]>
Commit | Line | Data |
---|---|---|
1 | #include "cache.h" | |
2 | #include "grep.h" | |
3 | #include "userdiff.h" | |
4 | #include "xdiff-interface.h" | |
5 | ||
6 | static struct grep_pat *create_grep_pat(const char *pat, size_t patlen, | |
7 | const char *origin, int no, | |
8 | enum grep_pat_token t, | |
9 | enum grep_header_field field) | |
10 | { | |
11 | struct grep_pat *p = xcalloc(1, sizeof(*p)); | |
12 | p->pattern = pat; | |
13 | p->patternlen = patlen; | |
14 | p->origin = origin; | |
15 | p->no = no; | |
16 | p->token = t; | |
17 | p->field = field; | |
18 | return p; | |
19 | } | |
20 | ||
21 | void append_header_grep_pattern(struct grep_opt *opt, | |
22 | enum grep_header_field field, const char *pat) | |
23 | { | |
24 | struct grep_pat *p = create_grep_pat(pat, strlen(pat), "header", 0, | |
25 | GREP_PATTERN_HEAD, field); | |
26 | *opt->header_tail = p; | |
27 | opt->header_tail = &p->next; | |
28 | p->next = NULL; | |
29 | } | |
30 | ||
31 | void append_grep_pattern(struct grep_opt *opt, const char *pat, | |
32 | const char *origin, int no, enum grep_pat_token t) | |
33 | { | |
34 | append_grep_pat(opt, pat, strlen(pat), origin, no, t); | |
35 | } | |
36 | ||
37 | void append_grep_pat(struct grep_opt *opt, const char *pat, size_t patlen, | |
38 | const char *origin, int no, enum grep_pat_token t) | |
39 | { | |
40 | struct grep_pat *p = create_grep_pat(pat, patlen, origin, no, t, 0); | |
41 | *opt->pattern_tail = p; | |
42 | opt->pattern_tail = &p->next; | |
43 | p->next = NULL; | |
44 | } | |
45 | ||
46 | struct grep_opt *grep_opt_dup(const struct grep_opt *opt) | |
47 | { | |
48 | struct grep_pat *pat; | |
49 | struct grep_opt *ret = xmalloc(sizeof(struct grep_opt)); | |
50 | *ret = *opt; | |
51 | ||
52 | ret->pattern_list = NULL; | |
53 | ret->pattern_tail = &ret->pattern_list; | |
54 | ||
55 | for(pat = opt->pattern_list; pat != NULL; pat = pat->next) | |
56 | { | |
57 | if(pat->token == GREP_PATTERN_HEAD) | |
58 | append_header_grep_pattern(ret, pat->field, | |
59 | pat->pattern); | |
60 | else | |
61 | append_grep_pat(ret, pat->pattern, pat->patternlen, | |
62 | pat->origin, pat->no, pat->token); | |
63 | } | |
64 | ||
65 | return ret; | |
66 | } | |
67 | ||
68 | static NORETURN void compile_regexp_failed(const struct grep_pat *p, | |
69 | const char *error) | |
70 | { | |
71 | char where[1024]; | |
72 | ||
73 | if (p->no) | |
74 | sprintf(where, "In '%s' at %d, ", p->origin, p->no); | |
75 | else if (p->origin) | |
76 | sprintf(where, "%s, ", p->origin); | |
77 | else | |
78 | where[0] = 0; | |
79 | ||
80 | die("%s'%s': %s", where, p->pattern, error); | |
81 | } | |
82 | ||
83 | #ifdef USE_LIBPCRE | |
84 | static void compile_pcre_regexp(struct grep_pat *p, const struct grep_opt *opt) | |
85 | { | |
86 | const char *error; | |
87 | int erroffset; | |
88 | int options = 0; | |
89 | ||
90 | if (opt->ignore_case) | |
91 | options |= PCRE_CASELESS; | |
92 | ||
93 | p->pcre_regexp = pcre_compile(p->pattern, options, &error, &erroffset, | |
94 | NULL); | |
95 | if (!p->pcre_regexp) | |
96 | compile_regexp_failed(p, error); | |
97 | ||
98 | p->pcre_extra_info = pcre_study(p->pcre_regexp, 0, &error); | |
99 | if (!p->pcre_extra_info && error) | |
100 | die("%s", error); | |
101 | } | |
102 | ||
103 | static int pcrematch(struct grep_pat *p, const char *line, const char *eol, | |
104 | regmatch_t *match, int eflags) | |
105 | { | |
106 | int ovector[30], ret, flags = 0; | |
107 | ||
108 | if (eflags & REG_NOTBOL) | |
109 | flags |= PCRE_NOTBOL; | |
110 | ||
111 | ret = pcre_exec(p->pcre_regexp, p->pcre_extra_info, line, eol - line, | |
112 | 0, flags, ovector, ARRAY_SIZE(ovector)); | |
113 | if (ret < 0 && ret != PCRE_ERROR_NOMATCH) | |
114 | die("pcre_exec failed with error code %d", ret); | |
115 | if (ret > 0) { | |
116 | ret = 0; | |
117 | match->rm_so = ovector[0]; | |
118 | match->rm_eo = ovector[1]; | |
119 | } | |
120 | ||
121 | return ret; | |
122 | } | |
123 | ||
124 | static void free_pcre_regexp(struct grep_pat *p) | |
125 | { | |
126 | pcre_free(p->pcre_regexp); | |
127 | pcre_free(p->pcre_extra_info); | |
128 | } | |
129 | #else /* !USE_LIBPCRE */ | |
130 | static void compile_pcre_regexp(struct grep_pat *p, const struct grep_opt *opt) | |
131 | { | |
132 | die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE"); | |
133 | } | |
134 | ||
135 | static int pcrematch(struct grep_pat *p, const char *line, const char *eol, | |
136 | regmatch_t *match, int eflags) | |
137 | { | |
