checkout $tree $path: do not clobber local changes in $path not in $tree
[git/jnareb-git.git] / grep.c
blob63c4280cac9a87f867451c0d9787d1fe21ea9c2d
1 #include "cache.h"
2 #include "grep.h"
3 #include "userdiff.h"
4 #include "xdiff-interface.h"
6 void append_header_grep_pattern(struct grep_opt *opt, enum grep_header_field field, const char *pat)
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;
20 void append_grep_pattern(struct grep_opt *opt, const char *pat,
21 const char *origin, int no, enum grep_pat_token t)
23 append_grep_pat(opt, pat, strlen(pat), origin, no, t);
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)
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;
40 struct grep_opt *grep_opt_dup(const struct grep_opt *opt)
42 struct grep_pat *pat;
43 struct grep_opt *ret = xmalloc(sizeof(struct grep_opt));
44 *ret = *opt;
46 ret->pattern_list = NULL;
47 ret->pattern_tail = &ret->pattern_list;
49 for(pat = opt->pattern_list; pat != NULL; pat = pat->next)
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);
59 return ret;
62 static void compile_regexp(struct grep_pat *p, struct grep_opt *opt)
64 int err;
66 p->word_regexp = opt->word_regexp;
67 p->ignore_case = opt->ignore_case;
68 p->fixed = opt->fixed;
70 if (p->fixed)
71 return;
73 err = regcomp(&p->regexp, p->pattern, opt->regflags);
74 if (err) {
75 char errbuf[1024];
76 char where[1024];
77 if (p->no)
78 sprintf(where, "In '%s' at %d, ",
79 p->origin, p->no);
80 else if (p->origin)
81 sprintf(where, "%s, ", p->origin);
82 else
83 where[0] = 0;
84 regerror(err, &p->regexp, errbuf, 1024);
85 regfree(&p->regexp);
86 die("%s'%s': %s", where, p->pattern, errbuf);
90 static struct grep_expr *compile_pattern_or(struct grep_pat **);
91 static struct grep_expr *compile_pattern_atom(struct grep_pat **list)
93 struct grep_pat *p;
94 struct grep_expr *x;
96 p = *list;
97 if (!p)
98 return NULL;
99 switch (p->token) {
100 case GREP_PATTERN: /* atom */
101 case GREP_PATTERN_HEAD:
102 case GREP_PATTERN_BODY:
103 x = xcalloc(1, sizeof (struct grep_expr));
104 x->node = GREP_NODE_ATOM;
105 x->u.atom = p;
106 *list = p->next;
107 return x;
108 case GREP_OPEN_PAREN:
109 *list = p->next;
110 x = compile_pattern_or(list);
111 if (!*list || (*list)->token != GREP_CLOSE_PAREN)
112 die("unmatched parenthesis");
113 *list = (*list)->next;
114 return x;
115 default:
116 return NULL;
120 static struct grep_expr *compile_pattern_not(struct grep_pat **list)
122 struct grep_pat *p;
123 struct grep_expr *x;
125 p = *list;
126 if (!p)
127 return NULL;
128 switch (p->token) {
129 case GREP_NOT:
130 if (!p->next)
131 die("--not not followed by pattern expression");
132 *list = p->next;
133 x = xcalloc(1, sizeof (struct grep_expr));
134 x->node = GREP_NODE_NOT;
135 x->u.unary = compile_pattern_not(list);
136 if (!x->u.unary)
137 die("--not followed by non pattern expression");
138 return x;
139 default:
140 return compile_pattern_atom(list);
144 static struct grep_expr *compile_pattern_and(struct grep_pat **list)
146 struct grep_pat *p;
147 struct grep_expr *x, *y, *z;
149 x = compile_pattern_not(list);
150 p = *list;
151 if (p && p->token == GREP_AND) {
152 if (!p->next)
153 die("--and not followed by pattern expression");
154 *list = p->next;
155 y = compile_pattern_and(list);
156 if (!y)
157 die("--and not followed by pattern expression");
158 z = xcalloc(1, sizeof (struct grep_expr));
159 z->node = GREP_NODE_AND;
160 z->u.binary.left = x;
161 z->u.binary.right = y;
162 return z;
164 return x;
167 static struct grep_expr *compile_pattern_or(struct grep_pat **list)
169 struct grep_pat *p;
