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));
10 p->patternlen = strlen(pat);
13 p->token = GREP_PATTERN_HEAD;
15 *opt->header_tail = p;
16 opt->header_tail = &p->next;
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));
31 p->patternlen = patlen;
35 *opt->pattern_tail = p;
36 opt->pattern_tail = &p->next;
40 struct grep_opt *grep_opt_dup(const struct grep_opt *opt)
43 struct grep_opt *ret = xmalloc(sizeof(struct grep_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,
55 append_grep_pat(ret, pat->pattern, pat->patternlen,
56 pat->origin, pat->no, pat->token);
62 static void compile_regexp(struct grep_pat *p, struct grep_opt *opt)
66 p->word_regexp = opt->word_regexp;
67 p->ignore_case = opt->ignore_case;
68 p->fixed = opt->fixed;
73 err = regcomp(&p->regexp, p->pattern, opt->regflags);
78 sprintf(where, "In '%s' at %d, ",
81 sprintf(where, "%s, ", p->origin);
84 regerror(err, &p->regexp, errbuf, 1024);
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)
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;
108 case GREP_OPEN_PAREN:
110 x = compile_pattern_or(list);
111 if (!*list || (*list)->token != GREP_CLOSE_PAREN)
112 die("unmatched parenthesis");
113 *list = (*list)->next;
120 static struct grep_expr *compile_pattern_not(struct grep_pat **list)
131 die("--not not followed by pattern expression");
133 x = xcalloc(1, sizeof (struct grep_expr));
134 x->node = GREP_NODE_NOT;
135 x->u.unary = compile_pattern_not(list);
137 die("--not followed by non pattern expression");
140 return compile_pattern_atom(list);
144 static struct grep_expr *compile_pattern_and(struct grep_pat **list)
147 struct grep_expr *x, *y, *z;
149 x = compile_pattern_not(list);
151 if (p && p->token == GREP_AND) {
153 die("--and not followed by pattern expression");
155 y = compile_pattern_and(list);
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;
167 static struct grep_expr *compile_pattern_or(struct grep_pat **list)
170 struct grep_expr *x, *y, *z;
172 x = compile_pattern_and(list);
174 if (x && p && p->token != GREP_CLOSE_PAREN) {
175 y = compile_pattern_or(list);
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;
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;
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;
208 static struct grep_expr *prep_header_patterns(struct grep_opt *opt)
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)
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) {
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;
240 header_group[p->field] = grep_or_expr(h, header_group[p->field]);
245 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) {
246 if (!header_group[fld])
249 header_expr = grep_true_expr();
250 header_expr = grep_or_expr(header_group[fld], header_expr);
255 void compile_grep_patterns(struct grep_opt *opt)
258 struct grep_expr *header_expr = prep_header_patterns(opt);
260 for (p = opt->pattern_list; p; p = p->next) {
262 case GREP_PATTERN: /* atom */
263 case GREP_PATTERN_HEAD:
264 case GREP_PATTERN_BODY:
265 compile_regexp(p, opt);
273 if (opt->all_match || header_expr)
275 else if (!opt->extended)
278 p = opt->pattern_list;
280 opt->pattern_expression = compile_pattern_expr(&p);
282 die("incomplete pattern expression: %s", p->pattern);
287 if (!opt->pattern_expression)
288 opt->pattern_expression = header_expr;
290 opt->pattern_expression = grep_or_expr(opt->pattern_expression,
295 static void free_pattern_expr(struct grep_expr *x)
302 free_pattern_expr(x->u.unary);
306 free_pattern_expr(x->u.binary.left);
307 free_pattern_expr(x->u.binary.right);
313 void free_grep_patterns(struct grep_opt *opt)
315 struct grep_pat *p, *n;
317 for (p = opt->pattern_list; p; p = n) {
320 case GREP_PATTERN: /* atom */
321 case GREP_PATTERN_HEAD:
322 case GREP_PATTERN_BODY:
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') {
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,
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));
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);
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,
382 if (p->ignore_case) {
