]>
Commit | Line | Data |
---|---|---|
1 | #include "cache.h" | |
2 | #include "grep.h" | |
3 | #include "userdiff.h" | |
4 | #include "xdiff-interface.h" | |
5 | ||
6 | void append_header_grep_pattern(struct grep_opt *opt, enum grep_header_field field, const char *pat) | |
7 | { | |
8 | struct grep_pat *p = xcalloc(1, sizeof(*p)); | |
9 | p->pattern = pat; | |
10 | p->patternlen = strlen(pat); | |
11 | p->origin = "header"; | |
12 | p->no = 0; | |
13 | p->token = GREP_PATTERN_HEAD; | |
14 | p->field = field; | |
15 | *opt->header_tail = p; | |
16 | opt->header_tail = &p->next; | |
17 | p->next = NULL; | |
18 | } | |
19 | ||
20 | void append_grep_pattern(struct grep_opt *opt, const char *pat, | |
21 | const char *origin, int no, enum grep_pat_token t) | |
22 | { | |
23 | append_grep_pat(opt, pat, strlen(pat), origin, no, t); | |
24 | } | |
25 | ||
26 | void append_grep_pat(struct grep_opt *opt, const char *pat, size_t patlen, | |
27 | const char *origin, int no, enum grep_pat_token t) | |
28 | { | |
29 | struct grep_pat *p = xcalloc(1, sizeof(*p)); | |
30 | p->pattern = pat; | |
31 | p->patternlen = patlen; | |
32 | p->origin = origin; | |
33 | p->no = no; | |
34 | p->token = t; | |
35 | *opt->pattern_tail = p; | |
36 | opt->pattern_tail = &p->next; | |
37 | p->next = NULL; | |
38 | } | |
39 | ||
40 | struct grep_opt *grep_opt_dup(const struct grep_opt *opt) | |
41 | { | |
42 | struct grep_pat *pat; | |
43 | struct grep_opt *ret = xmalloc(sizeof(struct grep_opt)); | |
44 | *ret = *opt; | |
45 | ||
46 | ret->pattern_list = NULL; | |
47 | ret->pattern_tail = &ret->pattern_list; | |
48 | ||
49 | for(pat = opt->pattern_list; pat != NULL; pat = pat->next) | |
50 | { | |
51 | if(pat->token == GREP_PATTERN_HEAD) | |
52 | append_header_grep_pattern(ret, pat->field, | |
53 | pat->pattern); | |
54 | else | |
55 | append_grep_pat(ret, pat->pattern, pat->patternlen, | |
56 | pat->origin, pat->no, pat->token); | |
57 | } | |
58 | ||
59 | return ret; | |
60 | } | |
61 | ||
62 | static NORETURN void compile_regexp_failed(const struct grep_pat *p, | |
63 | const char *error) | |
64 | { | |
65 | char where[1024]; | |
66 | ||
67 | if (p->no) | |
68 | sprintf(where, "In '%s' at %d, ", p->origin, p->no); | |
69 | else if (p->origin) | |
70 | sprintf(where, "%s, ", p->origin); | |
71 | else | |
72 | where[0] = 0; | |
73 | ||
74 | die("%s'%s': %s", where, p->pattern, error); | |
75 | } | |
76 | ||
77 | #ifdef USE_LIBPCRE | |
78 | static void compile_pcre_regexp(struct grep_pat *p, const struct grep_opt *opt) | |
79 | { | |
80 | const char *error; | |
81 | int erroffset; | |
82 | int options = 0; | |
83 | ||
84 | if (opt->ignore_case) | |
85 | options |= PCRE_CASELESS; | |
86 | ||
87 | p->pcre_regexp = pcre_compile(p->pattern, options, &error, &erroffset, | |
88 | NULL); | |
89 | if (!p->pcre_regexp) | |
90 | compile_regexp_failed(p, error); | |
91 | ||
92 | p->pcre_extra_info = pcre_study(p->pcre_regexp, 0, &error); | |
93 | if (!p->pcre_extra_info && error) | |
94 | die("%s", error); | |
95 | } | |
96 | ||
97 | static int pcrematch(struct grep_pat *p, const char *line, const char *eol, | |
98 | regmatch_t *match, int eflags) | |
99 | { | |
100 | int ovector[30], ret, flags = 0; | |
101 | ||
102 | if (eflags & REG_NOTBOL) | |
103 | flags |= PCRE_NOTBOL; | |
104 | ||
105 | ret = pcre_exec(p->pcre_regexp, p->pcre_extra_info, line, eol - line, | |
106 | 0, flags, ovector, ARRAY_SIZE(ovector)); | |
107 | if (ret < 0 && ret != PCRE_ERROR_NOMATCH) | |
108 | die("pcre_exec failed with error code %d", ret); | |
109 | if (ret > 0) { | |
110 | ret = 0; | |
111 | match->rm_so = ovector[0]; | |
112 | match->rm_eo = ovector[1]; | |
113 | } | |
114 | ||
115 | return ret; | |
116 | } | |
117 | ||
118 | static void free_pcre_regexp(struct grep_pat *p) | |
119 | { | |
120 | pcre_free(p->pcre_regexp); | |
121 | pcre_free(p->pcre_extra_info); | |
122 | } | |
123 | #else /* !USE_LIBPCRE */ | |
124 | static void compile_pcre_regexp(struct grep_pat *p, const struct grep_opt *opt) | |
125 | { | |
126 | die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE"); | |
127 | } | |
128 | ||
129 | static int pcrematch(struct grep_pat *p, const char *line, const char *eol, | |
130 | regmatch_t *match, int eflags) | |
131 | { | |
132 | return 1; | |
133 | } | |
134 | ||
135 | static void free_pcre_regexp(struct grep_pat *p) | |
136 | { | |
137 | } | |
138 | #endif /* !USE_LIBPCRE */ | |
139 | ||
140 | static void compile_regexp(struct grep_pat *p, struct grep_opt *opt) | |
141 | { | |
