]>
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; |
79286102 | 60 | p->fixed = opt->fixed; |
d7eb527d | 61 | |
c822255c RS |
62 | if (p->fixed) |
63 | return; | |
64 | ||
65 | err = regcomp(&p->regexp, p->pattern, opt->regflags); | |
83b5d2f5 JH |
66 | if (err) { |
67 | char errbuf[1024]; | |
68 | char where[1024]; | |
69 | if (p->no) | |
70 | sprintf(where, "In '%s' at %d, ", | |
71 | p->origin, p->no); | |
72 | else if (p->origin) | |
73 | sprintf(where, "%s, ", p->origin); | |
74 | else | |
75 | where[0] = 0; | |
76 | regerror(err, &p->regexp, errbuf, 1024); | |
77 | regfree(&p->regexp); | |
78 | die("%s'%s': %s", where, p->pattern, errbuf); | |
79 | } | |
80 | } | |
81 | ||
0ab7befa | 82 | static struct grep_expr *compile_pattern_or(struct grep_pat **); |
83b5d2f5 JH |
83 | static struct grep_expr *compile_pattern_atom(struct grep_pat **list) |
84 | { | |
85 | struct grep_pat *p; | |
86 | struct grep_expr *x; | |
87 | ||
88 | p = *list; | |
c922b01f LT |
89 | if (!p) |
90 | return NULL; | |
83b5d2f5 JH |
91 | switch (p->token) { |
92 | case GREP_PATTERN: /* atom */ | |
480c1ca6 JH |
93 | case GREP_PATTERN_HEAD: |
94 | case GREP_PATTERN_BODY: | |
83b5d2f5 JH |
95 | x = xcalloc(1, sizeof (struct grep_expr)); |
96 | x->node = GREP_NODE_ATOM; | |
97 | x->u.atom = p; | |
98 | *list = p->next; | |
99 | return x; | |
100 | case GREP_OPEN_PAREN: | |
101 | *list = p->next; | |
0ab7befa | 102 | x = compile_pattern_or(list); |
83b5d2f5 JH |
103 | if (!*list || (*list)->token != GREP_CLOSE_PAREN) |
104 | die("unmatched parenthesis"); | |
105 | *list = (*list)->next; | |
106 | return x; | |
107 | default: | |
108 | return NULL; | |
109 | } | |
110 | } | |
111 | ||
112 | static struct grep_expr *compile_pattern_not(struct grep_pat **list) | |
113 | { | |
114 | struct grep_pat *p; | |
115 | struct grep_expr *x; | |
116 | ||
117 | p = *list; | |
c922b01f LT |
118 | if (!p) |
119 | return NULL; | |
83b5d2f5 JH |
120 | switch (p->token) { |
121 | case GREP_NOT: | |
122 | if (!p->next) | |
123 | die("--not not followed by pattern expression"); | |
124 | *list = p->next; | |
125 | x = xcalloc(1, sizeof (struct grep_expr)); | |
126 | x->node = GREP_NODE_NOT; | |
127 | x->u.unary = compile_pattern_not(list); | |
128 | if (!x->u.unary) | |
129 | die("--not followed by non pattern expression"); | |
130 | return x; | |
131 | default: | |
132 | return compile_pattern_atom(list); | |
133 | } | |
134 | } | |
135 | ||
136 | static struct grep_expr *compile_pattern_and(struct grep_pat **list) | |
137 | { | |
138 | struct grep_pat *p; | |
139 | struct grep_expr *x, *y, *z; | |
140 | ||
141 | x = compile_pattern_not(list); | |
142 | p = *list; | |
143 | if (p && p->token == GREP_AND) { | |
144 | if (!p->next) | |
145 | die("--and not followed by pattern expression"); | |
146 | *list = p->next; | |
147 | y = compile_pattern_and(list); | |
148 | if (!y) | |
149 | die("--and not followed by pattern expression"); | |
150 | z = xcalloc(1, sizeof (struct grep_expr)); | |
151 | z->node = GREP_NODE_AND; | |
152 | z->u.binary.left = x; | |
153 | z->u.binary.right = y; | |
154 | return z; | |
155 | } | |
156 | return x; | |
157 | } | |
158 | ||
159 | static struct grep_expr *compile_pattern_or(struct grep_pat **list) | |
160 | { | |
161 | struct grep_pat *p; | |
162 | struct grep_expr *x, *y, *z; | |
163 | ||
164 | x = compile_pattern_and(list); | |
165 | p = *list; | |
166 | if (x && p && p->token != GREP_CLOSE_PAREN) { | |
167 | y = compile_pattern_or(list); | |
168 | if (!y) | |
169 | die("not a pattern expression %s", p->pattern); | |
170 | z = xcalloc(1, sizeof (struct grep_expr)); | |
171 | z->node = GREP_NODE_OR; | |
172 | z->u.binary.left = x; | |