385 hit = strcasestr(s, p->pattern);
391 hit = memmem(line, eol - line, p->pattern, p->patternlen);
394 match->rm_so = match->rm_eo = -1;
398 match->rm_so = hit - line;
399 match->rm_eo = match->rm_so + p->patternlen;
404 static int regmatch(const regex_t *preg, char *line, char *eol,
405 regmatch_t *match, int eflags)
409 match->rm_eo = eol - line;
410 eflags |= REG_STARTEND;
412 return regexec(preg, line, 1, match, eflags);
415 static int patmatch(struct grep_pat *p, char *line, char *eol,
416 regmatch_t *match, int eflags)
421 hit = !fixmatch(p, line, eol, match);
423 hit = !regmatch(&p->regexp, line, eol, match, eflags);
428 static int strip_timestamp(char *bol, char **eol_p)
433 while (bol < --eol) {
449 { "committer ", 10 },
452 static int match_one_pattern(struct grep_pat *p, char *bol, char *eol,
453 enum grep_context ctx,
454 regmatch_t *pmatch, int eflags)
458 const char *start = bol;
460 if ((p->token != GREP_PATTERN) &&
461 ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
464 if (p->token == GREP_PATTERN_HEAD) {
467 assert(p->field < ARRAY_SIZE(header_field));
468 field = header_field[p->field].field;
469 len = header_field[p->field].len;
470 if (strncmp(bol, field, len))
473 saved_ch = strip_timestamp(bol, &eol);
477 hit = patmatch(p, bol, eol, pmatch, eflags);
479 if (hit && p->word_regexp) {
480 if ((pmatch[0].rm_so < 0) ||
481 (eol - bol) < pmatch[0].rm_so ||
482 (pmatch[0].rm_eo < 0) ||
483 (eol - bol) < pmatch[0].rm_eo)
484 die("regexp returned nonsense");
486 /* Match beginning must be either beginning of the
487 * line, or at word boundary (i.e. the last char must
488 * not be a word char). Similarly, match end must be
489 * either end of the line, or at word boundary
490 * (i.e. the next char must not be a word char).
492 if ( ((pmatch[0].rm_so == 0) ||
493 !word_char(bol[pmatch[0].rm_so-1])) &&
494 ((pmatch[0].rm_eo == (eol-bol)) ||
495 !word_char(bol[pmatch[0].rm_eo])) )
500 /* Words consist of at least one character. */
501 if (pmatch->rm_so == pmatch->rm_eo)
504 if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
505 /* There could be more than one match on the
506 * line, and the first match might not be
507 * strict word match. But later ones could be!
508 * Forward to the next possible start, i.e. the
509 * next position following a non-word char.
511 bol = pmatch[0].rm_so + bol + 1;
512 while (word_char(bol[-1]) && bol < eol)
514 eflags |= REG_NOTBOL;
519 if (p->token == GREP_PATTERN_HEAD && saved_ch)
522 pmatch[0].rm_so += bol - start;
523 pmatch[0].rm_eo += bol - start;
528 static int match_expr_eval(struct grep_expr *x, char *bol, char *eol,
529 enum grep_context ctx, int collect_hits)
535 die("Not a valid grep expression");
541 h = match_one_pattern(x->u.atom, bol, eol, ctx, &match, 0);
544 h = !match_expr_eval(x->u.unary, bol, eol, ctx, 0);
547 if (!match_expr_eval(x->u.binary.left, bol, eol, ctx, 0))
549 h = match_expr_eval(x->u.binary.right, bol, eol, ctx, 0);
553 return (match_expr_eval(x->u.binary.left,
555 match_expr_eval(x->u.binary.right,
557 h = match_expr_eval(x->u.binary.left, bol, eol, ctx, 0);
558 x->u.binary.left->hit |= h;
559 h |= match_expr_eval(x->u.binary.right, bol, eol, ctx, 1);
562 die("Unexpected node type (internal error) %d", x->node);
569 static int match_expr(struct grep_opt *opt, char *bol, char *eol,
570 enum grep_context ctx, int collect_hits)
572 struct grep_expr *x = opt->pattern_expression;
573 return match_expr_eval(x, bol, eol, ctx, collect_hits);
576 static int match_line(struct grep_opt *opt, char *bol, char *eol,
577 enum grep_context ctx, int collect_hits)
583 return match_expr(opt, bol, eol, ctx, collect_hits);
585 /* we do not call with collect_hits without being extended */
586 for (p = opt->pattern_list; p; p = p->next) {
587 if (match_one_pattern(p, bol, eol, ctx, &match, 0))
593 static int match_next_pattern(struct grep_pat *p, char *bol, char *eol,
594 enum grep_context ctx,
595 regmatch_t *pmatch, int eflags)
599 if (!match_one_pattern(p, bol, eol, ctx, &match, eflags))
601 if (match.rm_so < 0 || match.rm_eo < 0)
603 if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) {
