3 #include "xdiff-interface.h"
5 void append_header_grep_pattern(struct grep_opt *opt, enum grep_header_field field, const char *pat)
7 struct grep_pat *p = xcalloc(1, sizeof(*p));
11 p->token = GREP_PATTERN_HEAD;
13 *opt->pattern_tail = p;
14 opt->pattern_tail = &p->next;
18 void append_grep_pattern(struct grep_opt *opt, const char *pat,
19 const char *origin, int no, enum grep_pat_token t)
21 struct grep_pat *p = xcalloc(1, sizeof(*p));
26 *opt->pattern_tail = p;
27 opt->pattern_tail = &p->next;
31 static int is_fixed(const char *s)
33 while (*s && !is_regex_special(*s))
38 static void compile_regexp(struct grep_pat *p, struct grep_opt *opt)
42 p->word_regexp = opt->word_regexp;
44 if (opt->fixed || is_fixed(p->pattern))
46 if (opt->regflags & REG_ICASE)
51 err = regcomp(&p->regexp, p->pattern, opt->regflags);
56 sprintf(where, "In '%s' at %d, ",
59 sprintf(where, "%s, ", p->origin);
62 regerror(err, &p->regexp, errbuf, 1024);
64 die("%s'%s': %s", where, p->pattern, errbuf);
68 static struct grep_expr *compile_pattern_or(struct grep_pat **);
69 static struct grep_expr *compile_pattern_atom(struct grep_pat **list)
78 case GREP_PATTERN: /* atom */
79 case GREP_PATTERN_HEAD:
80 case GREP_PATTERN_BODY:
81 x = xcalloc(1, sizeof (struct grep_expr));
82 x->node = GREP_NODE_ATOM;
88 x = compile_pattern_or(list);
89 if (!*list || (*list)->token != GREP_CLOSE_PAREN)
90 die("unmatched parenthesis");
91 *list = (*list)->next;
98 static struct grep_expr *compile_pattern_not(struct grep_pat **list)
109 die("--not not followed by pattern expression");
111 x = xcalloc(1, sizeof (struct grep_expr));
112 x->node = GREP_NODE_NOT;
113 x->u.unary = compile_pattern_not(list);
115 die("--not followed by non pattern expression");
118 return compile_pattern_atom(list);
122 static struct grep_expr *compile_pattern_and(struct grep_pat **list)
125 struct grep_expr *x, *y, *z;
127 x = compile_pattern_not(list);
129 if (p && p->token == GREP_AND) {
131 die("--and not followed by pattern expression");
133 y = compile_pattern_and(list);
135 die("--and not followed by pattern expression");
136 z = xcalloc(1, sizeof (struct grep_expr));
137 z->node = GREP_NODE_AND;
138 z->u.binary.left = x;
139 z->u.binary.right = y;
145 static struct grep_expr *compile_pattern_or(struct grep_pat **list)
148 struct grep_expr *x, *y, *z;
150 x = compile_pattern_and(list);
152 if (x && p && p->token != GREP_CLOSE_PAREN) {
153 y = compile_pattern_or(list);
155 die("not a pattern expression %s", p->pattern);
156 z = xcalloc(1, sizeof (struct grep_expr));
157 z->node = GREP_NODE_OR;
158 z->u.binary.left = x;
159 z->u.binary.right = y;
165 static struct grep_expr *compile_pattern_expr(struct grep_pat **list)
167 return compile_pattern_or(list);
170 void compile_grep_patterns(struct grep_opt *opt)
177 for (p = opt->pattern_list; p; p = p->next) {
179 case GREP_PATTERN: /* atom */
180 case GREP_PATTERN_HEAD:
181 case GREP_PATTERN_BODY:
182 compile_regexp(p, opt);
193 /* Then bundle them up in an expression.
194 * A classic recursive descent parser would do.