173 | z->u.binary.right = y; | |
174 | return z; | |
175 | } | |
176 | return x; | |
177 | } | |
178 | ||
179 | static struct grep_expr *compile_pattern_expr(struct grep_pat **list) | |
180 | { | |
181 | return compile_pattern_or(list); | |
182 | } | |
183 | ||
184 | void compile_grep_patterns(struct grep_opt *opt) | |
185 | { | |
186 | struct grep_pat *p; | |
187 | ||
0ab7befa JH |
188 | if (opt->all_match) |
189 | opt->extended = 1; | |
190 | ||
83b5d2f5 | 191 | for (p = opt->pattern_list; p; p = p->next) { |
480c1ca6 JH |
192 | switch (p->token) { |
193 | case GREP_PATTERN: /* atom */ | |
194 | case GREP_PATTERN_HEAD: | |
195 | case GREP_PATTERN_BODY: | |
c822255c | 196 | compile_regexp(p, opt); |
480c1ca6 JH |
197 | break; |
198 | default: | |
83b5d2f5 | 199 | opt->extended = 1; |
480c1ca6 JH |
200 | break; |
201 | } | |
83b5d2f5 JH |
202 | } |
203 | ||
204 | if (!opt->extended) | |
205 | return; | |
206 | ||
207 | /* Then bundle them up in an expression. | |
208 | * A classic recursive descent parser would do. | |
209 | */ | |
210 | p = opt->pattern_list; | |
ba150a3f MB |
211 | if (p) |
212 | opt->pattern_expression = compile_pattern_expr(&p); | |
83b5d2f5 JH |
213 | if (p) |
214 | die("incomplete pattern expression: %s", p->pattern); | |
215 | } | |
216 | ||
b48fb5b6 JH |
217 | static void free_pattern_expr(struct grep_expr *x) |
218 | { | |
219 | switch (x->node) { | |
220 | case GREP_NODE_ATOM: | |
221 | break; | |
222 | case GREP_NODE_NOT: | |
223 | free_pattern_expr(x->u.unary); | |
224 | break; | |
225 | case GREP_NODE_AND: | |
226 | case GREP_NODE_OR: | |
227 | free_pattern_expr(x->u.binary.left); | |
228 | free_pattern_expr(x->u.binary.right); | |
229 | break; | |
230 | } | |
231 | free(x); | |
232 | } | |
233 | ||
234 | void free_grep_patterns(struct grep_opt *opt) | |
235 | { | |
236 | struct grep_pat *p, *n; | |
237 | ||
238 | for (p = opt->pattern_list; p; p = n) { | |
239 | n = p->next; | |
240 | switch (p->token) { | |
241 | case GREP_PATTERN: /* atom */ | |
242 | case GREP_PATTERN_HEAD: | |
243 | case GREP_PATTERN_BODY: | |
244 | regfree(&p->regexp); | |
245 | break; | |
246 | default: | |
247 | break; | |
248 | } | |
249 | free(p); | |
250 | } | |
251 | ||
252 | if (!opt->extended) | |
253 | return; | |
254 | free_pattern_expr(opt->pattern_expression); | |
255 | } | |
256 | ||
83b5d2f5 JH |
257 | static char *end_of_line(char *cp, unsigned long *left) |
258 | { | |
259 | unsigned long l = *left; | |
260 | while (l && *cp != '\n') { | |
261 | l--; | |
262 | cp++; | |
263 | } | |
264 | *left = l; | |
265 | return cp; | |
266 | } | |
267 | ||
268 | static int word_char(char ch) | |
269 | { | |
270 | return isalnum(ch) || ch == '_'; | |
271 | } | |
272 | ||
83caecca RZ |
273 | static void show_name(struct grep_opt *opt, const char *name) |
274 | { | |
5b594f45 FK |
275 | opt->output(opt, name, strlen(name)); |
276 | opt->output(opt, opt->null_following_name ? "\0" : "\n", 1); | |
83caecca RZ |
277 | } |
278 | ||
5183bf67 BC |
279 | |
280 | static int fixmatch(const char *pattern, char *line, int ignore_case, regmatch_t *match) | |
83b5d2f5 | 281 | { |
5183bf67 BC |
282 | char *hit; |
283 | if (ignore_case) | |
284 | hit = strcasestr(line, pattern); | |
285 | else | |
286 | hit = strstr(line, pattern); | |
287 | ||
83b5d2f5 JH |
288 | if (!hit) { |
289 | match->rm_so = match->rm_eo = -1; | |
290 | return REG_NOMATCH; | |
291 | } | |
292 | else { | |
293 | match->rm_so = hit - line; | |
294 | match->rm_eo = match->rm_so + strlen(pattern); | |
295 | return 0; | |
296 | } | |
297 | } | |
298 | ||
a4d7d2c6 JH |
299 | static int strip_timestamp(char *bol, char **eol_p) |
300 | { | |
