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));
12 p->token = GREP_PATTERN_HEAD;
14 *opt->header_tail = p;
15 opt->header_tail = &p->next;
19 void append_grep_pattern(struct grep_opt *opt, const char *pat,
20 const char *origin, int no, enum grep_pat_token t)
22 struct grep_pat *p = xcalloc(1, sizeof(*p));
27 *opt->pattern_tail = p;
28 opt->pattern_tail = &p->next;
32 struct grep_opt *grep_opt_dup(const struct grep_opt *opt)
35 struct grep_opt *ret = xmalloc(sizeof(struct grep_opt));
38 ret->pattern_list = NULL;
39 ret->pattern_tail = &ret->pattern_list;
41 for(pat = opt->pattern_list; pat != NULL; pat = pat->next)
43 if(pat->token == GREP_PATTERN_HEAD)
44 append_header_grep_pattern(ret, pat->field,
47 append_grep_pattern(ret, pat->pattern, pat->origin,
54 static void compile_regexp(struct grep_pat *p, struct grep_opt *opt)
58 p->word_regexp = opt->word_regexp;
59 p->ignore_case = opt->ignore_case;
60 p->fixed = opt->fixed;
65 err = regcomp(&p->regexp, p->pattern, opt->regflags);
70 sprintf(where, "In '%s' at %d, ",
73 sprintf(where, "%s, ", p->origin);
76 regerror(err, &p->regexp, errbuf, 1024);
78 die("%s'%s': %s", where, p->pattern, errbuf);
82 static struct grep_expr *compile_pattern_or(struct grep_pat **);
83 static struct grep_expr *compile_pattern_atom(struct grep_pat **list)
92 case GREP_PATTERN: /* atom */
93 case GREP_PATTERN_HEAD:
94 case GREP_PATTERN_BODY:
95 x = xcalloc(1, sizeof (struct grep_expr));
96 x->node = GREP_NODE_ATOM;
100 case GREP_OPEN_PAREN:
102 x = compile_pattern_or(list);
103 if (!*list || (*list)->token != GREP_CLOSE_PAREN)
104 die("unmatched parenthesis");
105 *list = (*list)->next;
112 static struct grep_expr *compile_pattern_not(struct grep_pat **list)
123 die("--not not followed by pattern expression");
125 x = xcalloc(1, sizeof (struct grep_expr));
126 x->node = GREP_NODE_NOT;
127 x->u.unary = compile_pattern_not(list);
129 die("--not followed by non pattern expression");
132 return compile_pattern_atom(list);
136 static struct grep_expr *compile_pattern_and(struct grep_pat **list)
139 struct grep_expr *x, *y, *z;
141 x = compile_pattern_not(list);
143 if (p && p->token == GREP_AND) {
145 die("--and not followed by pattern expression");
147 y = compile_pattern_and(list);
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;
159 static struct grep_expr *compile_pattern_or(struct grep_pat **list)
162 struct grep_expr *x, *y, *z;
164 x = compile_pattern_and(list);
166 if (x && p && p->token != GREP_CLOSE_PAREN) {
167 y = compile_pattern_or(list);
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;
179 static struct grep_expr *compile_pattern_expr(struct grep_pat **list)
181 return compile_pattern_or(list);
184 void compile_grep_patterns(struct grep_opt *opt)
187 struct grep_expr *header_expr = NULL;
189 if (opt->header_list) {
190 p = opt->header_list;
191 header_expr = compile_pattern_expr(&p);
193 die("incomplete pattern expression: %s", p->pattern);
194 for (p = opt->header_list; p; p = p->next) {
196 case GREP_PATTERN: /* atom */
197 case GREP_PATTERN_HEAD:
198 case GREP_PATTERN_BODY:
199 compile_regexp(p, opt);
208 for (p = opt->pattern_list; p; p = p->next) {
210 case GREP_PATTERN: /* atom */
211 case GREP_PATTERN_HEAD:
212 case GREP_PATTERN_BODY:
213 compile_regexp(p, opt);
221 if (opt->all_match || header_expr)
223 else if (!opt->extended)
226 /* Then bundle them up in an expression.
