]>
Commit | Line | Data |
---|---|---|
83b5d2f5 | 1 | #include "cache.h" |
83b5d2f5 | 2 | #include "grep.h" |
60ecac98 | 3 | #include "userdiff.h" |
6bfce93e | 4 | #include "xdiff-interface.h" |
83b5d2f5 | 5 | |
a4d7d2c6 JH |
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->origin = "header"; | |
11 | p->no = 0; | |
12 | p->token = GREP_PATTERN_HEAD; | |
13 | p->field = field; | |
14 | *opt->pattern_tail = p; | |
15 | opt->pattern_tail = &p->next; | |
16 | p->next = NULL; | |
17 | } | |
18 | ||
83b5d2f5 JH |
19 | void append_grep_pattern(struct grep_opt *opt, const char *pat, |
20 | const char *origin, int no, enum grep_pat_token t) | |
21 | { | |
22 | struct grep_pat *p = xcalloc(1, sizeof(*p)); | |
23 | p->pattern = pat; | |
24 | p->origin = origin; | |
25 | p->no = no; | |
26 | p->token = t; | |
27 | *opt->pattern_tail = p; | |
28 | opt->pattern_tail = &p->next; | |
29 | p->next = NULL; | |
30 | } | |
31 | ||
5b594f45 FK |
32 | struct grep_opt *grep_opt_dup(const struct grep_opt *opt) |
33 | { | |
34 | struct grep_pat *pat; | |
35 | struct grep_opt *ret = xmalloc(sizeof(struct grep_opt)); | |
36 | *ret = *opt; | |
37 | ||
38 | ret->pattern_list = NULL; | |
39 | ret->pattern_tail = &ret->pattern_list; | |
40 | ||
41 | for(pat = opt->pattern_list; pat != NULL; pat = pat->next) | |
42 | { | |
43 | if(pat->token == GREP_PATTERN_HEAD) | |
44 | append_header_grep_pattern(ret, pat->field, | |
45 | pat->pattern); | |
46 | else | |
47 | append_grep_pattern(ret, pat->pattern, pat->origin, | |
48 | pat->no, pat->token); | |
49 | } | |
50 | ||
51 | return ret; | |
52 | } | |
53 | ||
83b5d2f5 JH |
54 | static void compile_regexp(struct grep_pat *p, struct grep_opt *opt) |
55 | { | |
c822255c RS |
56 | int err; |
57 | ||
d7eb527d | 58 | p->word_regexp = opt->word_regexp; |
5183bf67 | 59 | p->ignore_case = opt->ignore_case; |
d7eb527d | 60 | |
885d211e | 61 | if (opt->fixed) |
c822255c RS |
62 | p->fixed = 1; |
63 | if (opt->regflags & REG_ICASE) | |
64 | p->fixed = 0; | |
65 | if (p->fixed) | |
66 | return; | |
67 | ||
68 | err = regcomp(&p->regexp, p->pattern, opt->regflags); | |
83b5d2f5 JH |
69 | if (err) { |
70 | char errbuf[1024]; | |
71 | char where[1024]; | |
72 | if (p->no) | |
73 | sprintf(where, "In '%s' at %d, ", | |
74 | p->origin, p->no); | |
75 | else if (p->origin) | |
76 | sprintf(where, "%s, ", p->origin); | |
77 | else | |
78 | where[0] = 0; | |
79 | regerror(err, &p->regexp, errbuf, 1024); | |
80 | regfree(&p->regexp); | |
81 | die("%s'%s': %s", where, p->pattern, errbuf); | |
82 | } | |
83 | } | |
84 | ||
0ab7befa | 85 | static struct grep_expr *compile_pattern_or(struct grep_pat **); |
83b5d2f5 JH |
86 | static struct grep_expr *compile_pattern_atom(struct grep_pat **list) |
87 | { | |
88 | struct grep_pat *p; | |
89 | struct grep_expr *x; | |
90 | ||
91 | p = *list; | |
c922b01f LT |
92 | if (!p) |
93 | return NULL; | |
83b5d2f5 JH |
94 | switch (p->token) { |
95 | case GREP_PATTERN: /* atom */ | |
480c1ca6 JH |
96 | case GREP_PATTERN_HEAD: |
97 | case GREP_PATTERN_BODY: | |
83b5d2f5 JH |
98 | x = xcalloc(1, sizeof (struct grep_expr)); |
99 | x->node = GREP_NODE_ATOM; | |
100 | x->u.atom = p; | |
101 | *list = p->next; | |
102 | return x; | |
103 | case GREP_OPEN_PAREN: | |
104 | *list = p->next; | |
0ab7befa | 105 | x = compile_pattern_or(list); |
83b5d2f5 JH |
106 | if (!*list || (*list)->token != GREP_CLOSE_PAREN) |
107 | die("unmatched parenthesis"); | |
108 | *list = (*list)->next; | |
109 | return x; | |
110 | default: | |
111 | return NULL; | |
112 | } | |
113 | } | |
114 | ||
115 | static struct grep_expr *compile_pattern_not(struct grep_pat **list) | |
116 | { | |
117 | struct grep_pat *p; | |
118 | struct grep_expr *x; | |
119 | ||
120 | p = *list; | |
c922b01f LT |
121 | if (!p) |
122 | return NULL; | |
83b5d2f5 JH |
123 | switch (p->token) { |
124 | case GREP_NOT: | |
125 | if (!p->next) | |