301 | char *eol = *eol_p; | |
302 | int ch; | |
303 | ||
304 | while (bol < --eol) { | |
305 | if (*eol != '>') | |
306 | continue; | |
307 | *eol_p = ++eol; | |
308 | ch = *eol; | |
309 | *eol = '\0'; | |
310 | return ch; | |
311 | } | |
312 | return 0; | |
313 | } | |
314 | ||
315 | static struct { | |
316 | const char *field; | |
317 | size_t len; | |
318 | } header_field[] = { | |
319 | { "author ", 7 }, | |
320 | { "committer ", 10 }, | |
321 | }; | |
322 | ||
d7eb527d | 323 | static int match_one_pattern(struct grep_pat *p, char *bol, char *eol, |
79212772 RS |
324 | enum grep_context ctx, |
325 | regmatch_t *pmatch, int eflags) | |
83b5d2f5 JH |
326 | { |
327 | int hit = 0; | |
a4d7d2c6 | 328 | int saved_ch = 0; |
e701fadb | 329 | const char *start = bol; |
83b5d2f5 | 330 | |
480c1ca6 JH |
331 | if ((p->token != GREP_PATTERN) && |
332 | ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD))) | |
333 | return 0; | |
334 | ||
a4d7d2c6 JH |
335 | if (p->token == GREP_PATTERN_HEAD) { |
336 | const char *field; | |
337 | size_t len; | |
338 | assert(p->field < ARRAY_SIZE(header_field)); | |
339 | field = header_field[p->field].field; | |
340 | len = header_field[p->field].len; | |
341 | if (strncmp(bol, field, len)) | |
342 | return 0; | |
343 | bol += len; | |
344 | saved_ch = strip_timestamp(bol, &eol); | |
345 | } | |
346 | ||
83b5d2f5 | 347 | again: |
79212772 | 348 | if (p->fixed) |
5183bf67 | 349 | hit = !fixmatch(p->pattern, bol, p->ignore_case, pmatch); |
79212772 RS |
350 | else |
351 | hit = !regexec(&p->regexp, bol, 1, pmatch, eflags); | |
83b5d2f5 | 352 | |
d7eb527d | 353 | if (hit && p->word_regexp) { |
83b5d2f5 | 354 | if ((pmatch[0].rm_so < 0) || |
84201eae | 355 | (eol - bol) < pmatch[0].rm_so || |
83b5d2f5 JH |
356 | (pmatch[0].rm_eo < 0) || |
357 | (eol - bol) < pmatch[0].rm_eo) | |
358 | die("regexp returned nonsense"); | |
359 | ||
360 | /* Match beginning must be either beginning of the | |
361 | * line, or at word boundary (i.e. the last char must | |
362 | * not be a word char). Similarly, match end must be | |
363 | * either end of the line, or at word boundary | |
364 | * (i.e. the next char must not be a word char). | |
365 | */ | |
fb62eb7f | 366 | if ( ((pmatch[0].rm_so == 0) || |
83b5d2f5 JH |
367 | !word_char(bol[pmatch[0].rm_so-1])) && |
368 | ((pmatch[0].rm_eo == (eol-bol)) || | |
369 | !word_char(bol[pmatch[0].rm_eo])) ) | |
370 | ; | |
371 | else | |
372 | hit = 0; | |
373 | ||
84201eae RS |
374 | /* Words consist of at least one character. */ |
375 | if (pmatch->rm_so == pmatch->rm_eo) | |
376 | hit = 0; | |
377 | ||
83b5d2f5 JH |
378 | if (!hit && pmatch[0].rm_so + bol + 1 < eol) { |
379 | /* There could be more than one match on the | |
380 | * line, and the first match might not be | |
381 | * strict word match. But later ones could be! | |
fb62eb7f RS |
382 | * Forward to the next possible start, i.e. the |
383 | * next position following a non-word char. | |
83b5d2f5 JH |
384 | */ |
385 | bol = pmatch[0].rm_so + bol + 1; | |
fb62eb7f RS |
386 | while (word_char(bol[-1]) && bol < eol) |
387 | bol++; | |
dbb6a4ad | 388 | eflags |= REG_NOTBOL; |
fb62eb7f RS |
389 | if (bol < eol) |
390 | goto again; | |
83b5d2f5 JH |
391 | } |
392 | } | |
a4d7d2c6 JH |
393 | if (p->token == GREP_PATTERN_HEAD && saved_ch) |
394 | *eol = saved_ch; | |
e701fadb RS |
395 | if (hit) { |
396 | pmatch[0].rm_so += bol - start; | |
397 | pmatch[0].rm_eo += bol - start; | |
398 | } | |
83b5d2f5 JH |
399 | return hit; |
400 | } | |
401 | ||
d7eb527d RS |
402 | static int match_expr_eval(struct grep_expr *x, char *bol, char *eol, |
403 | enum grep_context ctx, int collect_hits) | |
83b5d2f5 | 404 | { |
