039 } | |
1040 EMIT_GLUE(); | |
1041 continue; | |
1042 } | |
1043 /* Try collating class like [. .] */ | |
1044 if (collclass != 0) | |
1045 { | |
1046 startc = collclass; /* allow [.a.]-x as a range */ | |
1047 /* Will emit the proper atom at the end of the | |
1048 * while loop. */ | |
1049 } | |
1050 } | |
1051 /* Try a range like 'a-x' or '\t-z' */ | |
1052 if (*regparse == '-') | |
1053 { | |
1054 emit_range = TRUE; | |
1055 startc = oldstartc; | |
1056 nfa_inc(®parse); | |
1057 continue; /* reading the end of the range */ | |
1058 } | |
1059 | |
1060 /* Now handle simple and escaped characters. | |
1061 * Only "\]", "\^", "\]" and "\\" are special in Vi. Vim | |
1062 * accepts "\t", "\e", etc., but only when the 'l' flag in | |
1063 * 'cpoptions' is not included. | |
1064 * Posix doesn't recognize backslash at all. | |
1065 */ | |
1066 if (*regparse == '\\' | |
1067 && !cpo_bsl | |
1068 && regparse + 1 <= endp | |
1069 && (vim_strchr(REGEXP_INRANGE, regparse[1]) != NULL | |
1070 || (!cpo_lit | |
1071 && vim_strchr(REGEXP_ABBR, regparse[1]) | |
1072 != NULL) | |
1073 ) | |
1074 ) | |
1075 { | |
1076 nfa_inc(®parse); | |
1077 | |
4507
9dbbddb2ed10
updated for version 7.3.1001
Bram Moolenaar <bram@vim.org>
parents:
4503
diff
changeset
|
1078 if (*regparse == 'n') |
4444 | 1079 startc = reg_string ? NL : NFA_NEWL; |
1080 else | |
1081 if (*regparse == 'd' | |
1082 || *regparse == 'o' | |
1083 || *regparse == 'x' | |
1084 || *regparse == 'u' | |
1085 || *regparse == 'U' | |
1086 ) | |
1087 { | |
1088 /* TODO(RE) This needs more testing */ | |
1089 startc = coll_get_char(); | |
1090 got_coll_char = TRUE; | |
1091 nfa_dec(®parse); | |
1092 } | |
1093 else | |
1094 { | |
1095 /* \r,\t,\e,\b */ | |
1096 startc = backslash_trans(*regparse); | |
1097 } | |
1098 } | |
1099 | |
1100 /* Normal printable char */ | |
1101 if (startc == -1) | |
1102 #ifdef FEAT_MBYTE | |
1103 startc = (*mb_ptr2char)(regparse); | |
1104 #else | |
1105 startc = *regparse; | |
1106 #endif | |
1107 | |
1108 /* Previous char was '-', so this char is end of range. */ | |
1109 if (emit_range) | |
1110 { | |
1111 endc = startc; startc = oldstartc; | |
1112 if (startc > endc) | |
1113 EMSG_RET_FAIL(_(e_invrange)); | |
1114 #ifdef FEAT_MBYTE | |
1115 if (has_mbyte && ((*mb_char2len)(startc) > 1 | |
1116 || (*mb_char2len)(endc) > 1)) | |
1117 { | |
1118 if (endc > startc + 256) | |
1119 EMSG_RET_FAIL(_(e_invrange)); | |
1120 /* Emit the range. "startc" was already emitted, so | |
1121 * skip it. */ | |
1122 for (c = startc + 1; c <= endc; c++) | |
1123 { | |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1124 EMIT(c); |
4444 | 1125 TRY_NEG(); |
1126 EMIT_GLUE(); | |
1127 } | |
1128 emit_range = FALSE; | |
1129 } | |
1130 else | |
1131 #endif | |
1132 { | |
1133 #ifdef EBCDIC | |
1134 int alpha_only = FALSE; | |
1135 | |
1136 /* for alphabetical range skip the gaps | |
1137 * 'i'-'j', 'r'-'s', 'I'-'J' and 'R'-'S'. */ | |
1138 if (isalpha(startc) && isalpha(endc)) | |
1139 alpha_only = TRUE; | |
1140 #endif | |
1141 /* Emit the range. "startc" was already emitted, so | |
1142 * skip it. */ | |
1143 for (c = startc + 1; c <= endc; c++) | |
1144 #ifdef EBCDIC | |
1145 if (!alpha_only || isalpha(startc)) | |
1146 #endif | |
1147 { | |
1148 EMIT(c); | |
1149 TRY_NEG(); | |
1150 EMIT_GLUE(); | |
1151 } | |
1152 emit_range = FALSE; | |
1153 } | |
1154 } | |
1155 else | |
1156 { | |
1157 /* | |
1158 * This char (startc) is not part of a range. Just | |
1159 * emit it. | |
1160 * | |
1161 * Normally, simply emit startc. But if we get char | |
1162 * code=0 from a collating char, then replace it with | |
1163 * 0x0a. | |
1164 * | |
1165 * This is needed to completely mimic the behaviour of | |
1166 * the backtracking engine. | |
1167 */ | |
1168 if (got_coll_char == TRUE && startc == 0) | |
1169 EMIT(0x0a); | |
1170 else | |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1171 EMIT(startc); |
4444 | 1172 TRY_NEG(); |
1173 EMIT_GLUE(); | |
1174 } | |
1175 | |
1176 nfa_inc(®parse); | |
1177 } /* while (p < endp) */ | |
1178 | |
1179 nfa_dec(®parse); | |
1180 if (*regparse == '-') /* if last, '-' is just a char */ | |
1181 { | |
1182 EMIT('-'); | |
1183 TRY_NEG(); | |
1184 EMIT_GLUE(); | |
1185 } | |
1186 nfa_inc(®parse); | |
1187 | |
1188 if (extra == ADD_NL) /* \_[] also matches \n */ | |
1189 { | |
1190 EMIT(reg_string ? NL : NFA_NEWL); | |
1191 TRY_NEG(); | |
1192 EMIT_GLUE(); | |
1193 } | |
1194 | |
1195 /* skip the trailing ] */ | |
1196 regparse = endp; | |
1197 nfa_inc(®parse); | |
1198 if (negated == TRUE) | |
1199 { | |
1200 /* Mark end of negated char range */ | |
1201 EMIT(NFA_END_NEG_RANGE); | |
1202 EMIT(NFA_CONCAT); | |
1203 } | |
1204 return OK; | |
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
1205 } /* if exists closing ] */ |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
1206 |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
1207 if (reg_strict) |
4444 | 1208 { |
1209 syntax_error = TRUE; | |
1210 EMSG_RET_FAIL(_(e_missingbracket)); | |
1211 } | |
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
1212 /* FALLTHROUGH */ |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
1213 |
4444 | 1214 default: |
1215 { | |
1216 #ifdef FEAT_MBYTE | |
1217 int plen; | |
1218 | |
1219 nfa_do_multibyte: | |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1220 /* Length of current char with composing chars. */ |
4535
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
1221 if (enc_utf8 && (clen != (plen = (*mb_ptr2len)(old_regparse)) |
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
1222 || utf_iscomposing(c))) |
4444 | 1223 { |
4535
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
1224 /* A base character plus composing characters, or just one |
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
1225 * or more composing characters. |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1226 * This requires creating a separate atom as if enclosing |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1227 * the characters in (), where NFA_COMPOSING is the ( and |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1228 * NFA_END_COMPOSING is the ). Note that right now we are |
4444 | 1229 * building the postfix form, not the NFA itself; |
1230 * a composing char could be: a, b, c, NFA_COMPOSING | |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1231 * where 'b' and 'c' are chars with codes > 256. */ |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1232 i = 0; |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1233 for (;;) |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1234 { |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1235 EMIT(c); |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1236 if (i > 0) |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1237 EMIT(NFA_CONCAT); |
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
1238 if ((i += utf_char2len(c)) >= plen) |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1239 break; |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1240 c = utf_ptr2char(old_regparse + i); |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1241 } |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
1242 EMIT(NFA_COMPOSING); |
4444 | 1243 regparse = old_regparse + plen; |
1244 } | |
1245 else | |
1246 #endif | |
1247 { | |
1248 c = no_Magic(c); | |
1249 EMIT(c); | |
1250 } | |
1251 return OK; | |
1252 } | |
1253 } | |
1254 | |
1255 #undef TRY_NEG | |
1256 #undef EMIT_GLUE | |
1257 | |
1258 return OK; | |
1259 } | |
1260 | |
1261 /* | |
1262 * Parse something followed by possible [*+=]. | |
1263 * | |
1264 * A piece is an atom, possibly followed by a multi, an indication of how many | |
1265 * times the atom can be matched. Example: "a*" matches any sequence of "a" | |
1266 * characters: "", "a", "aa", etc. | |
1267 * | |
1268 * piece ::= atom | |
1269 * or atom multi | |
1270 */ | |
1271 static int | |
1272 nfa_regpiece() | |
1273 { | |
1274 int i; | |
1275 int op; | |
1276 int ret; | |
1277 long minval, maxval; | |
1278 int greedy = TRUE; /* Braces are prefixed with '-' ? */ | |
1279 char_u *old_regparse, *new_regparse; | |
1280 int c2; | |
1281 int *old_post_ptr, *my_post_start; | |
1282 int old_regnpar; | |
1283 int quest; | |
1284 | |
1285 /* Save the current position in the regexp, so that we can use it if | |
1286 * <atom>{m,n} is next. */ | |
1287 old_regparse = regparse; | |
1288 /* Save current number of open parenthesis, so we can use it if | |
1289 * <atom>{m,n} is next */ | |
1290 old_regnpar = regnpar; | |
1291 /* store current pos in the postfix form, for \{m,n} involving 0s */ | |
1292 my_post_start = post_ptr; | |
1293 | |
1294 ret = nfa_regatom(); | |
1295 if (ret == FAIL) | |
1296 return FAIL; /* cascaded error */ | |
1297 | |
1298 op = peekchr(); | |
1299 if (re_multi_type(op) == NOT_MULTI) | |
1300 return OK; | |
1301 | |
1302 skipchr(); | |
1303 switch (op) | |
1304 { | |
1305 case Magic('*'): | |
1306 EMIT(NFA_STAR); | |
1307 break; | |
1308 | |
1309 case Magic('+'): | |
1310 /* | |
1311 * Trick: Normally, (a*)\+ would match the whole input "aaa". The | |
1312 * first and only submatch would be "aaa". But the backtracking | |
1313 * engine interprets the plus as "try matching one more time", and | |
1314 * a* matches a second time at the end of the input, the empty | |
1315 * string. | |
1316 * The submatch will the empty string. | |
1317 * | |
1318 * In order to be consistent with the old engine, we disable | |
1319 * NFA_PLUS, and replace <atom>+ with <atom><atom>* | |
1320 */ | |
1321 /* EMIT(NFA_PLUS); */ | |
1322 regnpar = old_regnpar; | |
1323 regparse = old_regparse; | |
1324 curchr = -1; | |
1325 if (nfa_regatom() == FAIL) | |
1326 return FAIL; | |
1327 EMIT(NFA_STAR); | |
1328 EMIT(NFA_CONCAT); | |
1329 skipchr(); /* skip the \+ */ | |
1330 break; | |
1331 | |
1332 case Magic('@'): | |
1333 op = no_Magic(getchr()); | |
1334 switch(op) | |
1335 { | |
1336 case '=': | |
1337 EMIT(NFA_PREV_ATOM_NO_WIDTH); | |
1338 break; | |
1339 case '!': | |
1340 case '<': | |
1341 case '>': | |
1342 /* Not supported yet */ | |
1343 return FAIL; | |
1344 default: | |
1345 syntax_error = TRUE; | |
4450 | 1346 EMSGN(_("E869: (NFA) Unknown operator '\\@%c'"), op); |
4444 | 1347 return FAIL; |
1348 } | |
1349 break; | |
1350 | |
1351 case Magic('?'): | |
1352 case Magic('='): | |
1353 EMIT(NFA_QUEST); | |
1354 break; | |
1355 | |
1356 case Magic('{'): | |
1357 /* a{2,5} will expand to 'aaa?a?a?' | |
1358 * a{-1,3} will expand to 'aa??a??', where ?? is the nongreedy | |
1359 * version of '?' | |
1360 * \v(ab){2,3} will expand to '(ab)(ab)(ab)?', where all the | |
1361 * parenthesis have the same id | |
1362 */ | |
1363 | |
1364 greedy = TRUE; | |
1365 c2 = peekchr(); | |
1366 if (c2 == '-' || c2 == Magic('-')) | |
1367 { | |
1368 skipchr(); | |
1369 greedy = FALSE; | |
1370 } | |
1371 if (!read_limits(&minval, &maxval)) | |
1372 { | |
1373 syntax_error = TRUE; | |
1374 EMSG_RET_FAIL(_("E870: (NFA regexp) Error reading repetition limits")); | |
1375 } | |
1376 /* <atom>{0,inf}, <atom>{0,} and <atom>{} are equivalent to | |
1377 * <atom>* */ | |
1378 if (minval == 0 && maxval == MAX_LIMIT && greedy) | |
1379 { | |
1380 EMIT(NFA_STAR); | |
1381 break; | |
1382 } | |
1383 | |
1384 if (maxval > NFA_BRACES_MAXLIMIT) | |
1385 { | |
1386 /* This would yield a huge automaton and use too much memory. | |
1387 * Revert to old engine */ | |
1388 return FAIL; | |
1389 } | |
1390 | |
1391 /* Special case: x{0} or x{-0} */ | |
1392 if (maxval == 0) | |
1393 { | |
1394 /* Ignore result of previous call to nfa_regatom() */ | |
1395 post_ptr = my_post_start; | |
1396 /* NFA_SKIP_CHAR has 0-length and works everywhere */ | |
1397 EMIT(NFA_SKIP_CHAR); | |
1398 return OK; | |
1399 } | |
1400 | |
1401 /* Ignore previous call to nfa_regatom() */ | |
1402 post_ptr = my_post_start; | |
1403 /* Save pos after the repeated atom and the \{} */ | |
1404 new_regparse = regparse; | |
1405 | |
1406 quest = (greedy == TRUE? NFA_QUEST : NFA_QUEST_NONGREEDY); | |
1407 for (i = 0; i < maxval; i++) | |