126 | die("--not not followed by pattern expression"); | |
127 | *list = p->next; | |
128 | x = xcalloc(1, sizeof (struct grep_expr)); | |
129 | x->node = GREP_NODE_NOT; | |
130 | x->u.unary = compile_pattern_not(list); | |
131 | if (!x->u.unary) | |
132 | die("--not followed by non pattern expression"); | |
133 | return x; | |
134 | default: | |
135 | return compile_pattern_atom(list); | |
136 | } | |
137 | } | |
138 | ||
139 | static struct grep_expr *compile_pattern_and(struct grep_pat **list) | |
140 | { | |
141 | struct grep_pat *p; | |
142 | struct grep_expr *x, *y, *z; | |
143 | ||
144 | x = compile_pattern_not(list); | |
145 | p = *list; | |
146 | if (p && p->token == GREP_AND) { | |
147 | if (!p->next) | |
148 | die("--and not followed by pattern expression"); | |
149 | *list = p->next; | |
150 | y = compile_pattern_and(list); | |
151 | if (!y) | |
152 | die("--and not followed by pattern expression"); | |
153 | z = xcalloc(1, sizeof (struct grep_expr)); | |
154 | z->node = GREP_NODE_AND; | |
155 | z->u.binary.left = x; | |
156 | z->u.binary.right = y; | |
157 | return z; | |
158 | } | |
159 | return x; | |
160 | } | |
161 | ||
162 | static struct grep_expr *compile_pattern_or(struct grep_pat **list) | |
163 | { | |
164 | struct grep_pat *p; | |
165 | struct grep_expr *x, *y, *z; | |
166 | ||
167 | x = compile_pattern_and(list); | |
168 | p = *list; | |
169 | if (x && p && p->token != GREP_CLOSE_PAREN) { | |
170 | y = compile_pattern_or(list); | |
171 | if (!y) | |
172 | die("not a pattern expression %s", p->pattern); | |
173 | z = xcalloc(1, sizeof (struct grep_expr)); | |
174 | z->node = GREP_NODE_OR; | |
175 | z->u.binary.left = x; | |
176 | z->u.binary.right = y; | |
177 | return z; | |
178 | } | |
179 | return x; | |
180 | } | |
181 | ||
182 | static struct grep_expr *compile_pattern_expr(struct grep_pat **list) | |
183 | { | |
184 | return compile_pattern_or(list); | |
185 | } | |
186 | ||
187 | void compile_grep_patterns(struct grep_opt *opt) | |
188 | { | |
189 | struct grep_pat *p; | |
190 | ||
0ab7befa JH |
191 | if (opt->all_match) |
192 | opt->extended = 1; | |
193 | ||
83b5d2f5 | 194 | for (p = opt->pattern_list; p; p = p->next) { |
480c1ca6 JH |
195 | switch (p->token) { |
196 | case GREP_PATTERN: /* atom */ | |
197 | case GREP_PATTERN_HEAD: | |
198 | case GREP_PATTERN_BODY: | |
c822255c | 199 | compile_regexp(p, opt); |
480c1ca6 JH |
200 | break; |
201 | default: | |
83b5d2f5 | 202 | opt->extended = 1; |
480c1ca6 JH |
203 | break; |
204 | } | |
83b5d2f5 JH |
205 | } |
206 | ||
207 | if (!opt->extended) | |
208 | return; | |
209 | ||
210 | /* Then bundle them up in an expression. | |
211 | * A classic recursive descent parser would do. | |
212 | */ | |
213 | p = opt->pattern_list; | |
ba150a3f MB |
214 | if (p) |
215 | opt->pattern_expression = compile_pattern_expr(&p); | |
83b5d2f5 JH |
216 | if (p) |
217 | die("incomplete pattern expression: %s", p->pattern); | |
218 | } | |
219 | ||
b48fb5b6 JH |
220 | static void free_pattern_expr(struct grep_expr *x) |
221 | { | |
222 | switch (x->node) { | |
223 | case GREP_NODE_ATOM: | |
224 | break; | |
225 | case GREP_NODE_NOT: | |
226 | free_pattern_expr(x->u.unary); | |
227 | break; | |
228 | case GREP_NODE_AND: | |
229 | case GREP_NODE_OR: | |
230 | free_pattern_expr(x->u.binary.left); | |
231 | free_pattern_expr(x->u.binary.right); | |
232 | break; | |
233 | } | |
234 | free(x); | |
235 | } | |
236 | ||
237 | void free_grep_patterns(struct grep_opt *opt) | |
238 | { | |
239 | struct grep_pat *p, *n; | |
240 | ||
241 | for (p = opt->pattern_list; p; p = n) { | |
242 | n = p->next; | |
243 | switch (p->token) { | |
244 | case GREP_PATTERN: /* atom */ | |
245 | case GREP_PATTERN_HEAD: | |
246 | case GREP_PATTERN_BODY: | |
247 | regfree(&p->regexp); | |
248 | break; | |
249 | default: | |
250 | break; | |
251 | } | |
252 | free(p); | |
253 | } | |
254 | ||
255 | if (!opt->extended) | |
256 | return; | |