138 | return 1; | |
139 | } | |
140 | ||
141 | static void free_pcre_regexp(struct grep_pat *p) | |
142 | { | |
143 | } | |
144 | #endif /* !USE_LIBPCRE */ | |
145 | ||
146 | static void compile_regexp(struct grep_pat *p, struct grep_opt *opt) | |
147 | { | |
148 | int err; | |
149 | ||
150 | p->word_regexp = opt->word_regexp; | |
151 | p->ignore_case = opt->ignore_case; | |
152 | p->fixed = opt->fixed; | |
153 | ||
154 | if (p->fixed) | |
155 | return; | |
156 | ||
157 | if (opt->pcre) { | |
158 | compile_pcre_regexp(p, opt); | |
159 | return; | |
160 | } | |
161 | ||
162 | err = regcomp(&p->regexp, p->pattern, opt->regflags); | |
163 | if (err) { | |
164 | char errbuf[1024]; | |
165 | regerror(err, &p->regexp, errbuf, 1024); | |
166 | regfree(&p->regexp); | |
167 | compile_regexp_failed(p, errbuf); | |
168 | } | |
169 | } | |
170 | ||
171 | static struct grep_expr *compile_pattern_or(struct grep_pat **); | |
172 | static struct grep_expr *compile_pattern_atom(struct grep_pat **list) | |
173 | { | |
174 | struct grep_pat *p; | |
175 | struct grep_expr *x; | |
176 | ||
177 | p = *list; | |
178 | if (!p) | |
179 | return NULL; | |
180 | switch (p->token) { | |
181 | case GREP_PATTERN: /* atom */ | |
182 | case GREP_PATTERN_HEAD: | |
183 | case GREP_PATTERN_BODY: | |
184 | x = xcalloc(1, sizeof (struct grep_expr)); | |
185 | x->node = GREP_NODE_ATOM; | |
186 | x->u.atom = p; | |
187 | *list = p->next; | |
188 | return x; | |
189 | case GREP_OPEN_PAREN: | |
190 | *list = p->next; | |
191 | x = compile_pattern_or(list); | |
192 | if (!*list || (*list)->token != GREP_CLOSE_PAREN) | |
193 | die("unmatched parenthesis"); | |
194 | *list = (*list)->next; | |
195 | return x; | |
196 | default: | |
197 | return NULL; | |
198 | } | |
199 | } | |
200 | ||
201 | static struct grep_expr *compile_pattern_not(struct grep_pat **list) | |
202 | { | |
203 | struct grep_pat *p; | |
204 | struct grep_expr *x; | |
205 | ||
206 | p = *list; | |
207 | if (!p) | |
208 | return NULL; | |
209 | switch (p->token) { | |
210 | case GREP_NOT: | |
211 | if (!p->next) | |
212 | die("--not not followed by pattern expression"); | |
213 | *list = p->next; | |
214 | x = xcalloc(1, sizeof (struct grep_expr)); | |
215 | x->node = GREP_NODE_NOT; | |
216 | x->u.unary = compile_pattern_not(list); | |
217 | if (!x->u.unary) | |
218 | die("--not followed by non pattern expression"); | |
219 | return x; | |
220 | default: | |
221 | return compile_pattern_atom(list); | |
222 | } | |
223 | } | |
224 | ||
225 | static struct grep_expr *compile_pattern_and(struct grep_pat **list) | |
226 | { | |
227 | struct grep_pat *p; | |
228 | struct grep_expr *x, *y, *z; | |
229 | ||
230 | x = compile_pattern_not(list); | |
231 | p = *list; | |
232 | if (p && p->token == GREP_AND) { | |
233 | if (!p->next) | |
234 | die("--and not followed by pattern expression"); | |
235 | *list = p->next; | |
236 | y = compile_pattern_and(list); | |
237 | if (!y) | |
238 | die("--and not followed by pattern expression"); | |
239 | z = xcalloc(1, sizeof (struct grep_expr)); | |
240 | z->node = GREP_NODE_AND; | |
241 | z->u.binary.left = x; | |
242 | z->u.binary.right = y; | |
243 | return z; | |
244 | } | |
245 | return x; | |
246 | } | |
247 | ||
248 | static struct grep_expr *compile_pattern_or(struct grep_pat **list) | |
249 | { | |
250 | struct grep_pat *p; | |
251 | struct grep_expr *x, *y, *z; | |
252 | ||
253 | x = compile_pattern_and(list); | |
254 | p = *list; | |
255 | if (x && p && p->token != GREP_CLOSE_PAREN) { | |
256 | y = compile_pattern_or(list); | |
257 | if (!y) | |
258 | die("not a pattern expression %s", p->pattern); | |
259 | z = xcalloc(1, sizeof (struct grep_expr)); | |
260 | z->node = GREP_NODE_OR; | |
261 | z->u.binary.left = x; | |
262 | z->u.binary.right = y; | |
263 | return z; | |
264 | } | |
265 | return x; | |
266 | } | |
267 | ||
268 | static struct grep_expr *compile_pattern_expr(struct grep_pat **list) | |
269 | { | |
270 | return compile_pattern_or(list); | |
271 | } | |
272 | ||
273 | static struct grep_expr *grep_true_expr(void) | |
274 | { | |
275 | struct grep_expr *z = xcalloc(1, sizeof(*z)); | |
276 | z->node = GREP_NODE_TRUE; | |
277 | return z; | |
278 | } | |
279 | ||
280 | static struct grep_expr *grep_or_expr(struct grep_expr *left, struct grep_expr *right) | |
281 | { | |
282 | struct grep_expr *z = xcalloc(1, sizeof(*z)); | |
283 | z->node = GREP_NODE_OR; | |
284 | z->u.binary.left = left; | |
285 | z->u.binary.right = right; | |
