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 isregexspecial(int c)
33 return isspecial(c) || c == '$' || c == '(' || c == ')' || c == '+' ||
34 c == '.' || c == '^' || c == '{' || c == '|';
37 static int is_fixed(const char *s)
39 while (!isregexspecial(*s))
44 static void compile_regexp(struct grep_pat *p, struct grep_opt *opt)
48 if (opt->fixed || is_fixed(p->pattern))
50 if (opt->regflags & REG_ICASE)
55 err = regcomp(&p->regexp, p->pattern, opt->regflags);
60 sprintf(where, "In '%s' at %d, ",
63 sprintf(where, "%s, ", p->origin);
66 regerror(err, &p->regexp, errbuf, 1024);
68 die("%s'%s': %s", where, p->pattern, errbuf);
72 static struct grep_expr *compile_pattern_or(struct grep_pat **);
73 static struct grep_expr *compile_pattern_atom(struct grep_pat **list)
80 case GREP_PATTERN: /* atom */
81 case GREP_PATTERN_HEAD:
82 case GREP_PATTERN_BODY:
83 x = xcalloc(1, sizeof (struct grep_expr));
84 x->node = GREP_NODE_ATOM;
90 x = compile_pattern_or(list);
93 if (!*list || (*list)->token != GREP_CLOSE_PAREN)
94 die("unmatched parenthesis");
95 *list = (*list)->next;
102 static struct grep_expr *compile_pattern_not(struct grep_pat **list)
111 die("--not not followed by pattern expression");
113 x = xcalloc(1, sizeof (struct grep_expr));
114 x->node = GREP_NODE_NOT;
115 x->u.unary = compile_pattern_not(list);
117 die("--not followed by non pattern expression");
120 return compile_pattern_atom(list);
124 static struct grep_expr *compile_pattern_and(struct grep_pat **list)
127 struct grep_expr *x, *y, *z;
129 x = compile_pattern_not(list);
131 if (p && p->token == GREP_AND) {
133 die("--and not followed by pattern expression");
135 y = compile_pattern_and(list);
137 die("--and not followed by pattern expression");
138 z = xcalloc(1, sizeof (struct grep_expr));
139 z->node = GREP_NODE_AND;
140 z->u.binary.left = x;
141 z->u.binary.right = y;
147 static struct grep_expr *compile_pattern_or(struct grep_pat **list)
150 struct grep_expr *x, *y, *z;
152 x = compile_pattern_and(list);
154 if (x && p && p->token != GREP_CLOSE_PAREN) {
155 y = compile_pattern_or(list);
157 die("not a pattern expression %s", p->pattern);
158 z = xcalloc(1, sizeof (struct grep_expr));
159 z->node = GREP_NODE_OR;
160 z->u.binary.left = x;
161 z->u.binary.right = y;
167 static struct grep_expr *compile_pattern_expr(struct grep_pat **list)
169 return compile_pattern_or(list);
172 void compile_grep_patterns(struct grep_opt *opt)
179 for (p = opt->pattern_list; p; p = p->next) {
181 case GREP_PATTERN: /* atom */
182 case GREP_PATTERN_HEAD:
183 case GREP_PATTERN_BODY:
184 compile_regexp(p, opt);
195 /* Then bundle them up in an expression.
196 * A classic recursive descent parser would do.