257 | free_pattern_expr(opt->pattern_expression); | |
258 | } | |
259 | ||
83b5d2f5 JH |
260 | static char *end_of_line(char *cp, unsigned long *left) |
261 | { | |
262 | unsigned long l = *left; | |
263 | while (l && *cp != '\n') { | |
264 | l--; | |
265 | cp++; | |
266 | } | |
267 | *left = l; | |
268 | return cp; | |
269 | } | |
270 | ||
271 | static int word_char(char ch) | |
272 | { | |
273 | return isalnum(ch) || ch == '_'; | |
274 | } | |
275 | ||
83caecca RZ |
276 | static void show_name(struct grep_opt *opt, const char *name) |
277 | { | |
5b594f45 FK |
278 | opt->output(opt, name, strlen(name)); |
279 | opt->output(opt, opt->null_following_name ? "\0" : "\n", 1); | |
83caecca RZ |
280 | } |
281 | ||
5183bf67 BC |
282 | |
283 | static int fixmatch(const char *pattern, char *line, int ignore_case, regmatch_t *match) | |
83b5d2f5 | 284 | { |
5183bf67 BC |
285 | char *hit; |
286 | if (ignore_case) | |
287 | hit = strcasestr(line, pattern); | |
288 | else | |
289 | hit = strstr(line, pattern); | |
290 | ||
83b5d2f5 JH |
291 | if (!hit) { |
292 | match->rm_so = match->rm_eo = -1; | |
293 | return REG_NOMATCH; | |
294 | } | |
295 | else { | |
296 | match->rm_so = hit - line; | |
297 | match->rm_eo = match->rm_so + strlen(pattern); | |
298 | return 0; | |
299 | } | |
300 | } | |
301 | ||
a4d7d2c6 JH |
302 | static int strip_timestamp(char *bol, char **eol_p) |
303 | { | |
304 | char *eol = *eol_p; | |
305 | int ch; | |
306 | ||
307 | while (bol < --eol) { | |
308 | if (*eol != '>') | |
309 | continue; | |
310 | *eol_p = ++eol; | |
311 | ch = *eol; | |
312 | *eol = '\0'; | |
313 | return ch; | |
314 | } | |
315 | return 0; | |
316 | } | |
317 | ||
318 | static struct { | |
319 | const char *field; | |
320 | size_t len; | |
321 | } header_field[] = { | |
322 | { "author ", 7 }, | |
323 | { "committer ", 10 }, | |
324 | }; | |
325 | ||
d7eb527d | 326 | static int match_one_pattern(struct grep_pat *p, char *bol, char *eol, |
79212772 RS |
327 | enum grep_context ctx, |
328 | regmatch_t *pmatch, int eflags) | |
83b5d2f5 JH |
329 | { |
330 | int hit = 0; | |
a4d7d2c6 | 331 | int saved_ch = 0; |
e701fadb | 332 | const char *start = bol; |
83b5d2f5 | 333 | |
480c1ca6 JH |
334 | if ((p->token != GREP_PATTERN) && |
335 | ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD))) | |
336 | return 0; | |
337 | ||
a4d7d2c6 JH |
338 | if (p->token == GREP_PATTERN_HEAD) { |
339 | const char *field; | |
340 | size_t len; | |
341 | assert(p->field < ARRAY_SIZE(header_field)); | |
342 | field = header_field[p->field].field; | |
343 | len = header_field[p->field].len; | |
344 | if (strncmp(bol, field, len)) | |
345 | return 0; | |
346 | bol += len; | |
347 | saved_ch = strip_timestamp(bol, &eol); | |
348 | } | |
349 | ||
83b5d2f5 | 350 | again: |
79212772 | 351 | if (p->fixed) |
5183bf67 | 352 | hit = !fixmatch(p->pattern, bol, p->ignore_case, pmatch); |
79212772 RS |
353 | else |
354 | hit = !regexec(&p->regexp, bol, 1, pmatch, eflags); | |
83b5d2f5 | 355 | |
d7eb527d | 356 | if (hit && p->word_regexp) { |
83b5d2f5 | 357 | if ((pmatch[0].rm_so < 0) || |
84201eae | 358 | (eol - bol) < pmatch[0].rm_so || |
83b5d2f5 JH |
359 | (pmatch[0].rm_eo < 0) || |
360 | (eol - bol) < pmatch[0].rm_eo) | |
361 | die("regexp returned nonsense"); | |
362 | ||
363 | /* Match beginning must be either beginning of the | |
364 | * line, or at word boundary (i.e. the last char must | |
365 | * not be a word char). Similarly, match end must be | |
366 | * either end of the line, or at word boundary | |
367 | * (i.e. the next char must not be a word char). | |
368 | */ | |
fb62eb7f | 369 | if ( ((pmatch[0].rm_so == 0) || |
83b5d2f5 JH |
370 | !word_char(bol[pmatch[0].rm_so-1])) && |
371 | ((pmatch[0].rm_eo == (eol-bol)) || | |
372 | !word_char(bol[pmatch[0].rm_eo])) ) | |
373 | ; | |
374 | else | |
375 | hit = 0; | |
376 | ||
84201eae RS |
377 | /* Words consist of at least one character. */ |