286 | return z; | |
287 | } | |
288 | ||
289 | static struct grep_expr *prep_header_patterns(struct grep_opt *opt) | |
290 | { | |
291 | struct grep_pat *p; | |
292 | struct grep_expr *header_expr; | |
293 | struct grep_expr *(header_group[GREP_HEADER_FIELD_MAX]); | |
294 | enum grep_header_field fld; | |
295 | ||
296 | if (!opt->header_list) | |
297 | return NULL; | |
298 | p = opt->header_list; | |
299 | for (p = opt->header_list; p; p = p->next) { | |
300 | if (p->token != GREP_PATTERN_HEAD) | |
301 | die("bug: a non-header pattern in grep header list."); | |
302 | if (p->field < 0 || GREP_HEADER_FIELD_MAX <= p->field) | |
303 | die("bug: unknown header field %d", p->field); | |
304 | compile_regexp(p, opt); | |
305 | } | |
306 | ||
307 | for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) | |
308 | header_group[fld] = NULL; | |
309 | ||
310 | for (p = opt->header_list; p; p = p->next) { | |
311 | struct grep_expr *h; | |
312 | struct grep_pat *pp = p; | |
313 | ||
314 | h = compile_pattern_atom(&pp); | |
315 | if (!h || pp != p->next) | |
316 | die("bug: malformed header expr"); | |
317 | if (!header_group[p->field]) { | |
318 | header_group[p->field] = h; | |
319 | continue; | |
320 | } | |
321 | header_group[p->field] = grep_or_expr(h, header_group[p->field]); | |
322 | } | |
323 | ||
324 | header_expr = NULL; | |
325 | ||
326 | for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) { | |
327 | if (!header_group[fld]) | |
328 | continue; | |
329 | if (!header_expr) | |
330 | header_expr = grep_true_expr(); | |
331 | header_expr = grep_or_expr(header_group[fld], header_expr); | |
332 | } | |
333 | return header_expr; | |
334 | } | |
335 | ||
336 | void compile_grep_patterns(struct grep_opt *opt) | |
337 | { | |
338 | struct grep_pat *p; | |
339 | struct grep_expr *header_expr = prep_header_patterns(opt); | |
340 | ||
341 | for (p = opt->pattern_list; p; p = p->next) { | |
342 | switch (p->token) { | |
343 | case GREP_PATTERN: /* atom */ | |
344 | case GREP_PATTERN_HEAD: | |
345 | case GREP_PATTERN_BODY: | |
346 | compile_regexp(p, opt); | |
347 | break; | |
348 | default: | |
349 | opt->extended = 1; | |
350 | break; | |
351 | } | |
352 | } | |
353 | ||
354 | if (opt->all_match || header_expr) | |
355 | opt->extended = 1; | |
356 | else if (!opt->extended) | |
357 | return; | |
358 | ||
359 | p = opt->pattern_list; | |
360 | if (p) | |
361 | opt->pattern_expression = compile_pattern_expr(&p); | |
362 | if (p) | |
363 | die("incomplete pattern expression: %s", p->pattern); | |
364 | ||
365 | if (!header_expr) | |
366 | return; | |
367 | ||
368 | if (!opt->pattern_expression) | |
369 | opt->pattern_expression = header_expr; | |
370 | else | |
371 | opt->pattern_expression = grep_or_expr(opt->pattern_expression, | |
372 | header_expr); | |
373 | opt->all_match = 1; | |
374 | } | |
375 | ||
376 | static void free_pattern_expr(struct grep_expr *x) | |
377 | { | |
378 | switch (x->node) { | |
379 | case GREP_NODE_TRUE: | |
380 | case GREP_NODE_ATOM: | |
381 | break; | |
382 | case GREP_NODE_NOT: | |
383 | free_pattern_expr(x->u.unary); | |
384 | break; | |
385 | case GREP_NODE_AND: | |
386 | case GREP_NODE_OR: | |
387 | free_pattern_expr(x->u.binary.left); | |
388 | free_pattern_expr(x->u.binary.right); | |
389 | break; | |
390 | } | |
391 | free(x); | |
392 | } | |
393 | ||
394 | void free_grep_patterns(struct grep_opt *opt) | |
395 | { | |
396 | struct grep_pat *p, *n; | |
397 | ||
398 | for (p = opt->pattern_list; p; p = n) { | |
399 | n = p->next; | |
400 | switch (p->token) { | |
401 | case GREP_PATTERN: /* atom */ | |
402 | case GREP_PATTERN_HEAD: | |
403 | case GREP_PATTERN_BODY: | |
404 | if (p->pcre_regexp) | |
405 | free_pcre_regexp(p); | |
406 | else | |
407 | regfree(&p->regexp); | |
408 | break; | |
409 | default: | |
410 | break; | |
411 | } | |
412 | free(p); | |
413 | } | |
414 | ||
415 | if (!opt->extended) | |
416 | return; | |
417 | free_pattern_expr(opt->pattern_expression); | |
418 | } | |
419 | ||
420 | static char *end_of_line(char *cp, unsigned long *left) | |
421 | { | |
422 | unsigned long l = *left; | |
423 | while (l && *cp != '\n') { | |
424 | l--; | |
425 | cp++; | |
426 | } | |
427 | *left = l; | |
428 | return cp; | |
429 | } | |
430 | ||
431 | static int word_char(char ch) | |
432 | { | |
433 | return isalnum(ch) || ch == '_'; | |
434 | } | |
435 | ||
436 | static void output_color(struct grep_opt *opt, const void *data, size_t size, | |
437 | const char *color) | |
438 | { | |
