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)
340 hit = strcasestr(s, pattern);
346 hit = memmem(line, eol - line, pattern, strlen(pattern));
349 match->rm_so = match->rm_eo = -1;
353 match->rm_so = hit - line;
354 match->rm_eo = match->rm_so + strlen(pattern);
359 static int strip_timestamp(char *bol, char **eol_p)
364 while (bol < --eol) {
380 { "committer ", 10 },
383 static int match_one_pattern(struct grep_pat *p, char *bol, char *eol,
384 enum grep_context ctx,
385 regmatch_t *pmatch, int eflags)
389 const char *start = bol;
391 if ((p->token != GREP_PATTERN) &&
392 ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
395 if (p->token == GREP_PATTERN_HEAD) {
398 assert(p->field < ARRAY_SIZE(header_field));
399 field = header_field[p->field].field;
400 len = header_field[p->field].len;
401 if (strncmp(bol, field, len))
404 saved_ch = strip_timestamp(bol, &eol);
409 hit = !fixmatch(p->pattern, bol, eol, p->ignore_case, pmatch);
411 hit = !regexec(&p->regexp, bol, 1, pmatch, eflags);
413 if (hit && p->word_regexp) {
414 if ((pmatch[0].rm_so < 0) ||
415 (eol - bol) < pmatch[0].rm_so ||
416 (pmatch[0].rm_eo < 0) ||
417 (eol - bol) < pmatch[0].rm_eo)
418 die("regexp returned nonsense");
420 /* Match beginning must be either beginning of the
421 * line, or at word boundary (i.e. the last char must
422 * not be a word char). Similarly, match end must be
423 * either end of the line, or at word boundary
424 * (i.e. the next char must not be a word char).
426 if ( ((pmatch[0].rm_so == 0) ||
427 !word_char(bol[pmatch[0].rm_so-1])) &&
428 ((pmatch[0].rm_eo == (eol-bol)) ||
429 !word_char(bol[pmatch[0].rm_eo])) )
434 /* Words consist of at least one character. */
435 if (pmatch->rm_so == pmatch->rm_eo)
438 if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
439 /* There could be more than one match on the
440 * line, and the first match might not be
441 * strict word match. But later ones could be!
442 * Forward to the next possible start, i.e. the
443 * next position following a non-word char.
445 bol = pmatch[0].rm_so + bol + 1;
446 while (word_char(bol[-1]) && bol < eol)
448 eflags |= REG_NOTBOL;
453 if (p->token == GREP_PATTERN_HEAD && saved_ch)
456 pmatch[0].rm_so += bol - start;
457 pmatch[0].rm_eo += bol - start;
462 static int match_expr_eval(struct grep_expr *x, char *bol, char *eol,
463 enum grep_context ctx, int collect_hits)
469 die("Not a valid grep expression");
472 h = match_one_pattern(x->u.atom, bol, eol, ctx, &match, 0);
475 h = !match_expr_eval(x->u.unary, bol, eol, ctx, 0);
478 if (!match_expr_eval(x->u.binary.left, bol, eol, ctx, 0))
480 h = match_expr_eval(x->u.binary.right, bol, eol, ctx, 0);
484 return (match_expr_eval(x->u.binary.left,
486 match_expr_eval(x->u.binary.right,
488 h = match_expr_eval(x->u.binary.left, bol, eol, ctx, 0);
489 x->u.binary.left->hit |= h;
490 h |= match_expr_eval(x->u.binary.right, bol, eol, ctx, 1);
493 die("Unexpected node type (internal error) %d", x->node);
500 static int match_expr(struct grep_opt *opt, char *bol, char *eol,
501 enum grep_context ctx, int collect_hits)
503 struct grep_expr *x = opt->pattern_expression;
504 return match_expr_eval(x, bol, eol, ctx, collect_hits);
507 static int match_line(struct grep_opt *opt, char *bol, char *eol,
508 enum grep_context ctx, int collect_hits)
514 return match_expr(opt, bol, eol, ctx, collect_hits);
516 /* we do not call with collect_hits without being extended */
517 for (p = opt->pattern_list; p; p = p->next) {
518 if (match_one_pattern(p, bol, eol, ctx, &match, 0))
524 static int match_next_pattern(struct grep_pat *p, char *bol, char *eol,
525 enum grep_context ctx,
526 regmatch_t *pmatch, int eflags)
530 if (!match_one_pattern(p, bol, eol, ctx, &match, eflags))
532 if (match.rm_so < 0 || match.rm_eo < 0)
534 if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) {
535 if (match.rm_so > pmatch->rm_so)
537 if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo)
540 pmatch->rm_so = match.rm_so;
541 pmatch->rm_eo = match.rm_eo;