1408 { | |
1409 /* Goto beginning of the repeated atom */ | |
1410 regparse = old_regparse; | |
1411 curchr = -1; | |
1412 /* Restore count of parenthesis */ | |
1413 regnpar = old_regnpar; | |
1414 old_post_ptr = post_ptr; | |
1415 if (nfa_regatom() == FAIL) | |
1416 return FAIL; | |
1417 /* after "minval" times, atoms are optional */ | |
1418 if (i + 1 > minval) | |
1419 EMIT(quest); | |
1420 if (old_post_ptr != my_post_start) | |
1421 EMIT(NFA_CONCAT); | |
1422 } | |
1423 | |
1424 /* Go to just after the repeated atom and the \{} */ | |
1425 regparse = new_regparse; | |
1426 curchr = -1; | |
1427 | |
1428 break; | |
1429 | |
1430 | |
1431 default: | |
1432 break; | |
1433 } /* end switch */ | |
1434 | |
1435 if (re_multi_type(peekchr()) != NOT_MULTI) | |
1436 { | |
1437 /* Can't have a multi follow a multi. */ | |
1438 syntax_error = TRUE; | |
1439 EMSG_RET_FAIL(_("E871: (NFA regexp) Can't have a multi follow a multi !")); | |
1440 } | |
1441 | |
1442 return OK; | |
1443 } | |
1444 | |
1445 /* | |
1446 * Parse one or more pieces, concatenated. It matches a match for the | |
1447 * first piece, followed by a match for the second piece, etc. Example: | |
1448 * "f[0-9]b", first matches "f", then a digit and then "b". | |
1449 * | |
1450 * concat ::= piece | |
1451 * or piece piece | |
1452 * or piece piece piece | |
1453 * etc. | |
1454 */ | |
1455 static int | |
1456 nfa_regconcat() | |
1457 { | |
1458 int cont = TRUE; | |
1459 int first = TRUE; | |
1460 | |
1461 while (cont) | |
1462 { | |
1463 switch (peekchr()) | |
1464 { | |
1465 case NUL: | |
1466 case Magic('|'): | |
1467 case Magic('&'): | |
1468 case Magic(')'): | |
1469 cont = FALSE; | |
1470 break; | |
1471 | |
1472 case Magic('Z'): | |
1473 #ifdef FEAT_MBYTE | |
1474 regflags |= RF_ICOMBINE; | |
1475 #endif | |
1476 skipchr_keepstart(); | |
1477 break; | |
1478 case Magic('c'): | |
1479 regflags |= RF_ICASE; | |
1480 skipchr_keepstart(); | |
1481 break; | |
1482 case Magic('C'): | |
1483 regflags |= RF_NOICASE; | |
1484 skipchr_keepstart(); | |
1485 break; | |
1486 case Magic('v'): | |
1487 reg_magic = MAGIC_ALL; | |
1488 skipchr_keepstart(); | |
1489 curchr = -1; | |
1490 break; | |
1491 case Magic('m'): | |
1492 reg_magic = MAGIC_ON; | |
1493 skipchr_keepstart(); | |
1494 curchr = -1; | |
1495 break; | |
1496 case Magic('M'): | |
1497 reg_magic = MAGIC_OFF; | |
1498 skipchr_keepstart(); | |
1499 curchr = -1; | |
1500 break; | |
1501 case Magic('V'): | |
1502 reg_magic = MAGIC_NONE; | |
1503 skipchr_keepstart(); | |
1504 curchr = -1; | |
1505 break; | |
1506 | |
1507 default: | |
1508 if (nfa_regpiece() == FAIL) | |
1509 return FAIL; | |
1510 if (first == FALSE) | |
1511 EMIT(NFA_CONCAT); | |
1512 else | |
1513 first = FALSE; | |
1514 break; | |
1515 } | |
1516 } | |
1517 | |
1518 return OK; | |
1519 } | |
1520 | |
1521 /* | |
1522 * Parse a branch, one or more concats, separated by "\&". It matches the | |
1523 * last concat, but only if all the preceding concats also match at the same | |
1524 * position. Examples: | |
1525 * "foobeep\&..." matches "foo" in "foobeep". | |
1526 * ".*Peter\&.*Bob" matches in a line containing both "Peter" and "Bob" | |
1527 * | |
1528 * branch ::= concat | |
1529 * or concat \& concat | |
1530 * or concat \& concat \& concat | |
1531 * etc. | |
1532 */ | |
1533 static int | |
1534 nfa_regbranch() | |
1535 { | |
1536 int ch; | |
1537 int *old_post_ptr; | |
1538 | |
1539 old_post_ptr = post_ptr; | |
1540 | |
1541 /* First branch, possibly the only one */ | |
1542 if (nfa_regconcat() == FAIL) | |
1543 return FAIL; | |
1544 | |
1545 ch = peekchr(); | |
1546 /* Try next concats */ | |
1547 while (ch == Magic('&')) | |
1548 { | |
1549 skipchr(); | |
1550 EMIT(NFA_NOPEN); | |
1551 EMIT(NFA_PREV_ATOM_NO_WIDTH); | |
1552 old_post_ptr = post_ptr; | |
1553 if (nfa_regconcat() == FAIL) | |
1554 return FAIL; | |
1555 /* if concat is empty, skip a input char. But do emit a node */ | |
1556 if (old_post_ptr == post_ptr) | |
1557 EMIT(NFA_SKIP_CHAR); | |
1558 EMIT(NFA_CONCAT); | |
1559 ch = peekchr(); | |
1560 } | |
1561 | |
1562 /* Even if a branch is empty, emit one node for it */ | |
1563 if (old_post_ptr == post_ptr) | |
1564 EMIT(NFA_SKIP_CHAR); | |
1565 | |
1566 return OK; | |
1567 } | |
1568 | |
1569 /* | |
1570 * Parse a pattern, one or more branches, separated by "\|". It matches | |
1571 * anything that matches one of the branches. Example: "foo\|beep" matches | |
1572 * "foo" and matches "beep". If more than one branch matches, the first one | |
1573 * is used. | |
1574 * | |
1575 * pattern ::= branch | |
1576 * or branch \| branch | |
1577 * or branch \| branch \| branch | |
1578 * etc. | |
1579 */ | |
1580 static int | |
1581 nfa_reg(paren) | |
1582 int paren; /* REG_NOPAREN, REG_PAREN, REG_NPAREN or REG_ZPAREN */ | |
1583 { | |
1584 int parno = 0; | |
1585 | |
1586 #ifdef FEAT_SYN_HL | |
1587 #endif | |
1588 if (paren == REG_PAREN) | |
1589 { | |
1590 if (regnpar >= NSUBEXP) /* Too many `(' */ | |
1591 { | |
1592 syntax_error = TRUE; | |
1593 EMSG_RET_FAIL(_("E872: (NFA regexp) Too many '('")); | |
1594 } | |
1595 parno = regnpar++; | |
1596 } | |
1597 | |
1598 if (nfa_regbranch() == FAIL) | |
1599 return FAIL; /* cascaded error */ | |
1600 | |
1601 while (peekchr() == Magic('|')) | |
1602 { | |
1603 skipchr(); | |
1604 if (nfa_regbranch() == FAIL) | |
1605 return FAIL; /* cascaded error */ | |
1606 EMIT(NFA_OR); | |
1607 } | |
1608 | |
1609 /* Check for proper termination. */ | |
1610 if (paren != REG_NOPAREN && getchr() != Magic(')')) | |
1611 { | |
1612 syntax_error = TRUE; | |
1613 if (paren == REG_NPAREN) | |
1614 EMSG2_RET_FAIL(_(e_unmatchedpp), reg_magic == MAGIC_ALL); | |
1615 else | |
1616 EMSG2_RET_FAIL(_(e_unmatchedp), reg_magic == MAGIC_ALL); | |
1617 } | |
1618 else if (paren == REG_NOPAREN && peekchr() != NUL) | |
1619 { | |
1620 syntax_error = TRUE; | |
1621 if (peekchr() == Magic(')')) | |
1622 EMSG2_RET_FAIL(_(e_unmatchedpar), reg_magic == MAGIC_ALL); | |
1623 else | |
1624 EMSG_RET_FAIL(_("E873: (NFA regexp) proper termination error")); | |
1625 } | |
1626 /* | |
1627 * Here we set the flag allowing back references to this set of | |
1628 * parentheses. | |
1629 */ | |
1630 if (paren == REG_PAREN) | |
1631 { | |
1632 had_endbrace[parno] = TRUE; /* have seen the close paren */ | |
1633 EMIT(NFA_MOPEN + parno); | |
1634 } | |
1635 | |
1636 return OK; | |
1637 } | |
1638 | |
1639 typedef struct | |
1640 { | |
1641 char_u *start[NSUBEXP]; | |
1642 char_u *end[NSUBEXP]; | |
1643 lpos_T startpos[NSUBEXP]; | |
1644 lpos_T endpos[NSUBEXP]; | |
1645 } regsub_T; | |
1646 | |
1647 static int nfa_regmatch __ARGS((nfa_state_T *start, regsub_T *submatch, regsub_T *m)); | |
1648 | |
1649 #ifdef DEBUG | |
1650 static char_u code[50]; | |
1651 | |
1652 static void | |
1653 nfa_set_code(c) | |
1654 int c; | |
1655 { | |
1656 int addnl = FALSE; | |
1657 | |
1658 if (c >= NFA_FIRST_NL && c <= NFA_LAST_NL) | |
1659 { | |
1660 addnl = TRUE; | |
1661 c -= ADD_NL; | |
1662 } | |
1663 | |
1664 STRCPY(code, ""); | |
1665 switch (c) | |
1666 { | |
1667 case NFA_MATCH: STRCPY(code, "NFA_MATCH "); break; | |
1668 case NFA_SPLIT: STRCPY(code, "NFA_SPLIT "); break; | |
1669 case NFA_CONCAT: STRCPY(code, "NFA_CONCAT "); break; | |
1670 case NFA_NEWL: STRCPY(code, "NFA_NEWL "); break; | |
1671 case NFA_ZSTART: STRCPY(code, "NFA_ZSTART"); break; | |
1672 case NFA_ZEND: STRCPY(code, "NFA_ZEND"); break; | |
1673 | |
1674 case NFA_PREV_ATOM_NO_WIDTH: | |
1675 STRCPY(code, "NFA_PREV_ATOM_NO_WIDTH"); break; | |
1676 case NFA_NOPEN: STRCPY(code, "NFA_MOPEN_INVISIBLE"); break; | |
1677 case NFA_NCLOSE: STRCPY(code, "NFA_MCLOSE_INVISIBLE"); break; | |
1678 case NFA_START_INVISIBLE: STRCPY(code, "NFA_START_INVISIBLE"); break; | |
1679 case NFA_END_INVISIBLE: STRCPY(code, "NFA_END_INVISIBLE"); break; | |
1680 | |
1681 case NFA_COMPOSING: STRCPY(code, "NFA_COMPOSING"); break; | |
1682 case NFA_END_COMPOSING: STRCPY(code, "NFA_END_COMPOSING"); break; | |
1683 | |
1684 case NFA_MOPEN + 0: | |
1685 case NFA_MOPEN + 1: | |
1686 case NFA_MOPEN + 2: | |
1687 case NFA_MOPEN + 3: | |
1688 case NFA_MOPEN + 4: | |
1689 case NFA_MOPEN + 5: | |
1690 case NFA_MOPEN + 6: | |
1691 case NFA_MOPEN + 7: | |
1692 case NFA_MOPEN + 8: | |
1693 case NFA_MOPEN + 9: | |
1694 STRCPY(code, "NFA_MOPEN(x)"); | |
1695 code[10] = c - NFA_MOPEN + '0'; | |
1696 break; | |
1697 case NFA_MCLOSE + 0: | |
1698 case NFA_MCLOSE + 1: | |
1699 case NFA_MCLOSE + 2: | |
1700 case NFA_MCLOSE + 3: | |
1701 case NFA_MCLOSE + 4: | |
1702 case NFA_MCLOSE + 5: | |
1703 case NFA_MCLOSE + 6: | |
1704 case NFA_MCLOSE + 7: | |
1705 case NFA_MCLOSE + 8: | |
1706 case NFA_MCLOSE + 9: | |
1707 STRCPY(code, "NFA_MCLOSE(x)"); | |
1708 code[11] = c - NFA_MCLOSE + '0'; | |
1709 break; | |
1710 case NFA_EOL: STRCPY(code, "NFA_EOL "); break; | |
1711 case NFA_BOL: STRCPY(code, "NFA_BOL "); break; | |
1712 case NFA_EOW: STRCPY(code, "NFA_EOW "); break; | |
1713 case NFA_BOW: STRCPY(code, "NFA_BOW "); break; | |
1714 case NFA_STAR: STRCPY(code, "NFA_STAR "); break; | |
1715 case NFA_PLUS: STRCPY(code, "NFA_PLUS "); break; | |
1716 case NFA_NOT: STRCPY(code, "NFA_NOT "); break; | |
1717 case NFA_SKIP_CHAR: STRCPY(code, "NFA_SKIP_CHAR"); break; | |
1718 case NFA_OR: STRCPY(code, "NFA_OR"); break; | |
1719 case NFA_QUEST: STRCPY(code, "NFA_QUEST"); break; | |
1720 case NFA_QUEST_NONGREEDY: STRCPY(code, "NFA_QUEST_NON_GREEDY"); break; | |
1721 case NFA_END_NEG_RANGE: STRCPY(code, "NFA_END_NEG_RANGE"); break; | |
1722 case NFA_CLASS_ALNUM: STRCPY(code, "NFA_CLASS_ALNUM"); break; | |
1723 case NFA_CLASS_ALPHA: STRCPY(code, "NFA_CLASS_ALPHA"); break; | |
1724 case NFA_CLASS_BLANK: STRCPY(code, "NFA_CLASS_BLANK"); break; | |
1725 case NFA_CLASS_CNTRL: STRCPY(code, "NFA_CLASS_CNTRL"); break; | |
1726 case NFA_CLASS_DIGIT: STRCPY(code, "NFA_CLASS_DIGIT"); break; | |
1727 case NFA_CLASS_GRAPH: STRCPY(code, "NFA_CLASS_GRAPH"); break; | |
1728 case NFA_CLASS_LOWER: STRCPY(code, "NFA_CLASS_LOWER"); break; | |
1729 case NFA_CLASS_PRINT: STRCPY(code, "NFA_CLASS_PRINT"); break; | |
1730 case NFA_CLASS_PUNCT: STRCPY(code, "NFA_CLASS_PUNCT"); break; | |
1731 case NFA_CLASS_SPACE: STRCPY(code, "NFA_CLASS_SPACE"); break; | |
1732 case NFA_CLASS_UPPER: STRCPY(code, "NFA_CLASS_UPPER"); break; | |
1733 case NFA_CLASS_XDIGIT: STRCPY(code, "NFA_CLASS_XDIGIT"); break; | |
1734 case NFA_CLASS_TAB: STRCPY(code, "NFA_CLASS_TAB"); break; | |
1735 case NFA_CLASS_RETURN: STRCPY(code, "NFA_CLASS_RETURN"); break; | |
1736 case NFA_CLASS_BACKSPACE: STRCPY(code, "NFA_CLASS_BACKSPACE"); break; | |
1737 case NFA_CLASS_ESCAPE: STRCPY(code, "NFA_CLASS_ESCAPE"); break; | |
1738 | |
1739 case NFA_ANY: STRCPY(code, "NFA_ANY"); break; | |
1740 case NFA_IDENT: STRCPY(code, "NFA_IDENT"); break; | |
1741 case NFA_SIDENT:STRCPY(code, "NFA_SIDENT"); break; | |
1742 case NFA_KWORD: STRCPY(code, "NFA_KWORD"); break; | |
1743 case NFA_SKWORD:STRCPY(code, "NFA_SKWORD"); break; | |
1744 case NFA_FNAME: STRCPY(code, "NFA_FNAME"); break; | |
1745 case NFA_SFNAME:STRCPY(code, "NFA_SFNAME"); break; | |
1746 case NFA_PRINT: STRCPY(code, "NFA_PRINT"); break; | |
1747 case NFA_SPRINT:STRCPY(code, "NFA_SPRINT"); break; | |
1748 case NFA_WHITE: STRCPY(code, "NFA_WHITE"); break; | |
1749 case NFA_NWHITE:STRCPY(code, "NFA_NWHITE"); break; | |
1750 case NFA_DIGIT: STRCPY(code, "NFA_DIGIT"); break; | |
1751 case NFA_NDIGIT:STRCPY(code, "NFA_NDIGIT"); break; | |
1752 case NFA_HEX: STRCPY(code, "NFA_HEX"); break; | |
1753 case NFA_NHEX: STRCPY(code, "NFA_NHEX"); break; | |
1754 case NFA_OCTAL: STRCPY(code, "NFA_OCTAL"); break; | |
1755 case NFA_NOCTAL:STRCPY(code, "NFA_NOCTAL"); break; | |
1756 case NFA_WORD: STRCPY(code, "NFA_WORD"); break; | |
1757 case NFA_NWORD: STRCPY(code, "NFA_NWORD"); break; | |
1758 case NFA_HEAD: STRCPY(code, "NFA_HEAD"); break; | |
1759 case NFA_NHEAD: STRCPY(code, "NFA_NHEAD"); break; | |
1760 case NFA_ALPHA: STRCPY(code, "NFA_ALPHA"); break; | |
1761 case NFA_NALPHA:STRCPY(code, "NFA_NALPHA"); break; | |
1762 case NFA_LOWER: STRCPY(code, "NFA_LOWER"); break; | |