142 | int err; | |
143 | ||
144 | p->word_regexp = opt->word_regexp; | |
145 | p->ignore_case = opt->ignore_case; | |
146 | p->fixed = opt->fixed; | |
147 | ||
148 | if (p->fixed) | |
149 | return; | |
150 | ||
151 | if (opt->pcre) { | |
152 | compile_pcre_regexp(p, opt); | |
153 | return; | |
154 | } | |
155 | ||
156 | err = regcomp(&p->regexp, p->pattern, opt->regflags); | |
157 | if (err) { | |
158 | char errbuf[1024]; | |
159 | regerror(err, &p->regexp, errbuf, 1024); | |
160 | regfree(&p->regexp); | |
161 | compile_regexp_failed(p, errbuf); | |
162 | } | |
163 | } | |
164 | ||
165 | static struct grep_expr *compile_pattern_or(struct grep_pat **); | |
166 | static struct grep_expr *compile_pattern_atom(struct grep_pat **list) | |
167 | { | |
168 | struct grep_pat *p; | |
169 | struct grep_expr *x; | |
170 | ||
171 | p = *list; | |
172 | if (!p) | |
173 | return NULL; | |
174 | switch (p->token) { | |
175 | case GREP_PATTERN: /* atom */ | |
176 | case GREP_PATTERN_HEAD: | |
177 | case GREP_PATTERN_BODY: | |
178 | x = xcalloc(1, sizeof (struct grep_expr)); | |
179 | x->node = GREP_NODE_ATOM; | |
180 | x->u.atom = p; | |
181 | *list = p->next; | |
182 | return x; | |
183 | case GREP_OPEN_PAREN: | |
184 | *list = p->next; | |
185 | x = compile_pattern_or(list); | |
186 | if (!*list || (*list)->token != GREP_CLOSE_PAREN) | |
187 | die("unmatched parenthesis"); | |
188 | *list = (*list)->next; | |
189 | return x; | |
190 | default: | |
191 | return NULL; | |
192 | } | |
193 | } | |
194 | ||
195 | static struct grep_expr *compile_pattern_not(struct grep_pat **list) | |
196 | { | |
197 | struct grep_pat *p; | |
198 | struct grep_expr *x; | |
199 | ||
200 | p = *list; | |
201 | if (!p) | |
202 | return NULL; | |
203 | switch (p->token) { | |
204 | case GREP_NOT: | |
205 | if (!p->next) | |
206 | die("--not not followed by pattern expression"); | |
207 | *list = p->next; | |
208 | x = xcalloc(1, sizeof (struct grep_expr)); | |
209 | x->node = GREP_NODE_NOT; | |
210 | x->u.unary = compile_pattern_not(list); | |
211 | if (!x->u.unary) | |
212 | die("--not followed by non pattern expression"); | |
213 | return x; | |
214 | default: | |
215 | return compile_pattern_atom(list); | |
216 | } | |
217 | } | |
218 | ||
219 | static struct grep_expr *compile_pattern_and(struct grep_pat **list) | |
220 | { | |
221 | struct grep_pat *p; | |
222 | struct grep_expr *x, *y, *z; | |
223 | ||
224 | x = compile_pattern_not(list); | |
225 | p = *list; | |
226 | if (p && p->token == GREP_AND) { | |
227 | if (!p->next) | |
228 | die("--and not followed by pattern expression"); | |
229 | *list = p->next; | |
230 | y = compile_pattern_and(list); | |
231 | if (!y) | |
232 | die("--and not followed by pattern expression"); | |
233 | z = xcalloc(1, sizeof (struct grep_expr)); | |
234 | z->node = GREP_NODE_AND; | |
235 | z->u.binary.left = x; | |
236 | z->u.binary.right = y; | |
237 | return z; | |
238 | } | |
239 | return x; | |
240 | } | |
241 | ||
242 | static struct grep_expr *compile_pattern_or(struct grep_pat **list) | |
243 | { | |
244 | struct grep_pat *p; | |
245 | struct grep_expr *x, *y, *z; | |
246 | ||
247 | x = compile_pattern_and(list); | |
248 | p = *list; | |
249 | if (x && p && p->token != GREP_CLOSE_PAREN) { | |
250 | y = compile_pattern_or(list); | |
251 | if (!y) | |
252 | die("not a pattern expression %s", p->pattern); | |
253 | z = xcalloc(1, sizeof (struct grep_expr)); | |
254 | z->node = GREP_NODE_OR; | |
255 | z->u.binary.left = x; | |
256 | z->u.binary.right = y; | |
257 | return z; | |
258 | } | |
259 | return x; | |
260 | } | |
261 | ||
262 | static struct grep_expr *compile_pattern_expr(struct grep_pat **list) | |
263 | { | |
264 | return compile_pattern_or(list); | |
265 | } | |
266 | ||
267 | static struct grep_expr *grep_true_expr(void) | |
268 | { | |
269 | struct grep_expr *z = xcalloc(1, sizeof(*z)); | |
270 | z->node = GREP_NODE_TRUE; | |
271 | return z; | |
272 | } | |
273 | ||
274 | static struct grep_expr *grep_or_expr(struct grep_expr *left, struct grep_expr *right) | |
275 | { | |
276 | struct grep_expr *z = xcalloc(1, sizeof(*z)); | |
277 | z->node = GREP_NODE_OR; | |
278 | z->u.binary.left = left; | |
279 | z->u.binary.right = right; | |
280 | return z; | |
281 | } | |
282 | ||
283 | static struct grep_expr *prep_header_patterns(struct grep_opt *opt) | |
284 | { | |
285 | struct grep_pat *p; | |
286 | struct grep_expr *header_expr; | |
287 | struct grep_expr *(header_group[GREP_HEADER_FIELD_MAX]); | |
288 | enum grep_header_field fld; | |
