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