1763 case NFA_NLOWER:STRCPY(code, "NFA_NLOWER"); break; | |
1764 case NFA_UPPER: STRCPY(code, "NFA_UPPER"); break; | |
1765 case NFA_NUPPER:STRCPY(code, "NFA_NUPPER"); break; | |
1766 | |
1767 default: | |
1768 STRCPY(code, "CHAR(x)"); | |
1769 code[5] = c; | |
1770 } | |
1771 | |
1772 if (addnl == TRUE) | |
1773 STRCAT(code, " + NEWLINE "); | |
1774 | |
1775 } | |
1776 | |
1777 #ifdef ENABLE_LOG | |
1778 static FILE *log_fd; | |
1779 | |
1780 /* | |
1781 * Print the postfix notation of the current regexp. | |
1782 */ | |
1783 static void | |
1784 nfa_postfix_dump(expr, retval) | |
1785 char_u *expr; | |
1786 int retval; | |
1787 { | |
1788 int *p; | |
1789 FILE *f; | |
1790 | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
1791 f = fopen(NFA_REGEXP_DUMP_LOG, "a"); |
4444 | 1792 if (f != NULL) |
1793 { | |
1794 fprintf(f, "\n-------------------------\n"); | |
1795 if (retval == FAIL) | |
1796 fprintf(f, ">>> NFA engine failed ... \n"); | |
1797 else if (retval == OK) | |
1798 fprintf(f, ">>> NFA engine succeeded !\n"); | |
1799 fprintf(f, "Regexp: \"%s\"\nPostfix notation (char): \"", expr); | |
4464 | 1800 for (p = post_start; *p && p < post_end; p++) |
4444 | 1801 { |
1802 nfa_set_code(*p); | |
1803 fprintf(f, "%s, ", code); | |
1804 } | |
1805 fprintf(f, "\"\nPostfix notation (int): "); | |
4464 | 1806 for (p = post_start; *p && p < post_end; p++) |
4444 | 1807 fprintf(f, "%d ", *p); |
1808 fprintf(f, "\n\n"); | |
1809 fclose(f); | |
1810 } | |
1811 } | |
1812 | |
1813 /* | |
1814 * Print the NFA starting with a root node "state". | |
1815 */ | |
1816 static void | |
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1817 nfa_print_state(debugf, state) |
4444 | 1818 FILE *debugf; |
1819 nfa_state_T *state; | |
1820 { | |
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1821 garray_T indent; |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1822 |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1823 ga_init2(&indent, 1, 64); |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1824 ga_append(&indent, '\0'); |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1825 nfa_print_state2(debugf, state, &indent); |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1826 ga_clear(&indent); |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1827 } |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1828 |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1829 static void |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1830 nfa_print_state2(debugf, state, indent) |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1831 FILE *debugf; |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1832 nfa_state_T *state; |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1833 garray_T *indent; |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1834 { |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1835 char_u *p; |
4444 | 1836 |
1837 if (state == NULL) | |
1838 return; | |
1839 | |
1840 fprintf(debugf, "(%2d)", abs(state->id)); | |
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1841 |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1842 /* Output indent */ |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1843 p = (char_u *)indent->ga_data; |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1844 if (indent->ga_len >= 3) |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1845 { |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1846 int last = indent->ga_len - 3; |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1847 char_u save[2]; |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1848 |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1849 STRNCPY(save, &p[last], 2); |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1850 STRNCPY(&p[last], "+-", 2); |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1851 fprintf(debugf, " %s", p); |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1852 STRNCPY(&p[last], save, 2); |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1853 } |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1854 else |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1855 fprintf(debugf, " %s", p); |
4444 | 1856 |
1857 nfa_set_code(state->c); | |
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1858 fprintf(debugf, "%s%s (%d) (id=%d)\n", |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1859 state->negated ? "NOT " : "", code, state->c, abs(state->id)); |
4444 | 1860 if (state->id < 0) |
1861 return; | |
1862 | |
1863 state->id = abs(state->id) * -1; | |
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1864 |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1865 /* grow indent for state->out */ |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1866 indent->ga_len -= 1; |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1867 if (state->out1) |
4537
5cc98a5898cf
updated for version 7.3.1016
Bram Moolenaar <bram@vim.org>
parents:
4535
diff
changeset
|
1868 ga_concat(indent, (char_u *)"| "); |
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1869 else |
4537
5cc98a5898cf
updated for version 7.3.1016
Bram Moolenaar <bram@vim.org>
parents:
4535
diff
changeset
|
1870 ga_concat(indent, (char_u *)" "); |
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1871 ga_append(indent, '\0'); |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1872 |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1873 nfa_print_state2(debugf, state->out, indent); |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1874 |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1875 /* replace last part of indent for state->out1 */ |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1876 indent->ga_len -= 3; |
4537
5cc98a5898cf
updated for version 7.3.1016
Bram Moolenaar <bram@vim.org>
parents:
4535
diff
changeset
|
1877 ga_concat(indent, (char_u *)" "); |
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1878 ga_append(indent, '\0'); |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1879 |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1880 nfa_print_state2(debugf, state->out1, indent); |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1881 |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1882 /* shrink indent */ |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1883 indent->ga_len -= 3; |
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1884 ga_append(indent, '\0'); |
4444 | 1885 } |
1886 | |
1887 /* | |
1888 * Print the NFA state machine. | |
1889 */ | |
1890 static void | |
1891 nfa_dump(prog) | |
1892 nfa_regprog_T *prog; | |
1893 { | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
1894 FILE *debugf = fopen(NFA_REGEXP_DUMP_LOG, "a"); |
4444 | 1895 |
1896 if (debugf != NULL) | |
1897 { | |
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
1898 nfa_print_state(debugf, prog->start); |
4444 | 1899 fclose(debugf); |
1900 } | |
1901 } | |
1902 #endif /* ENABLE_LOG */ | |
1903 #endif /* DEBUG */ | |
1904 | |
1905 /* | |
1906 * Parse r.e. @expr and convert it into postfix form. | |
1907 * Return the postfix string on success, NULL otherwise. | |
1908 */ | |
1909 static int * | |
1910 re2post() | |
1911 { | |
1912 if (nfa_reg(REG_NOPAREN) == FAIL) | |
1913 return NULL; | |
1914 EMIT(NFA_MOPEN); | |
1915 return post_start; | |
1916 } | |
1917 | |
1918 /* NB. Some of the code below is inspired by Russ's. */ | |
1919 | |
1920 /* | |
1921 * Represents an NFA state plus zero or one or two arrows exiting. | |
1922 * if c == MATCH, no arrows out; matching state. | |
1923 * If c == SPLIT, unlabeled arrows to out and out1 (if != NULL). | |
1924 * If c < 256, labeled arrow with character c to out. | |
1925 */ | |
1926 | |
1927 static nfa_state_T *state_ptr; /* points to nfa_prog->state */ | |
1928 | |
1929 /* | |
1930 * Allocate and initialize nfa_state_T. | |
1931 */ | |
1932 static nfa_state_T * | |
1933 new_state(c, out, out1) | |
1934 int c; | |
1935 nfa_state_T *out; | |
1936 nfa_state_T *out1; | |
1937 { | |
1938 nfa_state_T *s; | |
1939 | |
1940 if (istate >= nstate) | |
1941 return NULL; | |
1942 | |
1943 s = &state_ptr[istate++]; | |
1944 | |
1945 s->c = c; | |
1946 s->out = out; | |
1947 s->out1 = out1; | |
1948 | |
1949 s->id = istate; | |
1950 s->lastlist = 0; | |
1951 s->visits = 0; | |
1952 s->negated = FALSE; | |
1953 | |
1954 return s; | |
1955 } | |
1956 | |
1957 /* | |
1958 * A partially built NFA without the matching state filled in. | |
1959 * Frag_T.start points at the start state. | |
1960 * Frag_T.out is a list of places that need to be set to the | |
1961 * next state for this fragment. | |
1962 */ | |
1963 typedef union Ptrlist Ptrlist; | |
1964 struct Frag | |
1965 { | |
1966 nfa_state_T *start; | |
1967 Ptrlist *out; | |
1968 }; | |
1969 typedef struct Frag Frag_T; | |
1970 | |
1971 static Frag_T frag __ARGS((nfa_state_T *start, Ptrlist *out)); | |
1972 static Ptrlist *list1 __ARGS((nfa_state_T **outp)); | |
1973 static void patch __ARGS((Ptrlist *l, nfa_state_T *s)); | |
1974 static Ptrlist *append __ARGS((Ptrlist *l1, Ptrlist *l2)); | |
1975 static void st_push __ARGS((Frag_T s, Frag_T **p, Frag_T *stack_end)); | |
1976 static Frag_T st_pop __ARGS((Frag_T **p, Frag_T *stack)); | |
1977 | |
1978 /* | |
4456 | 1979 * Initialize a Frag_T struct and return it. |
4444 | 1980 */ |
1981 static Frag_T | |
1982 frag(start, out) | |
1983 nfa_state_T *start; | |
1984 Ptrlist *out; | |
1985 { | |
4456 | 1986 Frag_T n; |
1987 | |
1988 n.start = start; | |
1989 n.out = out; | |
4444 | 1990 return n; |
1991 } | |
1992 | |
1993 /* | |
1994 * Since the out pointers in the list are always | |
1995 * uninitialized, we use the pointers themselves | |
1996 * as storage for the Ptrlists. | |
1997 */ | |
1998 union Ptrlist | |
1999 { | |
2000 Ptrlist *next; | |
2001 nfa_state_T *s; | |
2002 }; | |
2003 | |
2004 /* | |
2005 * Create singleton list containing just outp. | |
2006 */ | |
2007 static Ptrlist * | |
2008 list1(outp) | |
2009 nfa_state_T **outp; | |
2010 { | |
2011 Ptrlist *l; | |
2012 | |
2013 l = (Ptrlist *)outp; | |
2014 l->next = NULL; | |
2015 return l; | |
2016 } | |
2017 | |
2018 /* | |
2019 * Patch the list of states at out to point to start. | |
2020 */ | |
2021 static void | |
2022 patch(l, s) | |
2023 Ptrlist *l; | |
2024 nfa_state_T *s; | |
2025 { | |
2026 Ptrlist *next; | |
2027 | |
2028 for (; l; l = next) | |
2029 { | |
2030 next = l->next; | |
2031 l->s = s; | |
2032 } | |
2033 } | |
2034 | |
2035 | |
2036 /* | |
2037 * Join the two lists l1 and l2, returning the combination. | |
2038 */ | |
2039 static Ptrlist * | |
2040 append(l1, l2) | |
2041 Ptrlist *l1; | |
2042 Ptrlist *l2; | |
2043 { | |
2044 Ptrlist *oldl1; | |
2045 | |
2046 oldl1 = l1; | |
2047 while (l1->next) | |
2048 l1 = l1->next; | |
2049 l1->next = l2; | |
2050 return oldl1; | |
2051 } | |
2052 | |
2053 /* | |
2054 * Stack used for transforming postfix form into NFA. | |
2055 */ | |
2056 static Frag_T empty; | |
2057 | |
2058 static void | |
2059 st_error(postfix, end, p) | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2060 int *postfix UNUSED; |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2061 int *end UNUSED; |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2062 int *p UNUSED; |
4444 | 2063 { |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2064 #ifdef NFA_REGEXP_ERROR_LOG |
4444 | 2065 FILE *df; |
2066 int *p2; | |
2067 | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2068 df = fopen(NFA_REGEXP_ERROR_LOG, "a"); |
4444 | 2069 if (df) |
2070 { | |
2071 fprintf(df, "Error popping the stack!\n"); | |
2072 #ifdef DEBUG | |
2073 fprintf(df, "Current regexp is \"%s\"\n", nfa_regengine.expr); | |
2074 #endif | |
2075 fprintf(df, "Postfix form is: "); | |
2076 #ifdef DEBUG | |
2077 for (p2 = postfix; p2 < end; p2++) | |
2078 { | |
2079 nfa_set_code(*p2); | |
2080 fprintf(df, "%s, ", code); | |
2081 } | |
2082 nfa_set_code(*p); | |
2083 fprintf(df, "\nCurrent position is: "); | |
2084 for (p2 = postfix; p2 <= p; p2 ++) | |
2085 { | |
2086 nfa_set_code(*p2); | |
2087 fprintf(df, "%s, ", code); | |