170 struct grep_expr *x, *y, *z;
172 x = compile_pattern_and(list);
173 p = *list;
174 if (x && p && p->token != GREP_CLOSE_PAREN) {
175 y = compile_pattern_or(list);
176 if (!y)
177 die("not a pattern expression %s", p->pattern);
178 z = xcalloc(1, sizeof (struct grep_expr));
179 z->node = GREP_NODE_OR;
180 z->u.binary.left = x;
181 z->u.binary.right = y;
182 return z;
184 return x;
187 static struct grep_expr *compile_pattern_expr(struct grep_pat **list)
189 return compile_pattern_or(list);
192 static struct grep_expr *grep_true_expr(void)
194 struct grep_expr *z = xcalloc(1, sizeof(*z));
195 z->node = GREP_NODE_TRUE;
196 return z;
199 static struct grep_expr *grep_or_expr(struct grep_expr *left, struct grep_expr *right)
201 struct grep_expr *z = xcalloc(1, sizeof(*z));
202 z->node = GREP_NODE_OR;
203 z->u.binary.left = left;
204 z->u.binary.right = right;
205 return z;
208 static struct grep_expr *prep_header_patterns(struct grep_opt *opt)
210 struct grep_pat *p;
211 struct grep_expr *header_expr;
212 struct grep_expr *(header_group[GREP_HEADER_FIELD_MAX]);
213 enum grep_header_field fld;
215 if (!opt->header_list)
216 return NULL;
217 p = opt->header_list;
218 for (p = opt->header_list; p; p = p->next) {
219 if (p->token != GREP_PATTERN_HEAD)
220 die("bug: a non-header pattern in grep header list.");
221 if (p->field < 0 || GREP_HEADER_FIELD_MAX <= p->field)
222 die("bug: unknown header field %d", p->field);
223 compile_regexp(p, opt);
226 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++)
227 header_group[fld] = NULL;
229 for (p = opt->header_list; p; p = p->next) {
230 struct grep_expr *h;
231 struct grep_pat *pp = p;
233 h = compile_pattern_atom(&pp);
234 if (!h || pp != p->next)
235 die("bug: malformed header expr");
236 if (!header_group[p->field]) {
237 header_group[p->field] = h;
238 continue;
240 header_group[p->field] = grep_or_expr(h, header_group[p->field]);
243 header_expr = NULL;
245 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) {
246 if (!header_group[fld])
247 continue;
248 if (!header_expr)
249 header_expr = grep_true_expr();
250 header_expr = grep_or_expr(header_group[fld], header_expr);
252 return header_expr;
255 void compile_grep_patterns(struct grep_opt *opt)
257 struct grep_pat *p;
258 struct grep_expr *header_expr = prep_header_patterns(opt);
260 for (p = opt->pattern_list; p; p = p->next) {
261 switch (p->token) {
262 case GREP_PATTERN: /* atom */
263 case GREP_PATTERN_HEAD:
264 case GREP_PATTERN_BODY:
265 compile_regexp(p, opt);
266 break;
267 default:
268 opt->extended = 1;
269 break;
273 if (opt->all_match || header_expr)
274 opt->extended = 1;
275 else if (!opt->extended)
276 return;
278 p = opt->pattern_list;
279 if (p)
280 opt->pattern_expression = compile_pattern_expr(&p);
281 if (p)
282 die("incomplete pattern expression: %s", p->pattern);
284 if (!header_expr)
285 return;
287 if (!opt->pattern_expression)
288 opt->pattern_expression = header_expr;
289 else
290 opt->pattern_expression = grep_or_expr(opt->pattern_expression,
291 header_expr);
292 opt->all_match = 1;
295 static void free_pattern_expr(struct grep_expr *x)
297 switch (x->node) {
298 case GREP_NODE_TRUE:
299 case GREP_NODE_ATOM:
300 break;
301 case GREP_NODE_NOT:
302 free_pattern_expr(x->u.unary);
303 break;
304 case GREP_NODE_AND:
305 case GREP_NODE_OR:
306 free_pattern_expr(x->u.binary.left);
307 free_pattern_expr(x->u.binary.right);
308 break;
310 free(x);
313 void free_grep_patterns(struct grep_opt *opt)
315 struct grep_pat *p, *n;
317 for (p = opt->pattern_list; p; p = n) {
318 n = p->next;
319 switch (p->token) {