439 | if (opt->color && color && color[0]) { | |
440 | opt->output(opt, color, strlen(color)); | |
441 | opt->output(opt, data, size); | |
442 | opt->output(opt, GIT_COLOR_RESET, strlen(GIT_COLOR_RESET)); | |
443 | } else | |
444 | opt->output(opt, data, size); | |
445 | } | |
446 | ||
447 | static void output_sep(struct grep_opt *opt, char sign) | |
448 | { | |
449 | if (opt->null_following_name) | |
450 | opt->output(opt, "\0", 1); | |
451 | else | |
452 | output_color(opt, &sign, 1, opt->color_sep); | |
453 | } | |
454 | ||
455 | static void show_name(struct grep_opt *opt, const char *name) | |
456 | { | |
457 | output_color(opt, name, strlen(name), opt->color_filename); | |
458 | opt->output(opt, opt->null_following_name ? "\0" : "\n", 1); | |
459 | } | |
460 | ||
461 | static int fixmatch(struct grep_pat *p, char *line, char *eol, | |
462 | regmatch_t *match) | |
463 | { | |
464 | char *hit; | |
465 | ||
466 | if (p->ignore_case) { | |
467 | char *s = line; | |
468 | do { | |
469 | hit = strcasestr(s, p->pattern); | |
470 | if (hit) | |
471 | break; | |
472 | s += strlen(s) + 1; | |
473 | } while (s < eol); | |
474 | } else | |
475 | hit = memmem(line, eol - line, p->pattern, p->patternlen); | |
476 | ||
477 | if (!hit) { | |
478 | match->rm_so = match->rm_eo = -1; | |
479 | return REG_NOMATCH; | |
480 | } | |
481 | else { | |
482 | match->rm_so = hit - line; | |
483 | match->rm_eo = match->rm_so + p->patternlen; | |
484 | return 0; | |
485 | } | |
486 | } | |
487 | ||
488 | static int regmatch(const regex_t *preg, char *line, char *eol, | |
489 | regmatch_t *match, int eflags) | |
490 | { | |
491 | #ifdef REG_STARTEND | |
492 | match->rm_so = 0; | |
493 | match->rm_eo = eol - line; | |
494 | eflags |= REG_STARTEND; | |
495 | #endif | |
496 | return regexec(preg, line, 1, match, eflags); | |
497 | } | |
498 | ||
499 | static int patmatch(struct grep_pat *p, char *line, char *eol, | |
500 | regmatch_t *match, int eflags) | |
501 | { | |
502 | int hit; | |
503 | ||
504 | if (p->fixed) | |
505 | hit = !fixmatch(p, line, eol, match); | |
506 | else if (p->pcre_regexp) | |
507 | hit = !pcrematch(p, line, eol, match, eflags); | |
508 | else | |
509 | hit = !regmatch(&p->regexp, line, eol, match, eflags); | |
510 | ||
511 | return hit; | |
512 | } | |
513 | ||
514 | static int strip_timestamp(char *bol, char **eol_p) | |
515 | { | |
516 | char *eol = *eol_p; | |
517 | int ch; | |
518 | ||
519 | while (bol < --eol) { | |
520 | if (*eol != '>') | |
521 | continue; | |
522 | *eol_p = ++eol; | |
523 | ch = *eol; | |
524 | *eol = '\0'; | |
525 | return ch; | |
526 | } | |
527 | return 0; | |
528 | } | |
529 | ||
530 | static struct { | |
531 | const char *field; | |
532 | size_t len; | |
533 | } header_field[] = { | |
534 | { "author ", 7 }, | |
535 | { "committer ", 10 }, | |
536 | }; | |
537 | ||
538 | static int match_one_pattern(struct grep_pat *p, char *bol, char *eol, | |
539 | enum grep_context ctx, | |
540 | regmatch_t *pmatch, int eflags) | |
541 | { | |
542 | int hit = 0; | |
543 | int saved_ch = 0; | |
544 | const char *start = bol; | |
545 | ||
546 | if ((p->token != GREP_PATTERN) && | |
547 | ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD))) | |
548 | return 0; | |
549 | ||
550 | if (p->token == GREP_PATTERN_HEAD) { | |
551 | const char *field; | |
552 | size_t len; | |
553 | assert(p->field < ARRAY_SIZE(header_field)); | |
554 | field = header_field[p->field].field; | |
555 | len = header_field[p->field].len; | |
556 | if (strncmp(bol, field, len)) | |
557 | return 0; | |
558 | bol += len; | |
559 | saved_ch = strip_timestamp(bol, &eol); | |
560 | } | |
561 | ||
562 | again: | |
563 | hit = patmatch(p, bol, eol, pmatch, eflags); | |
564 | ||
565 | if (hit && p->word_regexp) { | |
566 | if ((pmatch[0].rm_so < 0) || | |
567 | (eol - bol) < pmatch[0].rm_so || | |
568 | (pmatch[0].rm_eo < 0) || | |
569 | (eol - bol) < pmatch[0].rm_eo) | |
570 | die("regexp returned nonsense"); | |
571 | ||
572 | /* Match beginning must be either beginning of the | |
573 | * line, or at word boundary (i.e. the last char must | |
574 | * not be a word char). Similarly, match end must be | |
575 | * either end of the line, or at word boundary | |
576 | * (i.e. the next char must not be a word char). | |
577 | */ | |
578 | if ( ((pmatch[0].rm_so == 0) || | |
579 | !word_char(bol[pmatch[0].rm_so-1])) && | |
580 | ((pmatch[0].rm_eo == (eol-bol)) || | |
581 | !word_char(bol[pmatch[0].rm_eo])) ) | |
582 | ; | |
583 | else | |
