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 if (opt->fixed || is_fixed(p->pattern))
44 if (opt->regflags & REG_ICASE)
49 err = regcomp(&p->regexp, p->pattern, opt->regflags);
54 sprintf(where, "In '%s' at %d, ",
57 sprintf(where, "%s, ", p->origin);
60 regerror(err, &p->regexp, errbuf, 1024);
62 die("%s'%s': %s", where, p->pattern, errbuf);
66 static struct grep_expr *compile_pattern_or(struct grep_pat **);
67 static struct grep_expr *compile_pattern_atom(struct grep_pat **list)
74 case GREP_PATTERN: /* atom */
75 case GREP_PATTERN_HEAD:
76 case GREP_PATTERN_BODY:
77 x = xcalloc(1, sizeof (struct grep_expr));
78 x->node = GREP_NODE_ATOM;
84 x = compile_pattern_or(list);
87 if (!*list || (*list)->token != GREP_CLOSE_PAREN)
88 die("unmatched parenthesis");
89 *list = (*list)->next;
96 static struct grep_expr *compile_pattern_not(struct grep_pat **list)
105 die("--not not followed by pattern expression");
107 x = xcalloc(1, sizeof (struct grep_expr));
108 x->node = GREP_NODE_NOT;
109 x->u.unary = compile_pattern_not(list);
111 die("--not followed by non pattern expression");
114 return compile_pattern_atom(list);
118 static struct grep_expr *compile_pattern_and(struct grep_pat **list)
121 struct grep_expr *x, *y, *z;
123 x = compile_pattern_not(list);
125 if (p && p->token == GREP_AND) {
127 die("--and not followed by pattern expression");
129 y = compile_pattern_and(list);
131 die("--and not followed by pattern expression");
132 z = xcalloc(1, sizeof (struct grep_expr));
133 z->node = GREP_NODE_AND;
134 z->u.binary.left = x;
135 z->u.binary.right = y;
141 static struct grep_expr *compile_pattern_or(struct grep_pat **list)
144 struct grep_expr *x, *y, *z;
146 x = compile_pattern_and(list);
148 if (x && p && p->token != GREP_CLOSE_PAREN) {
149 y = compile_pattern_or(list);
151 die("not a pattern expression %s", p->pattern);
152 z = xcalloc(1, sizeof (struct grep_expr));
153 z->node = GREP_NODE_OR;
154 z->u.binary.left = x;
155 z->u.binary.right = y;
161 static struct grep_expr *compile_pattern_expr(struct grep_pat **list)
163 return compile_pattern_or(list);
166 void compile_grep_patterns(struct grep_opt *opt)
173 for (p = opt->pattern_list; p; p = p->next) {
175 case GREP_PATTERN: /* atom */
176 case GREP_PATTERN_HEAD:
177 case GREP_PATTERN_BODY:
178 compile_regexp(p, opt);
189 /* Then bundle them up in an expression.
190 * A classic recursive descent parser would do.