604 if (match.rm_so > pmatch->rm_so)
606 if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo)
609 pmatch->rm_so = match.rm_so;
610 pmatch->rm_eo = match.rm_eo;
614 static int next_match(struct grep_opt *opt, char *bol, char *eol,
615 enum grep_context ctx, regmatch_t *pmatch, int eflags)
620 pmatch->rm_so = pmatch->rm_eo = -1;
622 for (p = opt->pattern_list; p; p = p->next) {
624 case GREP_PATTERN: /* atom */
625 case GREP_PATTERN_HEAD:
626 case GREP_PATTERN_BODY:
627 hit |= match_next_pattern(p, bol, eol, ctx,
638 static void show_line(struct grep_opt *opt, char *bol, char *eol,
639 const char *name, unsigned lno, char sign)
641 int rest = eol - bol;
642 char *line_color = NULL;
644 if (opt->pre_context || opt->post_context) {
645 if (opt->last_shown == 0) {
646 if (opt->show_hunk_mark) {
647 output_color(opt, "--", 2, opt->color_sep);
648 opt->output(opt, "\n", 1);
650 } else if (lno > opt->last_shown + 1) {
651 output_color(opt, "--", 2, opt->color_sep);
652 opt->output(opt, "\n", 1);
655 opt->last_shown = lno;
658 output_color(opt, name, strlen(name), opt->color_filename);
659 output_sep(opt, sign);
663 snprintf(buf, sizeof(buf), "%d", lno);
664 output_color(opt, buf, strlen(buf), opt->color_lineno);
665 output_sep(opt, sign);
669 enum grep_context ctx = GREP_CONTEXT_BODY;
674 line_color = opt->color_selected;
675 else if (sign == '-')
676 line_color = opt->color_context;
677 else if (sign == '=')
678 line_color = opt->color_function;
680 while (next_match(opt, bol, eol, ctx, &match, eflags)) {
681 if (match.rm_so == match.rm_eo)
684 output_color(opt, bol, match.rm_so, line_color);
685 output_color(opt, bol + match.rm_so,
686 match.rm_eo - match.rm_so,
694 output_color(opt, bol, rest, line_color);
695 opt->output(opt, "\n", 1);
698 static int match_funcname(struct grep_opt *opt, char *bol, char *eol)
700 xdemitconf_t *xecfg = opt->priv;
701 if (xecfg && xecfg->find_func) {
703 return xecfg->find_func(bol, eol - bol, buf, 1,
704 xecfg->find_func_priv) >= 0;
709 if (isalpha(*bol) || *bol == '_' || *bol == '$')
714 static void show_funcname_line(struct grep_opt *opt, const char *name,
715 char *buf, char *bol, unsigned lno)
720 while (bol > buf && bol[-1] != '\n')
724 if (lno <= opt->last_shown)
727 if (match_funcname(opt, bol, eol)) {
728 show_line(opt, bol, eol, name, lno, '=');
734 static void show_pre_context(struct grep_opt *opt, const char *name, char *buf,
735 char *bol, unsigned lno)
737 unsigned cur = lno, from = 1, funcname_lno = 0;
738 int funcname_needed = opt->funcname;
740 if (opt->pre_context < lno)
741 from = lno - opt->pre_context;
742 if (from <= opt->last_shown)
743 from = opt->last_shown + 1;
746 while (bol > buf && cur > from) {
749 while (bol > buf && bol[-1] != '\n')
752 if (funcname_needed && match_funcname(opt, bol, eol)) {
758 /* We need to look even further back to find a function signature. */
759 if (opt->funcname && funcname_needed)
760 show_funcname_line(opt, name, buf, bol, cur);
764 char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-';
768 show_line(opt, bol, eol, name, cur, sign);
774 static int should_lookahead(struct grep_opt *opt)
779 return 0; /* punt for too complex stuff */
782 for (p = opt->pattern_list; p; p = p->next) {
783 if (p->token != GREP_PATTERN)
784 return 0; /* punt for "header only" and stuff */
789 static int look_ahead(struct grep_opt *opt,
790 unsigned long *left_p,
794 unsigned lno = *lno_p;
798 regoff_t earliest = -1;
800 for (p = opt->pattern_list; p; p = p->next) {
804 hit = patmatch(p, bol, bol + *left_p, &m, 0);
805 if (!hit || m.rm_so < 0 || m.rm_eo < 0)
807 if (earliest < 0 || m.rm_so < earliest)
812 *bol_p = bol + *left_p;
816 for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--)
817 ; /* find the beginning of the line */
820 for (sp = bol; sp < last_bol; sp++) {
824 *left_p -= last_bol - bol;
830 int grep_threads_ok(const struct grep_opt *opt)
832 /* If this condition is true, then we may use the attribute
833 * machinery in grep_buffer_1. The attribute code is not
834 * thread safe, so we disable the use of threads.