584 | hit = 0; | |
585 | ||
586 | /* Words consist of at least one character. */ | |
587 | if (pmatch->rm_so == pmatch->rm_eo) | |
588 | hit = 0; | |
589 | ||
590 | if (!hit && pmatch[0].rm_so + bol + 1 < eol) { | |
591 | /* There could be more than one match on the | |
592 | * line, and the first match might not be | |
593 | * strict word match. But later ones could be! | |
594 | * Forward to the next possible start, i.e. the | |
595 | * next position following a non-word char. | |
596 | */ | |
597 | bol = pmatch[0].rm_so + bol + 1; | |
598 | while (word_char(bol[-1]) && bol < eol) | |
599 | bol++; | |
600 | eflags |= REG_NOTBOL; | |
601 | if (bol < eol) | |
602 | goto again; | |
603 | } | |
604 | } | |
605 | if (p->token == GREP_PATTERN_HEAD && saved_ch) | |
606 | *eol = saved_ch; | |
607 | if (hit) { | |
608 | pmatch[0].rm_so += bol - start; | |
609 | pmatch[0].rm_eo += bol - start; | |
610 | } | |
611 | return hit; | |
612 | } | |
613 | ||
614 | static int match_expr_eval(struct grep_expr *x, char *bol, char *eol, | |
615 | enum grep_context ctx, int collect_hits) | |
616 | { | |
617 | int h = 0; | |
618 | regmatch_t match; | |
619 | ||
620 | if (!x) | |
621 | die("Not a valid grep expression"); | |
622 | switch (x->node) { | |
623 | case GREP_NODE_TRUE: | |
624 | h = 1; | |
625 | break; | |
626 | case GREP_NODE_ATOM: | |
627 | h = match_one_pattern(x->u.atom, bol, eol, ctx, &match, 0); | |
628 | break; | |
629 | case GREP_NODE_NOT: | |
630 | h = !match_expr_eval(x->u.unary, bol, eol, ctx, 0); | |
631 | break; | |
632 | case GREP_NODE_AND: | |
633 | if (!match_expr_eval(x->u.binary.left, bol, eol, ctx, 0)) | |
634 | return 0; | |
635 | h = match_expr_eval(x->u.binary.right, bol, eol, ctx, 0); | |
636 | break; | |
637 | case GREP_NODE_OR: | |
638 | if (!collect_hits) | |
639 | return (match_expr_eval(x->u.binary.left, | |
640 | bol, eol, ctx, 0) || | |
641 | match_expr_eval(x->u.binary.right, | |
642 | bol, eol, ctx, 0)); | |
643 | h = match_expr_eval(x->u.binary.left, bol, eol, ctx, 0); | |
644 | x->u.binary.left->hit |= h; | |
645 | h |= match_expr_eval(x->u.binary.right, bol, eol, ctx, 1); | |
646 | break; | |
647 | default: | |
648 | die("Unexpected node type (internal error) %d", x->node); | |
649 | } | |
650 | if (collect_hits) | |
651 | x->hit |= h; | |
652 | return h; | |
653 | } | |
654 | ||
655 | static int match_expr(struct grep_opt *opt, char *bol, char *eol, | |
656 | enum grep_context ctx, int collect_hits) | |
657 | { | |
658 | struct grep_expr *x = opt->pattern_expression; | |
659 | return match_expr_eval(x, bol, eol, ctx, collect_hits); | |
660 | } | |
661 | ||
662 | static int match_line(struct grep_opt *opt, char *bol, char *eol, | |
663 | enum grep_context ctx, int collect_hits) | |
664 | { | |
665 | struct grep_pat *p; | |
666 | regmatch_t match; | |
667 | ||
668 | if (opt->extended) | |
669 | return match_expr(opt, bol, eol, ctx, collect_hits); | |
670 | ||
671 | /* we do not call with collect_hits without being extended */ | |
672 | for (p = opt->pattern_list; p; p = p->next) { | |
673 | if (match_one_pattern(p, bol, eol, ctx, &match, 0)) | |
674 | return 1; | |
675 | } | |
676 | return 0; | |
677 | } | |
678 | ||
679 | static int match_next_pattern(struct grep_pat *p, char *bol, char *eol, | |
680 | enum grep_context ctx, | |
681 | regmatch_t *pmatch, int eflags) | |
682 | { | |
683 | regmatch_t match; | |
684 | ||
685 | if (!match_one_pattern(p, bol, eol, ctx, &match, eflags)) | |
686 | return 0; | |
687 | if (match.rm_so < 0 || match.rm_eo < 0) | |
688 | return 0; | |
689 | if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) { | |
690 | if (match.rm_so > pmatch->rm_so) | |
691 | return 1; | |
692 | if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo) | |
693 | return 1; | |
694 | } | |
695 | pmatch->rm_so = match.rm_so; | |
696 | pmatch->rm_eo = match.rm_eo; | |
697 | return 1; | |
698 | } | |
699 | ||
700 | static int next_match(struct grep_opt *opt, char *bol, char *eol, | |
701 | enum grep_context ctx, regmatch_t *pmatch, int eflags) | |
702 | { | |
703 | struct grep_pat *p; | |
704 | int hit = 0; | |
705 | ||
706 | pmatch->rm_so = pmatch->rm_eo = -1; | |
707 | if (bol < eol) { | |
708 | for (p = opt->pattern_list; p; p = p->next) { | |
709 | switch (p->token) { | |
710 | case GREP_PATTERN: /* atom */ | |
711 | case GREP_PATTERN_HEAD: | |
712 | case GREP_PATTERN_BODY: | |
713 | hit |= match_next_pattern(p, bol, eol, ctx, | |
714 | pmatch, eflags); | |
715 | break; | |