227 * A classic recursive descent parser would do.
229 p = opt->pattern_list;
231 opt->pattern_expression = compile_pattern_expr(&p);
233 die("incomplete pattern expression: %s", p->pattern);
238 if (opt->pattern_expression) {
240 z = xcalloc(1, sizeof(*z));
241 z->node = GREP_NODE_OR;
242 z->u.binary.left = opt->pattern_expression;
243 z->u.binary.right = header_expr;
244 opt->pattern_expression = z;
246 opt->pattern_expression = header_expr;
251 static void free_pattern_expr(struct grep_expr *x)
257 free_pattern_expr(x->u.unary);
261 free_pattern_expr(x->u.binary.left);
262 free_pattern_expr(x->u.binary.right);
268 void free_grep_patterns(struct grep_opt *opt)
270 struct grep_pat *p, *n;
272 for (p = opt->pattern_list; p; p = n) {
275 case GREP_PATTERN: /* atom */
276 case GREP_PATTERN_HEAD:
277 case GREP_PATTERN_BODY:
288 free_pattern_expr(opt->pattern_expression);
291 static char *end_of_line(char *cp, unsigned long *left)
293 unsigned long l = *left;
294 while (l && *cp != '\n') {
302 static int word_char(char ch)
304 return isalnum(ch) || ch == '_';
307 static void output_color(struct grep_opt *opt, const void *data, size_t size,
310 if (opt->color && color && color[0]) {
311 opt->output(opt, color, strlen(color));
312 opt->output(opt, data, size);
313 opt->output(opt, GIT_COLOR_RESET, strlen(GIT_COLOR_RESET));
315 opt->output(opt, data, size);
318 static void output_sep(struct grep_opt *opt, char sign)
320 if (opt->null_following_name)
321 opt->output(opt, "\0", 1);
323 output_color(opt, &sign, 1, opt->color_sep);
326 static void show_name(struct grep_opt *opt, const char *name)
328 output_color(opt, name, strlen(name), opt->color_filename);
329 opt->output(opt, opt->null_following_name ? "\0" : "\n", 1);
332 static int fixmatch(const char *pattern, char *line, char *eol,
333 int ignore_case, regmatch_t *match)
338 hit = strcasestr(line, pattern);
340 hit = memmem(line, eol - line, pattern, strlen(pattern));
343 match->rm_so = match->rm_eo = -1;
347 match->rm_so = hit - line;
348 match->rm_eo = match->rm_so + strlen(pattern);
353 static int strip_timestamp(char *bol, char **eol_p)
358 while (bol < --eol) {
374 { "committer ", 10 },
377 static int match_one_pattern(struct grep_pat *p, char *bol, char *eol,
378 enum grep_context ctx,
379 regmatch_t *pmatch, int eflags)
383 const char *start = bol;
385 if ((p->token != GREP_PATTERN) &&
386 ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
389 if (p->token == GREP_PATTERN_HEAD) {
392 assert(p->field < ARRAY_SIZE(header_field));
393 field = header_field[p->field].field;
394 len = header_field[p->field].len;
395 if (strncmp(bol, field, len))
398 saved_ch = strip_timestamp(bol, &eol);
403 hit = !fixmatch(p->pattern, bol, eol, p->ignore_case, pmatch);
405 hit = !regexec(&p->regexp, bol, 1, pmatch, eflags);
407 if (hit && p->word_regexp) {
408 if ((pmatch[0].rm_so < 0) ||
409 (eol - bol) < pmatch[0].rm_so ||
410 (pmatch[0].rm_eo < 0) ||
411 (eol - bol) < pmatch[0].rm_eo)
412 die("regexp returned nonsense");
414 /* Match beginning must be either beginning of the
415 * line, or at word boundary (i.e. the last char must
416 * not be a word char). Similarly, match end must be
417 * either end of the line, or at word boundary
418 * (i.e. the next char must not be a word char).