320 case GREP_PATTERN: /* atom */
321 case GREP_PATTERN_HEAD:
322 case GREP_PATTERN_BODY:
323 regfree(&p->regexp);
324 break;
325 default:
326 break;
328 free(p);
331 if (!opt->extended)
332 return;
333 free_pattern_expr(opt->pattern_expression);
336 static char *end_of_line(char *cp, unsigned long *left)
338 unsigned long l = *left;
339 while (l && *cp != '\n') {
340 l--;
341 cp++;
343 *left = l;
344 return cp;
347 static int word_char(char ch)
349 return isalnum(ch) || ch == '_';
352 static void output_color(struct grep_opt *opt, const void *data, size_t size,
353 const char *color)
355 if (opt->color && color && color[0]) {
356 opt->output(opt, color, strlen(color));
357 opt->output(opt, data, size);
358 opt->output(opt, GIT_COLOR_RESET, strlen(GIT_COLOR_RESET));
359 } else
360 opt->output(opt, data, size);
363 static void output_sep(struct grep_opt *opt, char sign)
365 if (opt->null_following_name)
366 opt->output(opt, "\0", 1);
367 else
368 output_color(opt, &sign, 1, opt->color_sep);
371 static void show_name(struct grep_opt *opt, const char *name)
373 output_color(opt, name, strlen(name), opt->color_filename);
374 opt->output(opt, opt->null_following_name ? "\0" : "\n", 1);
377 static int fixmatch(struct grep_pat *p, char *line, char *eol,
378 regmatch_t *match)
380 char *hit;
382 if (p->ignore_case) {
383 char *s = line;
384 do {
385 hit = strcasestr(s, p->pattern);
386 if (hit)
387 break;
388 s += strlen(s) + 1;
389 } while (s < eol);
390 } else
391 hit = memmem(line, eol - line, p->pattern, p->patternlen);
393 if (!hit) {
394 match->rm_so = match->rm_eo = -1;
395 return REG_NOMATCH;
397 else {
398 match->rm_so = hit - line;
399 match->rm_eo = match->rm_so + p->patternlen;
400 return 0;
404 static int regmatch(const regex_t *preg, char *line, char *eol,
405 regmatch_t *match, int eflags)
407 #ifdef REG_STARTEND
408 match->rm_so = 0;
409 match->rm_eo = eol - line;
410 eflags |= REG_STARTEND;
411 #endif
412 return regexec(preg, line, 1, match, eflags);
415 static int strip_timestamp(char *bol, char **eol_p)
417 char *eol = *eol_p;
418 int ch;
420 while (bol < --eol) {
421 if (*eol != '>')
422 continue;
423 *eol_p = ++eol;
424 ch = *eol;
425 *eol = '\0';
426 return ch;
428 return 0;
431 static struct {
432 const char *field;
433 size_t len;
434 } header_field[] = {
435 { "author ", 7 },
436 { "committer ", 10 },
439 static int match_one_pattern(struct grep_pat *p, char *bol, char *eol,
440 enum grep_context ctx,
441 regmatch_t *pmatch, int eflags)
443 int hit = 0;
444 int saved_ch = 0;
445 const char *start = bol;
447 if ((p->token != GREP_PATTERN) &&
448 ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
449 return 0;
451 if (p->token == GREP_PATTERN_HEAD) {
452 const char *field;
453 size_t len;
454 assert(p->field < ARRAY_SIZE(header_field));
455 field = header_field[p->field].field;
456 len = header_field[p->field].len;
457 if (strncmp(bol, field, len))
458 return 0;
459 bol += len;
460 saved_ch = strip_timestamp(bol, &eol);
463 again:
464 if (p->fixed)
465 hit = !fixmatch(p, bol, eol, pmatch);
466 else
467 hit = !regmatch(&p->regexp, bol, eol, pmatch, eflags);
469 if (hit && p->word_regexp) {
470 if ((pmatch[0].rm_so < 0) ||
471 (eol - bol) < pmatch[0].rm_so ||
472 (pmatch[0].rm_eo < 0) ||
473 (eol - bol) < pmatch[0].rm_eo)
474 die("regexp returned nonsense");
476 /* Match beginning must be either beginning of the
477 * line, or at word boundary (i.e. the last char must
478 * not be a word char). Similarly, match end must be
479 * either end of the line, or at word boundary
480 * (i.e. the next char must not be a word char).