198 p = opt->pattern_list;
199 opt->pattern_expression = compile_pattern_expr(&p);
201 die("incomplete pattern expression: %s", p->pattern);
204 static void free_pattern_expr(struct grep_expr *x)
210 free_pattern_expr(x->u.unary);
214 free_pattern_expr(x->u.binary.left);
215 free_pattern_expr(x->u.binary.right);
221 void free_grep_patterns(struct grep_opt *opt)
223 struct grep_pat *p, *n;
225 for (p = opt->pattern_list; p; p = n) {
228 case GREP_PATTERN: /* atom */
229 case GREP_PATTERN_HEAD:
230 case GREP_PATTERN_BODY:
241 free_pattern_expr(opt->pattern_expression);
244 static char *end_of_line(char *cp, unsigned long *left)
246 unsigned long l = *left;
247 while (l && *cp != '\n') {
255 static int word_char(char ch)
257 return isalnum(ch) || ch == '_';
260 static void show_line(struct grep_opt *opt, const char *bol, const char *eol,
261 const char *name, unsigned lno, char sign)
263 if (opt->null_following_name)
266 printf("%s%c", name, sign);
268 printf("%d%c", lno, sign);
269 printf("%.*s\n", (int)(eol-bol), bol);
272 static void show_name(struct grep_opt *opt, const char *name)
274 printf("%s%c", name, opt->null_following_name ? '\0' : '\n');
277 static int fixmatch(const char *pattern, char *line, regmatch_t *match)
279 char *hit = strstr(line, pattern);
281 match->rm_so = match->rm_eo = -1;
285 match->rm_so = hit - line;
286 match->rm_eo = match->rm_so + strlen(pattern);
291 static int strip_timestamp(char *bol, char **eol_p)
296 while (bol < --eol) {
312 { "committer ", 10 },
315 static int match_one_pattern(struct grep_opt *opt, struct grep_pat *p, char *bol, char *eol, enum grep_context ctx)
319 regmatch_t pmatch[10];
321 if ((p->token != GREP_PATTERN) &&
322 ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
325 if (p->token == GREP_PATTERN_HEAD) {
328 assert(p->field < ARRAY_SIZE(header_field));
329 field = header_field[p->field].field;
330 len = header_field[p->field].len;
331 if (strncmp(bol, field, len))
334 saved_ch = strip_timestamp(bol, &eol);
339 regex_t *exp = &p->regexp;
340 hit = !regexec(exp, bol, ARRAY_SIZE(pmatch),
344 hit = !fixmatch(p->pattern, bol, pmatch);
347 if (hit && opt->word_regexp) {
348 if ((pmatch[0].rm_so < 0) ||
349 (eol - bol) <= pmatch[0].rm_so ||
350 (pmatch[0].rm_eo < 0) ||
351 (eol - bol) < pmatch[0].rm_eo)
352 die("regexp returned nonsense");
354 /* Match beginning must be either beginning of the
355 * line, or at word boundary (i.e. the last char must
356 * not be a word char). Similarly, match end must be
357 * either end of the line, or at word boundary
358 * (i.e. the next char must not be a word char).
360 if ( ((pmatch[0].rm_so == 0) ||
361 !word_char(bol[pmatch[0].rm_so-1])) &&
362 ((pmatch[0].rm_eo == (eol-bol)) ||
363 !word_char(bol[pmatch[0].rm_eo])) )
368 if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
369 /* There could be more than one match on the
370 * line, and the first match might not be
371 * strict word match. But later ones could be!
372 * Forward to the next possible start, i.e. the
373 * next position following a non-word char.
375 bol = pmatch[0].rm_so + bol + 1;
376 while (word_char(bol[-1]) && bol < eol)
382 if (p->token == GREP_PATTERN_HEAD && saved_ch)
387 static int match_expr_eval(struct grep_opt *o,
389 char *bol, char *eol,
390 enum grep_context ctx,
397 h = match_one_pattern(o, x->u.atom, bol, eol, ctx);
400 h = !match_expr_eval(o, x->u.unary, bol, eol, ctx, 0);
404 return (match_expr_eval(o, x->u.binary.left,
406 match_expr_eval(o, x->u.binary.right,
408 h = match_expr_eval(o, x->u.binary.left, bol, eol, ctx, 0);
409 h &= match_expr_eval(o, x->u.binary.right, bol, eol, ctx, 0);
413 return (match_expr_eval(o, x->u.binary.left,
415 match_expr_eval(o, x->u.binary.right,
417 h = match_expr_eval(o, x->u.binary.left, bol, eol, ctx, 0);
418 x->u.binary.left->hit |= h;
419 h |= match_expr_eval(o, x->u.binary.right, bol, eol, ctx, 1);
422 die("Unexpected node type (internal error) %d", x->node);
429 static int match_expr(struct grep_opt *opt, char *bol, char *eol,