545 static int next_match(struct grep_opt *opt, char *bol, char *eol,
546 enum grep_context ctx, regmatch_t *pmatch, int eflags)
551 pmatch->rm_so = pmatch->rm_eo = -1;
553 for (p = opt->pattern_list; p; p = p->next) {
555 case GREP_PATTERN: /* atom */
556 case GREP_PATTERN_HEAD:
557 case GREP_PATTERN_BODY:
558 hit |= match_next_pattern(p, bol, eol, ctx,
569 static void show_line(struct grep_opt *opt, char *bol, char *eol,
570 const char *name, unsigned lno, char sign)
572 int rest = eol - bol;
573 char *line_color = NULL;
575 if (opt->pre_context || opt->post_context) {
576 if (opt->last_shown == 0) {
577 if (opt->show_hunk_mark) {
578 output_color(opt, "--", 2, opt->color_sep);
579 opt->output(opt, "\n", 1);
581 } else if (lno > opt->last_shown + 1) {
582 output_color(opt, "--", 2, opt->color_sep);
583 opt->output(opt, "\n", 1);
586 opt->last_shown = lno;
589 output_color(opt, name, strlen(name), opt->color_filename);
590 output_sep(opt, sign);
594 snprintf(buf, sizeof(buf), "%d", lno);
595 output_color(opt, buf, strlen(buf), opt->color_lineno);
596 output_sep(opt, sign);
600 enum grep_context ctx = GREP_CONTEXT_BODY;
605 line_color = opt->color_selected;
606 else if (sign == '-')
607 line_color = opt->color_context;
608 else if (sign == '=')
609 line_color = opt->color_function;
611 while (next_match(opt, bol, eol, ctx, &match, eflags)) {
612 if (match.rm_so == match.rm_eo)
615 output_color(opt, bol, match.rm_so, line_color);
616 output_color(opt, bol + match.rm_so,
617 match.rm_eo - match.rm_so,
625 output_color(opt, bol, rest, line_color);
626 opt->output(opt, "\n", 1);
629 static int match_funcname(struct grep_opt *opt, char *bol, char *eol)
631 xdemitconf_t *xecfg = opt->priv;
632 if (xecfg && xecfg->find_func) {
634 return xecfg->find_func(bol, eol - bol, buf, 1,
635 xecfg->find_func_priv) >= 0;
640 if (isalpha(*bol) || *bol == '_' || *bol == '$')
645 static void show_funcname_line(struct grep_opt *opt, const char *name,
646 char *buf, char *bol, unsigned lno)
651 while (bol > buf && bol[-1] != '\n')
655 if (lno <= opt->last_shown)
658 if (match_funcname(opt, bol, eol)) {
659 show_line(opt, bol, eol, name, lno, '=');
665 static void show_pre_context(struct grep_opt *opt, const char *name, char *buf,
666 char *bol, unsigned lno)
668 unsigned cur = lno, from = 1, funcname_lno = 0;
669 int funcname_needed = opt->funcname;
671 if (opt->pre_context < lno)
672 from = lno - opt->pre_context;
673 if (from <= opt->last_shown)
674 from = opt->last_shown + 1;
677 while (bol > buf && cur > from) {
680 while (bol > buf && bol[-1] != '\n')
683 if (funcname_needed && match_funcname(opt, bol, eol)) {
689 /* We need to look even further back to find a function signature. */
690 if (opt->funcname && funcname_needed)
691 show_funcname_line(opt, name, buf, bol, cur);
695 char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-';
699 show_line(opt, bol, eol, name, cur, sign);
705 static int should_lookahead(struct grep_opt *opt)
710 return 0; /* punt for too complex stuff */
713 for (p = opt->pattern_list; p; p = p->next) {
714 if (p->token != GREP_PATTERN)
715 return 0; /* punt for "header only" and stuff */
720 static int look_ahead(struct grep_opt *opt,
721 unsigned long *left_p,
725 unsigned lno = *lno_p;
729 regoff_t earliest = -1;
731 for (p = opt->pattern_list; p; p = p->next) {
736 hit = !fixmatch(p->pattern, bol, bol + *left_p,
742 hit = !regexec(&p->regexp, bol, 1, &m, REG_STARTEND);
744 hit = !regexec(&p->regexp, bol, 1, &m, 0);
747 if (!hit || m.rm_so < 0 || m.rm_eo < 0)
749 if (earliest < 0 || m.rm_so < earliest)
754 *bol_p = bol + *left_p;
758 for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--)
759 ; /* find the beginning of the line */
762 for (sp = bol; sp < last_bol; sp++) {
766 *left_p -= last_bol - bol;
772 int grep_threads_ok(const struct grep_opt *opt)
774 /* If this condition is true, then we may use the attribute
775 * machinery in grep_buffer_1. The attribute code is not
776 * thread safe, so we disable the use of threads.