192 p = opt->pattern_list;
193 opt->pattern_expression = compile_pattern_expr(&p);
195 die("incomplete pattern expression: %s", p->pattern);
198 static void free_pattern_expr(struct grep_expr *x)
204 free_pattern_expr(x->u.unary);
208 free_pattern_expr(x->u.binary.left);
209 free_pattern_expr(x->u.binary.right);
215 void free_grep_patterns(struct grep_opt *opt)
217 struct grep_pat *p, *n;
219 for (p = opt->pattern_list; p; p = n) {
222 case GREP_PATTERN: /* atom */
223 case GREP_PATTERN_HEAD:
224 case GREP_PATTERN_BODY:
235 free_pattern_expr(opt->pattern_expression);
238 static char *end_of_line(char *cp, unsigned long *left)
240 unsigned long l = *left;
241 while (l && *cp != '\n') {
249 static int word_char(char ch)
251 return isalnum(ch) || ch == '_';
254 static void show_line(struct grep_opt *opt, const char *bol, const char *eol,
255 const char *name, unsigned lno, char sign)
257 if (opt->null_following_name)
260 printf("%s%c", name, sign);
262 printf("%d%c", lno, sign);
263 printf("%.*s\n", (int)(eol-bol), bol);
266 static void show_name(struct grep_opt *opt, const char *name)
268 printf("%s%c", name, opt->null_following_name ? '\0' : '\n');
271 static int fixmatch(const char *pattern, char *line, regmatch_t *match)
273 char *hit = strstr(line, pattern);
275 match->rm_so = match->rm_eo = -1;
279 match->rm_so = hit - line;
280 match->rm_eo = match->rm_so + strlen(pattern);
285 static int strip_timestamp(char *bol, char **eol_p)
290 while (bol < --eol) {
306 { "committer ", 10 },
309 static int match_one_pattern(struct grep_opt *opt, struct grep_pat *p, char *bol, char *eol, enum grep_context ctx)
313 regmatch_t pmatch[10];
315 if ((p->token != GREP_PATTERN) &&
316 ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
319 if (p->token == GREP_PATTERN_HEAD) {
322 assert(p->field < ARRAY_SIZE(header_field));
323 field = header_field[p->field].field;
324 len = header_field[p->field].len;
325 if (strncmp(bol, field, len))
328 saved_ch = strip_timestamp(bol, &eol);
333 regex_t *exp = &p->regexp;
334 hit = !regexec(exp, bol, ARRAY_SIZE(pmatch),
338 hit = !fixmatch(p->pattern, bol, pmatch);
341 if (hit && opt->word_regexp) {
342 if ((pmatch[0].rm_so < 0) ||
343 (eol - bol) <= pmatch[0].rm_so ||
344 (pmatch[0].rm_eo < 0) ||
345 (eol - bol) < pmatch[0].rm_eo)
346 die("regexp returned nonsense");
348 /* Match beginning must be either beginning of the
349 * line, or at word boundary (i.e. the last char must
350 * not be a word char). Similarly, match end must be
351 * either end of the line, or at word boundary
352 * (i.e. the next char must not be a word char).
354 if ( ((pmatch[0].rm_so == 0) ||
355 !word_char(bol[pmatch[0].rm_so-1])) &&
356 ((pmatch[0].rm_eo == (eol-bol)) ||
357 !word_char(bol[pmatch[0].rm_eo])) )
362 if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
363 /* There could be more than one match on the
364 * line, and the first match might not be
365 * strict word match. But later ones could be!
366 * Forward to the next possible start, i.e. the
367 * next position following a non-word char.
369 bol = pmatch[0].rm_so + bol + 1;
370 while (word_char(bol[-1]) && bol < eol)
376 if (p->token == GREP_PATTERN_HEAD && saved_ch)
381 static int match_expr_eval(struct grep_opt *o,
383 char *bol, char *eol,
384 enum grep_context ctx,
391 h = match_one_pattern(o, x->u.atom, bol, eol, ctx);
394 h = !match_expr_eval(o, x->u.unary, bol, eol, ctx, 0);
397 if (!match_expr_eval(o, x->u.binary.left, bol, eol, ctx, 0))
399 h = match_expr_eval(o, x->u.binary.right, bol, eol, ctx, 0);
403 return (match_expr_eval(o, x->u.binary.left,
405 match_expr_eval(o, x->u.binary.right,
407 h = match_expr_eval(o, x->u.binary.left, bol, eol, ctx, 0);
408 x->u.binary.left->hit |= h;
409 h |= match_expr_eval(o, x->u.binary.right, bol, eol, ctx, 1);
412 die("Unexpected node type (internal error) %d", x->node);
419 static int match_expr(struct grep_opt *opt, char *bol, char *eol,
420 enum grep_context ctx, int collect_hits)