289 | ||
290 | if (!opt->header_list) | |
291 | return NULL; | |
292 | p = opt->header_list; | |
293 | for (p = opt->header_list; p; p = p->next) { | |
294 | if (p->token != GREP_PATTERN_HEAD) | |
295 | die("bug: a non-header pattern in grep header list."); | |
296 | if (p->field < 0 || GREP_HEADER_FIELD_MAX <= p->field) | |
297 | die("bug: unknown header field %d", p->field); | |
298 | compile_regexp(p, opt); | |
299 | } | |
300 | ||
301 | for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) | |
302 | header_group[fld] = NULL; | |
303 | ||
304 | for (p = opt->header_list; p; p = p->next) { | |
305 | struct grep_expr *h; | |
306 | struct grep_pat *pp = p; | |
307 | ||
308 | h = compile_pattern_atom(&pp); | |
309 | if (!h || pp != p->next) | |
310 | die("bug: malformed header expr"); | |
311 | if (!header_group[p->field]) { | |
312 | header_group[p->field] = h; | |
313 | continue; | |
314 | } | |
315 | header_group[p->field] = grep_or_expr(h, header_group[p->field]); | |
316 | } | |
317 | ||
318 | header_expr = NULL; | |
319 | ||
320 | for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) { | |
321 | if (!header_group[fld]) | |
322 | continue; | |
323 | if (!header_expr) | |
324 | header_expr = grep_true_expr(); | |
325 | header_expr = grep_or_expr(header_group[fld], header_expr); | |
326 | } | |
327 | return header_expr; | |
328 | } | |
329 | ||
330 | void compile_grep_patterns(struct grep_opt *opt) | |
331 | { | |
332 | struct grep_pat *p; | |
333 | struct grep_expr *header_expr = prep_header_patterns(opt); | |
334 | ||
335 | for (p = opt->pattern_list; p; p = p->next) { | |
336 | switch (p->token) { | |
337 | case GREP_PATTERN: /* atom */ | |
338 | case GREP_PATTERN_HEAD: | |
339 | case GREP_PATTERN_BODY: | |
340 | compile_regexp(p, opt); | |
341 | break; | |
342 | default: | |
343 | opt->extended = 1; | |
344 | break; | |
345 | } | |
346 | } | |
347 | ||
348 | if (opt->all_match || header_expr) | |
349 | opt->extended = 1; | |
350 | else if (!opt->extended) | |
351 | return; | |
352 | ||
353 | p = opt->pattern_list; | |
354 | if (p) | |
355 | opt->pattern_expression = compile_pattern_expr(&p); | |
356 | if (p) | |
357 | die("incomplete pattern expression: %s", p->pattern); | |
358 | ||
359 | if (!header_expr) | |
360 | return; | |
361 | ||
362 | if (!opt->pattern_expression) | |
363 | opt->pattern_expression = header_expr; | |
364 | else | |
365 | opt->pattern_expression = grep_or_expr(opt->pattern_expression, | |
366 | header_expr); | |
367 | opt->all_match = 1; | |
368 | } | |
369 | ||
370 | static void free_pattern_expr(struct grep_expr *x) | |
371 | { | |
372 | switch (x->node) { | |
373 | case GREP_NODE_TRUE: | |
374 | case GREP_NODE_ATOM: | |
375 | break; | |
376 | case GREP_NODE_NOT: | |
377 | free_pattern_expr(x->u.unary); | |
378 | break; | |
379 | case GREP_NODE_AND: | |
380 | case GREP_NODE_OR: | |
381 | free_pattern_expr(x->u.binary.left); | |
382 | free_pattern_expr(x->u.binary.right); | |
383 | break; | |
384 | } | |
385 | free(x); | |
386 | } | |
387 | ||
388 | void free_grep_patterns(struct grep_opt *opt) | |
389 | { | |
390 | struct grep_pat *p, *n; | |
391 | ||
392 | for (p = opt->pattern_list; p; p = n) { | |
393 | n = p->next; | |
394 | switch (p->token) { | |
395 | case GREP_PATTERN: /* atom */ | |
396 | case GREP_PATTERN_HEAD: | |
397 | case GREP_PATTERN_BODY: | |
398 | if (p->pcre_regexp) | |
399 | free_pcre_regexp(p); | |
400 | else | |
401 | regfree(&p->regexp); | |
402 | break; | |
403 | default: | |
404 | break; | |
405 | } | |
406 | free(p); | |
407 | } | |
408 | ||
409 | if (!opt->extended) | |
410 | return; | |
411 | free_pattern_expr(opt->pattern_expression); | |
412 | } | |
413 | ||
414 | static char *end_of_line(char *cp, unsigned long *left) | |
415 | { | |
416 | unsigned long l = *left; | |
417 | while (l && *cp != '\n') { | |
418 | l--; | |
419 | cp++; | |
420 | } | |
421 | *left = l; | |
422 | return cp; | |
423 | } | |
424 | ||
425 | static int word_char(char ch) | |
426 | { | |
427 | return isalnum(ch) || ch == '_'; | |
428 | } | |
429 | ||
430 | static void output_color(struct grep_opt *opt, const void *data, size_t size, | |
431 | const char *color) | |
432 | { | |
433 | if (opt->color && color && color[0]) { | |
434 | opt->output(opt, color, strlen(color)); | |
435 | opt->output(opt, data, size); | |
436 | opt->output(opt, GIT_COLOR_RESET, strlen(GIT_COLOR_RESET)); | |
437 | } else | |
438 | opt->output(opt, data, size); | |
439 | } | |
440 | ||
441 | static void output_sep(struct grep_opt *opt, char sign) | |