378 | if (pmatch->rm_so == pmatch->rm_eo) | |
379 | hit = 0; | |
380 | ||
83b5d2f5 JH |
381 | if (!hit && pmatch[0].rm_so + bol + 1 < eol) { |
382 | /* There could be more than one match on the | |
383 | * line, and the first match might not be | |
384 | * strict word match. But later ones could be! | |
fb62eb7f RS |
385 | * Forward to the next possible start, i.e. the |
386 | * next position following a non-word char. | |
83b5d2f5 JH |
387 | */ |
388 | bol = pmatch[0].rm_so + bol + 1; | |
fb62eb7f RS |
389 | while (word_char(bol[-1]) && bol < eol) |
390 | bol++; | |
dbb6a4ad | 391 | eflags |= REG_NOTBOL; |
fb62eb7f RS |
392 | if (bol < eol) |
393 | goto again; | |
83b5d2f5 JH |
394 | } |
395 | } | |
a4d7d2c6 JH |
396 | if (p->token == GREP_PATTERN_HEAD && saved_ch) |
397 | *eol = saved_ch; | |
e701fadb RS |
398 | if (hit) { |
399 | pmatch[0].rm_so += bol - start; | |
400 | pmatch[0].rm_eo += bol - start; | |
401 | } | |
83b5d2f5 JH |
402 | return hit; |
403 | } | |
404 | ||
d7eb527d RS |
405 | static int match_expr_eval(struct grep_expr *x, char *bol, char *eol, |
406 | enum grep_context ctx, int collect_hits) | |
83b5d2f5 | 407 | { |
0ab7befa | 408 | int h = 0; |
79212772 | 409 | regmatch_t match; |
0ab7befa | 410 | |
c922b01f LT |
411 | if (!x) |
412 | die("Not a valid grep expression"); | |
83b5d2f5 JH |
413 | switch (x->node) { |
414 | case GREP_NODE_ATOM: | |
79212772 | 415 | h = match_one_pattern(x->u.atom, bol, eol, ctx, &match, 0); |
83b5d2f5 JH |
416 | break; |
417 | case GREP_NODE_NOT: | |
d7eb527d | 418 | h = !match_expr_eval(x->u.unary, bol, eol, ctx, 0); |
0ab7befa | 419 | break; |
83b5d2f5 | 420 | case GREP_NODE_AND: |
d7eb527d | 421 | if (!match_expr_eval(x->u.binary.left, bol, eol, ctx, 0)) |
252d560d | 422 | return 0; |
d7eb527d | 423 | h = match_expr_eval(x->u.binary.right, bol, eol, ctx, 0); |
0ab7befa | 424 | break; |
83b5d2f5 | 425 | case GREP_NODE_OR: |
0ab7befa | 426 | if (!collect_hits) |
d7eb527d | 427 | return (match_expr_eval(x->u.binary.left, |
0ab7befa | 428 | bol, eol, ctx, 0) || |
d7eb527d | 429 | match_expr_eval(x->u.binary.right, |
0ab7befa | 430 | bol, eol, ctx, 0)); |
d7eb527d | 431 | h = match_expr_eval(x->u.binary.left, bol, eol, ctx, 0); |
0ab7befa | 432 | x->u.binary.left->hit |= h; |
d7eb527d | 433 | h |= match_expr_eval(x->u.binary.right, bol, eol, ctx, 1); |
0ab7befa JH |
434 | break; |
435 | default: | |
d7530708 | 436 | die("Unexpected node type (internal error) %d", x->node); |
83b5d2f5 | 437 | } |
0ab7befa JH |
438 | if (collect_hits) |
439 | x->hit |= h; | |
440 | return h; | |
83b5d2f5 JH |
441 | } |
442 | ||
480c1ca6 | 443 | static int match_expr(struct grep_opt *opt, char *bol, char *eol, |
0ab7befa | 444 | enum grep_context ctx, int collect_hits) |
83b5d2f5 JH |
445 | { |
446 | struct grep_expr *x = opt->pattern_expression; | |
d7eb527d | 447 | return match_expr_eval(x, bol, eol, ctx, collect_hits); |
83b5d2f5 JH |
448 | } |
449 | ||
480c1ca6 | 450 | static int match_line(struct grep_opt *opt, char *bol, char *eol, |
0ab7befa | 451 | enum grep_context ctx, int collect_hits) |
83b5d2f5 JH |
452 | { |
453 | struct grep_pat *p; | |
79212772 RS |
454 | regmatch_t match; |
455 | ||
83b5d2f5 | 456 | if (opt->extended) |
0ab7befa JH |
457 | return match_expr(opt, bol, eol, ctx, collect_hits); |
458 | ||
459 | /* we do not call with collect_hits without being extended */ | |
83b5d2f5 | 460 | for (p = opt->pattern_list; p; p = p->next) { |
79212772 | 461 | if (match_one_pattern(p, bol, eol, ctx, &match, 0)) |
83b5d2f5 JH |
462 | return 1; |
463 | } | |
464 | return 0; | |
465 | } | |
466 | ||
7e8f59d5 RS |
467 | static int match_next_pattern(struct grep_pat *p, char *bol, char *eol, |
468 | enum grep_context ctx, | |
469 | regmatch_t *pmatch, int eflags) | |
470 | { | |
471 | regmatch_t match; | |
472 | ||
473 | if (!match_one_pattern(p, bol, eol, ctx, &match, eflags)) | |