2088 } | |
2089 #else | |
2090 for (p2 = postfix; p2 < end; p2++) | |
2091 { | |
2092 fprintf(df, "%d, ", *p2); | |
2093 } | |
2094 fprintf(df, "\nCurrent position is: "); | |
2095 for (p2 = postfix; p2 <= p; p2 ++) | |
2096 { | |
2097 fprintf(df, "%d, ", *p2); | |
2098 } | |
2099 #endif | |
2100 fprintf(df, "\n--------------------------\n"); | |
2101 fclose(df); | |
2102 } | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2103 #endif |
4444 | 2104 EMSG(_("E874: (NFA) Could not pop the stack !")); |
2105 } | |
2106 | |
2107 /* | |
2108 * Push an item onto the stack. | |
2109 */ | |
2110 static void | |
2111 st_push(s, p, stack_end) | |
2112 Frag_T s; | |
2113 Frag_T **p; | |
2114 Frag_T *stack_end; | |
2115 { | |
2116 Frag_T *stackp = *p; | |
2117 | |
2118 if (stackp >= stack_end) | |
2119 return; | |
2120 *stackp = s; | |
2121 *p = *p + 1; | |
2122 } | |
2123 | |
2124 /* | |
2125 * Pop an item from the stack. | |
2126 */ | |
2127 static Frag_T | |
2128 st_pop(p, stack) | |
2129 Frag_T **p; | |
2130 Frag_T *stack; | |
2131 { | |
2132 Frag_T *stackp; | |
2133 | |
2134 *p = *p - 1; | |
2135 stackp = *p; | |
2136 if (stackp < stack) | |
2137 return empty; | |
2138 return **p; | |
2139 } | |
2140 | |
2141 /* | |
2142 * Convert a postfix form into its equivalent NFA. | |
2143 * Return the NFA start state on success, NULL otherwise. | |
2144 */ | |
2145 static nfa_state_T * | |
2146 post2nfa(postfix, end, nfa_calc_size) | |
2147 int *postfix; | |
2148 int *end; | |
2149 int nfa_calc_size; | |
2150 { | |
2151 int *p; | |
2152 int mopen; | |
2153 int mclose; | |
2154 Frag_T *stack = NULL; | |
2155 Frag_T *stackp = NULL; | |
2156 Frag_T *stack_end = NULL; | |
2157 Frag_T e1; | |
2158 Frag_T e2; | |
2159 Frag_T e; | |
2160 nfa_state_T *s; | |
2161 nfa_state_T *s1; | |
2162 nfa_state_T *matchstate; | |
4484 | 2163 nfa_state_T *ret = NULL; |
4444 | 2164 |
2165 if (postfix == NULL) | |
2166 return NULL; | |
2167 | |
4456 | 2168 #define PUSH(s) st_push((s), &stackp, stack_end) |
4444 | 2169 #define POP() st_pop(&stackp, stack); \ |
2170 if (stackp < stack) \ | |
2171 { \ | |
2172 st_error(postfix, end, p); \ | |
2173 return NULL; \ | |
2174 } | |
2175 | |
2176 if (nfa_calc_size == FALSE) | |
2177 { | |
2178 /* Allocate space for the stack. Max states on the stack : nstate */ | |
4462 | 2179 stack = (Frag_T *) lalloc((nstate + 1) * sizeof(Frag_T), TRUE); |
4444 | 2180 stackp = stack; |
4462 | 2181 stack_end = stack + (nstate + 1); |
4444 | 2182 } |
2183 | |
2184 for (p = postfix; p < end; ++p) | |
2185 { | |
2186 switch (*p) | |
2187 { | |
2188 case NFA_CONCAT: | |
2189 /* Catenation. | |
2190 * Pay attention: this operator does not exist | |
2191 * in the r.e. itself (it is implicit, really). | |
2192 * It is added when r.e. is translated to postfix | |
2193 * form in re2post(). | |
2194 * | |
2195 * No new state added here. */ | |
2196 if (nfa_calc_size == TRUE) | |
2197 { | |
4458 | 2198 /* nstate += 0; */ |
4444 | 2199 break; |
2200 } | |
2201 e2 = POP(); | |
2202 e1 = POP(); | |
2203 patch(e1.out, e2.start); | |
2204 PUSH(frag(e1.start, e2.out)); | |
2205 break; | |
2206 | |
2207 case NFA_NOT: | |
2208 /* Negation of a character */ | |
2209 if (nfa_calc_size == TRUE) | |
2210 { | |
4458 | 2211 /* nstate += 0; */ |
4444 | 2212 break; |
2213 } | |
2214 e1 = POP(); | |
2215 e1.start->negated = TRUE; | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2216 #ifdef FEAT_MBYTE |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2217 if (e1.start->c == NFA_COMPOSING) |
4444 | 2218 e1.start->out1->negated = TRUE; |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2219 #endif |
4444 | 2220 PUSH(e1); |
2221 break; | |
2222 | |
2223 case NFA_OR: | |
2224 /* Alternation */ | |
2225 if (nfa_calc_size == TRUE) | |
2226 { | |
4458 | 2227 nstate++; |
4444 | 2228 break; |
2229 } | |
2230 e2 = POP(); | |
2231 e1 = POP(); | |
2232 s = new_state(NFA_SPLIT, e1.start, e2.start); | |
2233 if (s == NULL) | |
4484 | 2234 goto theend; |
4444 | 2235 PUSH(frag(s, append(e1.out, e2.out))); |
2236 break; | |
2237 | |
2238 case NFA_STAR: | |
2239 /* Zero or more */ | |
2240 if (nfa_calc_size == TRUE) | |
2241 { | |
4458 | 2242 nstate++; |
4444 | 2243 break; |
2244 } | |
2245 e = POP(); | |
2246 s = new_state(NFA_SPLIT, e.start, NULL); | |
2247 if (s == NULL) | |
4484 | 2248 goto theend; |
4444 | 2249 patch(e.out, s); |
2250 PUSH(frag(s, list1(&s->out1))); | |
2251 break; | |
2252 | |
2253 case NFA_QUEST: | |
2254 /* one or zero atoms=> greedy match */ | |
2255 if (nfa_calc_size == TRUE) | |
2256 { | |
4458 | 2257 nstate++; |
4444 | 2258 break; |
2259 } | |
2260 e = POP(); | |
2261 s = new_state(NFA_SPLIT, e.start, NULL); | |
2262 if (s == NULL) | |
4484 | 2263 goto theend; |
4444 | 2264 PUSH(frag(s, append(e.out, list1(&s->out1)))); |
2265 break; | |
2266 | |
2267 case NFA_QUEST_NONGREEDY: | |
2268 /* zero or one atoms => non-greedy match */ | |
2269 if (nfa_calc_size == TRUE) | |
2270 { | |
4458 | 2271 nstate++; |
4444 | 2272 break; |
2273 } | |
2274 e = POP(); | |
2275 s = new_state(NFA_SPLIT, NULL, e.start); | |
2276 if (s == NULL) | |
4484 | 2277 goto theend; |
4444 | 2278 PUSH(frag(s, append(e.out, list1(&s->out)))); |
2279 break; | |
2280 | |
2281 case NFA_PLUS: | |
2282 /* One or more */ | |
2283 if (nfa_calc_size == TRUE) | |
2284 { | |
4458 | 2285 nstate++; |
4444 | 2286 break; |
2287 } | |
2288 e = POP(); | |
2289 s = new_state(NFA_SPLIT, e.start, NULL); | |
2290 if (s == NULL) | |
4484 | 2291 goto theend; |
4444 | 2292 patch(e.out, s); |
2293 PUSH(frag(e.start, list1(&s->out1))); | |
2294 break; | |
2295 | |
2296 case NFA_SKIP_CHAR: | |
2297 /* Symbol of 0-length, Used in a repetition | |
2298 * with max/min count of 0 */ | |
2299 if (nfa_calc_size == TRUE) | |
2300 { | |
4458 | 2301 nstate++; |
4444 | 2302 break; |
2303 } | |
2304 s = new_state(NFA_SKIP_CHAR, NULL, NULL); | |
2305 if (s == NULL) | |
4484 | 2306 goto theend; |
4444 | 2307 PUSH(frag(s, list1(&s->out))); |
2308 break; | |
2309 | |
2310 case NFA_PREV_ATOM_NO_WIDTH: | |
2311 /* The \@= operator: match the preceding atom with 0 width. | |
2312 * Surrounds the preceding atom with START_INVISIBLE and | |
2313 * END_INVISIBLE, similarly to MOPEN. | |
2314 */ | |
2315 /* TODO: Maybe this drops the speed? */ | |
4484 | 2316 goto theend; |
4444 | 2317 |
2318 if (nfa_calc_size == TRUE) | |
2319 { | |
2320 nstate += 2; | |
2321 break; | |
2322 } | |
2323 e = POP(); | |
2324 s1 = new_state(NFA_END_INVISIBLE, NULL, NULL); | |
2325 if (s1 == NULL) | |
4484 | 2326 goto theend; |
4444 | 2327 patch(e.out, s1); |
2328 | |
2329 s = new_state(NFA_START_INVISIBLE, e.start, s1); | |
2330 if (s == NULL) | |
4484 | 2331 goto theend; |
4444 | 2332 PUSH(frag(s, list1(&s1->out))); |
2333 break; | |
2334 | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2335 #ifdef FEAT_MBYTE |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2336 case NFA_COMPOSING: /* char with composing char */ |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2337 #if 0 |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2338 /* TODO */ |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2339 if (regflags & RF_ICOMBINE) |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2340 { |
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
2341 /* use the base character only */ |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2342 } |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2343 #endif |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2344 /* FALLTHROUGH */ |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2345 #endif |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2346 |
4444 | 2347 case NFA_MOPEN + 0: /* Submatch */ |
2348 case NFA_MOPEN + 1: | |
2349 case NFA_MOPEN + 2: | |
2350 case NFA_MOPEN + 3: | |
2351 case NFA_MOPEN + 4: | |
2352 case NFA_MOPEN + 5: | |
2353 case NFA_MOPEN + 6: | |
2354 case NFA_MOPEN + 7: | |
2355 case NFA_MOPEN + 8: | |
2356 case NFA_MOPEN + 9: | |
2357 case NFA_NOPEN: /* \%( "Invisible Submatch" */ | |
2358 if (nfa_calc_size == TRUE) | |
2359 { | |
2360 nstate += 2; | |
2361 break; | |
2362 } | |
2363 | |
2364 mopen = *p; | |
2365 switch (*p) | |
2366 { | |
2367 case NFA_NOPEN: | |
2368 mclose = NFA_NCLOSE; | |
2369 break; | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2370 #ifdef FEAT_MBYTE |
4444 | 2371 case NFA_COMPOSING: |
2372 mclose = NFA_END_COMPOSING; | |
2373 break; | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2374 #endif |
4444 | 2375 default: |
2376 /* NFA_MOPEN(0) ... NFA_MOPEN(9) */ | |
2377 mclose = *p + NSUBEXP; | |
2378 break; | |
2379 } | |
2380 | |
2381 /* Allow "NFA_MOPEN" as a valid postfix representation for | |
2382 * the empty regexp "". In this case, the NFA will be | |
2383 * NFA_MOPEN -> NFA_MCLOSE. Note that this also allows | |
2384 * empty groups of parenthesis, and empty mbyte chars */ | |
2385 if (stackp == stack) | |
2386 { | |
2387 s = new_state(mopen, NULL, NULL); | |
2388 if (s == NULL) | |
4484 | 2389 goto theend; |
4444 | 2390 s1 = new_state(mclose, NULL, NULL); |
2391 if (s1 == NULL) | |
4484 | 2392 goto theend; |
4444 | 2393 patch(list1(&s->out), s1); |
2394 PUSH(frag(s, list1(&s1->out))); | |
2395 break; | |
2396 } | |
2397 | |
2398 /* At least one node was emitted before NFA_MOPEN, so | |
2399 * at least one node will be between NFA_MOPEN and NFA_MCLOSE */ | |
2400 e = POP(); | |
2401 s = new_state(mopen, e.start, NULL); /* `(' */ | |
2402 if (s == NULL) | |
4484 | 2403 goto theend; |
4444 | 2404 |
2405 s1 = new_state(mclose, NULL, NULL); /* `)' */ | |
2406 if (s1 == NULL) | |
4484 | 2407 goto theend; |
4444 | 2408 patch(e.out, s1); |
2409 | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2410 #ifdef FEAT_MBYTE |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2411 if (mopen == NFA_COMPOSING) |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2412 /* COMPOSING->out1 = END_COMPOSING */ |
4444 | 2413 patch(list1(&s->out1), s1); |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2414 #endif |
4444 | 2415 |
2416 PUSH(frag(s, list1(&s1->out))); | |
2417 break; | |
2418 | |
2419 case NFA_ZSTART: | |
2420 case NFA_ZEND: | |
2421 default: | |
2422 /* Operands */ | |
2423 if (nfa_calc_size == TRUE) | |
2424 { | |
4458 | 2425 nstate++; |
4444 | 2426 break; |
2427 } | |
2428 s = new_state(*p, NULL, NULL); | |
2429 if (s == NULL) | |
4484 | 2430 goto theend; |
4444 | 2431 PUSH(frag(s, list1(&s->out))); |
2432 break; | |
2433 | |
2434 } /* switch(*p) */ | |
2435 | |
2436 } /* for(p = postfix; *p; ++p) */ | |
2437 | |
2438 if (nfa_calc_size == TRUE) | |
2439 { | |
4458 | 2440 nstate++; |
4484 | 2441 goto theend; /* Return value when counting size is ignored anyway */ |
4444 | 2442 } |
2443 | |
2444 e = POP(); | |
2445 if (stackp != stack) | |
2446 EMSG_RET_NULL(_("E875: (NFA regexp) (While converting from postfix to NFA), too many states left on stack")); | |
2447 | |
2448 if (istate >= nstate) | |
2449 EMSG_RET_NULL(_("E876: (NFA regexp) Not enough space to store the whole NFA ")); | |
2450 | |
2451 matchstate = &state_ptr[istate++]; /* the match state */ | |
2452 matchstate->c = NFA_MATCH; | |
2453 matchstate->out = matchstate->out1 = NULL; | |
2454 | |
2455 patch(e.out, matchstate); | |
4484 | 2456 ret = e.start; |
2457 | |
2458 theend: | |
2459 vim_free(stack); | |
2460 return ret; | |
4444 | 2461 |
2462 #undef POP1 | |
2463 #undef PUSH1 | |
2464 #undef POP2 | |
2465 #undef PUSH2 | |
2466 #undef POP | |
2467 #undef PUSH | |
2468 } | |
2469 | |
2470 /**************************************************************** | |
2471 * NFA execution code. | |
2472 ****************************************************************/ | |
2473 | |
2474 /* thread_T contains runtime information of a NFA state */ | |
2475 struct thread | |
2476 { | |
2477 nfa_state_T *state; | |
2478 regsub_T sub; /* submatch info */ | |
2479 }; | |
2480 | |
2481 typedef struct | |
2482 { | |
2483 thread_T *t; | |
2484 int n; | |
2485 } List; | |
2486 | |
2487 static void addstate __ARGS((List *l, nfa_state_T *state, regsub_T *m, int off, int lid, int *match)); | |