442 | { | |
443 | if (opt->null_following_name) | |
444 | opt->output(opt, "\0", 1); | |
445 | else | |
446 | output_color(opt, &sign, 1, opt->color_sep); | |
447 | } | |
448 | ||
449 | static void show_name(struct grep_opt *opt, const char *name) | |
450 | { | |
451 | output_color(opt, name, strlen(name), opt->color_filename); | |
452 | opt->output(opt, opt->null_following_name ? "\0" : "\n", 1); | |
453 | } | |
454 | ||
455 | static int fixmatch(struct grep_pat *p, char *line, char *eol, | |
456 | regmatch_t *match) | |
457 | { | |
458 | char *hit; | |
459 | ||
460 | if (p->ignore_case) { | |
461 | char *s = line; | |
462 | do { | |
463 | hit = strcasestr(s, p->pattern); | |
464 | if (hit) | |
465 | break; | |
466 | s += strlen(s) + 1; | |
467 | } while (s < eol); | |
468 | } else | |
469 | hit = memmem(line, eol - line, p->pattern, p->patternlen); | |
470 | ||
471 | if (!hit) { | |
472 | match->rm_so = match->rm_eo = -1; | |
473 | return REG_NOMATCH; | |
474 | } | |
475 | else { | |
476 | match->rm_so = hit - line; | |
477 | match->rm_eo = match->rm_so + p->patternlen; | |
478 | return 0; | |
479 | } | |
480 | } | |
481 | ||
482 | static int regmatch(const regex_t *preg, char *line, char *eol, | |
483 | regmatch_t *match, int eflags) | |
484 | { | |
485 | #ifdef REG_STARTEND | |
486 | match->rm_so = 0; | |
487 | match->rm_eo = eol - line; | |
488 | eflags |= REG_STARTEND; | |
489 | #endif | |
490 | return regexec(preg, line, 1, match, eflags); | |
491 | } | |
492 | ||
493 | static int patmatch(struct grep_pat *p, char *line, char *eol, | |
494 | regmatch_t *match, int eflags) | |
495 | { | |
496 | int hit; | |
497 | ||
498 | if (p->fixed) | |
499 | hit = !fixmatch(p, line, eol, match); | |
500 | else if (p->pcre_regexp) | |
501 | hit = !pcrematch(p, line, eol, match, eflags); | |
502 | else | |
503 | hit = !regmatch(&p->regexp, line, eol, match, eflags); | |
504 | ||
505 | return hit; | |
506 | } | |
507 | ||
508 | static int strip_timestamp(char *bol, char **eol_p) | |
509 | { | |
510 | char *eol = *eol_p; | |
511 | int ch; | |
512 | ||
513 | while (bol < --eol) { | |
514 | if (*eol != '>') | |
515 | continue; | |
516 | *eol_p = ++eol; | |
517 | ch = *eol; | |
518 | *eol = '\0'; | |
519 | return ch; | |
520 | } | |
521 | return 0; | |
522 | } | |
523 | ||
524 | static struct { | |
525 | const char *field; | |
526 | size_t len; | |
527 | } header_field[] = { | |
528 | { "author ", 7 }, | |
529 | { "committer ", 10 }, | |
530 | }; | |
531 | ||
532 | static int match_one_pattern(struct grep_pat *p, char *bol, char *eol, | |
533 | enum grep_context ctx, | |
534 | regmatch_t *pmatch, int eflags) | |
535 | { | |
536 | int hit = 0; | |
537 | int saved_ch = 0; | |
538 | const char *start = bol; | |
539 | ||
540 | if ((p->token != GREP_PATTERN) && | |
541 | ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD))) | |
542 | return 0; | |
543 | ||
544 | if (p->token == GREP_PATTERN_HEAD) { | |
545 | const char *field; | |
546 | size_t len; | |
547 | assert(p->field < ARRAY_SIZE(header_field)); | |
548 | field = header_field[p->field].field; | |
549 | len = header_field[p->field].len; | |
550 | if (strncmp(bol, field, len)) | |
551 | return 0; | |
552 | bol += len; | |
553 | saved_ch = strip_timestamp(bol, &eol); | |
554 | } | |
555 | ||
556 | again: | |
557 | hit = patmatch(p, bol, eol, pmatch, eflags); | |
558 | ||
559 | if (hit && p->word_regexp) { | |
560 | if ((pmatch[0].rm_so < 0) || | |
561 | (eol - bol) < pmatch[0].rm_so || | |
562 | (pmatch[0].rm_eo < 0) || | |
563 | (eol - bol) < pmatch[0].rm_eo) | |
564 | die("regexp returned nonsense"); | |
565 | ||
566 | /* Match beginning must be either beginning of the | |
567 | * line, or at word boundary (i.e. the last char must | |
568 | * not be a word char). Similarly, match end must be | |
569 | * either end of the line, or at word boundary | |
570 | * (i.e. the next char must not be a word char). | |
571 | */ | |
572 | if ( ((pmatch[0].rm_so == 0) || | |
573 | !word_char(bol[pmatch[0].rm_so-1])) && | |
574 | ((pmatch[0].rm_eo == (eol-bol)) || | |
575 | !word_char(bol[pmatch[0].rm_eo])) ) | |
576 | ; | |
577 | else | |
578 | hit = 0; | |
579 | ||
580 | /* Words consist of at least one character. */ | |
581 | if (pmatch->rm_so == pmatch->rm_eo) | |
582 | hit = 0; | |
583 | ||
584 | if (!hit && pmatch[0].rm_so + bol + 1 < eol) { | |
585 | /* There could be more than one match on the | |
586 | * line, and the first match might not be | |
587 | * strict word match. But later ones could be! | |