474 | return 0; | |
475 | if (match.rm_so < 0 || match.rm_eo < 0) | |
476 | return 0; | |
477 | if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) { | |
478 | if (match.rm_so > pmatch->rm_so) | |
479 | return 1; | |
480 | if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo) | |
481 | return 1; | |
482 | } | |
483 | pmatch->rm_so = match.rm_so; | |
484 | pmatch->rm_eo = match.rm_eo; | |
485 | return 1; | |
486 | } | |
487 | ||
488 | static int next_match(struct grep_opt *opt, char *bol, char *eol, | |
489 | enum grep_context ctx, regmatch_t *pmatch, int eflags) | |
490 | { | |
491 | struct grep_pat *p; | |
492 | int hit = 0; | |
493 | ||
494 | pmatch->rm_so = pmatch->rm_eo = -1; | |
495 | if (bol < eol) { | |
496 | for (p = opt->pattern_list; p; p = p->next) { | |
497 | switch (p->token) { | |
498 | case GREP_PATTERN: /* atom */ | |
499 | case GREP_PATTERN_HEAD: | |
500 | case GREP_PATTERN_BODY: | |
501 | hit |= match_next_pattern(p, bol, eol, ctx, | |
502 | pmatch, eflags); | |
503 | break; | |
504 | default: | |
505 | break; | |
506 | } | |
507 | } | |
508 | } | |
509 | return hit; | |
510 | } | |
511 | ||
512 | static void show_line(struct grep_opt *opt, char *bol, char *eol, | |
513 | const char *name, unsigned lno, char sign) | |
514 | { | |
515 | int rest = eol - bol; | |
5b594f45 | 516 | char sign_str[1]; |
7e8f59d5 | 517 | |
5b594f45 | 518 | sign_str[0] = sign; |
ed24e401 | 519 | if (opt->pre_context || opt->post_context) { |
046802d0 RS |
520 | if (opt->last_shown == 0) { |
521 | if (opt->show_hunk_mark) | |
5b594f45 | 522 | opt->output(opt, "--\n", 3); |
046802d0 RS |
523 | else |
524 | opt->show_hunk_mark = 1; | |
ed24e401 | 525 | } else if (lno > opt->last_shown + 1) |
5b594f45 | 526 | opt->output(opt, "--\n", 3); |
5dd06d38 RS |
527 | } |
528 | opt->last_shown = lno; | |
529 | ||
7e8f59d5 | 530 | if (opt->null_following_name) |
5b594f45 FK |
531 | sign_str[0] = '\0'; |
532 | if (opt->pathname) { | |
533 | opt->output(opt, name, strlen(name)); | |
534 | opt->output(opt, sign_str, 1); | |
535 | } | |
536 | if (opt->linenum) { | |
537 | char buf[32]; | |
538 | snprintf(buf, sizeof(buf), "%d", lno); | |
539 | opt->output(opt, buf, strlen(buf)); | |
540 | opt->output(opt, sign_str, 1); | |
541 | } | |
7e8f59d5 RS |
542 | if (opt->color) { |
543 | regmatch_t match; | |
544 | enum grep_context ctx = GREP_CONTEXT_BODY; | |
545 | int ch = *eol; | |
546 | int eflags = 0; | |
547 | ||
548 | *eol = '\0'; | |
549 | while (next_match(opt, bol, eol, ctx, &match, eflags)) { | |
1f5b9cc4 RS |
550 | if (match.rm_so == match.rm_eo) |
551 | break; | |
5b594f45 FK |
552 | |
553 | opt->output(opt, bol, match.rm_so); | |
554 | opt->output(opt, opt->color_match, | |
555 | strlen(opt->color_match)); | |
556 | opt->output(opt, bol + match.rm_so, | |
557 | (int)(match.rm_eo - match.rm_so)); | |
558 | opt->output(opt, GIT_COLOR_RESET, | |
559 | strlen(GIT_COLOR_RESET)); | |
7e8f59d5 RS |
560 | bol += match.rm_eo; |
561 | rest -= match.rm_eo; | |
562 | eflags = REG_NOTBOL; | |
563 | } | |
564 | *eol = ch; | |
565 | } | |
5b594f45 FK |
566 | opt->output(opt, bol, rest); |
567 | opt->output(opt, "\n", 1); | |
7e8f59d5 RS |
568 | } |
569 | ||
60ecac98 | 570 | static int match_funcname(struct grep_opt *opt, char *bol, char *eol) |
2944e4e6 | 571 | { |
60ecac98 RS |
572 | xdemitconf_t *xecfg = opt->priv; |
573 | if (xecfg && xecfg->find_func) { | |
574 | char buf[1]; | |
575 | return xecfg->find_func(bol, eol - bol, buf, 1, | |
576 | xecfg->find_func_priv) >= 0; | |
577 | } | |
578 | ||
2944e4e6 RS |
579 | if (bol == eol) |
580 | return 0; | |
581 | if (isalpha(*bol) || *bol == '_' || *bol == '$') | |
582 | return 1; | |
583 | return 0; | |
584 | } | |
585 | ||
586 | static void show_funcname_line(struct grep_opt *opt, const char *name, | |
587 | char *buf, char *bol, unsigned lno) | |
588 | { | |
589 | while (bol > buf) { | |
590 | char *eol = --bol; | |
591 | ||