420 if ( ((pmatch[0].rm_so == 0) ||
421 !word_char(bol[pmatch[0].rm_so-1])) &&
422 ((pmatch[0].rm_eo == (eol-bol)) ||
423 !word_char(bol[pmatch[0].rm_eo])) )
428 /* Words consist of at least one character. */
429 if (pmatch->rm_so == pmatch->rm_eo)
432 if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
433 /* There could be more than one match on the
434 * line, and the first match might not be
435 * strict word match. But later ones could be!
436 * Forward to the next possible start, i.e. the
437 * next position following a non-word char.
439 bol = pmatch[0].rm_so + bol + 1;
440 while (word_char(bol[-1]) && bol < eol)
442 eflags |= REG_NOTBOL;
447 if (p->token == GREP_PATTERN_HEAD && saved_ch)
450 pmatch[0].rm_so += bol - start;
451 pmatch[0].rm_eo += bol - start;
456 static int match_expr_eval(struct grep_expr *x, char *bol, char *eol,
457 enum grep_context ctx, int collect_hits)
463 die("Not a valid grep expression");
466 h = match_one_pattern(x->u.atom, bol, eol, ctx, &match, 0);
469 h = !match_expr_eval(x->u.unary, bol, eol, ctx, 0);
472 if (!match_expr_eval(x->u.binary.left, bol, eol, ctx, 0))
474 h = match_expr_eval(x->u.binary.right, bol, eol, ctx, 0);
478 return (match_expr_eval(x->u.binary.left,
480 match_expr_eval(x->u.binary.right,
482 h = match_expr_eval(x->u.binary.left, bol, eol, ctx, 0);
483 x->u.binary.left->hit |= h;
484 h |= match_expr_eval(x->u.binary.right, bol, eol, ctx, 1);
487 die("Unexpected node type (internal error) %d", x->node);
494 static int match_expr(struct grep_opt *opt, char *bol, char *eol,
495 enum grep_context ctx, int collect_hits)
497 struct grep_expr *x = opt->pattern_expression;
498 return match_expr_eval(x, bol, eol, ctx, collect_hits);
501 static int match_line(struct grep_opt *opt, char *bol, char *eol,
502 enum grep_context ctx, int collect_hits)
508 return match_expr(opt, bol, eol, ctx, collect_hits);
510 /* we do not call with collect_hits without being extended */
511 for (p = opt->pattern_list; p; p = p->next) {
512 if (match_one_pattern(p, bol, eol, ctx, &match, 0))
518 static int match_next_pattern(struct grep_pat *p, char *bol, char *eol,
519 enum grep_context ctx,
520 regmatch_t *pmatch, int eflags)
524 if (!match_one_pattern(p, bol, eol, ctx, &match, eflags))
526 if (match.rm_so < 0 || match.rm_eo < 0)
528 if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) {
529 if (match.rm_so > pmatch->rm_so)
531 if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo)
534 pmatch->rm_so = match.rm_so;
535 pmatch->rm_eo = match.rm_eo;
539 static int next_match(struct grep_opt *opt, char *bol, char *eol,
540 enum grep_context ctx, regmatch_t *pmatch, int eflags)
545 pmatch->rm_so = pmatch->rm_eo = -1;
547 for (p = opt->pattern_list; p; p = p->next) {
549 case GREP_PATTERN: /* atom */
550 case GREP_PATTERN_HEAD:
551 case GREP_PATTERN_BODY:
552 hit |= match_next_pattern(p, bol, eol, ctx,
563 static void show_line(struct grep_opt *opt, char *bol, char *eol,
564 const char *name, unsigned lno, char sign)
566 int rest = eol - bol;
567 char *line_color = NULL;
569 if (opt->pre_context || opt->post_context) {
570 if (opt->last_shown == 0) {
571 if (opt->show_hunk_mark) {
572 output_color(opt, "--", 2, opt->color_sep);
573 opt->output(opt, "\n", 1);
575 } else if (lno > opt->last_shown + 1) {
