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 void compile_regexp(struct grep_pat *p, struct grep_opt *opt)
33 int err = regcomp(&p->regexp, p->pattern, opt->regflags);
38 sprintf(where, "In '%s' at %d, ",
41 sprintf(where, "%s, ", p->origin);
44 regerror(err, &p->regexp, errbuf, 1024);
46 die("%s'%s': %s", where, p->pattern, errbuf);
50 static struct grep_expr *compile_pattern_or(struct grep_pat **);
51 static struct grep_expr *compile_pattern_atom(struct grep_pat **list)
58 case GREP_PATTERN: /* atom */
59 case GREP_PATTERN_HEAD:
60 case GREP_PATTERN_BODY:
61 x = xcalloc(1, sizeof (struct grep_expr));
62 x->node = GREP_NODE_ATOM;
68 x = compile_pattern_or(list);
71 if (!*list || (*list)->token != GREP_CLOSE_PAREN)
72 die("unmatched parenthesis");
73 *list = (*list)->next;
80 static struct grep_expr *compile_pattern_not(struct grep_pat **list)
89 die("--not not followed by pattern expression");
91 x = xcalloc(1, sizeof (struct grep_expr));
92 x->node = GREP_NODE_NOT;
93 x->u.unary = compile_pattern_not(list);
95 die("--not followed by non pattern expression");
98 return compile_pattern_atom(list);
102 static struct grep_expr *compile_pattern_and(struct grep_pat **list)
105 struct grep_expr *x, *y, *z;
107 x = compile_pattern_not(list);
109 if (p && p->token == GREP_AND) {
111 die("--and not followed by pattern expression");
113 y = compile_pattern_and(list);
115 die("--and not followed by pattern expression");
116 z = xcalloc(1, sizeof (struct grep_expr));
117 z->node = GREP_NODE_AND;
118 z->u.binary.left = x;
119 z->u.binary.right = y;
125 static struct grep_expr *compile_pattern_or(struct grep_pat **list)
128 struct grep_expr *x, *y, *z;
130 x = compile_pattern_and(list);
132 if (x && p && p->token != GREP_CLOSE_PAREN) {
133 y = compile_pattern_or(list);
135 die("not a pattern expression %s", p->pattern);
136 z = xcalloc(1, sizeof (struct grep_expr));
137 z->node = GREP_NODE_OR;
138 z->u.binary.left = x;
139 z->u.binary.right = y;
145 static struct grep_expr *compile_pattern_expr(struct grep_pat **list)
147 return compile_pattern_or(list);
150 void compile_grep_patterns(struct grep_opt *opt)
157 for (p = opt->pattern_list; p; p = p->next) {
159 case GREP_PATTERN: /* atom */
160 case GREP_PATTERN_HEAD:
161 case GREP_PATTERN_BODY:
163 compile_regexp(p, opt);
174 /* Then bundle them up in an expression.
175 * A classic recursive descent parser would do.
177 p = opt->pattern_list;
178 opt->pattern_expression = compile_pattern_expr(&p);
180 die("incomplete pattern expression: %s", p->pattern);
183 static void free_pattern_expr(struct grep_expr *x)
189 free_pattern_expr(x->u.unary);
193 free_pattern_expr(x->u.binary.left);
194 free_pattern_expr(x->u.binary.right);
200 void free_grep_patterns(struct grep_opt *opt)
202 struct grep_pat *p, *n;
204 for (p = opt->pattern_list; p; p = n) {
207 case GREP_PATTERN: /* atom */
208 case GREP_PATTERN_HEAD:
209 case GREP_PATTERN_BODY:
220 free_pattern_expr(opt->pattern_expression);
223 static char *end_of_line(char *cp, unsigned long *left)
225 unsigned long l = *left;
226 while (l && *cp != '\n') {
234 static int word_char(char ch)
236 return isalnum(ch) || ch == '_';
239 static void show_line(struct grep_opt *opt, const char *bol, const char *eol,
240 const char *name, unsigned lno, char sign)
242 if (opt->null_following_name)
245 printf("%s%c", name, sign);
247 printf("%d%c", lno, sign);
248 printf("%.*s\n", (int)(eol-bol), bol);
251 static void show_name(struct grep_opt *opt, const char *name)
253 printf("%s%c", name, opt->null_following_name ? '\0' : '\n');
256 static int fixmatch(const char *pattern, char *line, regmatch_t *match)
258 char *hit = strstr(line, pattern);
260 match->rm_so = match->rm_eo = -1;
264 match->rm_so = hit - line;
265 match->rm_eo = match->rm_so + strlen(pattern);
270 static int strip_timestamp(char *bol, char **eol_p)
275 while (bol < --eol) {
291 { "committer ", 10 },
294 static int match_one_pattern(struct grep_opt *opt, struct grep_pat *p, char *bol, char *eol, enum grep_context ctx)
298 regmatch_t pmatch[10];
300 if ((p->token != GREP_PATTERN) &&
301 ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
304 if (p->token == GREP_PATTERN_HEAD) {
307 assert(p->field < ARRAY_SIZE(header_field));
308 field = header_field[p->field].field;
309 len = header_field[p->field].len;
310 if (strncmp(bol, field, len))
313 saved_ch = strip_timestamp(bol, &eol);
318 regex_t *exp = &p->regexp;
319 hit = !regexec(exp, bol, ARRAY_SIZE(pmatch),
323 hit = !fixmatch(p->pattern, bol, pmatch);
326 if (hit && opt->word_regexp) {
327 if ((pmatch[0].rm_so < 0) ||
328 (eol - bol) <= pmatch[0].rm_so ||
329 (pmatch[0].rm_eo < 0) ||
330 (eol - bol) < pmatch[0].rm_eo)
331 die("regexp returned nonsense");
333 /* Match beginning must be either beginning of the
334 * line, or at word boundary (i.e. the last char must
335 * not be a word char). Similarly, match end must be
336 * either end of the line, or at word boundary
337 * (i.e. the next char must not be a word char).