2488 | |
2489 static void | |
2490 addstate(l, state, m, off, lid, match) | |
2491 List *l; /* runtime state list */ | |
2492 nfa_state_T *state; /* state to update */ | |
2493 regsub_T *m; /* pointers to subexpressions */ | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2494 int off; /* byte offset, when -1 go to next line */ |
4444 | 2495 int lid; |
2496 int *match; /* found match? */ | |
2497 { | |
2498 regsub_T save; | |
2499 int subidx = 0; | |
4537
5cc98a5898cf
updated for version 7.3.1016
Bram Moolenaar <bram@vim.org>
parents:
4535
diff
changeset
|
2500 thread_T *lastthread; |
4444 | 2501 |
2502 if (l == NULL || state == NULL) | |
2503 return; | |
2504 | |
2505 switch (state->c) | |
2506 { | |
2507 case NFA_SPLIT: | |
2508 case NFA_NOT: | |
2509 case NFA_NOPEN: | |
2510 case NFA_NCLOSE: | |
2511 case NFA_MCLOSE: | |
2512 case NFA_MCLOSE + 1: | |
2513 case NFA_MCLOSE + 2: | |
2514 case NFA_MCLOSE + 3: | |
2515 case NFA_MCLOSE + 4: | |
2516 case NFA_MCLOSE + 5: | |
2517 case NFA_MCLOSE + 6: | |
2518 case NFA_MCLOSE + 7: | |
2519 case NFA_MCLOSE + 8: | |
2520 case NFA_MCLOSE + 9: | |
2521 /* Do not remember these nodes in list "thislist" or "nextlist" */ | |
2522 break; | |
2523 | |
2524 default: | |
2525 if (state->lastlist == lid) | |
2526 { | |
2527 if (++state->visits > 2) | |
2528 return; | |
2529 } | |
2530 else | |
2531 { | |
2532 /* add the state to the list */ | |
2533 state->lastlist = lid; | |
4537
5cc98a5898cf
updated for version 7.3.1016
Bram Moolenaar <bram@vim.org>
parents:
4535
diff
changeset
|
2534 lastthread = &l->t[l->n++]; |
5cc98a5898cf
updated for version 7.3.1016
Bram Moolenaar <bram@vim.org>
parents:
4535
diff
changeset
|
2535 lastthread->state = state; |
5cc98a5898cf
updated for version 7.3.1016
Bram Moolenaar <bram@vim.org>
parents:
4535
diff
changeset
|
2536 lastthread->sub = *m; |
4444 | 2537 } |
2538 } | |
2539 | |
2540 #ifdef ENABLE_LOG | |
2541 nfa_set_code(state->c); | |
2542 fprintf(log_fd, "> Adding state %d to list. Character %s, code %d\n", | |
2543 abs(state->id), code, state->c); | |
2544 #endif | |
2545 switch (state->c) | |
2546 { | |
2547 case NFA_MATCH: | |
2548 *match = TRUE; | |
2549 break; | |
2550 | |
2551 case NFA_SPLIT: | |
2552 addstate(l, state->out, m, off, lid, match); | |
2553 addstate(l, state->out1, m, off, lid, match); | |
2554 break; | |
2555 | |
2556 case NFA_SKIP_CHAR: | |
2557 addstate(l, state->out, m, off, lid, match); | |
2558 break; | |
2559 | |
2560 #if 0 | |
2561 case NFA_END_NEG_RANGE: | |
2562 /* Nothing to handle here. nfa_regmatch() will take care of it */ | |
2563 break; | |
2564 | |
2565 case NFA_NOT: | |
2566 EMSG(_("E999: (NFA regexp internal error) Should not process NOT node !")); | |
2567 #ifdef ENABLE_LOG | |
2568 fprintf(f, "\n\n>>> E999: Added state NFA_NOT to a list ... Something went wrong ! Why wasn't it processed already? \n\n"); | |
2569 #endif | |
2570 break; | |
2571 | |
2572 case NFA_COMPOSING: | |
2573 /* nfa_regmatch() will match all the bytes of this composing char. */ | |
2574 break; | |
2575 #endif | |
2576 | |
2577 case NFA_NOPEN: | |
2578 case NFA_NCLOSE: | |
2579 addstate(l, state->out, m, off, lid, match); | |
2580 break; | |
2581 | |
2582 /* If this state is reached, then a recursive call of nfa_regmatch() | |
2583 * succeeded. the next call saves the found submatches in the | |
2584 * first state after the "invisible" branch. */ | |
2585 #if 0 | |
2586 case NFA_END_INVISIBLE: | |
2587 break; | |
2588 #endif | |
2589 | |
2590 case NFA_MOPEN + 0: | |
2591 case NFA_MOPEN + 1: | |
2592 case NFA_MOPEN + 2: | |
2593 case NFA_MOPEN + 3: | |
2594 case NFA_MOPEN + 4: | |
2595 case NFA_MOPEN + 5: | |
2596 case NFA_MOPEN + 6: | |
2597 case NFA_MOPEN + 7: | |
2598 case NFA_MOPEN + 8: | |
2599 case NFA_MOPEN + 9: | |
2600 case NFA_ZSTART: | |
2601 subidx = state->c - NFA_MOPEN; | |
2602 if (state->c == NFA_ZSTART) | |
2603 subidx = 0; | |
2604 | |
2605 if (REG_MULTI) | |
2606 { | |
2607 save.startpos[subidx] = m->startpos[subidx]; | |
2608 save.endpos[subidx] = m->endpos[subidx]; | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2609 if (off == -1) |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2610 { |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2611 m->startpos[subidx].lnum = reglnum + 1; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2612 m->startpos[subidx].col = 0; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2613 } |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2614 else |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2615 { |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2616 m->startpos[subidx].lnum = reglnum; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2617 m->startpos[subidx].col = |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2618 (colnr_T)(reginput - regline + off); |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2619 } |
4444 | 2620 } |
2621 else | |
2622 { | |
2623 save.start[subidx] = m->start[subidx]; | |
2624 save.end[subidx] = m->end[subidx]; | |
2625 m->start[subidx] = reginput + off; | |
2626 } | |
2627 | |
2628 addstate(l, state->out, m, off, lid, match); | |
2629 | |
2630 if (REG_MULTI) | |
2631 { | |
2632 m->startpos[subidx] = save.startpos[subidx]; | |
2633 m->endpos[subidx] = save.endpos[subidx]; | |
2634 } | |
2635 else | |
2636 { | |
2637 m->start[subidx] = save.start[subidx]; | |
2638 m->end[subidx] = save.end[subidx]; | |
2639 } | |
2640 break; | |
2641 | |
2642 case NFA_MCLOSE + 0: | |
2643 if (nfa_has_zend == TRUE) | |
2644 { | |
2645 addstate(l, state->out, m, off, lid, match); | |
2646 break; | |
2647 } | |
2648 case NFA_MCLOSE + 1: | |
2649 case NFA_MCLOSE + 2: | |
2650 case NFA_MCLOSE + 3: | |
2651 case NFA_MCLOSE + 4: | |
2652 case NFA_MCLOSE + 5: | |
2653 case NFA_MCLOSE + 6: | |
2654 case NFA_MCLOSE + 7: | |
2655 case NFA_MCLOSE + 8: | |
2656 case NFA_MCLOSE + 9: | |
2657 case NFA_ZEND: | |
2658 subidx = state->c - NFA_MCLOSE; | |
2659 if (state->c == NFA_ZEND) | |
2660 subidx = 0; | |
2661 | |
2662 if (REG_MULTI) | |
2663 { | |
2664 save.startpos[subidx] = m->startpos[subidx]; | |
2665 save.endpos[subidx] = m->endpos[subidx]; | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2666 if (off == -1) |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2667 { |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2668 m->endpos[subidx].lnum = reglnum + 1; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2669 m->endpos[subidx].col = 0; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2670 } |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2671 else |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2672 { |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2673 m->endpos[subidx].lnum = reglnum; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2674 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
|
2675 } |
4444 | 2676 } |
2677 else | |
2678 { | |
2679 save.start[subidx] = m->start[subidx]; | |
2680 save.end[subidx] = m->end[subidx]; | |
2681 m->end[subidx] = reginput + off; | |
2682 } | |
2683 | |
2684 addstate(l, state->out, m, off, lid, match); | |
2685 | |
2686 if (REG_MULTI) | |
2687 { | |
2688 m->startpos[subidx] = save.startpos[subidx]; | |
2689 m->endpos[subidx] = save.endpos[subidx]; | |
2690 } | |
2691 else | |
2692 { | |
2693 m->start[subidx] = save.start[subidx]; | |
2694 m->end[subidx] = save.end[subidx]; | |
2695 } | |
2696 break; | |
2697 } | |
2698 } | |
2699 | |
2700 /* | |
2701 * Check character class "class" against current character c. | |
2702 */ | |
2703 static int | |
2704 check_char_class(class, c) | |
2705 int class; | |
2706 int c; | |
2707 { | |
2708 switch (class) | |
2709 { | |
2710 case NFA_CLASS_ALNUM: | |
4464 | 2711 if (c >= 1 && c <= 255 && isalnum(c)) |
4444 | 2712 return OK; |
2713 break; | |
2714 case NFA_CLASS_ALPHA: | |
4464 | 2715 if (c >= 1 && c <= 255 && isalpha(c)) |
4444 | 2716 return OK; |
2717 break; | |
2718 case NFA_CLASS_BLANK: | |
2719 if (c == ' ' || c == '\t') | |
2720 return OK; | |
2721 break; | |
2722 case NFA_CLASS_CNTRL: | |
4464 | 2723 if (c >= 1 && c <= 255 && iscntrl(c)) |
4444 | 2724 return OK; |
2725 break; | |
2726 case NFA_CLASS_DIGIT: | |
2727 if (VIM_ISDIGIT(c)) | |
2728 return OK; | |
2729 break; | |
2730 case NFA_CLASS_GRAPH: | |
4464 | 2731 if (c >= 1 && c <= 255 && isgraph(c)) |
4444 | 2732 return OK; |
2733 break; | |
2734 case NFA_CLASS_LOWER: | |
2735 if (MB_ISLOWER(c)) | |
2736 return OK; | |
2737 break; | |
2738 case NFA_CLASS_PRINT: | |
2739 if (vim_isprintc(c)) | |
2740 return OK; | |
2741 break; | |
2742 case NFA_CLASS_PUNCT: | |
4464 | 2743 if (c >= 1 && c <= 255 && ispunct(c)) |
4444 | 2744 return OK; |
2745 break; | |
2746 case NFA_CLASS_SPACE: | |
2747 if ((c >=9 && c <= 13) || (c == ' ')) | |
2748 return OK; | |
2749 break; | |
2750 case NFA_CLASS_UPPER: | |
2751 if (MB_ISUPPER(c)) | |
2752 return OK; | |
2753 break; | |
2754 case NFA_CLASS_XDIGIT: | |
2755 if (vim_isxdigit(c)) | |
2756 return OK; | |
2757 break; | |
2758 case NFA_CLASS_TAB: | |
2759 if (c == '\t') | |
2760 return OK; | |
2761 break; | |
2762 case NFA_CLASS_RETURN: | |
2763 if (c == '\r') | |
2764 return OK; | |
2765 break; | |
2766 case NFA_CLASS_BACKSPACE: | |
2767 if (c == '\b') | |
2768 return OK; | |
2769 break; | |
2770 case NFA_CLASS_ESCAPE: | |
2771 if (c == '\033') | |
2772 return OK; | |
2773 break; | |
2774 | |
2775 default: | |
2776 /* should not be here :P */ | |
2777 EMSG_RET_FAIL(_("E877: (NFA regexp) Invalid character class ")); | |
2778 } | |
2779 return FAIL; | |
2780 } | |
2781 | |
2782 /* | |
2783 * Set all NFA nodes' list ID equal to -1. | |
2784 */ | |
2785 static void | |
2786 nfa_set_neg_listids(start) | |
2787 nfa_state_T *start; | |
2788 { | |
2789 if (start == NULL) | |
2790 return; | |
2791 if (start->lastlist >= 0) | |
2792 { | |
2793 start->lastlist = -1; | |
2794 nfa_set_neg_listids(start->out); | |
2795 nfa_set_neg_listids(start->out1); | |
2796 } | |
2797 } | |
2798 | |
2799 /* | |
2800 * Set all NFA nodes' list ID equal to 0. | |
2801 */ | |
2802 static void | |
2803 nfa_set_null_listids(start) | |
2804 nfa_state_T *start; | |
2805 { | |
2806 if (start == NULL) | |
2807 return; | |
2808 if (start->lastlist == -1) | |
2809 { | |
2810 start->lastlist = 0; | |
2811 nfa_set_null_listids(start->out); | |
2812 nfa_set_null_listids(start->out1); | |
2813 } | |
2814 } | |
2815 | |
2816 /* | |
2817 * Save list IDs for all NFA states in "list". | |
2818 */ | |
2819 static void | |
2820 nfa_save_listids(start, list) | |
2821 nfa_state_T *start; | |
2822 int *list; | |
2823 { | |
2824 if (start == NULL) | |
2825 return; | |
2826 if (start->lastlist != -1) | |
2827 { | |
2828 list[abs(start->id)] = start->lastlist; | |
2829 start->lastlist = -1; | |
2830 nfa_save_listids(start->out, list); | |
2831 nfa_save_listids(start->out1, list); | |
2832 } | |
2833 } | |
2834 | |
2835 /* | |
2836 * Restore list IDs from "list" to all NFA states. | |
2837 */ | |
2838 static void | |
2839 nfa_restore_listids(start, list) | |
2840 nfa_state_T *start; | |
2841 int *list; | |
2842 { | |
2843 if (start == NULL) | |
2844 return; | |
2845 if (start->lastlist == -1) | |
2846 { | |
2847 start->lastlist = list[abs(start->id)]; | |
2848 nfa_restore_listids(start->out, list); | |
2849 nfa_restore_listids(start->out1, list); | |
2850 } | |
2851 } | |
2852 | |
2853 /* | |
2854 * Main matching routine. | |
2855 * | |
2856 * Run NFA to determine whether it matches reginput. | |
2857 * | |
2858 * Return TRUE if there is a match, FALSE otherwise. | |
2859 * Note: Caller must ensure that: start != NULL. | |
2860 */ | |
2861 static int | |
2862 nfa_regmatch(start, submatch, m) | |
2863 nfa_state_T *start; | |
2864 regsub_T *submatch; | |
2865 regsub_T *m; | |
2866 { | |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
2867 int c; |
4444 | 2868 int n; |
2869 int i = 0; | |
2870 int result; | |
2871 int size = 0; | |
2872 int match = FALSE; | |
2873 int flag = 0; | |
2874 int old_reglnum = -1; | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2875 int go_to_nextline; |