592 | while (bol > buf && bol[-1] != '\n') | |
593 | bol--; | |
594 | lno--; | |
595 | ||
596 | if (lno <= opt->last_shown) | |
597 | break; | |
598 | ||
60ecac98 | 599 | if (match_funcname(opt, bol, eol)) { |
2944e4e6 RS |
600 | show_line(opt, bol, eol, name, lno, '='); |
601 | break; | |
602 | } | |
603 | } | |
604 | } | |
605 | ||
49de3216 RS |
606 | static void show_pre_context(struct grep_opt *opt, const char *name, char *buf, |
607 | char *bol, unsigned lno) | |
608 | { | |
2944e4e6 RS |
609 | unsigned cur = lno, from = 1, funcname_lno = 0; |
610 | int funcname_needed = opt->funcname; | |
49de3216 RS |
611 | |
612 | if (opt->pre_context < lno) | |
613 | from = lno - opt->pre_context; | |
614 | if (from <= opt->last_shown) | |
615 | from = opt->last_shown + 1; | |
616 | ||
617 | /* Rewind. */ | |
618 | while (bol > buf && cur > from) { | |
2944e4e6 RS |
619 | char *eol = --bol; |
620 | ||
49de3216 RS |
621 | while (bol > buf && bol[-1] != '\n') |
622 | bol--; | |
623 | cur--; | |
60ecac98 | 624 | if (funcname_needed && match_funcname(opt, bol, eol)) { |
2944e4e6 RS |
625 | funcname_lno = cur; |
626 | funcname_needed = 0; | |
627 | } | |
49de3216 RS |
628 | } |
629 | ||
2944e4e6 RS |
630 | /* We need to look even further back to find a function signature. */ |
631 | if (opt->funcname && funcname_needed) | |
632 | show_funcname_line(opt, name, buf, bol, cur); | |
633 | ||
49de3216 RS |
634 | /* Back forward. */ |
635 | while (cur < lno) { | |
2944e4e6 | 636 | char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-'; |
49de3216 RS |
637 | |
638 | while (*eol != '\n') | |
639 | eol++; | |
2944e4e6 | 640 | show_line(opt, bol, eol, name, cur, sign); |
49de3216 RS |
641 | bol = eol + 1; |
642 | cur++; | |
643 | } | |
644 | } | |
645 | ||
a26345b6 JH |
646 | static int should_lookahead(struct grep_opt *opt) |
647 | { | |
648 | struct grep_pat *p; | |
649 | ||
650 | if (opt->extended) | |
651 | return 0; /* punt for too complex stuff */ | |
652 | if (opt->invert) | |
653 | return 0; | |
654 | for (p = opt->pattern_list; p; p = p->next) { | |
655 | if (p->token != GREP_PATTERN) | |
656 | return 0; /* punt for "header only" and stuff */ | |
657 | } | |
658 | return 1; | |
659 | } | |
660 | ||
661 | static int look_ahead(struct grep_opt *opt, | |
662 | unsigned long *left_p, | |
663 | unsigned *lno_p, | |
664 | char **bol_p) | |
665 | { | |
666 | unsigned lno = *lno_p; | |
667 | char *bol = *bol_p; | |
668 | struct grep_pat *p; | |
669 | char *sp, *last_bol; | |
670 | regoff_t earliest = -1; | |
671 | ||
672 | for (p = opt->pattern_list; p; p = p->next) { | |
673 | int hit; | |
674 | regmatch_t m; | |
675 | ||
676 | if (p->fixed) | |
e2d2e383 | 677 | hit = !fixmatch(p->pattern, bol, p->ignore_case, &m); |
24072c02 BK |
678 | else { |
679 | #ifdef REG_STARTEND | |
680 | m.rm_so = 0; | |
681 | m.rm_eo = *left_p; | |
682 | hit = !regexec(&p->regexp, bol, 1, &m, REG_STARTEND); | |
683 | #else | |
a26345b6 | 684 | hit = !regexec(&p->regexp, bol, 1, &m, 0); |
24072c02 BK |
685 | #endif |
686 | } | |
a26345b6 JH |
687 | if (!hit || m.rm_so < 0 || m.rm_eo < 0) |
688 | continue; | |
689 | if (earliest < 0 || m.rm_so < earliest) | |
690 | earliest = m.rm_so; | |
691 | } | |
692 | ||
693 | if (earliest < 0) { | |
694 | *bol_p = bol + *left_p; | |
695 | *left_p = 0; | |
696 | return 1; | |
697 | } | |
698 | for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--) | |
699 | ; /* find the beginning of the line */ | |
700 | last_bol = sp; | |
701 | ||
702 | for (sp = bol; sp < last_bol; sp++) { | |
703 | if (*sp == '\n') | |
704 | lno++; | |
705 | } | |
706 | *left_p -= last_bol - bol; | |
707 | *bol_p = last_bol; | |
708 | *lno_p = lno; | |
709 | return 0; | |
710 | } | |
711 | ||
5b594f45 FK |
712 | int grep_threads_ok(const struct grep_opt *opt) |
713 | { | |
714 | /* If this condition is true, then we may use the attribute | |