482 if ( ((pmatch[0].rm_so == 0) ||
483 !word_char(bol[pmatch[0].rm_so-1])) &&
484 ((pmatch[0].rm_eo == (eol-bol)) ||
485 !word_char(bol[pmatch[0].rm_eo])) )
487 else
488 hit = 0;
490 /* Words consist of at least one character. */
491 if (pmatch->rm_so == pmatch->rm_eo)
492 hit = 0;
494 if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
495 /* There could be more than one match on the
496 * line, and the first match might not be
497 * strict word match. But later ones could be!
498 * Forward to the next possible start, i.e. the
499 * next position following a non-word char.
501 bol = pmatch[0].rm_so + bol + 1;
502 while (word_char(bol[-1]) && bol < eol)
503 bol++;
504 eflags |= REG_NOTBOL;
505 if (bol < eol)
506 goto again;
509 if (p->token == GREP_PATTERN_HEAD && saved_ch)
510 *eol = saved_ch;
511 if (hit) {
512 pmatch[0].rm_so += bol - start;
513 pmatch[0].rm_eo += bol - start;
515 return hit;
518 static int match_expr_eval(struct grep_expr *x, char *bol, char *eol,
519 enum grep_context ctx, int collect_hits)
521 int h = 0;
522 regmatch_t match;
524 if (!x)
525 die("Not a valid grep expression");
526 switch (x->node) {
527 case GREP_NODE_TRUE:
528 h = 1;
529 break;
530 case GREP_NODE_ATOM:
531 h = match_one_pattern(x->u.atom, bol, eol, ctx, &match, 0);
532 break;
533 case GREP_NODE_NOT:
534 h = !match_expr_eval(x->u.unary, bol, eol, ctx, 0);
535 break;
536 case GREP_NODE_AND:
537 if (!match_expr_eval(x->u.binary.left, bol, eol, ctx, 0))
538 return 0;
539 h = match_expr_eval(x->u.binary.right, bol, eol, ctx, 0);
540 break;
541 case GREP_NODE_OR:
542 if (!collect_hits)
543 return (match_expr_eval(x->u.binary.left,
544 bol, eol, ctx, 0) ||
545 match_expr_eval(x->u.binary.right,
546 bol, eol, ctx, 0));
547 h = match_expr_eval(x->u.binary.left, bol, eol, ctx, 0);
548 x->u.binary.left->hit |= h;
549 h |= match_expr_eval(x->u.binary.right, bol, eol, ctx, 1);
550 break;
551 default:
552 die("Unexpected node type (internal error) %d", x->node);
554 if (collect_hits)
555 x->hit |= h;
556 return h;
559 static int match_expr(struct grep_opt *opt, char *bol, char *eol,
560 enum grep_context ctx, int collect_hits)
562 struct grep_expr *x = opt->pattern_expression;
563 return match_expr_eval(x, bol, eol, ctx, collect_hits);
566 static int match_line(struct grep_opt *opt, char *bol, char *eol,
567 enum grep_context ctx, int collect_hits)
569 struct grep_pat *p;
570 regmatch_t match;
572 if (opt->extended)
573 return match_expr(opt, bol, eol, ctx, collect_hits);
575 /* we do not call with collect_hits without being extended */
576 for (p = opt->pattern_list; p; p = p->next) {
577 if (match_one_pattern(p, bol, eol, ctx, &match, 0))
578 return 1;
580 return 0;
583 static int match_next_pattern(struct grep_pat *p, char *bol, char *eol,
584 enum grep_context ctx,
585 regmatch_t *pmatch, int eflags)
587 regmatch_t match;
589 if (!match_one_pattern(p, bol, eol, ctx, &match, eflags))
590 return 0;
591 if (match.rm_so < 0 || match.rm_eo < 0)
592 return 0;
593 if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) {
594 if (match.rm_so > pmatch->rm_so)
595 return 1;
596 if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo)
597 return 1;
599 pmatch->rm_so = match.rm_so;
600 pmatch->rm_eo = match.rm_eo;
601 return 1;
604 static int next_match(struct grep_opt *opt, char *bol, char *eol,
605 enum grep_context ctx, regmatch_t *pmatch, int eflags)
607 struct grep_pat *p;
608 int hit = 0;
610 pmatch->rm_so = pmatch->rm_eo = -1;
611 if (bol < eol) {