588 | * Forward to the next possible start, i.e. the | |
589 | * next position following a non-word char. | |
590 | */ | |
591 | bol = pmatch[0].rm_so + bol + 1; | |
592 | while (word_char(bol[-1]) && bol < eol) | |
593 | bol++; | |
594 | eflags |= REG_NOTBOL; | |
595 | if (bol < eol) | |
596 | goto again; | |
597 | } | |
598 | } | |
599 | if (p->token == GREP_PATTERN_HEAD && saved_ch) | |
600 | *eol = saved_ch; | |
601 | if (hit) { | |
602 | pmatch[0].rm_so += bol - start; | |
603 | pmatch[0].rm_eo += bol - start; | |
604 | } | |
605 | return hit; | |
606 | } | |
607 | ||
608 | static int match_expr_eval(struct grep_expr *x, char *bol, char *eol, | |
609 | enum grep_context ctx, int collect_hits) | |
610 | { | |
611 | int h = 0; | |
612 | regmatch_t match; | |
613 | ||
614 | if (!x) | |
615 | die("Not a valid grep expression"); | |
616 | switch (x->node) { | |
617 | case GREP_NODE_TRUE: | |
618 | h = 1; | |
619 | break; | |
620 | case GREP_NODE_ATOM: | |
621 | h = match_one_pattern(x->u.atom, bol, eol, ctx, &match, 0); | |
622 | break; | |
623 | case GREP_NODE_NOT: | |
624 | h = !match_expr_eval(x->u.unary, bol, eol, ctx, 0); | |
625 | break; | |
626 | case GREP_NODE_AND: | |
627 | if (!match_expr_eval(x->u.binary.left, bol, eol, ctx, 0)) | |
628 | return 0; | |
629 | h = match_expr_eval(x->u.binary.right, bol, eol, ctx, 0); | |
630 | break; | |
631 | case GREP_NODE_OR: | |
632 | if (!collect_hits) | |
633 | return (match_expr_eval(x->u.binary.left, | |
634 | bol, eol, ctx, 0) || | |
635 | match_expr_eval(x->u.binary.right, | |
636 | bol, eol, ctx, 0)); | |
637 | h = match_expr_eval(x->u.binary.left, bol, eol, ctx, 0); | |
638 | x->u.binary.left->hit |= h; | |
639 | h |= match_expr_eval(x->u.binary.right, bol, eol, ctx, 1); | |
640 | break; | |
641 | default: | |
642 | die("Unexpected node type (internal error) %d", x->node); | |
643 | } | |
644 | if (collect_hits) | |
645 | x->hit |= h; | |
646 | return h; | |
647 | } | |
648 | ||
649 | static int match_expr(struct grep_opt *opt, char *bol, char *eol, | |
650 | enum grep_context ctx, int collect_hits) | |
651 | { | |
652 | struct grep_expr *x = opt->pattern_expression; | |
653 | return match_expr_eval(x, bol, eol, ctx, collect_hits); | |
654 | } | |
655 | ||
656 | static int match_line(struct grep_opt *opt, char *bol, char *eol, | |
657 | enum grep_context ctx, int collect_hits) | |
658 | { | |
659 | struct grep_pat *p; | |
660 | regmatch_t match; | |
661 | ||
662 | if (opt->extended) | |
663 | return match_expr(opt, bol, eol, ctx, collect_hits); | |
664 | ||
665 | /* we do not call with collect_hits without being extended */ | |
666 | for (p = opt->pattern_list; p; p = p->next) { | |
667 | if (match_one_pattern(p, bol, eol, ctx, &match, 0)) | |
668 | return 1; | |
669 | } | |
670 | return 0; | |
671 | } | |
672 | ||
673 | static int match_next_pattern(struct grep_pat *p, char *bol, char *eol, | |
674 | enum grep_context ctx, | |
675 | regmatch_t *pmatch, int eflags) | |
676 | { | |
677 | regmatch_t match; | |
678 | ||
679 | if (!match_one_pattern(p, bol, eol, ctx, &match, eflags)) | |
680 | return 0; | |
681 | if (match.rm_so < 0 || match.rm_eo < 0) | |
682 | return 0; | |
683 | if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) { | |
684 | if (match.rm_so > pmatch->rm_so) | |
685 | return 1; | |
686 | if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo) | |
687 | return 1; | |
688 | } | |
689 | pmatch->rm_so = match.rm_so; | |
690 | pmatch->rm_eo = match.rm_eo; | |
691 | return 1; | |
692 | } | |
693 | ||
694 | static int next_match(struct grep_opt *opt, char *bol, char *eol, | |
695 | enum grep_context ctx, regmatch_t *pmatch, int eflags) | |
696 | { | |
697 | struct grep_pat *p; | |
698 | int hit = 0; | |
699 | ||
700 | pmatch->rm_so = pmatch->rm_eo = -1; | |
701 | if (bol < eol) { | |
702 | for (p = opt->pattern_list; p; p = p->next) { | |
703 | switch (p->token) { | |
704 | case GREP_PATTERN: /* atom */ | |
705 | case GREP_PATTERN_HEAD: | |
706 | case GREP_PATTERN_BODY: | |
707 | hit |= match_next_pattern(p, bol, eol, ctx, | |
708 | pmatch, eflags); | |
709 | break; | |
710 | default: | |
711 | break; | |
712 | } | |
713 | } | |
714 | } | |
715 | return hit; | |
716 | } | |
717 | ||
718 | static void show_line(struct grep_opt *opt, char *bol, char *eol, | |
719 | const char *name, unsigned lno, char sign) | |
720 | { | |
721 | int rest = eol - bol; | |
722 | char *line_color = NULL; | |
723 | ||
724 | if (opt->file_break && opt->last_shown == 0) { | |