0ab7befa | 405 | int h = 0; |
79212772 | 406 | regmatch_t match; |
0ab7befa | 407 | |
c922b01f LT |
408 | if (!x) |
409 | die("Not a valid grep expression"); | |
83b5d2f5 JH |
410 | switch (x->node) { |
411 | case GREP_NODE_ATOM: | |
79212772 | 412 | h = match_one_pattern(x->u.atom, bol, eol, ctx, &match, 0); |
83b5d2f5 JH |
413 | break; |
414 | case GREP_NODE_NOT: | |
d7eb527d | 415 | h = !match_expr_eval(x->u.unary, bol, eol, ctx, 0); |
0ab7befa | 416 | break; |
83b5d2f5 | 417 | case GREP_NODE_AND: |
d7eb527d | 418 | if (!match_expr_eval(x->u.binary.left, bol, eol, ctx, 0)) |
252d560d | 419 | return 0; |
d7eb527d | 420 | h = match_expr_eval(x->u.binary.right, bol, eol, ctx, 0); |
0ab7befa | 421 | break; |
83b5d2f5 | 422 | case GREP_NODE_OR: |
0ab7befa | 423 | if (!collect_hits) |
d7eb527d | 424 | return (match_expr_eval(x->u.binary.left, |
0ab7befa | 425 | bol, eol, ctx, 0) || |
d7eb527d | 426 | match_expr_eval(x->u.binary.right, |
0ab7befa | 427 | bol, eol, ctx, 0)); |
d7eb527d | 428 | h = match_expr_eval(x->u.binary.left, bol, eol, ctx, 0); |
0ab7befa | 429 | x->u.binary.left->hit |= h; |
d7eb527d | 430 | h |= match_expr_eval(x->u.binary.right, bol, eol, ctx, 1); |
0ab7befa JH |
431 | break; |
432 | default: | |
d7530708 | 433 | die("Unexpected node type (internal error) %d", x->node); |
83b5d2f5 | 434 | } |
0ab7befa JH |
435 | if (collect_hits) |
436 | x->hit |= h; | |
437 | return h; | |
83b5d2f5 JH |
438 | } |
439 | ||
480c1ca6 | 440 | static int match_expr(struct grep_opt *opt, char *bol, char *eol, |
0ab7befa | 441 | enum grep_context ctx, int collect_hits) |
83b5d2f5 JH |
442 | { |
443 | struct grep_expr *x = opt->pattern_expression; | |
d7eb527d | 444 | return match_expr_eval(x, bol, eol, ctx, collect_hits); |
83b5d2f5 JH |
445 | } |
446 | ||
480c1ca6 | 447 | static int match_line(struct grep_opt *opt, char *bol, char *eol, |
0ab7befa | 448 | enum grep_context ctx, int collect_hits) |
83b5d2f5 JH |
449 | { |
450 | struct grep_pat *p; | |
79212772 RS |
451 | regmatch_t match; |
452 | ||
83b5d2f5 | 453 | if (opt->extended) |
0ab7befa JH |
454 | return match_expr(opt, bol, eol, ctx, collect_hits); |
455 | ||
456 | /* we do not call with collect_hits without being extended */ | |
83b5d2f5 | 457 | for (p = opt->pattern_list; p; p = p->next) { |
79212772 | 458 | if (match_one_pattern(p, bol, eol, ctx, &match, 0)) |
83b5d2f5 JH |
459 | return 1; |
460 | } | |
461 | return 0; | |
462 | } | |
463 | ||
7e8f59d5 RS |
464 | static int match_next_pattern(struct grep_pat *p, char *bol, char *eol, |
465 | enum grep_context ctx, | |
466 | regmatch_t *pmatch, int eflags) | |
467 | { | |
468 | regmatch_t match; | |
469 | ||
470 | if (!match_one_pattern(p, bol, eol, ctx, &match, eflags)) | |
471 | return 0; | |
472 | if (match.rm_so < 0 || match.rm_eo < 0) | |
473 | return 0; | |
474 | if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) { | |
475 | if (match.rm_so > pmatch->rm_so) | |
476 | return 1; | |
477 | if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo) | |
478 | return 1; | |
479 | } | |
480 | pmatch->rm_so = match.rm_so; | |
481 | pmatch->rm_eo = match.rm_eo; | |
482 | return 1; | |
483 | } | |
484 | ||
485 | static int next_match(struct grep_opt *opt, char *bol, char *eol, | |
486 | enum grep_context ctx, regmatch_t *pmatch, int eflags) | |
487 | { | |
488 | struct grep_pat *p; | |
489 | int hit = 0; | |
490 | ||
491 | pmatch->rm_so = pmatch->rm_eo = -1; | |
492 | if (bol < eol) { | |
493 | for (p = opt->pattern_list; p; p = p->next) { | |
494 | switch (p->token) { | |
495 | case GREP_PATTERN: /* atom */ | |
496 | case GREP_PATTERN_HEAD: | |
497 | case GREP_PATTERN_BODY: | |