715 | * machinery in grep_buffer_1. The attribute code is not | |
716 | * thread safe, so we disable the use of threads. | |
717 | */ | |
718 | if (opt->funcname && !opt->unmatch_name_only && !opt->status_only && | |
719 | !opt->name_only) | |
720 | return 0; | |
721 | ||
722 | /* If we are showing hunk marks, we should not do it for the | |
723 | * first match. The synchronization problem we get for this | |
724 | * constraint is not yet solved, so we disable threading in | |
725 | * this case. | |
726 | */ | |
727 | if (opt->pre_context || opt->post_context) | |
728 | return 0; | |
729 | ||
730 | return 1; | |
731 | } | |
732 | ||
733 | static void std_output(struct grep_opt *opt, const void *buf, size_t size) | |
734 | { | |
735 | fwrite(buf, size, 1, stdout); | |
736 | } | |
737 | ||
0ab7befa JH |
738 | static int grep_buffer_1(struct grep_opt *opt, const char *name, |
739 | char *buf, unsigned long size, int collect_hits) | |
83b5d2f5 JH |
740 | { |
741 | char *bol = buf; | |
742 | unsigned long left = size; | |
743 | unsigned lno = 1; | |
83b5d2f5 | 744 | unsigned last_hit = 0; |
83b5d2f5 | 745 | int binary_match_only = 0; |
83b5d2f5 | 746 | unsigned count = 0; |
a26345b6 | 747 | int try_lookahead = 0; |
480c1ca6 | 748 | enum grep_context ctx = GREP_CONTEXT_HEAD; |
60ecac98 | 749 | xdemitconf_t xecfg; |
83b5d2f5 | 750 | |
5dd06d38 RS |
751 | opt->last_shown = 0; |
752 | ||
5b594f45 FK |
753 | if (!opt->output) |
754 | opt->output = std_output; | |
755 | ||
83b5d2f5 JH |
756 | if (buffer_is_binary(buf, size)) { |
757 | switch (opt->binary) { | |
758 | case GREP_BINARY_DEFAULT: | |
759 | binary_match_only = 1; | |
760 | break; | |
761 | case GREP_BINARY_NOMATCH: | |
762 | return 0; /* Assume unmatch */ | |
763 | break; | |
764 | default: | |
765 | break; | |
766 | } | |
767 | } | |
768 | ||
60ecac98 RS |
769 | memset(&xecfg, 0, sizeof(xecfg)); |
770 | if (opt->funcname && !opt->unmatch_name_only && !opt->status_only && | |
771 | !opt->name_only && !binary_match_only && !collect_hits) { | |
772 | struct userdiff_driver *drv = userdiff_find_by_path(name); | |
773 | if (drv && drv->funcname.pattern) { | |
774 | const struct userdiff_funcname *pe = &drv->funcname; | |
775 | xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags); | |
776 | opt->priv = &xecfg; | |
777 | } | |
778 | } | |
a26345b6 | 779 | try_lookahead = should_lookahead(opt); |
60ecac98 | 780 | |
83b5d2f5 JH |
781 | while (left) { |
782 | char *eol, ch; | |
0ab7befa | 783 | int hit; |
83b5d2f5 | 784 | |
a26345b6 JH |
785 | /* |
786 | * look_ahead() skips quicly to the line that possibly | |
787 | * has the next hit; don't call it if we need to do | |
788 | * something more than just skipping the current line | |
789 | * in response to an unmatch for the current line. E.g. | |
790 | * inside a post-context window, we will show the current | |
791 | * line as a context around the previous hit when it | |
792 | * doesn't hit. | |
793 | */ | |
794 | if (try_lookahead | |
795 | && !(last_hit | |
796 | && lno <= last_hit + opt->post_context) | |
797 | && look_ahead(opt, &left, &lno, &bol)) | |
798 | break; | |
83b5d2f5 JH |
799 | eol = end_of_line(bol, &left); |
800 | ch = *eol; | |
801 | *eol = 0; | |
802 | ||
480c1ca6 JH |
803 | if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol)) |
804 | ctx = GREP_CONTEXT_BODY; | |
805 | ||
0ab7befa | 806 | hit = match_line(opt, bol, eol, ctx, collect_hits); |
83b5d2f5 JH |
807 | *eol = ch; |
808 | ||
0ab7befa JH |
809 | if (collect_hits) |
810 | goto next_line; | |
811 | ||
83b5d2f5 JH |
812 | /* "grep -v -e foo -e bla" should list lines |
813 | * that do not have either, so inversion should | |
814 | * be done outside. | |
815 | */ | |
816 | if (opt->invert) | |
817 | hit = !hit; | |
818 | if (opt->unmatch_name_only) { | |
819 | if (hit) | |
820 | return 0; | |
821 | goto next_line; | |
822 | } | |
823 | if (hit) { | |
824 | count++; | |
825 | if (opt->status_only) | |