612 for (p = opt->pattern_list; p; p = p->next) {
613 switch (p->token) {
614 case GREP_PATTERN: /* atom */
615 case GREP_PATTERN_HEAD:
616 case GREP_PATTERN_BODY:
617 hit |= match_next_pattern(p, bol, eol, ctx,
618 pmatch, eflags);
619 break;
620 default:
621 break;
625 return hit;
628 static void show_line(struct grep_opt *opt, char *bol, char *eol,
629 const char *name, unsigned lno, char sign)
631 int rest = eol - bol;
632 char *line_color = NULL;
634 if (opt->pre_context || opt->post_context) {
635 if (opt->last_shown == 0) {
636 if (opt->show_hunk_mark) {
637 output_color(opt, "--", 2, opt->color_sep);
638 opt->output(opt, "\n", 1);
640 } else if (lno > opt->last_shown + 1) {
641 output_color(opt, "--", 2, opt->color_sep);
642 opt->output(opt, "\n", 1);
645 opt->last_shown = lno;
647 if (opt->pathname) {
648 output_color(opt, name, strlen(name), opt->color_filename);
649 output_sep(opt, sign);
651 if (opt->linenum) {
652 char buf[32];
653 snprintf(buf, sizeof(buf), "%d", lno);
654 output_color(opt, buf, strlen(buf), opt->color_lineno);
655 output_sep(opt, sign);
657 if (opt->color) {
658 regmatch_t match;
659 enum grep_context ctx = GREP_CONTEXT_BODY;
660 int ch = *eol;
661 int eflags = 0;
663 if (sign == ':')
664 line_color = opt->color_selected;
665 else if (sign == '-')
666 line_color = opt->color_context;
667 else if (sign == '=')
668 line_color = opt->color_function;
669 *eol = '\0';
670 while (next_match(opt, bol, eol, ctx, &match, eflags)) {
671 if (match.rm_so == match.rm_eo)
672 break;
674 output_color(opt, bol, match.rm_so, line_color);
675 output_color(opt, bol + match.rm_so,
676 match.rm_eo - match.rm_so,
677 opt->color_match);
678 bol += match.rm_eo;
679 rest -= match.rm_eo;
680 eflags = REG_NOTBOL;
682 *eol = ch;
684 output_color(opt, bol, rest, line_color);
685 opt->output(opt, "\n", 1);
688 static int match_funcname(struct grep_opt *opt, char *bol, char *eol)
690 xdemitconf_t *xecfg = opt->priv;
691 if (xecfg && xecfg->find_func) {
692 char buf[1];
693 return xecfg->find_func(bol, eol - bol, buf, 1,
694 xecfg->find_func_priv) >= 0;
697 if (bol == eol)
698 return 0;
699 if (isalpha(*bol) || *bol == '_' || *bol == '$')
700 return 1;
701 return 0;
704 static void show_funcname_line(struct grep_opt *opt, const char *name,
705 char *buf, char *bol, unsigned lno)
707 while (bol > buf) {
708 char *eol = --bol;
710 while (bol > buf && bol[-1] != '\n')
711 bol--;
712 lno--;
714 if (lno <= opt->last_shown)
715 break;
717 if (match_funcname(opt, bol, eol)) {
718 show_line(opt, bol, eol, name, lno, '=');
719 break;
724 static void show_pre_context(struct grep_opt *opt, const char *name, char *buf,
725 char *bol, unsigned lno)
727 unsigned cur = lno, from = 1, funcname_lno = 0;
728 int funcname_needed = opt->funcname;
730 if (opt->pre_context < lno)
731 from = lno - opt->pre_context;
732 if (from <= opt->last_shown)
733 from = opt->last_shown + 1;
735 /* Rewind. */
736 while (bol > buf && cur > from) {
737 char *eol = --bol;
739 while (bol > buf && bol[-1] != '\n')
740 bol--;
741 cur--;
742 if (funcname_needed && match_funcname(opt, bol, eol)) {
743 funcname_lno = cur;
744 funcname_needed = 0;
748 /* We need to look even further back to find a function signature. */
749 if (opt->funcname && funcname_needed)
750 show_funcname_line(opt, name, buf, bol, cur);
752 /* Back forward. */
753 while (cur < lno) {
754 char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-';
756 while (*eol != '\n')
757 eol++;