196 p = opt->pattern_list;
198 opt->pattern_expression = compile_pattern_expr(&p);
200 die("incomplete pattern expression: %s", p->pattern);
203 static void free_pattern_expr(struct grep_expr *x)
209 free_pattern_expr(x->u.unary);
213 free_pattern_expr(x->u.binary.left);
214 free_pattern_expr(x->u.binary.right);
220 void free_grep_patterns(struct grep_opt *opt)
222 struct grep_pat *p, *n;
224 for (p = opt->pattern_list; p; p = n) {
227 case GREP_PATTERN: /* atom */
228 case GREP_PATTERN_HEAD:
229 case GREP_PATTERN_BODY:
240 free_pattern_expr(opt->pattern_expression);
243 static char *end_of_line(char *cp, unsigned long *left)
245 unsigned long l = *left;
246 while (l && *cp != '\n') {
254 static int word_char(char ch)
256 return isalnum(ch) || ch == '_';
259 static void show_name(struct grep_opt *opt, const char *name)
261 printf("%s%c", name, opt->null_following_name ? '\0' : '\n');
264 static int fixmatch(const char *pattern, char *line, regmatch_t *match)
266 char *hit = strstr(line, pattern);
268 match->rm_so = match->rm_eo = -1;
272 match->rm_so = hit - line;
273 match->rm_eo = match->rm_so + strlen(pattern);
278 static int strip_timestamp(char *bol, char **eol_p)
283 while (bol < --eol) {
299 { "committer ", 10 },
302 static int match_one_pattern(struct grep_pat *p, char *bol, char *eol,
303 enum grep_context ctx,
304 regmatch_t *pmatch, int eflags)
308 const char *start = bol;
310 if ((p->token != GREP_PATTERN) &&
311 ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
314 if (p->token == GREP_PATTERN_HEAD) {
317 assert(p->field < ARRAY_SIZE(header_field));
318 field = header_field[p->field].field;
319 len = header_field[p->field].len;
320 if (strncmp(bol, field, len))
323 saved_ch = strip_timestamp(bol, &eol);
328 hit = !fixmatch(p->pattern, bol, pmatch);
330 hit = !regexec(&p->regexp, bol, 1, pmatch, eflags);
332 if (hit && p->word_regexp) {
333 if ((pmatch[0].rm_so < 0) ||
334 (eol - bol) <= pmatch[0].rm_so ||
335 (pmatch[0].rm_eo < 0) ||
336 (eol - bol) < pmatch[0].rm_eo)
337 die("regexp returned nonsense");
339 /* Match beginning must be either beginning of the
340 * line, or at word boundary (i.e. the last char must
341 * not be a word char). Similarly, match end must be
342 * either end of the line, or at word boundary
343 * (i.e. the next char must not be a word char).
345 if ( ((pmatch[0].rm_so == 0) ||
346 !word_char(bol[pmatch[0].rm_so-1])) &&
347 ((pmatch[0].rm_eo == (eol-bol)) ||
348 !word_char(bol[pmatch[0].rm_eo])) )
353 if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
354 /* There could be more than one match on the
355 * line, and the first match might not be
356 * strict word match. But later ones could be!
357 * Forward to the next possible start, i.e. the
358 * next position following a non-word char.
360 bol = pmatch[0].rm_so + bol + 1;
361 while (word_char(bol[-1]) && bol < eol)
363 eflags |= REG_NOTBOL;
368 if (p->token == GREP_PATTERN_HEAD && saved_ch)
371 pmatch[0].rm_so += bol - start;
372 pmatch[0].rm_eo += bol - start;
377 static int match_expr_eval(struct grep_expr *x, char *bol, char *eol,
378 enum grep_context ctx, int collect_hits)
384 die("Not a valid grep expression");
387 h = match_one_pattern(x->u.atom, bol, eol, ctx, &match, 0);
390 h = !match_expr_eval(x->u.unary, bol, eol, ctx, 0);
393 if (!match_expr_eval(x->u.binary.left, bol, eol, ctx, 0))
395 h = match_expr_eval(x->u.binary.right, bol, eol, ctx, 0);
399 return (match_expr_eval(x->u.binary.left,
401 match_expr_eval(x->u.binary.right,
403 h = match_expr_eval(x->u.binary.left, bol, eol, ctx, 0);
404 x->u.binary.left->hit |= h;
405 h |= match_expr_eval(x->u.binary.right, bol, eol, ctx, 1);
408 die("Unexpected node type (internal error) %d", x->node);
415 static int match_expr(struct grep_opt *opt, char *bol, char *eol,
416 enum grep_context ctx, int collect_hits)
418 struct grep_expr *x = opt->pattern_expression;
419 return match_expr_eval(x, bol, eol, ctx, collect_hits);
422 static int match_line(struct grep_opt *opt, char *bol, char *eol,
423 enum grep_context ctx, int collect_hits)
429 return match_expr(opt, bol, eol, ctx, collect_hits);
431 /* we do not call with collect_hits without being extended */
432 for (p = opt->pattern_list; p; p = p->next) {
433 if (match_one_pattern(p, bol, eol, ctx, &match, 0))
439 static int match_next_pattern(struct grep_pat *p, char *bol, char *eol,
440 enum grep_context ctx,
441 regmatch_t *pmatch, int eflags)
445 if (!match_one_pattern(p, bol, eol, ctx, &match, eflags))
447 if (match.rm_so < 0 || match.rm_eo < 0)
449 if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) {