422 struct grep_expr *x = opt->pattern_expression;
423 return match_expr_eval(opt, x, bol, eol, ctx, collect_hits);
426 static int match_line(struct grep_opt *opt, char *bol, char *eol,
427 enum grep_context ctx, int collect_hits)
431 return match_expr(opt, bol, eol, ctx, collect_hits);
433 /* we do not call with collect_hits without being extended */
434 for (p = opt->pattern_list; p; p = p->next) {
435 if (match_one_pattern(opt, p, bol, eol, ctx))
441 static int grep_buffer_1(struct grep_opt *opt, const char *name,
442 char *buf, unsigned long size, int collect_hits)
445 unsigned long left = size;
447 struct pre_context_line {
450 } *prev = NULL, *pcl;
451 unsigned last_hit = 0;
452 unsigned last_shown = 0;
453 int binary_match_only = 0;
454 const char *hunk_mark = "";
456 enum grep_context ctx = GREP_CONTEXT_HEAD;
458 if (buffer_is_binary(buf, size)) {
459 switch (opt->binary) {
460 case GREP_BINARY_DEFAULT:
461 binary_match_only = 1;
463 case GREP_BINARY_NOMATCH:
464 return 0; /* Assume unmatch */
471 if (opt->pre_context)
472 prev = xcalloc(opt->pre_context, sizeof(*prev));
473 if (opt->pre_context || opt->post_context)
480 eol = end_of_line(bol, &left);
484 if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
485 ctx = GREP_CONTEXT_BODY;
487 hit = match_line(opt, bol, eol, ctx, collect_hits);
493 /* "grep -v -e foo -e bla" should list lines
494 * that do not have either, so inversion should
499 if (opt->unmatch_name_only) {
506 if (opt->status_only)
508 if (binary_match_only) {
509 printf("Binary file %s matches\n", name);
512 if (opt->name_only) {
513 show_name(opt, name);
516 /* Hit at this line. If we haven't shown the
517 * pre-context lines, we would need to show them.
518 * When asked to do "count", this still show
519 * the context which is nonsense, but the user
520 * deserves to get that ;-).
522 if (opt->pre_context) {
524 if (opt->pre_context < lno)
525 from = lno - opt->pre_context;
528 if (from <= last_shown)
529 from = last_shown + 1;
530 if (last_shown && from != last_shown + 1)
531 fputs(hunk_mark, stdout);
533 pcl = &prev[lno-from-1];
534 show_line(opt, pcl->bol, pcl->eol,
540 if (last_shown && lno != last_shown + 1)
541 fputs(hunk_mark, stdout);
543 show_line(opt, bol, eol, name, lno, ':');
544 last_shown = last_hit = lno;
547 lno <= last_hit + opt->post_context) {
548 /* If the last hit is within the post context,
549 * we need to show this line.
551 if (last_shown && lno != last_shown + 1)
552 fputs(hunk_mark, stdout);
553 show_line(opt, bol, eol, name, lno, '-');
556 if (opt->pre_context) {
557 memmove(prev+1, prev,
558 (opt->pre_context-1) * sizeof(*prev));
575 if (opt->status_only)
577 if (opt->unmatch_name_only) {
578 /* We did not see any hit, so we want to show this */
579 show_name(opt, name);
584 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
585 * which feels mostly useless but sometimes useful. Maybe
586 * make it another option? For now suppress them.
588 if (opt->count && count)
589 printf("%s%c%u\n", name,
590 opt->null_following_name ? '\0' : ':', count);
594 static void clr_hit_marker(struct grep_expr *x)
596 /* All-hit markers are meaningful only at the very top level
601 if (x->node != GREP_NODE_OR)
603 x->u.binary.left->hit = 0;
604 x = x->u.binary.right;
608 static int chk_hit_marker(struct grep_expr *x)
610 /* Top level nodes have hit markers. See if they all are hits */
612 if (x->node != GREP_NODE_OR)
614 if (!x->u.binary.left->hit)
616 x = x->u.binary.right;
620 int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size)
623 * we do not have to do the two-pass grep when we do not check
624 * buffer-wide "all-match".
627 return grep_buffer_1(opt, name, buf, size, 0);
629 /* Otherwise the toplevel "or" terms hit a bit differently.
630 * We first clear hit markers from them.
632 clr_hit_marker(opt->pattern_expression);
633 grep_buffer_1(opt, name, buf, size, 1);
635 if (!chk_hit_marker(opt->pattern_expression))
638 return grep_buffer_1(opt, name, buf, size, 0);