758 show_line(opt, bol, eol, name, cur, sign);
759 bol = eol + 1;
760 cur++;
764 static int should_lookahead(struct grep_opt *opt)
766 struct grep_pat *p;
768 if (opt->extended)
769 return 0; /* punt for too complex stuff */
770 if (opt->invert)
771 return 0;
772 for (p = opt->pattern_list; p; p = p->next) {
773 if (p->token != GREP_PATTERN)
774 return 0; /* punt for "header only" and stuff */
776 return 1;
779 static int look_ahead(struct grep_opt *opt,
780 unsigned long *left_p,
781 unsigned *lno_p,
782 char **bol_p)
784 unsigned lno = *lno_p;
785 char *bol = *bol_p;
786 struct grep_pat *p;
787 char *sp, *last_bol;
788 regoff_t earliest = -1;
790 for (p = opt->pattern_list; p; p = p->next) {
791 int hit;
792 regmatch_t m;
794 if (p->fixed)
795 hit = !fixmatch(p, bol, bol + *left_p, &m);
796 else
797 hit = !regmatch(&p->regexp, bol, bol + *left_p, &m, 0);
798 if (!hit || m.rm_so < 0 || m.rm_eo < 0)
799 continue;
800 if (earliest < 0 || m.rm_so < earliest)
801 earliest = m.rm_so;
804 if (earliest < 0) {
805 *bol_p = bol + *left_p;
806 *left_p = 0;
807 return 1;
809 for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--)
810 ; /* find the beginning of the line */
811 last_bol = sp;
813 for (sp = bol; sp < last_bol; sp++) {
814 if (*sp == '\n')
815 lno++;
817 *left_p -= last_bol - bol;
818 *bol_p = last_bol;
819 *lno_p = lno;
820 return 0;
823 int grep_threads_ok(const struct grep_opt *opt)
825 /* If this condition is true, then we may use the attribute
826 * machinery in grep_buffer_1. The attribute code is not
827 * thread safe, so we disable the use of threads.
829 if (opt->funcname && !opt->unmatch_name_only && !opt->status_only &&
830 !opt->name_only)
831 return 0;
833 return 1;
836 static void std_output(struct grep_opt *opt, const void *buf, size_t size)
838 fwrite(buf, size, 1, stdout);
841 static int grep_buffer_1(struct grep_opt *opt, const char *name,
842 char *buf, unsigned long size, int collect_hits)
844 char *bol = buf;
845 unsigned long left = size;
846 unsigned lno = 1;
847 unsigned last_hit = 0;
848 int binary_match_only = 0;
849 unsigned count = 0;
850 int try_lookahead = 0;
851 enum grep_context ctx = GREP_CONTEXT_HEAD;
852 xdemitconf_t xecfg;
854 if (!opt->output)
855 opt->output = std_output;
857 if (opt->last_shown && (opt->pre_context || opt->post_context) &&
858 opt->output == std_output)
859 opt->show_hunk_mark = 1;
860 opt->last_shown = 0;
862 switch (opt->binary) {
863 case GREP_BINARY_DEFAULT:
864 if (buffer_is_binary(buf, size))
865 binary_match_only = 1;
866 break;
867 case GREP_BINARY_NOMATCH:
868 if (buffer_is_binary(buf, size))
869 return 0; /* Assume unmatch */
870 break;
871 case GREP_BINARY_TEXT:
872 break;
873 default:
874 die("bug: unknown binary handling mode");
877 memset(&xecfg, 0, sizeof(xecfg));
878 if (opt->funcname && !opt->unmatch_name_only && !opt->status_only &&
879 !opt->name_only && !binary_match_only && !collect_hits) {
880 struct userdiff_driver *drv = userdiff_find_by_path(name);
881 if (drv && drv->funcname.pattern) {
882 const struct userdiff_funcname *pe = &drv->funcname;
883 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
884 opt->priv = &xecfg;
887 try_lookahead = should_lookahead(opt);
889 while (left) {
890 char *eol, ch;
891 int hit;
894 * look_ahead() skips quicly to the line that possibly
895 * has the next hit; don't call it if we need to do
896 * something more than just skipping the current line
897 * in response to an unmatch for the current line. E.g.