430 enum grep_context ctx, int collect_hits)
432 struct grep_expr *x = opt->pattern_expression;
433 return match_expr_eval(opt, x, bol, eol, ctx, collect_hits);
436 static int match_line(struct grep_opt *opt, char *bol, char *eol,
437 enum grep_context ctx, int collect_hits)
441 return match_expr(opt, bol, eol, ctx, collect_hits);
443 /* we do not call with collect_hits without being extended */
444 for (p = opt->pattern_list; p; p = p->next) {
445 if (match_one_pattern(opt, p, bol, eol, ctx))
451 static int grep_buffer_1(struct grep_opt *opt, const char *name,
452 char *buf, unsigned long size, int collect_hits)
455 unsigned long left = size;
457 struct pre_context_line {
460 } *prev = NULL, *pcl;
461 unsigned last_hit = 0;
462 unsigned last_shown = 0;
463 int binary_match_only = 0;
464 const char *hunk_mark = "";
466 enum grep_context ctx = GREP_CONTEXT_HEAD;
468 if (buffer_is_binary(buf, size)) {
469 switch (opt->binary) {
470 case GREP_BINARY_DEFAULT:
471 binary_match_only = 1;
473 case GREP_BINARY_NOMATCH:
474 return 0; /* Assume unmatch */
481 if (opt->pre_context)
482 prev = xcalloc(opt->pre_context, sizeof(*prev));
483 if (opt->pre_context || opt->post_context)
490 eol = end_of_line(bol, &left);
494 if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
495 ctx = GREP_CONTEXT_BODY;
497 hit = match_line(opt, bol, eol, ctx, collect_hits);
503 /* "grep -v -e foo -e bla" should list lines
504 * that do not have either, so inversion should
509 if (opt->unmatch_name_only) {
516 if (opt->status_only)
518 if (binary_match_only) {
519 printf("Binary file %s matches\n", name);
522 if (opt->name_only) {
523 show_name(opt, name);
526 /* Hit at this line. If we haven't shown the
527 * pre-context lines, we would need to show them.
528 * When asked to do "count", this still show
529 * the context which is nonsense, but the user
530 * deserves to get that ;-).
532 if (opt->pre_context) {
534 if (opt->pre_context < lno)
535 from = lno - opt->pre_context;
538 if (from <= last_shown)
539 from = last_shown + 1;
540 if (last_shown && from != last_shown + 1)
541 fputs(hunk_mark, stdout);
543 pcl = &prev[lno-from-1];
544 show_line(opt, pcl->bol, pcl->eol,
550 if (last_shown && lno != last_shown + 1)
551 fputs(hunk_mark, stdout);
553 show_line(opt, bol, eol, name, lno, ':');
554 last_shown = last_hit = lno;
557 lno <= last_hit + opt->post_context) {
558 /* If the last hit is within the post context,
559 * we need to show this line.
561 if (last_shown && lno != last_shown + 1)
562 fputs(hunk_mark, stdout);
563 show_line(opt, bol, eol, name, lno, '-');
566 if (opt->pre_context) {
567 memmove(prev+1, prev,
568 (opt->pre_context-1) * sizeof(*prev));
585 if (opt->status_only)
587 if (opt->unmatch_name_only) {
588 /* We did not see any hit, so we want to show this */
589 show_name(opt, name);
594 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
595 * which feels mostly useless but sometimes useful. Maybe
596 * make it another option? For now suppress them.
598 if (opt->count && count)
599 printf("%s%c%u\n", name,
600 opt->null_following_name ? '\0' : ':', count);
604 static void clr_hit_marker(struct grep_expr *x)
606 /* All-hit markers are meaningful only at the very top level
611 if (x->node != GREP_NODE_OR)
613 x->u.binary.left->hit = 0;
614 x = x->u.binary.right;
618 static int chk_hit_marker(struct grep_expr *x)
620 /* Top level nodes have hit markers. See if they all are hits */
622 if (x->node != GREP_NODE_OR)
624 if (!x->u.binary.left->hit)
626 x = x->u.binary.right;
630 int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size)
633 * we do not have to do the two-pass grep when we do not check
634 * buffer-wide "all-match".
637 return grep_buffer_1(opt, name, buf, size, 0);
639 /* Otherwise the toplevel "or" terms hit a bit differently.
640 * We first clear hit markers from them.
642 clr_hit_marker(opt->pattern_expression);
643 grep_buffer_1(opt, name, buf, size, 1);
645 if (!chk_hit_marker(opt->pattern_expression))
648 return grep_buffer_1(opt, name, buf, size, 0);