4444 | 2876 thread_T *t; |
2877 char_u *old_reginput = NULL; | |
2878 char_u *old_regline = NULL; | |
2879 List list[3]; | |
2880 List *listtbl[2][2]; | |
2881 List *ll; | |
2882 int listid = 1; | |
2883 List *thislist; | |
2884 List *nextlist; | |
2885 List *neglist; | |
2886 int *listids = NULL; | |
2887 int j = 0; | |
4460 | 2888 #ifdef NFA_REGEXP_DEBUG_LOG |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2889 FILE *debug = fopen(NFA_REGEXP_DEBUG_LOG, "a"); |
4444 | 2890 |
2891 if (debug == NULL) | |
2892 { | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2893 EMSG2(_("(NFA) COULD NOT OPEN %s !"), NFA_REGEXP_DEBUG_LOG); |
4444 | 2894 return FALSE; |
2895 } | |
2896 #endif | |
2897 | |
2898 /* Allocate memory for the lists of nodes */ | |
2899 size = (nstate + 1) * sizeof(thread_T); | |
2900 list[0].t = (thread_T *)lalloc(size, TRUE); | |
2901 list[1].t = (thread_T *)lalloc(size, TRUE); | |
2902 list[2].t = (thread_T *)lalloc(size, TRUE); | |
2903 if (list[0].t == NULL || list[1].t == NULL || list[2].t == NULL) | |
2904 goto theend; | |
2905 vim_memset(list[0].t, 0, size); | |
2906 vim_memset(list[1].t, 0, size); | |
2907 vim_memset(list[2].t, 0, size); | |
2908 | |
2909 #ifdef ENABLE_LOG | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
2910 log_fd = fopen(NFA_REGEXP_RUN_LOG, "a"); |
4444 | 2911 if (log_fd != NULL) |
2912 { | |
2913 fprintf(log_fd, "**********************************\n"); | |
2914 nfa_set_code(start->c); | |
2915 fprintf(log_fd, " RUNNING nfa_regmatch() starting with state %d, code %s\n", | |
2916 abs(start->id), code); | |
2917 fprintf(log_fd, "**********************************\n"); | |
2918 } | |
2919 else | |
2920 { | |
2921 EMSG(_("Could not open temporary log file for writing, displaying on stderr ... ")); | |
2922 log_fd = stderr; | |
2923 } | |
2924 #endif | |
2925 | |
2926 thislist = &list[0]; | |
2927 thislist->n = 0; | |
2928 nextlist = &list[1]; | |
2929 nextlist->n = 0; | |
2930 neglist = &list[2]; | |
2931 neglist->n = 0; | |
2932 #ifdef ENABLE_LOG | |
2933 fprintf(log_fd, "(---) STARTSTATE\n"); | |
2934 #endif | |
2935 addstate(thislist, start, m, 0, listid, &match); | |
2936 | |
2937 /* There are two cases when the NFA advances: 1. input char matches the | |
2938 * NFA node and 2. input char does not match the NFA node, but the next | |
2939 * node is NFA_NOT. The following macro calls addstate() according to | |
2940 * these rules. It is used A LOT, so use the "listtbl" table for speed */ | |
2941 listtbl[0][0] = NULL; | |
2942 listtbl[0][1] = neglist; | |
2943 listtbl[1][0] = nextlist; | |
2944 listtbl[1][1] = NULL; | |
2945 #define ADD_POS_NEG_STATE(node) \ | |
2946 ll = listtbl[result ? 1 : 0][node->negated]; \ | |
2947 if (ll != NULL) \ | |
2948 addstate(ll, node->out , &t->sub, n, listid + 1, &match); | |
2949 | |
2950 | |
2951 /* | |
2952 * Run for each character. | |
2953 */ | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2954 for (;;) |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2955 { |
4444 | 2956 #ifdef FEAT_MBYTE |
2957 if (has_mbyte) | |
2958 { | |
2959 c = (*mb_ptr2char)(reginput); | |
2960 n = (*mb_ptr2len)(reginput); | |
2961 } | |
2962 else | |
2963 #endif | |
2964 { | |
2965 c = *reginput; | |
2966 n = 1; | |
2967 } | |
2968 if (c == NUL) | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2969 { |
4444 | 2970 n = 0; |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2971 go_to_nextline = FALSE; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
2972 } |
4444 | 2973 |
2974 /* swap lists */ | |
2975 thislist = &list[flag]; | |
2976 nextlist = &list[flag ^= 1]; | |
2977 nextlist->n = 0; /* `clear' nextlist */ | |
2978 listtbl[1][0] = nextlist; | |
2979 ++listid; | |
2980 | |
2981 #ifdef ENABLE_LOG | |
2982 fprintf(log_fd, "------------------------------------------\n"); | |
2983 fprintf(log_fd, ">>> Reginput is \"%s\"\n", reginput); | |
2984 fprintf(log_fd, ">>> Advanced one character ... Current char is %c (code %d) \n", c, (int)c); | |
2985 fprintf(log_fd, ">>> Thislist has %d states available: ", thislist->n); | |
4537
5cc98a5898cf
updated for version 7.3.1016
Bram Moolenaar <bram@vim.org>
parents:
4535
diff
changeset
|
2986 for (i = 0; i < thislist->n; i++) |
4444 | 2987 fprintf(log_fd, "%d ", abs(thislist->t[i].state->id)); |
2988 fprintf(log_fd, "\n"); | |
2989 #endif | |
2990 | |
4460 | 2991 #ifdef NFA_REGEXP_DEBUG_LOG |
4444 | 2992 fprintf(debug, "\n-------------------\n"); |
2993 #endif | |
4480 | 2994 /* |
2995 * If the state lists are empty we can stop. | |
2996 */ | |
2997 if (thislist->n == 0 && neglist->n == 0) | |
2998 break; | |
4444 | 2999 |
3000 /* compute nextlist */ | |
3001 for (i = 0; i < thislist->n || neglist->n > 0; ++i) | |
3002 { | |
3003 if (neglist->n > 0) | |
3004 { | |
3005 t = &neglist->t[0]; | |
4468 | 3006 neglist->n--; |
4444 | 3007 i--; |
3008 } | |
3009 else | |
3010 t = &thislist->t[i]; | |
3011 | |
4460 | 3012 #ifdef NFA_REGEXP_DEBUG_LOG |
4444 | 3013 nfa_set_code(t->state->c); |
3014 fprintf(debug, "%s, ", code); | |
3015 #endif | |
3016 #ifdef ENABLE_LOG | |
3017 nfa_set_code(t->state->c); | |
3018 fprintf(log_fd, "(%d) %s, code %d ... \n", abs(t->state->id), | |
3019 code, (int)t->state->c); | |
3020 #endif | |
3021 | |
3022 /* | |
3023 * Handle the possible codes of the current state. | |
3024 * The most important is NFA_MATCH. | |
3025 */ | |
3026 switch (t->state->c) | |
3027 { | |
3028 case NFA_MATCH: | |
3029 match = TRUE; | |
3030 *submatch = t->sub; | |
3031 #ifdef ENABLE_LOG | |
3032 for (j = 0; j < 4; j++) | |
3033 if (REG_MULTI) | |
3034 fprintf(log_fd, "\n *** group %d, start: c=%d, l=%d, end: c=%d, l=%d", | |
3035 j, | |
3036 t->sub.startpos[j].col, | |
3037 (int)t->sub.startpos[j].lnum, | |
3038 t->sub.endpos[j].col, | |
3039 (int)t->sub.endpos[j].lnum); | |
3040 else | |
3041 fprintf(log_fd, "\n *** group %d, start: \"%s\", end: \"%s\"", | |
3042 j, | |
3043 (char *)t->sub.start[j], | |
3044 (char *)t->sub.end[j]); | |
3045 fprintf(log_fd, "\n"); | |
3046 #endif | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3047 /* 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
|
3048 * states at this position. */ |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3049 goto nextchar; |
4444 | 3050 |
3051 case NFA_END_INVISIBLE: | |
3052 /* This is only encountered after a NFA_START_INVISIBLE node. | |
3053 * They surround a zero-width group, used with "\@=" and "\&". | |
3054 * If we got here, it means that the current "invisible" group | |
3055 * finished successfully, so return control to the parent | |
3056 * nfa_regmatch(). Submatches are stored in *m, and used in | |
3057 * the parent call. */ | |
3058 if (start->c == NFA_MOPEN + 0) | |
3059 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
3060 &match); | |
3061 else | |
3062 { | |
3063 *m = t->sub; | |
3064 match = TRUE; | |
3065 } | |
3066 break; | |
3067 | |
3068 case NFA_START_INVISIBLE: | |
3069 /* Save global variables, and call nfa_regmatch() to check if | |
3070 * the current concat matches at this position. The concat | |
3071 * ends with the node NFA_END_INVISIBLE */ | |
3072 old_reginput = reginput; | |
3073 old_regline = regline; | |
3074 old_reglnum = reglnum; | |
3075 if (listids == NULL) | |
3076 { | |
3077 listids = (int *) lalloc(sizeof(int) * nstate, TRUE); | |
3078 if (listids == NULL) | |
3079 { | |
3080 EMSG(_("E878: (NFA) Could not allocate memory for branch traversal!")); | |
3081 return 0; | |
3082 } | |
3083 } | |
3084 #ifdef ENABLE_LOG | |
3085 if (log_fd != stderr) | |
3086 fclose(log_fd); | |
3087 log_fd = NULL; | |
3088 #endif | |
3089 /* Have to clear the listid field of the NFA nodes, so that | |
3090 * nfa_regmatch() and addstate() can run properly after | |
3091 * recursion. */ | |
3092 nfa_save_listids(start, listids); | |
3093 nfa_set_null_listids(start); | |
3094 result = nfa_regmatch(t->state->out, submatch, m); | |
3095 nfa_set_neg_listids(start); | |
3096 nfa_restore_listids(start, listids); | |
3097 | |
3098 #ifdef ENABLE_LOG | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
3099 log_fd = fopen(NFA_REGEXP_RUN_LOG, "a"); |
4444 | 3100 if (log_fd != NULL) |
3101 { | |
3102 fprintf(log_fd, "****************************\n"); | |
3103 fprintf(log_fd, "FINISHED RUNNING nfa_regmatch() recursively\n"); | |
3104 fprintf(log_fd, "MATCH = %s\n", result == TRUE ? "OK" : "FALSE"); | |
3105 fprintf(log_fd, "****************************\n"); | |
3106 } | |
3107 else | |
3108 { | |
3109 EMSG(_("Could not open temporary log file for writing, displaying on stderr ... ")); | |
3110 log_fd = stderr; | |
3111 } | |
3112 #endif | |
3113 if (result == TRUE) | |
3114 { | |
3115 /* Restore position in input text */ | |
3116 reginput = old_reginput; | |
3117 regline = old_regline; | |
3118 reglnum = old_reglnum; | |
3119 /* Copy submatch info from the recursive call */ | |
3120 if (REG_MULTI) | |
3121 for (j = 1; j < NSUBEXP; j++) | |
3122 { | |
3123 t->sub.startpos[j] = m->startpos[j]; | |
3124 t->sub.endpos[j] = m->endpos[j]; | |
3125 } | |
3126 else | |
3127 for (j = 1; j < NSUBEXP; j++) | |
3128 { | |
3129 t->sub.start[j] = m->start[j]; | |
3130 t->sub.end[j] = m->end[j]; | |
3131 } | |
3132 /* t->state->out1 is the corresponding END_INVISIBLE node */ | |
3133 addstate(thislist, t->state->out1->out, &t->sub, 0, listid, | |
3134 &match); | |
3135 } | |
3136 else | |
3137 { | |
3138 /* continue with next input char */ | |
3139 reginput = old_reginput; | |
3140 } | |
3141 break; | |
3142 | |
3143 case NFA_BOL: | |
3144 if (reginput == regline) | |
3145 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
3146 &match); | |
3147 break; | |
3148 | |
3149 case NFA_EOL: | |
3150 if (c == NUL) | |
3151 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
3152 &match); | |
3153 break; | |
3154 | |
3155 case NFA_BOW: | |
3156 { | |
3157 int bow = TRUE; | |
3158 | |
3159 if (c == NUL) | |
3160 bow = FALSE; | |
3161 #ifdef FEAT_MBYTE | |
3162 else if (has_mbyte) | |
3163 { | |
3164 int this_class; | |
3165 | |
3166 /* Get class of current and previous char (if it exists). */ | |
4503 | 3167 this_class = mb_get_class_buf(reginput, reg_buf); |
4444 | 3168 if (this_class <= 1) |
3169 bow = FALSE; | |
3170 else if (reg_prev_class() == this_class) | |
3171 bow = FALSE; | |
3172 } | |
3173 #endif | |
4503 | 3174 else if (!vim_iswordc_buf(c, reg_buf) |
3175 || (reginput > regline | |
3176 && vim_iswordc_buf(reginput[-1], reg_buf))) | |
4444 | 3177 bow = FALSE; |
3178 if (bow) | |
3179 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
3180 &match); | |
3181 break; | |
3182 } | |
3183 | |
3184 case NFA_EOW: | |
3185 { | |
3186 int eow = TRUE; | |
3187 | |
3188 if (reginput == regline) | |
3189 eow = FALSE; | |
3190 #ifdef FEAT_MBYTE | |
3191 else if (has_mbyte) | |
3192 { | |
3193 int this_class, prev_class; | |
3194 | |
3195 /* Get class of current and previous char (if it exists). */ | |
4503 | 3196 this_class = mb_get_class_buf(reginput, reg_buf); |
4444 | 3197 prev_class = reg_prev_class(); |
3198 if (this_class == prev_class | |
3199 || prev_class == 0 || prev_class == 1) | |
3200 eow = FALSE; | |
3201 } | |
3202 #endif | |
4503 | 3203 else if (!vim_iswordc_buf(reginput[-1], reg_buf) |
3204 || (reginput[0] != NUL && vim_iswordc_buf(c, reg_buf))) | |
4444 | 3205 eow = FALSE; |
3206 if (eow) | |
3207 addstate(thislist, t->state->out, &t->sub, 0, listid, | |
3208 &match); | |
3209 break; | |
3210 } | |
3211 | |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3212 #ifdef FEAT_MBYTE |
4444 | 3213 case NFA_COMPOSING: |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3214 { |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
3215 int mc = c; |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
3216 int len = 0; |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
3217 nfa_state_T *end; |
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
3218 nfa_state_T *sta; |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3219 |
4444 | 3220 result = OK; |
3221 sta = t->state->out; | |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3222 len = 0; |