725 | if (opt->show_hunk_mark) | |
726 | opt->output(opt, "\n", 1); | |
727 | } else if (opt->pre_context || opt->post_context) { | |
728 | if (opt->last_shown == 0) { | |
729 | if (opt->show_hunk_mark) { | |
730 | output_color(opt, "--", 2, opt->color_sep); | |
731 | opt->output(opt, "\n", 1); | |
732 | } | |
733 | } else if (lno > opt->last_shown + 1) { | |
734 | output_color(opt, "--", 2, opt->color_sep); | |
735 | opt->output(opt, "\n", 1); | |
736 | } | |
737 | } | |
738 | opt->last_shown = lno; | |
739 | ||
740 | if (opt->pathname) { | |
741 | output_color(opt, name, strlen(name), opt->color_filename); | |
742 | output_sep(opt, sign); | |
743 | } | |
744 | if (opt->linenum) { | |
745 | char buf[32]; | |
746 | snprintf(buf, sizeof(buf), "%d", lno); | |
747 | output_color(opt, buf, strlen(buf), opt->color_lineno); | |
748 | output_sep(opt, sign); | |
749 | } | |
750 | if (opt->color) { | |
751 | regmatch_t match; | |
752 | enum grep_context ctx = GREP_CONTEXT_BODY; | |
753 | int ch = *eol; | |
754 | int eflags = 0; | |
755 | ||
756 | if (sign == ':') | |
757 | line_color = opt->color_selected; | |
758 | else if (sign == '-') | |
759 | line_color = opt->color_context; | |
760 | else if (sign == '=') | |
761 | line_color = opt->color_function; | |
762 | *eol = '\0'; | |
763 | while (next_match(opt, bol, eol, ctx, &match, eflags)) { | |
764 | if (match.rm_so == match.rm_eo) | |
765 | break; | |
766 | ||
767 | output_color(opt, bol, match.rm_so, line_color); | |
768 | output_color(opt, bol + match.rm_so, | |
769 | match.rm_eo - match.rm_so, | |
770 | opt->color_match); | |
771 | bol += match.rm_eo; | |
772 | rest -= match.rm_eo; | |
773 | eflags = REG_NOTBOL; | |
774 | } | |
775 | *eol = ch; | |
776 | } | |
777 | output_color(opt, bol, rest, line_color); | |
778 | opt->output(opt, "\n", 1); | |
779 | } | |
780 | ||
781 | static int match_funcname(struct grep_opt *opt, char *bol, char *eol) | |
782 | { | |
783 | xdemitconf_t *xecfg = opt->priv; | |
784 | if (xecfg && xecfg->find_func) { | |
785 | char buf[1]; | |
786 | return xecfg->find_func(bol, eol - bol, buf, 1, | |
787 | xecfg->find_func_priv) >= 0; | |
788 | } | |
789 | ||
790 | if (bol == eol) | |
791 | return 0; | |
792 | if (isalpha(*bol) || *bol == '_' || *bol == '$') | |
793 | return 1; | |
794 | return 0; | |
795 | } | |
796 | ||
797 | static void show_funcname_line(struct grep_opt *opt, const char *name, | |
798 | char *buf, char *bol, unsigned lno) | |
799 | { | |
800 | while (bol > buf) { | |
801 | char *eol = --bol; | |
802 | ||
803 | while (bol > buf && bol[-1] != '\n') | |
804 | bol--; | |
805 | lno--; | |
806 | ||
807 | if (lno <= opt->last_shown) | |
808 | break; | |
809 | ||
810 | if (match_funcname(opt, bol, eol)) { | |
811 | show_line(opt, bol, eol, name, lno, '='); | |
812 | break; | |
813 | } | |
814 | } | |
815 | } | |
816 | ||
817 | static void show_pre_context(struct grep_opt *opt, const char *name, char *buf, | |
818 | char *bol, unsigned lno) | |
819 | { | |
820 | unsigned cur = lno, from = 1, funcname_lno = 0; | |
821 | int funcname_needed = opt->funcname; | |
822 | ||
823 | if (opt->pre_context < lno) | |
824 | from = lno - opt->pre_context; | |
825 | if (from <= opt->last_shown) | |
826 | from = opt->last_shown + 1; | |
827 | ||
828 | /* Rewind. */ | |
829 | while (bol > buf && cur > from) { | |
830 | char *eol = --bol; | |
831 | ||
832 | while (bol > buf && bol[-1] != '\n') | |
833 | bol--; | |
834 | cur--; | |
835 | if (funcname_needed && match_funcname(opt, bol, eol)) { | |
836 | funcname_lno = cur; | |
837 | funcname_needed = 0; | |
838 | } | |
839 | } | |
840 | ||
841 | /* We need to look even further back to find a function signature. */ | |
842 | if (opt->funcname && funcname_needed) | |
843 | show_funcname_line(opt, name, buf, bol, cur); | |
844 | ||
845 | /* Back forward. */ | |
846 | while (cur < lno) { | |
847 | char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-'; | |
848 | ||
849 | while (*eol != '\n') | |
850 | eol++; | |
851 | show_line(opt, bol, eol, name, cur, sign); | |
852 | bol = eol + 1; | |
853 | cur++; | |
854 | } | |
855 | } | |
856 | ||
857 | static int should_lookahead(struct grep_opt *opt) | |
858 | { | |
859 | struct grep_pat *p; | |
860 | ||
861 | if (opt->extended) | |
862 | return 0; /* punt for too complex stuff */ | |
863 | if (opt->invert) | |
864 | return 0; | |
865 | for (p = opt->pattern_list; p; p = p->next) { | |
866 | if (p->token != GREP_PATTERN) | |
867 | return 0; /* punt for "header only" and stuff */ | |
868 | } | |