576 output_color(opt, "--", 2, opt->color_sep);
577 opt->output(opt, "\n", 1);
580 opt->last_shown = lno;
583 output_color(opt, name, strlen(name), opt->color_filename);
584 output_sep(opt, sign);
588 snprintf(buf, sizeof(buf), "%d", lno);
589 output_color(opt, buf, strlen(buf), opt->color_lineno);
590 output_sep(opt, sign);
594 enum grep_context ctx = GREP_CONTEXT_BODY;
599 line_color = opt->color_selected;
600 else if (sign == '-')
601 line_color = opt->color_context;
602 else if (sign == '=')
603 line_color = opt->color_function;
605 while (next_match(opt, bol, eol, ctx, &match, eflags)) {
606 if (match.rm_so == match.rm_eo)
609 output_color(opt, bol, match.rm_so, line_color);
610 output_color(opt, bol + match.rm_so,
611 match.rm_eo - match.rm_so,
619 output_color(opt, bol, rest, line_color);
620 opt->output(opt, "\n", 1);
623 static int match_funcname(struct grep_opt *opt, char *bol, char *eol)
625 xdemitconf_t *xecfg = opt->priv;
626 if (xecfg && xecfg->find_func) {
628 return xecfg->find_func(bol, eol - bol, buf, 1,
629 xecfg->find_func_priv) >= 0;
634 if (isalpha(*bol) || *bol == '_' || *bol == '$')
639 static void show_funcname_line(struct grep_opt *opt, const char *name,
640 char *buf, char *bol, unsigned lno)
645 while (bol > buf && bol[-1] != '\n')
649 if (lno <= opt->last_shown)
652 if (match_funcname(opt, bol, eol)) {
653 show_line(opt, bol, eol, name, lno, '=');
659 static void show_pre_context(struct grep_opt *opt, const char *name, char *buf,
660 char *bol, unsigned lno)
662 unsigned cur = lno, from = 1, funcname_lno = 0;
663 int funcname_needed = opt->funcname;
665 if (opt->pre_context < lno)
666 from = lno - opt->pre_context;
667 if (from <= opt->last_shown)
668 from = opt->last_shown + 1;
671 while (bol > buf && cur > from) {
674 while (bol > buf && bol[-1] != '\n')
677 if (funcname_needed && match_funcname(opt, bol, eol)) {
683 /* We need to look even further back to find a function signature. */
684 if (opt->funcname && funcname_needed)
685 show_funcname_line(opt, name, buf, bol, cur);
689 char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-';
693 show_line(opt, bol, eol, name, cur, sign);
699 static int should_lookahead(struct grep_opt *opt)
704 return 0; /* punt for too complex stuff */
707 for (p = opt->pattern_list; p; p = p->next) {
708 if (p->token != GREP_PATTERN)
709 return 0; /* punt for "header only" and stuff */
714 static int look_ahead(struct grep_opt *opt,
715 unsigned long *left_p,
719 unsigned lno = *lno_p;
723 regoff_t earliest = -1;
725 for (p = opt->pattern_list; p; p = p->next) {
730 hit = !fixmatch(p->pattern, bol, bol + *left_p,
736 hit = !regexec(&p->regexp, bol, 1, &m, REG_STARTEND);
738 hit = !regexec(&p->regexp, bol, 1, &m, 0);
741 if (!hit || m.rm_so < 0 || m.rm_eo < 0)
743 if (earliest < 0 || m.rm_so < earliest)
748 *bol_p = bol + *left_p;
752 for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--)
753 ; /* find the beginning of the line */
756 for (sp = bol; sp < last_bol; sp++) {
760 *left_p -= last_bol - bol;
766 int grep_threads_ok(const struct grep_opt *opt)
768 /* If this condition is true, then we may use the attribute
769 * machinery in grep_buffer_1. The attribute code is not
770 * thread safe, so we disable the use of threads.