498 | hit |= match_next_pattern(p, bol, eol, ctx, | |
499 | pmatch, eflags); | |
500 | break; | |
501 | default: | |
502 | break; | |
503 | } | |
504 | } | |
505 | } | |
506 | return hit; | |
507 | } | |
508 | ||
509 | static void show_line(struct grep_opt *opt, char *bol, char *eol, | |
510 | const char *name, unsigned lno, char sign) | |
511 | { | |
512 | int rest = eol - bol; | |
5b594f45 | 513 | char sign_str[1]; |
7e8f59d5 | 514 | |
5b594f45 | 515 | sign_str[0] = sign; |
ed24e401 | 516 | if (opt->pre_context || opt->post_context) { |
046802d0 RS |
517 | if (opt->last_shown == 0) { |
518 | if (opt->show_hunk_mark) | |
5b594f45 | 519 | opt->output(opt, "--\n", 3); |
046802d0 RS |
520 | else |
521 | opt->show_hunk_mark = 1; | |
ed24e401 | 522 | } else if (lno > opt->last_shown + 1) |
5b594f45 | 523 | opt->output(opt, "--\n", 3); |
5dd06d38 RS |
524 | } |
525 | opt->last_shown = lno; | |
526 | ||
7e8f59d5 | 527 | if (opt->null_following_name) |
5b594f45 FK |
528 | sign_str[0] = '\0'; |
529 | if (opt->pathname) { | |
530 | opt->output(opt, name, strlen(name)); | |
531 | opt->output(opt, sign_str, 1); | |
532 | } | |
533 | if (opt->linenum) { | |
534 | char buf[32]; | |
535 | snprintf(buf, sizeof(buf), "%d", lno); | |
536 | opt->output(opt, buf, strlen(buf)); | |
537 | opt->output(opt, sign_str, 1); | |
538 | } | |
7e8f59d5 RS |
539 | if (opt->color) { |
540 | regmatch_t match; | |
541 | enum grep_context ctx = GREP_CONTEXT_BODY; | |
542 | int ch = *eol; | |
543 | int eflags = 0; | |
544 | ||
545 | *eol = '\0'; | |
546 | while (next_match(opt, bol, eol, ctx, &match, eflags)) { | |
1f5b9cc4 RS |
547 | if (match.rm_so == match.rm_eo) |
548 | break; | |
5b594f45 FK |
549 | |
550 | opt->output(opt, bol, match.rm_so); | |
551 | opt->output(opt, opt->color_match, | |
552 | strlen(opt->color_match)); | |
553 | opt->output(opt, bol + match.rm_so, | |
554 | (int)(match.rm_eo - match.rm_so)); | |
555 | opt->output(opt, GIT_COLOR_RESET, | |
556 | strlen(GIT_COLOR_RESET)); | |
7e8f59d5 RS |
557 | bol += match.rm_eo; |
558 | rest -= match.rm_eo; | |
559 | eflags = REG_NOTBOL; | |
560 | } | |
561 | *eol = ch; | |
562 | } | |
5b594f45 FK |
563 | opt->output(opt, bol, rest); |
564 | opt->output(opt, "\n", 1); | |
7e8f59d5 RS |
565 | } |
566 | ||
60ecac98 | 567 | static int match_funcname(struct grep_opt *opt, char *bol, char *eol) |
2944e4e6 | 568 | { |
60ecac98 RS |
569 | xdemitconf_t *xecfg = opt->priv; |
570 | if (xecfg && xecfg->find_func) { | |
571 | char buf[1]; | |
572 | return xecfg->find_func(bol, eol - bol, buf, 1, | |
573 | xecfg->find_func_priv) >= 0; | |
574 | } | |
575 | ||
2944e4e6 RS |
576 | if (bol == eol) |
577 | return 0; | |
578 | if (isalpha(*bol) || *bol == '_' || *bol == '$') | |
579 | return 1; | |
580 | return 0; | |
581 | } | |
582 | ||
583 | static void show_funcname_line(struct grep_opt *opt, const char *name, | |
584 | char *buf, char *bol, unsigned lno) | |
585 | { | |
586 | while (bol > buf) { | |
587 | char *eol = --bol; | |
588 | ||
589 | while (bol > buf && bol[-1] != '\n') | |
590 | bol--; | |
591 | lno--; | |
592 | ||
593 | if (lno <= opt->last_shown) | |
594 | break; | |
595 | ||
60ecac98 | 596 | if (match_funcname(opt, bol, eol)) { |
2944e4e6 RS |
597 | show_line(opt, bol, eol, name, lno, '='); |
598 | break; | |
599 | } | |
600 | } | |
601 | } | |
602 | ||
49de3216 RS |
603 | static void show_pre_context(struct grep_opt *opt, const char *name, char *buf, |
604 | char *bol, unsigned lno) | |
605 | { | |
2944e4e6 RS |
606 | unsigned cur = lno, from = 1, funcname_lno = 0; |
607 | int funcname_needed = opt->funcname; | |
49de3216 RS |
608 | |
609 | if (opt->pre_context < lno) | |
610 | from = lno - opt->pre_context; | |