898 * inside a post-context window, we will show the current
899 * line as a context around the previous hit when it
900 * doesn't hit.
902 if (try_lookahead
903 && !(last_hit
904 && lno <= last_hit + opt->post_context)
905 && look_ahead(opt, &left, &lno, &bol))
906 break;
907 eol = end_of_line(bol, &left);
908 ch = *eol;
909 *eol = 0;
911 if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
912 ctx = GREP_CONTEXT_BODY;
914 hit = match_line(opt, bol, eol, ctx, collect_hits);
915 *eol = ch;
917 if (collect_hits)
918 goto next_line;
920 /* "grep -v -e foo -e bla" should list lines
921 * that do not have either, so inversion should
922 * be done outside.
924 if (opt->invert)
925 hit = !hit;
926 if (opt->unmatch_name_only) {
927 if (hit)
928 return 0;
929 goto next_line;
931 if (hit) {
932 count++;
933 if (opt->status_only)
934 return 1;
935 if (opt->name_only) {
936 show_name(opt, name);
937 return 1;
939 if (opt->count)
940 goto next_line;
941 if (binary_match_only) {
942 opt->output(opt, "Binary file ", 12);
943 output_color(opt, name, strlen(name),
944 opt->color_filename);
945 opt->output(opt, " matches\n", 9);
946 return 1;
948 /* Hit at this line. If we haven't shown the
949 * pre-context lines, we would need to show them.
951 if (opt->pre_context)
952 show_pre_context(opt, name, buf, bol, lno);
953 else if (opt->funcname)
954 show_funcname_line(opt, name, buf, bol, lno);
955 show_line(opt, bol, eol, name, lno, ':');
956 last_hit = lno;
958 else if (last_hit &&
959 lno <= last_hit + opt->post_context) {
960 /* If the last hit is within the post context,
961 * we need to show this line.
963 show_line(opt, bol, eol, name, lno, '-');
966 next_line:
967 bol = eol + 1;
968 if (!left)
969 break;
970 left--;
971 lno++;
974 if (collect_hits)
975 return 0;
977 if (opt->status_only)
978 return 0;
979 if (opt->unmatch_name_only) {
980 /* We did not see any hit, so we want to show this */
981 show_name(opt, name);
982 return 1;
985 xdiff_clear_find_func(&xecfg);
986 opt->priv = NULL;
988 /* NEEDSWORK:
989 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
990 * which feels mostly useless but sometimes useful. Maybe
991 * make it another option? For now suppress them.
993 if (opt->count && count) {
994 char buf[32];
995 output_color(opt, name, strlen(name), opt->color_filename);
996 output_sep(opt, ':');
997 snprintf(buf, sizeof(buf), "%u\n", count);
998 opt->output(opt, buf, strlen(buf));
999 return 1;
1001 return !!last_hit;
1004 static void clr_hit_marker(struct grep_expr *x)
1006 /* All-hit markers are meaningful only at the very top level
1007 * OR node.
1009 while (1) {
1010 x->hit = 0;
1011 if (x->node != GREP_NODE_OR)
1012 return;
1013 x->u.binary.left->hit = 0;
1014 x = x->u.binary.right;
1018 static int chk_hit_marker(struct grep_expr *x)
1020 /* Top level nodes have hit markers. See if they all are hits */
1021 while (1) {
1022 if (x->node != GREP_NODE_OR)
1023 return x->hit;
1024 if (!x->u.binary.left->hit)
1025 return 0;
1026 x = x->u.binary.right;
1030 int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size)
1033 * we do not have to do the two-pass grep when we do not check
1034 * buffer-wide "all-match".
1036 if (!opt->all_match)
1037 return grep_buffer_1(opt, name, buf, size, 0);
1039 /* Otherwise the toplevel "or" terms hit a bit differently.
1040 * We first clear hit markers from them.
1042 clr_hit_marker(opt->pattern_expression);
1043 grep_buffer_1(opt, name, buf, size, 1);
1045 if (!chk_hit_marker(opt->pattern_expression))
1046 return 0;
1048 return grep_buffer_1(opt, name, buf, size, 0);