716 | default: | |
717 | break; | |
718 | } | |
719 | } | |
720 | } | |
721 | return hit; | |
722 | } | |
723 | ||
724 | static void show_line(struct grep_opt *opt, char *bol, char *eol, | |
725 | const char *name, unsigned lno, char sign) | |
726 | { | |
727 | int rest = eol - bol; | |
728 | char *line_color = NULL; | |
729 | ||
730 | if (opt->pre_context || opt->post_context) { | |
731 | if (opt->last_shown == 0) { | |
732 | if (opt->show_hunk_mark) { | |
733 | output_color(opt, "--", 2, opt->color_sep); | |
734 | opt->output(opt, "\n", 1); | |
735 | } | |
736 | } else if (lno > opt->last_shown + 1) { | |
737 | output_color(opt, "--", 2, opt->color_sep); | |
738 | opt->output(opt, "\n", 1); | |
739 | } | |
740 | } | |
741 | opt->last_shown = lno; | |
742 | ||
743 | if (opt->pathname) { | |
744 | output_color(opt, name, strlen(name), opt->color_filename); | |
745 | output_sep(opt, sign); | |
746 | } | |
747 | if (opt->linenum) { | |
748 | char buf[32]; | |
749 | snprintf(buf, sizeof(buf), "%d", lno); | |
750 | output_color(opt, buf, strlen(buf), opt->color_lineno); | |
751 | output_sep(opt, sign); | |
752 | } | |
753 | if (opt->color) { | |
754 | regmatch_t match; | |
755 | enum grep_context ctx = GREP_CONTEXT_BODY; | |
756 | int ch = *eol; | |
757 | int eflags = 0; | |
758 | ||
759 | if (sign == ':') | |
760 | line_color = opt->color_selected; | |
761 | else if (sign == '-') | |
762 | line_color = opt->color_context; | |
763 | else if (sign == '=') | |
764 | line_color = opt->color_function; | |
765 | *eol = '\0'; | |
766 | while (next_match(opt, bol, eol, ctx, &match, eflags)) { | |
767 | if (match.rm_so == match.rm_eo) | |
768 | break; | |
769 | ||
770 | output_color(opt, bol, match.rm_so, line_color); | |
771 | output_color(opt, bol + match.rm_so, | |
772 | match.rm_eo - match.rm_so, | |
773 | opt->color_match); | |
774 | bol += match.rm_eo; | |
775 | rest -= match.rm_eo; | |
776 | eflags = REG_NOTBOL; | |
777 | } | |
778 | *eol = ch; | |
779 | } | |
780 | output_color(opt, bol, rest, line_color); | |
781 | opt->output(opt, "\n", 1); | |
782 | } | |
783 | ||
784 | static int match_funcname(struct grep_opt *opt, char *bol, char *eol) | |
785 | { | |
786 | xdemitconf_t *xecfg = opt->priv; | |
787 | if (xecfg && xecfg->find_func) { | |
788 | char buf[1]; | |
789 | return xecfg->find_func(bol, eol - bol, buf, 1, | |
790 | xecfg->find_func_priv) >= 0; | |
791 | } | |
792 | ||
793 | if (bol == eol) | |
794 | return 0; | |
795 | if (isalpha(*bol) || *bol == '_' || *bol == '$') | |
796 | return 1; | |
797 | return 0; | |
798 | } | |
799 | ||
800 | static void show_funcname_line(struct grep_opt *opt, const char *name, | |
801 | char *buf, char *bol, unsigned lno) | |
802 | { | |
803 | while (bol > buf) { | |
804 | char *eol = --bol; | |
805 | ||
806 | while (bol > buf && bol[-1] != '\n') | |
807 | bol--; | |
808 | lno--; | |
809 | ||
810 | if (lno <= opt->last_shown) | |
811 | break; | |
812 | ||
813 | if (match_funcname(opt, bol, eol)) { | |
814 | show_line(opt, bol, eol, name, lno, '='); | |
815 | break; | |
816 | } | |
817 | } | |
818 | } | |
819 | ||
820 | static void show_pre_context(struct grep_opt *opt, const char *name, char *buf, | |
821 | char *bol, unsigned lno) | |
822 | { | |
823 | unsigned cur = lno, from = 1, funcname_lno = 0; | |
824 | int funcname_needed = opt->funcname; | |
825 | ||
826 | if (opt->pre_context < lno) | |
827 | from = lno - opt->pre_context; | |
828 | if (from <= opt->last_shown) | |
829 | from = opt->last_shown + 1; | |
830 | ||
831 | /* Rewind. */ | |
832 | while (bol > buf && cur > from) { | |
833 | char *eol = --bol; | |
834 | ||
835 | while (bol > buf && bol[-1] != '\n') | |
836 | bol--; | |
837 | cur--; | |
838 | if (funcname_needed && match_funcname(opt, bol, eol)) { | |
839 | funcname_lno = cur; | |
840 | funcname_needed = 0; | |
841 | } | |
842 | } | |
843 | ||
844 | /* We need to look even further back to find a function signature. */ | |
845 | if (opt->funcname && funcname_needed) | |
846 | show_funcname_line(opt, name, buf, bol, cur); | |
847 | ||
848 | /* Back forward. */ | |
849 | while (cur < lno) { | |
850 | char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-'; | |
851 | ||
852 | while (*eol != '\n') | |
853 | eol++; | |
854 | show_line(opt, bol, eol, name, cur, sign); | |
855 | bol = eol + 1; | |
856 | cur++; | |
857 | } | |
858 | } | |
859 | ||
860 | static int should_lookahead(struct grep_opt *opt) | |
861 | { | |
862 | struct grep_pat *p; | |
863 | ||