826 | return 1; | |
827 | if (binary_match_only) { | |
5b594f45 FK |
828 | opt->output(opt, "Binary file ", 12); |
829 | opt->output(opt, name, strlen(name)); | |
830 | opt->output(opt, " matches\n", 9); | |
83b5d2f5 JH |
831 | return 1; |
832 | } | |
833 | if (opt->name_only) { | |
83caecca | 834 | show_name(opt, name); |
83b5d2f5 JH |
835 | return 1; |
836 | } | |
837 | /* Hit at this line. If we haven't shown the | |
838 | * pre-context lines, we would need to show them. | |
839 | * When asked to do "count", this still show | |
840 | * the context which is nonsense, but the user | |
841 | * deserves to get that ;-). | |
842 | */ | |
49de3216 RS |
843 | if (opt->pre_context) |
844 | show_pre_context(opt, name, buf, bol, lno); | |
2944e4e6 RS |
845 | else if (opt->funcname) |
846 | show_funcname_line(opt, name, buf, bol, lno); | |
83b5d2f5 JH |
847 | if (!opt->count) |
848 | show_line(opt, bol, eol, name, lno, ':'); | |
5dd06d38 | 849 | last_hit = lno; |
83b5d2f5 JH |
850 | } |
851 | else if (last_hit && | |
852 | lno <= last_hit + opt->post_context) { | |
853 | /* If the last hit is within the post context, | |
854 | * we need to show this line. | |
855 | */ | |
83b5d2f5 | 856 | show_line(opt, bol, eol, name, lno, '-'); |
83b5d2f5 | 857 | } |
83b5d2f5 JH |
858 | |
859 | next_line: | |
860 | bol = eol + 1; | |
861 | if (!left) | |
862 | break; | |
863 | left--; | |
864 | lno++; | |
865 | } | |
866 | ||
0ab7befa JH |
867 | if (collect_hits) |
868 | return 0; | |
b48fb5b6 | 869 | |
83b5d2f5 JH |
870 | if (opt->status_only) |
871 | return 0; | |
872 | if (opt->unmatch_name_only) { | |
873 | /* We did not see any hit, so we want to show this */ | |
83caecca | 874 | show_name(opt, name); |
83b5d2f5 JH |
875 | return 1; |
876 | } | |
877 | ||
60ecac98 RS |
878 | xdiff_clear_find_func(&xecfg); |
879 | opt->priv = NULL; | |
880 | ||
83b5d2f5 JH |
881 | /* NEEDSWORK: |
882 | * The real "grep -c foo *.c" gives many "bar.c:0" lines, | |
883 | * which feels mostly useless but sometimes useful. Maybe | |
884 | * make it another option? For now suppress them. | |
885 | */ | |
5b594f45 FK |
886 | if (opt->count && count) { |
887 | char buf[32]; | |
888 | opt->output(opt, name, strlen(name)); | |
889 | snprintf(buf, sizeof(buf), "%c%u\n", | |
890 | opt->null_following_name ? '\0' : ':', count); | |
891 | opt->output(opt, buf, strlen(buf)); | |
892 | } | |
83b5d2f5 JH |
893 | return !!last_hit; |
894 | } | |
895 | ||
0ab7befa JH |
896 | static void clr_hit_marker(struct grep_expr *x) |
897 | { | |
898 | /* All-hit markers are meaningful only at the very top level | |
899 | * OR node. | |
900 | */ | |
901 | while (1) { | |
902 | x->hit = 0; | |
903 | if (x->node != GREP_NODE_OR) | |
904 | return; | |
905 | x->u.binary.left->hit = 0; | |
906 | x = x->u.binary.right; | |
907 | } | |
908 | } | |
909 | ||
910 | static int chk_hit_marker(struct grep_expr *x) | |
911 | { | |
912 | /* Top level nodes have hit markers. See if they all are hits */ | |
913 | while (1) { | |
914 | if (x->node != GREP_NODE_OR) | |
915 | return x->hit; | |
916 | if (!x->u.binary.left->hit) | |
917 | return 0; | |
918 | x = x->u.binary.right; | |
919 | } | |
920 | } | |
921 | ||
922 | int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size) | |
923 | { | |
924 | /* | |
925 | * we do not have to do the two-pass grep when we do not check | |
926 | * buffer-wide "all-match". | |
927 | */ | |
928 | if (!opt->all_match) | |
929 | return grep_buffer_1(opt, name, buf, size, 0); | |
930 | ||
931 | /* Otherwise the toplevel "or" terms hit a bit differently. | |
932 | * We first clear hit markers from them. | |
933 | */ | |
934 | clr_hit_marker(opt->pattern_expression); | |
935 | grep_buffer_1(opt, name, buf, size, 1); | |
936 | ||
937 | if (!chk_hit_marker(opt->pattern_expression)) | |
938 | return 0; | |
939 | ||
940 | return grep_buffer_1(opt, name, buf, size, 0); | |
941 | } |