778 if (opt->funcname && !opt->unmatch_name_only && !opt->status_only &&
785 static void std_output(struct grep_opt *opt, const void *buf, size_t size)
787 fwrite(buf, size, 1, stdout);
790 static int grep_buffer_1(struct grep_opt *opt, const char *name,
791 char *buf, unsigned long size, int collect_hits)
794 unsigned long left = size;
796 unsigned last_hit = 0;
797 int binary_match_only = 0;
799 int try_lookahead = 0;
800 enum grep_context ctx = GREP_CONTEXT_HEAD;
804 opt->output = std_output;
806 if (opt->last_shown && (opt->pre_context || opt->post_context) &&
807 opt->output == std_output)
808 opt->show_hunk_mark = 1;
811 switch (opt->binary) {
812 case GREP_BINARY_DEFAULT:
813 if (buffer_is_binary(buf, size))
814 binary_match_only = 1;
816 case GREP_BINARY_NOMATCH:
817 if (buffer_is_binary(buf, size))
818 return 0; /* Assume unmatch */
820 case GREP_BINARY_TEXT:
823 die("bug: unknown binary handling mode");
826 memset(&xecfg, 0, sizeof(xecfg));
827 if (opt->funcname && !opt->unmatch_name_only && !opt->status_only &&
828 !opt->name_only && !binary_match_only && !collect_hits) {
829 struct userdiff_driver *drv = userdiff_find_by_path(name);
830 if (drv && drv->funcname.pattern) {
831 const struct userdiff_funcname *pe = &drv->funcname;
832 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
836 try_lookahead = should_lookahead(opt);
843 * look_ahead() skips quicly to the line that possibly
844 * has the next hit; don't call it if we need to do
845 * something more than just skipping the current line
846 * in response to an unmatch for the current line. E.g.
847 * inside a post-context window, we will show the current
848 * line as a context around the previous hit when it
853 && lno <= last_hit + opt->post_context)
854 && look_ahead(opt, &left, &lno, &bol))
856 eol = end_of_line(bol, &left);
860 if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
861 ctx = GREP_CONTEXT_BODY;
863 hit = match_line(opt, bol, eol, ctx, collect_hits);
869 /* "grep -v -e foo -e bla" should list lines
870 * that do not have either, so inversion should
875 if (opt->unmatch_name_only) {
882 if (opt->status_only)
884 if (opt->name_only) {
885 show_name(opt, name);
890 if (binary_match_only) {
891 opt->output(opt, "Binary file ", 12);
892 output_color(opt, name, strlen(name),
893 opt->color_filename);
894 opt->output(opt, " matches\n", 9);
897 /* Hit at this line. If we haven't shown the
898 * pre-context lines, we would need to show them.
900 if (opt->pre_context)
901 show_pre_context(opt, name, buf, bol, lno);
902 else if (opt->funcname)
903 show_funcname_line(opt, name, buf, bol, lno);
904 show_line(opt, bol, eol, name, lno, ':');
908 lno <= last_hit + opt->post_context) {
909 /* If the last hit is within the post context,
910 * we need to show this line.
912 show_line(opt, bol, eol, name, lno, '-');
926 if (opt->status_only)
928 if (opt->unmatch_name_only) {
929 /* We did not see any hit, so we want to show this */
930 show_name(opt, name);
934 xdiff_clear_find_func(&xecfg);
938 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
939 * which feels mostly useless but sometimes useful. Maybe
940 * make it another option? For now suppress them.
942 if (opt->count && count) {
944 output_color(opt, name, strlen(name), opt->color_filename);
945 output_sep(opt, ':');
946 snprintf(buf, sizeof(buf), "%u\n", count);
947 opt->output(opt, buf, strlen(buf));
953 static void clr_hit_marker(struct grep_expr *x)
955 /* All-hit markers are meaningful only at the very top level
960 if (x->node != GREP_NODE_OR)
962 x->u.binary.left->hit = 0;
963 x = x->u.binary.right;
967 static int chk_hit_marker(struct grep_expr *x)
969 /* Top level nodes have hit markers. See if they all are hits */
971 if (x->node != GREP_NODE_OR)
973 if (!x->u.binary.left->hit)
975 x = x->u.binary.right;
979 int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size)
982 * we do not have to do the two-pass grep when we do not check
983 * buffer-wide "all-match".
986 return grep_buffer_1(opt, name, buf, size, 0);
988 /* Otherwise the toplevel "or" terms hit a bit differently.
989 * We first clear hit markers from them.
991 clr_hit_marker(opt->pattern_expression);
992 grep_buffer_1(opt, name, buf, size, 1);
994 if (!chk_hit_marker(opt->pattern_expression))
997 return grep_buffer_1(opt, name, buf, size, 0);