864 | if (opt->extended) | |
865 | return 0; /* punt for too complex stuff */ | |
866 | if (opt->invert) | |
867 | return 0; | |
868 | for (p = opt->pattern_list; p; p = p->next) { | |
869 | if (p->token != GREP_PATTERN) | |
870 | return 0; /* punt for "header only" and stuff */ | |
871 | } | |
872 | return 1; | |
873 | } | |
874 | ||
875 | static int look_ahead(struct grep_opt *opt, | |
876 | unsigned long *left_p, | |
877 | unsigned *lno_p, | |
878 | char **bol_p) | |
879 | { | |
880 | unsigned lno = *lno_p; | |
881 | char *bol = *bol_p; | |
882 | struct grep_pat *p; | |
883 | char *sp, *last_bol; | |
884 | regoff_t earliest = -1; | |
885 | ||
886 | for (p = opt->pattern_list; p; p = p->next) { | |
887 | int hit; | |
888 | regmatch_t m; | |
889 | ||
890 | hit = patmatch(p, bol, bol + *left_p, &m, 0); | |
891 | if (!hit || m.rm_so < 0 || m.rm_eo < 0) | |
892 | continue; | |
893 | if (earliest < 0 || m.rm_so < earliest) | |
894 | earliest = m.rm_so; | |
895 | } | |
896 | ||
897 | if (earliest < 0) { | |
898 | *bol_p = bol + *left_p; | |
899 | *left_p = 0; | |
900 | return 1; | |
901 | } | |
902 | for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--) | |
903 | ; /* find the beginning of the line */ | |
904 | last_bol = sp; | |
905 | ||
906 | for (sp = bol; sp < last_bol; sp++) { | |
907 | if (*sp == '\n') | |
908 | lno++; | |
909 | } | |
910 | *left_p -= last_bol - bol; | |
911 | *bol_p = last_bol; | |
912 | *lno_p = lno; | |
913 | return 0; | |
914 | } | |
915 | ||
916 | int grep_threads_ok(const struct grep_opt *opt) | |
917 | { | |
918 | /* If this condition is true, then we may use the attribute | |
919 | * machinery in grep_buffer_1. The attribute code is not | |
920 | * thread safe, so we disable the use of threads. | |
921 | */ | |
922 | if (opt->funcname && !opt->unmatch_name_only && !opt->status_only && | |
923 | !opt->name_only) | |
924 | return 0; | |
925 | ||
926 | return 1; | |
927 | } | |
928 | ||
929 | static void std_output(struct grep_opt *opt, const void *buf, size_t size) | |
930 | { | |
931 | fwrite(buf, size, 1, stdout); | |
932 | } | |
933 | ||
934 | static int grep_buffer_1(struct grep_opt *opt, const char *name, | |
935 | char *buf, unsigned long size, int collect_hits) | |
936 | { | |
937 | char *bol = buf; | |
938 | unsigned long left = size; | |
939 | unsigned lno = 1; | |
940 | unsigned last_hit = 0; | |
941 | int binary_match_only = 0; | |
942 | unsigned count = 0; | |
943 | int try_lookahead = 0; | |
944 | enum grep_context ctx = GREP_CONTEXT_HEAD; | |
945 | xdemitconf_t xecfg; | |
946 | ||
947 | if (!opt->output) | |
948 | opt->output = std_output; | |
949 | ||
950 | if (opt->last_shown && (opt->pre_context || opt->post_context) && | |
951 | opt->output == std_output) | |
952 | opt->show_hunk_mark = 1; | |
953 | opt->last_shown = 0; | |
954 | ||
955 | switch (opt->binary) { | |
956 | case GREP_BINARY_DEFAULT: | |
957 | if (buffer_is_binary(buf, size)) | |
958 | binary_match_only = 1; | |
959 | break; | |
960 | case GREP_BINARY_NOMATCH: | |
961 | if (buffer_is_binary(buf, size)) | |
962 | return 0; /* Assume unmatch */ | |
963 | break; | |
964 | case GREP_BINARY_TEXT: | |
965 | break; | |
966 | default: | |
967 | die("bug: unknown binary handling mode"); | |
968 | } | |
969 | ||
970 | memset(&xecfg, 0, sizeof(xecfg)); | |
971 | if (opt->funcname && !opt->unmatch_name_only && !opt->status_only && | |
972 | !opt->name_only && !binary_match_only && !collect_hits) { | |
973 | struct userdiff_driver *drv = userdiff_find_by_path(name); | |
974 | if (drv && drv->funcname.pattern) { | |
975 | const struct userdiff_funcname *pe = &drv->funcname; | |
976 | xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags); | |
977 | opt->priv = &xecfg; | |
978 | } | |
979 | } | |
980 | try_lookahead = should_lookahead(opt); | |
981 | ||
982 | while (left) { | |
983 | char *eol, ch; | |
984 | int hit; | |
985 | ||
986 | /* | |
987 | * look_ahead() skips quickly to the line that possibly | |
988 | * has the next hit; don't call it if we need to do | |
989 | * something more than just skipping the current line | |
990 | * in response to an unmatch for the current line. E.g. | |
991 | * inside a post-context window, we will show the current | |
992 | * line as a context around the previous hit when it | |
993 | * doesn't hit. | |
994 | */ | |
995 | if (try_lookahead | |
996 | && !(last_hit | |
997 | && lno <= last_hit + opt->post_context) | |
998 | && look_ahead(opt, &left, &lno, &bol)) | |