869 | return 1; | |
870 | } | |
871 | ||
872 | static int look_ahead(struct grep_opt *opt, | |
873 | unsigned long *left_p, | |
874 | unsigned *lno_p, | |
875 | char **bol_p) | |
876 | { | |
877 | unsigned lno = *lno_p; | |
878 | char *bol = *bol_p; | |
879 | struct grep_pat *p; | |
880 | char *sp, *last_bol; | |
881 | regoff_t earliest = -1; | |
882 | ||
883 | for (p = opt->pattern_list; p; p = p->next) { | |
884 | int hit; | |
885 | regmatch_t m; | |
886 | ||
887 | hit = patmatch(p, bol, bol + *left_p, &m, 0); | |
888 | if (!hit || m.rm_so < 0 || m.rm_eo < 0) | |
889 | continue; | |
890 | if (earliest < 0 || m.rm_so < earliest) | |
891 | earliest = m.rm_so; | |
892 | } | |
893 | ||
894 | if (earliest < 0) { | |
895 | *bol_p = bol + *left_p; | |
896 | *left_p = 0; | |
897 | return 1; | |
898 | } | |
899 | for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--) | |
900 | ; /* find the beginning of the line */ | |
901 | last_bol = sp; | |
902 | ||
903 | for (sp = bol; sp < last_bol; sp++) { | |
904 | if (*sp == '\n') | |
905 | lno++; | |
906 | } | |
907 | *left_p -= last_bol - bol; | |
908 | *bol_p = last_bol; | |
909 | *lno_p = lno; | |
910 | return 0; | |
911 | } | |
912 | ||
913 | int grep_threads_ok(const struct grep_opt *opt) | |
914 | { | |
915 | /* If this condition is true, then we may use the attribute | |
916 | * machinery in grep_buffer_1. The attribute code is not | |
917 | * thread safe, so we disable the use of threads. | |
918 | */ | |
919 | if (opt->funcname && !opt->unmatch_name_only && !opt->status_only && | |
920 | !opt->name_only) | |
921 | return 0; | |
922 | ||
923 | return 1; | |
924 | } | |
925 | ||
926 | static void std_output(struct grep_opt *opt, const void *buf, size_t size) | |
927 | { | |
928 | fwrite(buf, size, 1, stdout); | |
929 | } | |
930 | ||
931 | static int grep_buffer_1(struct grep_opt *opt, const char *name, | |
932 | char *buf, unsigned long size, int collect_hits) | |
933 | { | |
934 | char *bol = buf; | |
935 | unsigned long left = size; | |
936 | unsigned lno = 1; | |
937 | unsigned last_hit = 0; | |
938 | int binary_match_only = 0; | |
939 | unsigned count = 0; | |
940 | int try_lookahead = 0; | |
941 | enum grep_context ctx = GREP_CONTEXT_HEAD; | |
942 | xdemitconf_t xecfg; | |
943 | ||
944 | if (!opt->output) | |
945 | opt->output = std_output; | |
946 | ||
947 | if (opt->pre_context || opt->post_context || opt->file_break) { | |
948 | /* Show hunk marks, except for the first file. */ | |
949 | if (opt->last_shown) | |
950 | opt->show_hunk_mark = 1; | |
951 | /* | |
952 | * If we're using threads then we can't easily identify | |
953 | * the first file. Always put hunk marks in that case | |
954 | * and skip the very first one later in work_done(). | |
955 | */ | |
956 | if (opt->output != std_output) | |
957 | opt->show_hunk_mark = 1; | |
958 | } | |
959 | opt->last_shown = 0; | |
960 | ||
961 | switch (opt->binary) { | |
962 | case GREP_BINARY_DEFAULT: | |
963 | if (buffer_is_binary(buf, size)) | |
964 | binary_match_only = 1; | |
965 | break; | |
966 | case GREP_BINARY_NOMATCH: | |
967 | if (buffer_is_binary(buf, size)) | |
968 | return 0; /* Assume unmatch */ | |
969 | break; | |
970 | case GREP_BINARY_TEXT: | |
971 | break; | |
972 | default: | |
973 | die("bug: unknown binary handling mode"); | |
974 | } | |
975 | ||
976 | memset(&xecfg, 0, sizeof(xecfg)); | |
977 | if (opt->funcname && !opt->unmatch_name_only && !opt->status_only && | |
978 | !opt->name_only && !binary_match_only && !collect_hits) { | |
979 | struct userdiff_driver *drv = userdiff_find_by_path(name); | |
980 | if (drv && drv->funcname.pattern) { | |
981 | const struct userdiff_funcname *pe = &drv->funcname; | |
982 | xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags); | |
983 | opt->priv = &xecfg; | |
984 | } | |
985 | } | |
986 | try_lookahead = should_lookahead(opt); | |
987 | ||
988 | while (left) { | |
989 | char *eol, ch; | |
990 | int hit; | |
991 | ||
992 | /* | |
993 | * look_ahead() skips quickly to the line that possibly | |
994 | * has the next hit; don't call it if we need to do | |
995 | * something more than just skipping the current line | |
996 | * in response to an unmatch for the current line. E.g. | |
997 | * inside a post-context window, we will show the current | |
998 | * line as a context around the previous hit when it | |
999 | * doesn't hit. | |
1000 | */ | |
1001 | if (try_lookahead | |
1002 | && !(last_hit | |
1003 | && lno <= last_hit + opt->post_context) | |