4535
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
3223 if (utf_iscomposing(sta->c)) |
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
3224 { |
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
3225 /* Only match composing character(s), ignore base |
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
3226 * character. Used for ".{composing}" and "{composing}" |
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
3227 * (no preceding character). */ |
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
3228 len += mb_char2len(c); |
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
3229 } |
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3230 if (ireg_icombine) |
4444 | 3231 { |
4535
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
3232 /* If \Z was present, then ignore composing characters. |
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
3233 * When ignoring the base character this always matches. */ |
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3234 /* TODO: How about negated? */ |
4535
45f97c349537
updated for version 7.3.1015
Bram Moolenaar <bram@vim.org>
parents:
4533
diff
changeset
|
3235 if (len == 0 && sta->c != c) |
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3236 result = FAIL; |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3237 len = n; |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3238 while (sta->c != NFA_END_COMPOSING) |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3239 sta = sta->out; |
4444 | 3240 } |
4529
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3241 else |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3242 while (sta->c != NFA_END_COMPOSING && len < n) |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3243 { |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3244 if (len > 0) |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3245 mc = mb_ptr2char(reginput + len); |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3246 if (mc != sta->c) |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3247 break; |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3248 len += mb_char2len(mc); |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3249 sta = sta->out; |
432a6b8c7d93
updated for version 7.3.1012
Bram Moolenaar <bram@vim.org>
parents:
4527
diff
changeset
|
3250 } |
4444 | 3251 |
3252 /* if input char length doesn't match regexp char length */ | |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3253 if (len < n || sta->c != NFA_END_COMPOSING) |
4525
36ddcf4cecbc
updated for version 7.3.1010
Bram Moolenaar <bram@vim.org>
parents:
4517
diff
changeset
|
3254 result = FAIL; |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3255 end = t->state->out1; /* NFA_END_COMPOSING */ |
4444 | 3256 ADD_POS_NEG_STATE(end); |
3257 break; | |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3258 } |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3259 #endif |
4444 | 3260 |
3261 case NFA_NEWL: | |
3262 if (!reg_line_lbr && REG_MULTI | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3263 && c == NUL && reglnum <= reg_maxline) |
4444 | 3264 { |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3265 go_to_nextline = TRUE; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3266 /* 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
|
3267 * at the start of the next line. */ |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3268 addstate(nextlist, t->state->out, &t->sub, -1, |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3269 listid + 1, &match); |
4444 | 3270 } |
3271 break; | |
3272 | |
3273 case NFA_CLASS_ALNUM: | |
3274 case NFA_CLASS_ALPHA: | |
3275 case NFA_CLASS_BLANK: | |
3276 case NFA_CLASS_CNTRL: | |
3277 case NFA_CLASS_DIGIT: | |
3278 case NFA_CLASS_GRAPH: | |
3279 case NFA_CLASS_LOWER: | |
3280 case NFA_CLASS_PRINT: | |
3281 case NFA_CLASS_PUNCT: | |
3282 case NFA_CLASS_SPACE: | |
3283 case NFA_CLASS_UPPER: | |
3284 case NFA_CLASS_XDIGIT: | |
3285 case NFA_CLASS_TAB: | |
3286 case NFA_CLASS_RETURN: | |
3287 case NFA_CLASS_BACKSPACE: | |
3288 case NFA_CLASS_ESCAPE: | |
3289 result = check_char_class(t->state->c, c); | |
3290 ADD_POS_NEG_STATE(t->state); | |
3291 break; | |
3292 | |
3293 case NFA_END_NEG_RANGE: | |
3294 /* This follows a series of negated nodes, like: | |
3295 * CHAR(x), NFA_NOT, CHAR(y), NFA_NOT etc. */ | |
3296 if (c > 0) | |
3297 addstate(nextlist, t->state->out, &t->sub, n, listid + 1, | |
3298 &match); | |
3299 break; | |
3300 | |
3301 case NFA_ANY: | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3302 /* Any char except '\0', (end of input) does not match. */ |
4444 | 3303 if (c > 0) |
3304 addstate(nextlist, t->state->out, &t->sub, n, listid + 1, | |
3305 &match); | |
3306 break; | |
3307 | |
3308 /* | |
3309 * Character classes like \a for alpha, \d for digit etc. | |
3310 */ | |
3311 case NFA_IDENT: /* \i */ | |
3312 result = vim_isIDc(c); | |
3313 ADD_POS_NEG_STATE(t->state); | |
3314 break; | |
3315 | |
3316 case NFA_SIDENT: /* \I */ | |
3317 result = !VIM_ISDIGIT(c) && vim_isIDc(c); | |
3318 ADD_POS_NEG_STATE(t->state); | |
3319 break; | |
3320 | |
3321 case NFA_KWORD: /* \k */ | |
4503 | 3322 result = vim_iswordp_buf(reginput, reg_buf); |
4444 | 3323 ADD_POS_NEG_STATE(t->state); |
3324 break; | |
3325 | |
3326 case NFA_SKWORD: /* \K */ | |
4503 | 3327 result = !VIM_ISDIGIT(c) && vim_iswordp_buf(reginput, reg_buf); |
4444 | 3328 ADD_POS_NEG_STATE(t->state); |
3329 break; | |
3330 | |
3331 case NFA_FNAME: /* \f */ | |
3332 result = vim_isfilec(c); | |
3333 ADD_POS_NEG_STATE(t->state); | |
3334 break; | |
3335 | |
3336 case NFA_SFNAME: /* \F */ | |
3337 result = !VIM_ISDIGIT(c) && vim_isfilec(c); | |
3338 ADD_POS_NEG_STATE(t->state); | |
3339 break; | |
3340 | |
3341 case NFA_PRINT: /* \p */ | |
4470 | 3342 result = ptr2cells(reginput) == 1; |
4444 | 3343 ADD_POS_NEG_STATE(t->state); |
3344 break; | |
3345 | |
3346 case NFA_SPRINT: /* \P */ | |
4470 | 3347 result = !VIM_ISDIGIT(c) && ptr2cells(reginput) == 1; |
4444 | 3348 ADD_POS_NEG_STATE(t->state); |
3349 break; | |
3350 | |
3351 case NFA_WHITE: /* \s */ | |
3352 result = vim_iswhite(c); | |
3353 ADD_POS_NEG_STATE(t->state); | |
3354 break; | |
3355 | |
3356 case NFA_NWHITE: /* \S */ | |
3357 result = c != NUL && !vim_iswhite(c); | |
3358 ADD_POS_NEG_STATE(t->state); | |
3359 break; | |
3360 | |
3361 case NFA_DIGIT: /* \d */ | |
3362 result = ri_digit(c); | |
3363 ADD_POS_NEG_STATE(t->state); | |
3364 break; | |
3365 | |
3366 case NFA_NDIGIT: /* \D */ | |
3367 result = c != NUL && !ri_digit(c); | |
3368 ADD_POS_NEG_STATE(t->state); | |
3369 break; | |
3370 | |
3371 case NFA_HEX: /* \x */ | |
3372 result = ri_hex(c); | |
3373 ADD_POS_NEG_STATE(t->state); | |
3374 break; | |
3375 | |
3376 case NFA_NHEX: /* \X */ | |
3377 result = c != NUL && !ri_hex(c); | |
3378 ADD_POS_NEG_STATE(t->state); | |
3379 break; | |
3380 | |
3381 case NFA_OCTAL: /* \o */ | |
3382 result = ri_octal(c); | |
3383 ADD_POS_NEG_STATE(t->state); | |
3384 break; | |
3385 | |
3386 case NFA_NOCTAL: /* \O */ | |
3387 result = c != NUL && !ri_octal(c); | |
3388 ADD_POS_NEG_STATE(t->state); | |
3389 break; | |
3390 | |
3391 case NFA_WORD: /* \w */ | |
3392 result = ri_word(c); | |
3393 ADD_POS_NEG_STATE(t->state); | |
3394 break; | |
3395 | |
3396 case NFA_NWORD: /* \W */ | |
3397 result = c != NUL && !ri_word(c); | |
3398 ADD_POS_NEG_STATE(t->state); | |
3399 break; | |
3400 | |
3401 case NFA_HEAD: /* \h */ | |
3402 result = ri_head(c); | |
3403 ADD_POS_NEG_STATE(t->state); | |
3404 break; | |
3405 | |
3406 case NFA_NHEAD: /* \H */ | |
3407 result = c != NUL && !ri_head(c); | |
3408 ADD_POS_NEG_STATE(t->state); | |
3409 break; | |
3410 | |
3411 case NFA_ALPHA: /* \a */ | |
3412 result = ri_alpha(c); | |
3413 ADD_POS_NEG_STATE(t->state); | |
3414 break; | |
3415 | |
3416 case NFA_NALPHA: /* \A */ | |
3417 result = c != NUL && !ri_alpha(c); | |
3418 ADD_POS_NEG_STATE(t->state); | |
3419 break; | |
3420 | |
3421 case NFA_LOWER: /* \l */ | |
3422 result = ri_lower(c); | |
3423 ADD_POS_NEG_STATE(t->state); | |
3424 break; | |
3425 | |
3426 case NFA_NLOWER: /* \L */ | |
3427 result = c != NUL && !ri_lower(c); | |
3428 ADD_POS_NEG_STATE(t->state); | |
3429 break; | |
3430 | |
3431 case NFA_UPPER: /* \u */ | |
3432 result = ri_upper(c); | |
3433 ADD_POS_NEG_STATE(t->state); | |
3434 break; | |
3435 | |
3436 case NFA_NUPPER: /* \U */ | |
3437 result = c != NUL && !ri_upper(c); | |
3438 ADD_POS_NEG_STATE(t->state); | |
3439 break; | |
3440 | |
4482 | 3441 case NFA_MOPEN + 0: |
3442 case NFA_MOPEN + 1: | |
3443 case NFA_MOPEN + 2: | |
3444 case NFA_MOPEN + 3: | |
3445 case NFA_MOPEN + 4: | |
3446 case NFA_MOPEN + 5: | |
3447 case NFA_MOPEN + 6: | |
3448 case NFA_MOPEN + 7: | |
3449 case NFA_MOPEN + 8: | |
3450 case NFA_MOPEN + 9: | |
3451 /* handled below */ | |
3452 break; | |
3453 | |
3454 case NFA_SKIP_CHAR: | |
3455 case NFA_ZSTART: | |
3456 /* TODO: should not happen? */ | |
3457 break; | |
3458 | |
4444 | 3459 default: /* regular character */ |
4482 | 3460 /* TODO: put this in #ifdef later */ |
3461 if (t->state->c < -256) | |
3462 EMSGN("INTERNAL: Negative state char: %ld", t->state->c); | |
4444 | 3463 result = (no_Magic(t->state->c) == c); |
4482 | 3464 |
4444 | 3465 if (!result) |
3466 result = ireg_ic == TRUE | |
3467 && MB_TOLOWER(t->state->c) == MB_TOLOWER(c); | |
4527
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3468 #ifdef FEAT_MBYTE |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3469 /* If there is a composing character which is not being |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3470 * ignored there can be no match. Match with composing |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3471 * character uses NFA_COMPOSING above. */ |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3472 if (result && enc_utf8 && !ireg_icombine |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3473 && n != utf_char2len(c)) |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3474 result = FALSE; |
55bcaa1d2749
updated for version 7.3.1011
Bram Moolenaar <bram@vim.org>
parents:
4525
diff
changeset
|
3475 #endif |
4444 | 3476 ADD_POS_NEG_STATE(t->state); |
3477 break; | |
3478 } | |
3479 | |
3480 } /* for (thislist = thislist; thislist->state; thislist++) */ | |
3481 | |
3482 /* The first found match is the leftmost one, but there may be a | |
3483 * longer one. Keep running the NFA, but don't start from the | |
3484 * beginning. Also, do not add the start state in recursive calls of | |
3485 * nfa_regmatch(), because recursive calls should only start in the | |
3486 * first position. */ | |
3487 if (match == FALSE && start->c == NFA_MOPEN + 0) | |
3488 { | |
3489 #ifdef ENABLE_LOG | |
3490 fprintf(log_fd, "(---) STARTSTATE\n"); | |
3491 #endif | |
3492 addstate(nextlist, start, m, n, listid + 1, &match); | |
3493 } | |
3494 | |
3495 #ifdef ENABLE_LOG | |
3496 fprintf(log_fd, ">>> Thislist had %d states available: ", thislist->n); | |
3497 for (i = 0; i< thislist->n; i++) | |
3498 fprintf(log_fd, "%d ", abs(thislist->t[i].state->id)); | |
3499 fprintf(log_fd, "\n"); | |
3500 #endif | |
3501 | |
3502 nextchar: | |
4515
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3503 /* 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
|
3504 * finish. */ |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3505 if (n != 0) |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3506 reginput += n; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3507 else if (go_to_nextline) |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3508 reg_nextline(); |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3509 else |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3510 break; |
90e9917d4114
updated for version 7.3.1005
Bram Moolenaar <bram@vim.org>
parents:
4507
diff
changeset
|
3511 } |
4444 | 3512 |
3513 #ifdef ENABLE_LOG | |
3514 if (log_fd != stderr) | |
3515 fclose(log_fd); | |
3516 log_fd = NULL; | |
3517 #endif | |
3518 | |
3519 theend: | |
3520 /* Free memory */ | |