611 | if (from <= opt->last_shown) | |
612 | from = opt->last_shown + 1; | |
613 | ||
614 | /* Rewind. */ | |
615 | while (bol > buf && cur > from) { | |
2944e4e6 RS |
616 | char *eol = --bol; |
617 | ||
49de3216 RS |
618 | while (bol > buf && bol[-1] != '\n') |
619 | bol--; | |
620 | cur--; | |
60ecac98 | 621 | if (funcname_needed && match_funcname(opt, bol, eol)) { |
2944e4e6 RS |
622 | funcname_lno = cur; |
623 | funcname_needed = 0; | |
624 | } | |
49de3216 RS |
625 | } |
626 | ||
2944e4e6 RS |
627 | /* We need to look even further back to find a function signature. */ |
628 | if (opt->funcname && funcname_needed) | |
629 | show_funcname_line(opt, name, buf, bol, cur); | |
630 | ||
49de3216 RS |
631 | /* Back forward. */ |
632 | while (cur < lno) { | |
2944e4e6 | 633 | char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-'; |
49de3216 RS |
634 | |
635 | while (*eol != '\n') | |
636 | eol++; | |
2944e4e6 | 637 | show_line(opt, bol, eol, name, cur, sign); |
49de3216 RS |
638 | bol = eol + 1; |
639 | cur++; | |
640 | } | |
641 | } | |
642 | ||
a26345b6 JH |
643 | static int should_lookahead(struct grep_opt *opt) |
644 | { | |
645 | struct grep_pat *p; | |
646 | ||
647 | if (opt->extended) | |
648 | return 0; /* punt for too complex stuff */ | |
649 | if (opt->invert) | |
650 | return 0; | |
651 | for (p = opt->pattern_list; p; p = p->next) { | |
652 | if (p->token != GREP_PATTERN) | |
653 | return 0; /* punt for "header only" and stuff */ | |
654 | } | |
655 | return 1; | |
656 | } | |
657 | ||
658 | static int look_ahead(struct grep_opt *opt, | |
659 | unsigned long *left_p, | |
660 | unsigned *lno_p, | |
661 | char **bol_p) | |
662 | { | |
663 | unsigned lno = *lno_p; | |
664 | char *bol = *bol_p; | |
665 | struct grep_pat *p; | |
666 | char *sp, *last_bol; | |
667 | regoff_t earliest = -1; | |
668 | ||
669 | for (p = opt->pattern_list; p; p = p->next) { | |
670 | int hit; | |
671 | regmatch_t m; | |
672 | ||
673 | if (p->fixed) | |
e2d2e383 | 674 | hit = !fixmatch(p->pattern, bol, p->ignore_case, &m); |
24072c02 BK |
675 | else { |
676 | #ifdef REG_STARTEND | |
677 | m.rm_so = 0; | |
678 | m.rm_eo = *left_p; | |
679 | hit = !regexec(&p->regexp, bol, 1, &m, REG_STARTEND); | |
680 | #else | |
a26345b6 | 681 | hit = !regexec(&p->regexp, bol, 1, &m, 0); |
24072c02 BK |
682 | #endif |
683 | } | |
a26345b6 JH |
684 | if (!hit || m.rm_so < 0 || m.rm_eo < 0) |
685 | continue; | |
686 | if (earliest < 0 || m.rm_so < earliest) | |
687 | earliest = m.rm_so; | |
688 | } | |
689 | ||
690 | if (earliest < 0) { | |
691 | *bol_p = bol + *left_p; | |
692 | *left_p = 0; | |
693 | return 1; | |
694 | } | |
695 | for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--) | |
696 | ; /* find the beginning of the line */ | |
697 | last_bol = sp; | |
698 | ||
699 | for (sp = bol; sp < last_bol; sp++) { | |
700 | if (*sp == '\n') | |
701 | lno++; | |
702 | } | |
703 | *left_p -= last_bol - bol; | |
704 | *bol_p = last_bol; | |
705 | *lno_p = lno; | |
706 | return 0; | |
707 | } | |
708 | ||
5b594f45 FK |
709 | int grep_threads_ok(const struct grep_opt *opt) |
710 | { | |
711 | /* If this condition is true, then we may use the attribute | |
712 | * machinery in grep_buffer_1. The attribute code is not | |
713 | * thread safe, so we disable the use of threads. | |
714 | */ | |
715 | if (opt->funcname && !opt->unmatch_name_only && !opt->status_only && | |
716 | !opt->name_only) | |
717 | return 0; | |
718 | ||
719 | /* If we are showing hunk marks, we should not do it for the | |
720 | * first match. The synchronization problem we get for this | |
721 | * constraint is not yet solved, so we disable threading in | |