1004 | && look_ahead(opt, &left, &lno, &bol)) | |
1005 | break; | |
1006 | eol = end_of_line(bol, &left); | |
1007 | ch = *eol; | |
1008 | *eol = 0; | |
1009 | ||
1010 | if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol)) | |
1011 | ctx = GREP_CONTEXT_BODY; | |
1012 | ||
1013 | hit = match_line(opt, bol, eol, ctx, collect_hits); | |
1014 | *eol = ch; | |
1015 | ||
1016 | if (collect_hits) | |
1017 | goto next_line; | |
1018 | ||
1019 | /* "grep -v -e foo -e bla" should list lines | |
1020 | * that do not have either, so inversion should | |
1021 | * be done outside. | |
1022 | */ | |
1023 | if (opt->invert) | |
1024 | hit = !hit; | |
1025 | if (opt->unmatch_name_only) { | |
1026 | if (hit) | |
1027 | return 0; | |
1028 | goto next_line; | |
1029 | } | |
1030 | if (hit) { | |
1031 | count++; | |
1032 | if (opt->status_only) | |
1033 | return 1; | |
1034 | if (opt->name_only) { | |
1035 | show_name(opt, name); | |
1036 | return 1; | |
1037 | } | |
1038 | if (opt->count) | |
1039 | goto next_line; | |
1040 | if (binary_match_only) { | |
1041 | opt->output(opt, "Binary file ", 12); | |
1042 | output_color(opt, name, strlen(name), | |
1043 | opt->color_filename); | |
1044 | opt->output(opt, " matches\n", 9); | |
1045 | return 1; | |
1046 | } | |
1047 | /* Hit at this line. If we haven't shown the | |
1048 | * pre-context lines, we would need to show them. | |
1049 | */ | |
1050 | if (opt->pre_context) | |
1051 | show_pre_context(opt, name, buf, bol, lno); | |
1052 | else if (opt->funcname) | |
1053 | show_funcname_line(opt, name, buf, bol, lno); | |
1054 | show_line(opt, bol, eol, name, lno, ':'); | |
1055 | last_hit = lno; | |
1056 | } | |
1057 | else if (last_hit && | |
1058 | lno <= last_hit + opt->post_context) { | |
1059 | /* If the last hit is within the post context, | |
1060 | * we need to show this line. | |
1061 | */ | |
1062 | show_line(opt, bol, eol, name, lno, '-'); | |
1063 | } | |
1064 | ||
1065 | next_line: | |
1066 | bol = eol + 1; | |
1067 | if (!left) | |
1068 | break; | |
1069 | left--; | |
1070 | lno++; | |
1071 | } | |
1072 | ||
1073 | if (collect_hits) | |
1074 | return 0; | |
1075 | ||
1076 | if (opt->status_only) | |
1077 | return 0; | |
1078 | if (opt->unmatch_name_only) { | |
1079 | /* We did not see any hit, so we want to show this */ | |
1080 | show_name(opt, name); | |
1081 | return 1; | |
1082 | } | |
1083 | ||
1084 | xdiff_clear_find_func(&xecfg); | |
1085 | opt->priv = NULL; | |
1086 | ||
1087 | /* NEEDSWORK: | |
1088 | * The real "grep -c foo *.c" gives many "bar.c:0" lines, | |
1089 | * which feels mostly useless but sometimes useful. Maybe | |
1090 | * make it another option? For now suppress them. | |
1091 | */ | |
1092 | if (opt->count && count) { | |
1093 | char buf[32]; | |
1094 | output_color(opt, name, strlen(name), opt->color_filename); | |
1095 | output_sep(opt, ':'); | |
1096 | snprintf(buf, sizeof(buf), "%u\n", count); | |
1097 | opt->output(opt, buf, strlen(buf)); | |
1098 | return 1; | |
1099 | } | |
1100 | return !!last_hit; | |
1101 | } | |
1102 | ||
1103 | static void clr_hit_marker(struct grep_expr *x) | |
1104 | { | |
1105 | /* All-hit markers are meaningful only at the very top level | |
1106 | * OR node. | |
1107 | */ | |
1108 | while (1) { | |
1109 | x->hit = 0; | |
1110 | if (x->node != GREP_NODE_OR) | |
1111 | return; | |
1112 | x->u.binary.left->hit = 0; | |
1113 | x = x->u.binary.right; | |
1114 | } | |
1115 | } | |
1116 | ||
1117 | static int chk_hit_marker(struct grep_expr *x) | |
1118 | { | |
1119 | /* Top level nodes have hit markers. See if they all are hits */ | |
1120 | while (1) { | |
1121 | if (x->node != GREP_NODE_OR) | |
1122 | return x->hit; | |
1123 | if (!x->u.binary.left->hit) | |
1124 | return 0; | |
1125 | x = x->u.binary.right; | |
1126 | } | |
1127 | } | |
1128 | ||
1129 | int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size) | |
1130 | { | |
1131 | /* | |
1132 | * we do not have to do the two-pass grep when we do not check | |
1133 | * buffer-wide "all-match". | |
1134 | */ | |
1135 | if (!opt->all_match) | |
1136 | return grep_buffer_1(opt, name, buf, size, 0); | |
1137 | ||
1138 | /* Otherwise the toplevel "or" terms hit a bit differently. | |
1139 | * We first clear hit markers from them. | |
1140 | */ | |
1141 | clr_hit_marker(opt->pattern_expression); | |
1142 | grep_buffer_1(opt, name, buf, size, 1); | |
1143 | ||
1144 | if (!chk_hit_marker(opt->pattern_expression)) | |
1145 | return 0; | |
1146 | ||
1147 | return grep_buffer_1(opt, name, buf, size, 0); | |
1148 | } |