999 | break; | |
1000 | eol = end_of_line(bol, &left); | |
1001 | ch = *eol; | |
1002 | *eol = 0; | |
1003 | ||
1004 | if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol)) | |
1005 | ctx = GREP_CONTEXT_BODY; | |
1006 | ||
1007 | hit = match_line(opt, bol, eol, ctx, collect_hits); | |
1008 | *eol = ch; | |
1009 | ||
1010 | if (collect_hits) | |
1011 | goto next_line; | |
1012 | ||
1013 | /* "grep -v -e foo -e bla" should list lines | |
1014 | * that do not have either, so inversion should | |
1015 | * be done outside. | |
1016 | */ | |
1017 | if (opt->invert) | |
1018 | hit = !hit; | |
1019 | if (opt->unmatch_name_only) { | |
1020 | if (hit) | |
1021 | return 0; | |
1022 | goto next_line; | |
1023 | } | |
1024 | if (hit) { | |
1025 | count++; | |
1026 | if (opt->status_only) | |
1027 | return 1; | |
1028 | if (opt->name_only) { | |
1029 | show_name(opt, name); | |
1030 | return 1; | |
1031 | } | |
1032 | if (opt->count) | |
1033 | goto next_line; | |
1034 | if (binary_match_only) { | |
1035 | opt->output(opt, "Binary file ", 12); | |
1036 | output_color(opt, name, strlen(name), | |
1037 | opt->color_filename); | |
1038 | opt->output(opt, " matches\n", 9); | |
1039 | return 1; | |
1040 | } | |
1041 | /* Hit at this line. If we haven't shown the | |
1042 | * pre-context lines, we would need to show them. | |
1043 | */ | |
1044 | if (opt->pre_context) | |
1045 | show_pre_context(opt, name, buf, bol, lno); | |
1046 | else if (opt->funcname) | |
1047 | show_funcname_line(opt, name, buf, bol, lno); | |
1048 | show_line(opt, bol, eol, name, lno, ':'); | |
1049 | last_hit = lno; | |
1050 | } | |
1051 | else if (last_hit && | |
1052 | lno <= last_hit + opt->post_context) { | |
1053 | /* If the last hit is within the post context, | |
1054 | * we need to show this line. | |
1055 | */ | |
1056 | show_line(opt, bol, eol, name, lno, '-'); | |
1057 | } | |
1058 | ||
1059 | next_line: | |
1060 | bol = eol + 1; | |
1061 | if (!left) | |
1062 | break; | |
1063 | left--; | |
1064 | lno++; | |
1065 | } | |
1066 | ||
1067 | if (collect_hits) | |
1068 | return 0; | |
1069 | ||
1070 | if (opt->status_only) | |
1071 | return 0; | |
1072 | if (opt->unmatch_name_only) { | |
1073 | /* We did not see any hit, so we want to show this */ | |
1074 | show_name(opt, name); | |
1075 | return 1; | |
1076 | } | |
1077 | ||
1078 | xdiff_clear_find_func(&xecfg); | |
1079 | opt->priv = NULL; | |
1080 | ||
1081 | /* NEEDSWORK: | |
1082 | * The real "grep -c foo *.c" gives many "bar.c:0" lines, | |
1083 | * which feels mostly useless but sometimes useful. Maybe | |
1084 | * make it another option? For now suppress them. | |
1085 | */ | |
1086 | if (opt->count && count) { | |
1087 | char buf[32]; | |
1088 | output_color(opt, name, strlen(name), opt->color_filename); | |
1089 | output_sep(opt, ':'); | |
1090 | snprintf(buf, sizeof(buf), "%u\n", count); | |
1091 | opt->output(opt, buf, strlen(buf)); | |
1092 | return 1; | |
1093 | } | |
1094 | return !!last_hit; | |
1095 | } | |
1096 | ||
1097 | static void clr_hit_marker(struct grep_expr *x) | |
1098 | { | |
1099 | /* All-hit markers are meaningful only at the very top level | |
1100 | * OR node. | |
1101 | */ | |
1102 | while (1) { | |
1103 | x->hit = 0; | |
1104 | if (x->node != GREP_NODE_OR) | |
1105 | return; | |
1106 | x->u.binary.left->hit = 0; | |
1107 | x = x->u.binary.right; | |
1108 | } | |
1109 | } | |
1110 | ||
1111 | static int chk_hit_marker(struct grep_expr *x) | |
1112 | { | |
1113 | /* Top level nodes have hit markers. See if they all are hits */ | |
1114 | while (1) { | |
1115 | if (x->node != GREP_NODE_OR) | |
1116 | return x->hit; | |
1117 | if (!x->u.binary.left->hit) | |
1118 | return 0; | |
1119 | x = x->u.binary.right; | |
1120 | } | |
1121 | } | |
1122 | ||
1123 | int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size) | |
1124 | { | |
1125 | /* | |
1126 | * we do not have to do the two-pass grep when we do not check | |
1127 | * buffer-wide "all-match". | |
1128 | */ | |
1129 | if (!opt->all_match) | |
1130 | return grep_buffer_1(opt, name, buf, size, 0); | |
1131 | ||
1132 | /* Otherwise the toplevel "or" terms hit a bit differently. | |
1133 | * We first clear hit markers from them. | |
1134 | */ | |
1135 | clr_hit_marker(opt->pattern_expression); | |
1136 | grep_buffer_1(opt, name, buf, size, 1); | |
1137 | ||
1138 | if (!chk_hit_marker(opt->pattern_expression)) | |
1139 | return 0; | |
1140 | ||
1141 | return grep_buffer_1(opt, name, buf, size, 0); | |
1142 | } |