836 if (opt->funcname && !opt->unmatch_name_only && !opt->status_only &&
843 static void std_output(struct grep_opt *opt, const void *buf, size_t size)
845 fwrite(buf, size, 1, stdout);
848 static int grep_buffer_1(struct grep_opt *opt, const char *name,
849 char *buf, unsigned long size, int collect_hits)
852 unsigned long left = size;
854 unsigned last_hit = 0;
855 int binary_match_only = 0;
857 int try_lookahead = 0;
858 enum grep_context ctx = GREP_CONTEXT_HEAD;
862 opt->output = std_output;
864 if (opt->last_shown && (opt->pre_context || opt->post_context) &&
865 opt->output == std_output)
866 opt->show_hunk_mark = 1;
869 switch (opt->binary) {
870 case GREP_BINARY_DEFAULT:
871 if (buffer_is_binary(buf, size))
872 binary_match_only = 1;
874 case GREP_BINARY_NOMATCH:
875 if (buffer_is_binary(buf, size))
876 return 0; /* Assume unmatch */
878 case GREP_BINARY_TEXT:
881 die("bug: unknown binary handling mode");
884 memset(&xecfg, 0, sizeof(xecfg));
885 if (opt->funcname && !opt->unmatch_name_only && !opt->status_only &&
886 !opt->name_only && !binary_match_only && !collect_hits) {
887 struct userdiff_driver *drv = userdiff_find_by_path(name);
888 if (drv && drv->funcname.pattern) {
889 const struct userdiff_funcname *pe = &drv->funcname;
890 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
894 try_lookahead = should_lookahead(opt);
901 * look_ahead() skips quickly to the line that possibly
902 * has the next hit; don't call it if we need to do
903 * something more than just skipping the current line
904 * in response to an unmatch for the current line. E.g.
905 * inside a post-context window, we will show the current
906 * line as a context around the previous hit when it
911 && lno <= last_hit + opt->post_context)
912 && look_ahead(opt, &left, &lno, &bol))
914 eol = end_of_line(bol, &left);
918 if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
919 ctx = GREP_CONTEXT_BODY;
921 hit = match_line(opt, bol, eol, ctx, collect_hits);
927 /* "grep -v -e foo -e bla" should list lines
928 * that do not have either, so inversion should
933 if (opt->unmatch_name_only) {
940 if (opt->status_only)
942 if (opt->name_only) {
943 show_name(opt, name);
948 if (binary_match_only) {
949 opt->output(opt, "Binary file ", 12);
950 output_color(opt, name, strlen(name),
951 opt->color_filename);
952 opt->output(opt, " matches\n", 9);
955 /* Hit at this line. If we haven't shown the
956 * pre-context lines, we would need to show them.
958 if (opt->pre_context)
959 show_pre_context(opt, name, buf, bol, lno);
960 else if (opt->funcname)
961 show_funcname_line(opt, name, buf, bol, lno);
962 show_line(opt, bol, eol, name, lno, ':');
966 lno <= last_hit + opt->post_context) {
967 /* If the last hit is within the post context,
968 * we need to show this line.
970 show_line(opt, bol, eol, name, lno, '-');
984 if (opt->status_only)
986 if (opt->unmatch_name_only) {
987 /* We did not see any hit, so we want to show this */
988 show_name(opt, name);
992 xdiff_clear_find_func(&xecfg);
996 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
997 * which feels mostly useless but sometimes useful. Maybe
998 * make it another option? For now suppress them.
1000 if (opt->count && count) {
1002 output_color(opt, name, strlen(name), opt->color_filename);
1003 output_sep(opt, ':');
1004 snprintf(buf, sizeof(buf), "%u\n", count);
1005 opt->output(opt, buf, strlen(buf));
1011 static void clr_hit_marker(struct grep_expr *x)
1013 /* All-hit markers are meaningful only at the very top level
1018 if (x->node != GREP_NODE_OR)
1020 x->u.binary.left->hit = 0;
1021 x = x->u.binary.right;
1025 static int chk_hit_marker(struct grep_expr *x)
1027 /* Top level nodes have hit markers. See if they all are hits */
1029 if (x->node != GREP_NODE_OR)
1031 if (!x->u.binary.left->hit)
1033 x = x->u.binary.right;
1037 int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size)
1040 * we do not have to do the two-pass grep when we do not check
1041 * buffer-wide "all-match".
1043 if (!opt->all_match)
1044 return grep_buffer_1(opt, name, buf, size, 0);
1046 /* Otherwise the toplevel "or" terms hit a bit differently.
1047 * We first clear hit markers from them.
1049 clr_hit_marker(opt->pattern_expression);
1050 grep_buffer_1(opt, name, buf, size, 1);
1052 if (!chk_hit_marker(opt->pattern_expression))
1055 return grep_buffer_1(opt, name, buf, size, 0);