722 | * this case. | |
723 | */ | |
724 | if (opt->pre_context || opt->post_context) | |
725 | return 0; | |
726 | ||
727 | return 1; | |
728 | } | |
729 | ||
730 | static void std_output(struct grep_opt *opt, const void *buf, size_t size) | |
731 | { | |
732 | fwrite(buf, size, 1, stdout); | |
733 | } | |
734 | ||
0ab7befa JH |
735 | static int grep_buffer_1(struct grep_opt *opt, const char *name, |
736 | char *buf, unsigned long size, int collect_hits) | |
83b5d2f5 JH |
737 | { |
738 | char *bol = buf; | |
739 | unsigned long left = size; | |
740 | unsigned lno = 1; | |
83b5d2f5 | 741 | unsigned last_hit = 0; |
83b5d2f5 | 742 | int binary_match_only = 0; |
83b5d2f5 | 743 | unsigned count = 0; |
a26345b6 | 744 | int try_lookahead = 0; |
480c1ca6 | 745 | enum grep_context ctx = GREP_CONTEXT_HEAD; |
60ecac98 | 746 | xdemitconf_t xecfg; |
83b5d2f5 | 747 | |
5dd06d38 RS |
748 | opt->last_shown = 0; |
749 | ||
5b594f45 FK |
750 | if (!opt->output) |
751 | opt->output = std_output; | |
752 | ||
83b5d2f5 JH |
753 | if (buffer_is_binary(buf, size)) { |
754 | switch (opt->binary) { | |
755 | case GREP_BINARY_DEFAULT: | |
756 | binary_match_only = 1; | |
757 | break; | |
758 | case GREP_BINARY_NOMATCH: | |
759 | return 0; /* Assume unmatch */ | |
760 | break; | |
761 | default: | |
762 | break; | |
763 | } | |
764 | } | |
765 | ||
60ecac98 RS |
766 | memset(&xecfg, 0, sizeof(xecfg)); |
767 | if (opt->funcname && !opt->unmatch_name_only && !opt->status_only && | |
768 | !opt->name_only && !binary_match_only && !collect_hits) { | |
769 | struct userdiff_driver *drv = userdiff_find_by_path(name); | |
770 | if (drv && drv->funcname.pattern) { | |
771 | const struct userdiff_funcname *pe = &drv->funcname; | |
772 | xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags); | |
773 | opt->priv = &xecfg; | |
774 | } | |
775 | } | |
a26345b6 | 776 | try_lookahead = should_lookahead(opt); |
60ecac98 | 777 | |
83b5d2f5 JH |
778 | while (left) { |
779 | char *eol, ch; | |
0ab7befa | 780 | int hit; |
83b5d2f5 | 781 | |
a26345b6 JH |
782 | /* |
783 | * look_ahead() skips quicly to the line that possibly | |
784 | * has the next hit; don't call it if we need to do | |
785 | * something more than just skipping the current line | |
786 | * in response to an unmatch for the current line. E.g. | |
787 | * inside a post-context window, we will show the current | |
788 | * line as a context around the previous hit when it | |
789 | * doesn't hit. | |
790 | */ | |
791 | if (try_lookahead | |
792 | && !(last_hit | |
793 | && lno <= last_hit + opt->post_context) | |
794 | && look_ahead(opt, &left, &lno, &bol)) | |
795 | break; | |
83b5d2f5 JH |
796 | eol = end_of_line(bol, &left); |
797 | ch = *eol; | |
798 | *eol = 0; | |
799 | ||
480c1ca6 JH |
800 | if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol)) |
801 | ctx = GREP_CONTEXT_BODY; | |
802 | ||
0ab7befa | 803 | hit = match_line(opt, bol, eol, ctx, collect_hits); |
83b5d2f5 JH |
804 | *eol = ch; |
805 | ||
0ab7befa JH |
806 | if (collect_hits) |
807 | goto next_line; | |
808 | ||
83b5d2f5 JH |
809 | /* "grep -v -e foo -e bla" should list lines |
810 | * that do not have either, so inversion should | |
811 | * be done outside. | |
812 | */ | |
813 | if (opt->invert) | |
814 | hit = !hit; | |
815 | if (opt->unmatch_name_only) { | |
816 | if (hit) | |
817 | return 0; | |
818 | goto next_line; | |
819 | } | |
820 | if (hit) { | |
821 | count++; | |
822 | if (opt->status_only) | |
823 | return 1; | |
824 | if (binary_match_only) { | |
5b594f45 FK |
825 | opt->output(opt, "Binary file ", 12); |
826 | opt->output(opt, name, strlen(name)); | |
827 | opt->output(opt, " matches\n", 9); | |
83b5d2f5 JH |