3521 vim_free(list[0].t); | |
3522 vim_free(list[1].t); | |
3523 vim_free(list[2].t); | |
3524 list[0].t = list[1].t = list[2].t = NULL; | |
3525 if (listids != NULL) | |
3526 vim_free(listids); | |
3527 #undef ADD_POS_NEG_STATE | |
4460 | 3528 #ifdef NFA_REGEXP_DEBUG_LOG |
4444 | 3529 fclose(debug); |
3530 #endif | |
3531 | |
3532 return match; | |
3533 } | |
3534 | |
3535 /* | |
3536 * Try match of "prog" with at regline["col"]. | |
3537 * Returns 0 for failure, number of lines contained in the match otherwise. | |
3538 */ | |
3539 static long | |
3540 nfa_regtry(start, col) | |
3541 nfa_state_T *start; | |
3542 colnr_T col; | |
3543 { | |
3544 int i; | |
3545 regsub_T sub, m; | |
3546 #ifdef ENABLE_LOG | |
3547 FILE *f; | |
3548 #endif | |
3549 | |
3550 reginput = regline + col; | |
3551 need_clear_subexpr = TRUE; | |
3552 | |
3553 #ifdef ENABLE_LOG | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
3554 f = fopen(NFA_REGEXP_RUN_LOG, "a"); |
4444 | 3555 if (f != NULL) |
3556 { | |
3557 fprintf(f, "\n\n\n\n\n\n\t\t=======================================================\n"); | |
3558 fprintf(f, " =======================================================\n"); | |
3559 #ifdef DEBUG | |
3560 fprintf(f, "\tRegexp is \"%s\"\n", nfa_regengine.expr); | |
3561 #endif | |
3562 fprintf(f, "\tInput text is \"%s\" \n", reginput); | |
3563 fprintf(f, " =======================================================\n\n\n\n\n\n\n"); | |
4533
6a2005efa02b
updated for version 7.3.1014
Bram Moolenaar <bram@vim.org>
parents:
4531
diff
changeset
|
3564 nfa_print_state(f, start); |
4444 | 3565 fprintf(f, "\n\n"); |
3566 fclose(f); | |
3567 } | |
3568 else | |
3569 EMSG(_("Could not open temporary log file for writing ")); | |
3570 #endif | |
3571 | |
3572 if (REG_MULTI) | |
3573 { | |
3574 /* Use 0xff to set lnum to -1 */ | |
3575 vim_memset(sub.startpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
3576 vim_memset(sub.endpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
3577 vim_memset(m.startpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
3578 vim_memset(m.endpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
3579 } | |
3580 else | |
3581 { | |
3582 vim_memset(sub.start, 0, sizeof(char_u *) * NSUBEXP); | |
3583 vim_memset(sub.end, 0, sizeof(char_u *) * NSUBEXP); | |
3584 vim_memset(m.start, 0, sizeof(char_u *) * NSUBEXP); | |
3585 vim_memset(m.end, 0, sizeof(char_u *) * NSUBEXP); | |
3586 } | |
3587 | |
3588 if (nfa_regmatch(start, &sub, &m) == FALSE) | |
3589 return 0; | |
3590 | |
3591 cleanup_subexpr(); | |
3592 if (REG_MULTI) | |
3593 { | |
3594 for (i = 0; i < NSUBEXP; i++) | |
3595 { | |
3596 reg_startpos[i] = sub.startpos[i]; | |
3597 reg_endpos[i] = sub.endpos[i]; | |
3598 } | |
3599 | |
3600 if (reg_startpos[0].lnum < 0) | |
3601 { | |
3602 reg_startpos[0].lnum = 0; | |
3603 reg_startpos[0].col = col; | |
3604 } | |
3605 if (reg_endpos[0].lnum < 0) | |
3606 { | |
3607 reg_endpos[0].lnum = reglnum; | |
3608 reg_endpos[0].col = (int)(reginput - regline); | |
3609 } | |
3610 else | |
3611 /* Use line number of "\ze". */ | |
3612 reglnum = reg_endpos[0].lnum; | |
3613 } | |
3614 else | |
3615 { | |
3616 for (i = 0; i < NSUBEXP; i++) | |
3617 { | |
3618 reg_startp[i] = sub.start[i]; | |
3619 reg_endp[i] = sub.end[i]; | |
3620 } | |
3621 | |
3622 if (reg_startp[0] == NULL) | |
3623 reg_startp[0] = regline + col; | |
3624 if (reg_endp[0] == NULL) | |
3625 reg_endp[0] = reginput; | |
3626 } | |
3627 | |
3628 return 1 + reglnum; | |
3629 } | |
3630 | |
3631 /* | |
3632 * Match a regexp against a string ("line" points to the string) or multiple | |
3633 * lines ("line" is NULL, use reg_getline()). | |
3634 * | |
3635 * Returns 0 for failure, number of lines contained in the match otherwise. | |
3636 */ | |
3637 static long | |
3638 nfa_regexec_both(line, col) | |
3639 char_u *line; | |
3640 colnr_T col; /* column to start looking for match */ | |
3641 { | |
3642 nfa_regprog_T *prog; | |
3643 long retval = 0L; | |
3644 int i; | |
3645 | |
3646 if (REG_MULTI) | |
3647 { | |
3648 prog = (nfa_regprog_T *)reg_mmatch->regprog; | |
3649 line = reg_getline((linenr_T)0); /* relative to the cursor */ | |
3650 reg_startpos = reg_mmatch->startpos; | |
3651 reg_endpos = reg_mmatch->endpos; | |
3652 } | |
3653 else | |
3654 { | |
3655 prog = (nfa_regprog_T *)reg_match->regprog; | |
3656 reg_startp = reg_match->startp; | |
3657 reg_endp = reg_match->endp; | |
3658 } | |
3659 | |
3660 /* Be paranoid... */ | |
3661 if (prog == NULL || line == NULL) | |
3662 { | |
3663 EMSG(_(e_null)); | |
3664 goto theend; | |
3665 } | |
3666 | |
3667 /* If the start column is past the maximum column: no need to try. */ | |
3668 if (ireg_maxcol > 0 && col >= ireg_maxcol) | |
3669 goto theend; | |
3670 | |
3671 /* If pattern contains "\c" or "\C": overrule value of ireg_ic */ | |
3672 if (prog->regflags & RF_ICASE) | |
3673 ireg_ic = TRUE; | |
3674 else if (prog->regflags & RF_NOICASE) | |
3675 ireg_ic = FALSE; | |
3676 | |
3677 #ifdef FEAT_MBYTE | |
3678 /* If pattern contains "\Z" overrule value of ireg_icombine */ | |
3679 if (prog->regflags & RF_ICOMBINE) | |
3680 ireg_icombine = TRUE; | |
3681 #endif | |
3682 | |
3683 regline = line; | |
3684 reglnum = 0; /* relative to line */ | |
3685 | |
3686 nstate = prog->nstate; | |
3687 | |
3688 for (i = 0; i < nstate; ++i) | |
3689 { | |
3690 prog->state[i].id = i; | |
3691 prog->state[i].lastlist = 0; | |
3692 prog->state[i].visits = 0; | |
3693 } | |
3694 | |
3695 retval = nfa_regtry(prog->start, col); | |
3696 | |
3697 theend: | |
3698 return retval; | |
3699 } | |
3700 | |
3701 /* | |
3702 * Compile a regular expression into internal code for the NFA matcher. | |
3703 * Returns the program in allocated space. Returns NULL for an error. | |
3704 */ | |
3705 static regprog_T * | |
3706 nfa_regcomp(expr, re_flags) | |
3707 char_u *expr; | |
3708 int re_flags; | |
3709 { | |
3710 nfa_regprog_T *prog; | |
4458 | 3711 size_t prog_size; |
4444 | 3712 int *postfix; |
3713 | |
3714 if (expr == NULL) | |
3715 return NULL; | |
3716 | |
3717 #ifdef DEBUG | |
3718 nfa_regengine.expr = expr; | |
3719 #endif | |
3720 | |
3721 init_class_tab(); | |
3722 | |
3723 if (nfa_regcomp_start(expr, re_flags) == FAIL) | |
3724 return NULL; | |
3725 | |
3726 /* Space for compiled regexp */ | |
3727 prog_size = sizeof(nfa_regprog_T) + sizeof(nfa_state_T) * nstate_max; | |
3728 prog = (nfa_regprog_T *)lalloc(prog_size, TRUE); | |
3729 if (prog == NULL) | |
3730 goto fail; | |
3731 vim_memset(prog, 0, prog_size); | |
3732 | |
3733 /* Build postfix form of the regexp. Needed to build the NFA | |
3734 * (and count its size) */ | |
3735 postfix = re2post(); | |
3736 if (postfix == NULL) | |
3737 goto fail; /* Cascaded (syntax?) error */ | |
3738 | |
3739 /* | |
3740 * In order to build the NFA, we parse the input regexp twice: | |
3741 * 1. first pass to count size (so we can allocate space) | |
3742 * 2. second to emit code | |
3743 */ | |
3744 #ifdef ENABLE_LOG | |
3745 { | |
4531
1be43c095aff
updated for version 7.3.1013
Bram Moolenaar <bram@vim.org>
parents:
4529
diff
changeset
|
3746 FILE *f = fopen(NFA_REGEXP_RUN_LOG, "a"); |
4444 | 3747 |
3748 if (f != NULL) | |
3749 { | |
3750 fprintf(f, "\n*****************************\n\n\n\n\tCompiling regexp \"%s\" ... hold on !\n", expr); | |
3751 fclose(f); | |
3752 } | |
3753 } | |
3754 #endif | |
3755 | |
3756 /* | |
3757 * PASS 1 | |
3758 * Count number of NFA states in "nstate". Do not build the NFA. | |
3759 */ | |
3760 post2nfa(postfix, post_ptr, TRUE); | |
3761 state_ptr = prog->state; | |
3762 | |
3763 /* | |
3764 * PASS 2 | |
3765 * Build the NFA | |
3766 */ | |
3767 prog->start = post2nfa(postfix, post_ptr, FALSE); | |
3768 if (prog->start == NULL) | |
3769 goto fail; | |
3770 | |
3771 prog->regflags = regflags; | |
3772 prog->engine = &nfa_regengine; | |
3773 prog->nstate = nstate; | |
3774 #ifdef ENABLE_LOG | |
3775 nfa_postfix_dump(expr, OK); | |
3776 nfa_dump(prog); | |
3777 #endif | |
3778 | |
3779 out: | |
3780 vim_free(post_start); | |
3781 post_start = post_ptr = post_end = NULL; | |
3782 state_ptr = NULL; | |
3783 return (regprog_T *)prog; | |
3784 | |
3785 fail: | |
3786 vim_free(prog); | |
3787 prog = NULL; | |
3788 #ifdef ENABLE_LOG | |
3789 nfa_postfix_dump(expr, FAIL); | |
3790 #endif | |
3791 #ifdef DEBUG | |
3792 nfa_regengine.expr = NULL; | |
3793 #endif | |
3794 goto out; | |
3795 } | |
3796 | |
3797 | |
3798 /* | |
3799 * Match a regexp against a string. | |
3800 * "rmp->regprog" is a compiled regexp as returned by nfa_regcomp(). | |
3801 * Uses curbuf for line count and 'iskeyword'. | |
3802 * | |
3803 * Return TRUE if there is a match, FALSE if not. | |
3804 */ | |
3805 static int | |
3806 nfa_regexec(rmp, line, col) | |
3807 regmatch_T *rmp; | |
3808 char_u *line; /* string to match against */ | |
3809 colnr_T col; /* column to start looking for match */ | |
3810 { | |
3811 reg_match = rmp; | |
3812 reg_mmatch = NULL; | |
3813 reg_maxline = 0; | |
3814 reg_line_lbr = FALSE; | |
3815 reg_buf = curbuf; | |
3816 reg_win = NULL; | |
3817 ireg_ic = rmp->rm_ic; | |
3818 #ifdef FEAT_MBYTE | |
3819 ireg_icombine = FALSE; | |
3820 #endif | |
3821 ireg_maxcol = 0; | |
3822 return (nfa_regexec_both(line, col) != 0); | |
3823 } | |
3824 | |
3825 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ | |
3826 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
3827 | |
3828 static int nfa_regexec_nl __ARGS((regmatch_T *rmp, char_u *line, colnr_T col)); | |
3829 | |
3830 /* | |
3831 * Like nfa_regexec(), but consider a "\n" in "line" to be a line break. | |
3832 */ | |
3833 static int | |
3834 nfa_regexec_nl(rmp, line, col) | |
3835 regmatch_T *rmp; | |
3836 char_u *line; /* string to match against */ | |
3837 colnr_T col; /* column to start looking for match */ | |
3838 { | |
3839 reg_match = rmp; | |
3840 reg_mmatch = NULL; | |
3841 reg_maxline = 0; | |
3842 reg_line_lbr = TRUE; | |
3843 reg_buf = curbuf; | |
3844 reg_win = NULL; | |
3845 ireg_ic = rmp->rm_ic; | |
3846 #ifdef FEAT_MBYTE | |
3847 ireg_icombine = FALSE; | |
3848 #endif | |
3849 ireg_maxcol = 0; | |
3850 return (nfa_regexec_both(line, col) != 0); | |
3851 } | |
3852 #endif | |
3853 | |
3854 | |
3855 /* | |
3856 * Match a regexp against multiple lines. | |
3857 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
3858 * Uses curbuf for line count and 'iskeyword'. | |
3859 * | |
3860 * Return zero if there is no match. Return number of lines contained in the | |
3861 * match otherwise. | |
3862 * | |
3863 * Note: the body is the same as bt_regexec() except for nfa_regexec_both() | |
3864 * | |
3865 * ! Also NOTE : match may actually be in another line. e.g.: | |
3866 * when r.e. is \nc, cursor is at 'a' and the text buffer looks like | |
3867 * | |
3868 * +-------------------------+ | |
3869 * |a | | |
3870 * |b | | |
3871 * |c | | |
3872 * | | | |
3873 * +-------------------------+ | |
3874 * | |
3875 * then nfa_regexec_multi() returns 3. while the original | |
3876 * vim_regexec_multi() returns 0 and a second call at line 2 will return 2. | |
3877 * | |
3878 * FIXME if this behavior is not compatible. | |
3879 */ | |
3880 static long | |
3881 nfa_regexec_multi(rmp, win, buf, lnum, col, tm) | |
3882 regmmatch_T *rmp; | |
3883 win_T *win; /* window in which to search or NULL */ | |
3884 buf_T *buf; /* buffer in which to search */ | |
3885 linenr_T lnum; /* nr of line to start looking for match */ | |
3886 colnr_T col; /* column to start looking for match */ | |
3887 proftime_T *tm UNUSED; /* timeout limit or NULL */ | |
3888 { | |
3889 reg_match = NULL; | |
3890 reg_mmatch = rmp; | |
3891 reg_buf = buf; | |
3892 reg_win = win; | |
3893 reg_firstlnum = lnum; | |
3894 reg_maxline = reg_buf->b_ml.ml_line_count - lnum; | |
3895 reg_line_lbr = FALSE; | |
3896 ireg_ic = rmp->rmm_ic; | |
3897 #ifdef FEAT_MBYTE | |
3898 ireg_icombine = FALSE; | |
3899 #endif | |
3900 ireg_maxcol = rmp->rmm_maxcol; | |
3901 | |
4503 | 3902 return nfa_regexec_both(NULL, col); |
4444 | 3903 } |
3904 | |
3905 #ifdef DEBUG | |
3906 # undef ENABLE_LOG | |
3907 #endif |