772 if (opt->funcname && !opt->unmatch_name_only && !opt->status_only &&
779 static void std_output(struct grep_opt *opt, const void *buf, size_t size)
781 fwrite(buf, size, 1, stdout);
784 static int grep_buffer_1(struct grep_opt *opt, const char *name,
785 char *buf, unsigned long size, int collect_hits)
788 unsigned long left = size;
790 unsigned last_hit = 0;
791 int binary_match_only = 0;
793 int try_lookahead = 0;
794 enum grep_context ctx = GREP_CONTEXT_HEAD;
798 opt->output = std_output;
800 if (opt->last_shown && (opt->pre_context || opt->post_context) &&
801 opt->output == std_output)
802 opt->show_hunk_mark = 1;
805 switch (opt->binary) {
806 case GREP_BINARY_DEFAULT:
807 if (buffer_is_binary(buf, size))
808 binary_match_only = 1;
810 case GREP_BINARY_NOMATCH:
811 if (buffer_is_binary(buf, size))
812 return 0; /* Assume unmatch */
814 case GREP_BINARY_TEXT:
817 die("bug: unknown binary handling mode");
820 memset(&xecfg, 0, sizeof(xecfg));
821 if (opt->funcname && !opt->unmatch_name_only && !opt->status_only &&
822 !opt->name_only && !binary_match_only && !collect_hits) {
823 struct userdiff_driver *drv = userdiff_find_by_path(name);
824 if (drv && drv->funcname.pattern) {
825 const struct userdiff_funcname *pe = &drv->funcname;
826 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
830 try_lookahead = should_lookahead(opt);
837 * look_ahead() skips quicly to the line that possibly
838 * has the next hit; don't call it if we need to do
839 * something more than just skipping the current line
840 * in response to an unmatch for the current line. E.g.
841 * inside a post-context window, we will show the current
842 * line as a context around the previous hit when it
847 && lno <= last_hit + opt->post_context)
848 && look_ahead(opt, &left, &lno, &bol))
850 eol = end_of_line(bol, &left);
854 if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
855 ctx = GREP_CONTEXT_BODY;
857 hit = match_line(opt, bol, eol, ctx, collect_hits);
863 /* "grep -v -e foo -e bla" should list lines
864 * that do not have either, so inversion should
869 if (opt->unmatch_name_only) {
876 if (opt->status_only)
878 if (opt->name_only) {
879 show_name(opt, name);
884 if (binary_match_only) {
885 opt->output(opt, "Binary file ", 12);
886 output_color(opt, name, strlen(name),
887 opt->color_filename);
888 opt->output(opt, " matches\n", 9);
891 /* Hit at this line. If we haven't shown the
892 * pre-context lines, we would need to show them.
894 if (opt->pre_context)
895 show_pre_context(opt, name, buf, bol, lno);
896 else if (opt->funcname)
897 show_funcname_line(opt, name, buf, bol, lno);
898 show_line(opt, bol, eol, name, lno, ':');
902 lno <= last_hit + opt->post_context) {
903 /* If the last hit is within the post context,
904 * we need to show this line.
906 show_line(opt, bol, eol, name, lno, '-');
920 if (opt->status_only)
922 if (opt->unmatch_name_only) {
923 /* We did not see any hit, so we want to show this */
924 show_name(opt, name);
928 xdiff_clear_find_func(&xecfg);
932 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
933 * which feels mostly useless but sometimes useful. Maybe
934 * make it another option? For now suppress them.
936 if (opt->count && count) {
938 output_color(opt, name, strlen(name), opt->color_filename);
939 output_sep(opt, ':');
940 snprintf(buf, sizeof(buf), "%u\n", count);
941 opt->output(opt, buf, strlen(buf));
947 static void clr_hit_marker(struct grep_expr *x)
949 /* All-hit markers are meaningful only at the very top level
954 if (x->node != GREP_NODE_OR)
956 x->u.binary.left->hit = 0;
957 x = x->u.binary.right;
961 static int chk_hit_marker(struct grep_expr *x)
963 /* Top level nodes have hit markers. See if they all are hits */
965 if (x->node != GREP_NODE_OR)
967 if (!x->u.binary.left->hit)
969 x = x->u.binary.right;
973 int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size)
976 * we do not have to do the two-pass grep when we do not check
977 * buffer-wide "all-match".
980 return grep_buffer_1(opt, name, buf, size, 0);
982 /* Otherwise the toplevel "or" terms hit a bit differently.
983 * We first clear hit markers from them.
985 clr_hit_marker(opt->pattern_expression);
986 grep_buffer_1(opt, name, buf, size, 1);
988 if (!chk_hit_marker(opt->pattern_expression))
991 return grep_buffer_1(opt, name, buf, size, 0);