828 | return 1; |
829 | } | |
830 | if (opt->name_only) { | |
83caecca | 831 | show_name(opt, name); |
83b5d2f5 JH |
832 | return 1; |
833 | } | |
834 | /* Hit at this line. If we haven't shown the | |
835 | * pre-context lines, we would need to show them. | |
836 | * When asked to do "count", this still show | |
837 | * the context which is nonsense, but the user | |
838 | * deserves to get that ;-). | |
839 | */ | |
49de3216 RS |
840 | if (opt->pre_context) |
841 | show_pre_context(opt, name, buf, bol, lno); | |
2944e4e6 RS |
842 | else if (opt->funcname) |
843 | show_funcname_line(opt, name, buf, bol, lno); | |
83b5d2f5 JH |
844 | if (!opt->count) |
845 | show_line(opt, bol, eol, name, lno, ':'); | |
5dd06d38 | 846 | last_hit = lno; |
83b5d2f5 JH |
847 | } |
848 | else if (last_hit && | |
849 | lno <= last_hit + opt->post_context) { | |
850 | /* If the last hit is within the post context, | |
851 | * we need to show this line. | |
852 | */ | |
83b5d2f5 | 853 | show_line(opt, bol, eol, name, lno, '-'); |
83b5d2f5 | 854 | } |
83b5d2f5 JH |
855 | |
856 | next_line: | |
857 | bol = eol + 1; | |
858 | if (!left) | |
859 | break; | |
860 | left--; | |
861 | lno++; | |
862 | } | |
863 | ||
0ab7befa JH |
864 | if (collect_hits) |
865 | return 0; | |
b48fb5b6 | 866 | |
83b5d2f5 JH |
867 | if (opt->status_only) |
868 | return 0; | |
869 | if (opt->unmatch_name_only) { | |
870 | /* We did not see any hit, so we want to show this */ | |
83caecca | 871 | show_name(opt, name); |
83b5d2f5 JH |
872 | return 1; |
873 | } | |
874 | ||
60ecac98 RS |
875 | xdiff_clear_find_func(&xecfg); |
876 | opt->priv = NULL; | |
877 | ||
83b5d2f5 JH |
878 | /* NEEDSWORK: |
879 | * The real "grep -c foo *.c" gives many "bar.c:0" lines, | |
880 | * which feels mostly useless but sometimes useful. Maybe | |
881 | * make it another option? For now suppress them. | |
882 | */ | |
5b594f45 FK |
883 | if (opt->count && count) { |
884 | char buf[32]; | |
885 | opt->output(opt, name, strlen(name)); | |
886 | snprintf(buf, sizeof(buf), "%c%u\n", | |
887 | opt->null_following_name ? '\0' : ':', count); | |
888 | opt->output(opt, buf, strlen(buf)); | |
889 | } | |
83b5d2f5 JH |
890 | return !!last_hit; |
891 | } | |
892 | ||
0ab7befa JH |
893 | static void clr_hit_marker(struct grep_expr *x) |
894 | { | |
895 | /* All-hit markers are meaningful only at the very top level | |
896 | * OR node. | |
897 | */ | |
898 | while (1) { | |
899 | x->hit = 0; | |
900 | if (x->node != GREP_NODE_OR) | |
901 | return; | |
902 | x->u.binary.left->hit = 0; | |
903 | x = x->u.binary.right; | |
904 | } | |
905 | } | |
906 | ||
907 | static int chk_hit_marker(struct grep_expr *x) | |
908 | { | |
909 | /* Top level nodes have hit markers. See if they all are hits */ | |
910 | while (1) { | |
911 | if (x->node != GREP_NODE_OR) | |
912 | return x->hit; | |
913 | if (!x->u.binary.left->hit) | |
914 | return 0; | |
915 | x = x->u.binary.right; | |
916 | } | |
917 | } | |
918 | ||
919 | int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size) | |
920 | { | |
921 | /* | |
922 | * we do not have to do the two-pass grep when we do not check | |
923 | * buffer-wide "all-match". | |
924 | */ | |
925 | if (!opt->all_match) | |
926 | return grep_buffer_1(opt, name, buf, size, 0); | |
927 | ||
928 | /* Otherwise the toplevel "or" terms hit a bit differently. | |
929 | * We first clear hit markers from them. | |
930 | */ | |
931 | clr_hit_marker(opt->pattern_expression); | |
932 | grep_buffer_1(opt, name, buf, size, 1); | |
933 | ||
934 | if (!chk_hit_marker(opt->pattern_expression)) | |
935 | return 0; | |
936 | ||
937 | return grep_buffer_1(opt, name, buf, size, 0); | |
938 | } |