450 if (match.rm_so > pmatch->rm_so)
452 if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo)
455 pmatch->rm_so = match.rm_so;
456 pmatch->rm_eo = match.rm_eo;
460 static int next_match(struct grep_opt *opt, char *bol, char *eol,
461 enum grep_context ctx, regmatch_t *pmatch, int eflags)
466 pmatch->rm_so = pmatch->rm_eo = -1;
468 for (p = opt->pattern_list; p; p = p->next) {
470 case GREP_PATTERN: /* atom */
471 case GREP_PATTERN_HEAD:
472 case GREP_PATTERN_BODY:
473 hit |= match_next_pattern(p, bol, eol, ctx,
484 static void show_line(struct grep_opt *opt, char *bol, char *eol,
485 const char *name, unsigned lno, char sign)
487 int rest = eol - bol;
489 if (opt->null_following_name)
492 printf("%s%c", name, sign);
494 printf("%d%c", lno, sign);
497 enum grep_context ctx = GREP_CONTEXT_BODY;
502 while (next_match(opt, bol, eol, ctx, &match, eflags)) {
503 if (match.rm_so == match.rm_eo)
505 printf("%.*s%s%.*s%s",
506 (int)match.rm_so, bol,
508 (int)(match.rm_eo - match.rm_so), bol + match.rm_so,
516 printf("%.*s\n", rest, bol);
519 static int grep_buffer_1(struct grep_opt *opt, const char *name,
520 char *buf, unsigned long size, int collect_hits)
523 unsigned long left = size;
525 struct pre_context_line {
528 } *prev = NULL, *pcl;
529 unsigned last_hit = 0;
530 unsigned last_shown = 0;
531 int binary_match_only = 0;
532 const char *hunk_mark = "";
534 enum grep_context ctx = GREP_CONTEXT_HEAD;
536 if (buffer_is_binary(buf, size)) {
537 switch (opt->binary) {
538 case GREP_BINARY_DEFAULT:
539 binary_match_only = 1;
541 case GREP_BINARY_NOMATCH:
542 return 0; /* Assume unmatch */
549 if (opt->pre_context)
550 prev = xcalloc(opt->pre_context, sizeof(*prev));
551 if (opt->pre_context || opt->post_context)
558 eol = end_of_line(bol, &left);
562 if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
563 ctx = GREP_CONTEXT_BODY;
565 hit = match_line(opt, bol, eol, ctx, collect_hits);
571 /* "grep -v -e foo -e bla" should list lines
572 * that do not have either, so inversion should
577 if (opt->unmatch_name_only) {
584 if (opt->status_only)
586 if (binary_match_only) {
587 printf("Binary file %s matches\n", name);
590 if (opt->name_only) {
591 show_name(opt, name);
594 /* Hit at this line. If we haven't shown the
595 * pre-context lines, we would need to show them.
596 * When asked to do "count", this still show
597 * the context which is nonsense, but the user
598 * deserves to get that ;-).
600 if (opt->pre_context) {
602 if (opt->pre_context < lno)
603 from = lno - opt->pre_context;
606 if (from <= last_shown)
607 from = last_shown + 1;
608 if (last_shown && from != last_shown + 1)
609 fputs(hunk_mark, stdout);
611 pcl = &prev[lno-from-1];
612 show_line(opt, pcl->bol, pcl->eol,
618 if (last_shown && lno != last_shown + 1)
619 fputs(hunk_mark, stdout);
621 show_line(opt, bol, eol, name, lno, ':');
622 last_shown = last_hit = lno;
625 lno <= last_hit + opt->post_context) {
626 /* If the last hit is within the post context,
627 * we need to show this line.
629 if (last_shown && lno != last_shown + 1)
630 fputs(hunk_mark, stdout);
631 show_line(opt, bol, eol, name, lno, '-');
634 if (opt->pre_context) {
635 memmove(prev+1, prev,
636 (opt->pre_context-1) * sizeof(*prev));
653 if (opt->status_only)
655 if (opt->unmatch_name_only) {
656 /* We did not see any hit, so we want to show this */
657 show_name(opt, name);
662 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
663 * which feels mostly useless but sometimes useful. Maybe
664 * make it another option? For now suppress them.
666 if (opt->count && count)
667 printf("%s%c%u\n", name,
668 opt->null_following_name ? '\0' : ':', count);
672 static void clr_hit_marker(struct grep_expr *x)
674 /* All-hit markers are meaningful only at the very top level
679 if (x->node != GREP_NODE_OR)
681 x->u.binary.left->hit = 0;
682 x = x->u.binary.right;
686 static int chk_hit_marker(struct grep_expr *x)
688 /* Top level nodes have hit markers. See if they all are hits */
690 if (x->node != GREP_NODE_OR)
692 if (!x->u.binary.left->hit)
694 x = x->u.binary.right;
698 int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size)
701 * we do not have to do the two-pass grep when we do not check
702 * buffer-wide "all-match".
705 return grep_buffer_1(opt, name, buf, size, 0);
707 /* Otherwise the toplevel "or" terms hit a bit differently.
708 * We first clear hit markers from them.
710 clr_hit_marker(opt->pattern_expression);
711 grep_buffer_1(opt, name, buf, size, 1);
713 if (!chk_hit_marker(opt->pattern_expression))
716 return grep_buffer_1(opt, name, buf, size, 0);