339 if ( ((pmatch[0].rm_so == 0) ||
340 !word_char(bol[pmatch[0].rm_so-1])) &&
341 ((pmatch[0].rm_eo == (eol-bol)) ||
342 !word_char(bol[pmatch[0].rm_eo])) )
347 if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
348 /* There could be more than one match on the
349 * line, and the first match might not be
350 * strict word match. But later ones could be!
351 * Forward to the next possible start, i.e. the
352 * next position following a non-word char.
354 bol = pmatch[0].rm_so + bol + 1;
355 while (word_char(bol[-1]) && bol < eol)
361 if (p->token == GREP_PATTERN_HEAD && saved_ch)
366 static int match_expr_eval(struct grep_opt *o,
368 char *bol, char *eol,
369 enum grep_context ctx,
376 h = match_one_pattern(o, x->u.atom, bol, eol, ctx);
379 h = !match_expr_eval(o, x->u.unary, bol, eol, ctx, 0);
383 return (match_expr_eval(o, x->u.binary.left,
385 match_expr_eval(o, x->u.binary.right,
387 h = match_expr_eval(o, x->u.binary.left, bol, eol, ctx, 0);
388 h &= match_expr_eval(o, x->u.binary.right, bol, eol, ctx, 0);
392 return (match_expr_eval(o, x->u.binary.left,
394 match_expr_eval(o, x->u.binary.right,
396 h = match_expr_eval(o, x->u.binary.left, bol, eol, ctx, 0);
397 x->u.binary.left->hit |= h;
398 h |= match_expr_eval(o, x->u.binary.right, bol, eol, ctx, 1);
401 die("Unexpected node type (internal error) %d", x->node);
408 static int match_expr(struct grep_opt *opt, char *bol, char *eol,
409 enum grep_context ctx, int collect_hits)
411 struct grep_expr *x = opt->pattern_expression;
412 return match_expr_eval(opt, x, bol, eol, ctx, collect_hits);
415 static int match_line(struct grep_opt *opt, char *bol, char *eol,
416 enum grep_context ctx, int collect_hits)
420 return match_expr(opt, bol, eol, ctx, collect_hits);
422 /* we do not call with collect_hits without being extended */
423 for (p = opt->pattern_list; p; p = p->next) {
424 if (match_one_pattern(opt, p, bol, eol, ctx))
430 static int grep_buffer_1(struct grep_opt *opt, const char *name,
431 char *buf, unsigned long size, int collect_hits)
434 unsigned long left = size;
436 struct pre_context_line {
439 } *prev = NULL, *pcl;
440 unsigned last_hit = 0;
441 unsigned last_shown = 0;
442 int binary_match_only = 0;
443 const char *hunk_mark = "";
445 enum grep_context ctx = GREP_CONTEXT_HEAD;
447 if (buffer_is_binary(buf, size)) {
448 switch (opt->binary) {
449 case GREP_BINARY_DEFAULT:
450 binary_match_only = 1;
452 case GREP_BINARY_NOMATCH:
453 return 0; /* Assume unmatch */
460 if (opt->pre_context)
461 prev = xcalloc(opt->pre_context, sizeof(*prev));
462 if (opt->pre_context || opt->post_context)
469 eol = end_of_line(bol, &left);
473 if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
474 ctx = GREP_CONTEXT_BODY;
476 hit = match_line(opt, bol, eol, ctx, collect_hits);
482 /* "grep -v -e foo -e bla" should list lines
483 * that do not have either, so inversion should
488 if (opt->unmatch_name_only) {
495 if (opt->status_only)
497 if (binary_match_only) {
498 printf("Binary file %s matches\n", name);
501 if (opt->name_only) {
502 show_name(opt, name);
505 /* Hit at this line. If we haven't shown the
506 * pre-context lines, we would need to show them.
507 * When asked to do "count", this still show
508 * the context which is nonsense, but the user
509 * deserves to get that ;-).
511 if (opt->pre_context) {
513 if (opt->pre_context < lno)
514 from = lno - opt->pre_context;
517 if (from <= last_shown)
518 from = last_shown + 1;
519 if (last_shown && from != last_shown + 1)
520 fputs(hunk_mark, stdout);
522 pcl = &prev[lno-from-1];
523 show_line(opt, pcl->bol, pcl->eol,
529 if (last_shown && lno != last_shown + 1)
530 fputs(hunk_mark, stdout);
532 show_line(opt, bol, eol, name, lno, ':');
533 last_shown = last_hit = lno;
536 lno <= last_hit + opt->post_context) {
537 /* If the last hit is within the post context,
538 * we need to show this line.
540 if (last_shown && lno != last_shown + 1)
541 fputs(hunk_mark, stdout);
542 show_line(opt, bol, eol, name, lno, '-');
545 if (opt->pre_context) {
546 memmove(prev+1, prev,
547 (opt->pre_context-1) * sizeof(*prev));
564 if (opt->status_only)
566 if (opt->unmatch_name_only) {
567 /* We did not see any hit, so we want to show this */
568 show_name(opt, name);
573 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
574 * which feels mostly useless but sometimes useful. Maybe
575 * make it another option? For now suppress them.
577 if (opt->count && count)
578 printf("%s%c%u\n", name,
579 opt->null_following_name ? '\0' : ':', count);
583 static void clr_hit_marker(struct grep_expr *x)
585 /* All-hit markers are meaningful only at the very top level
590 if (x->node != GREP_NODE_OR)
592 x->u.binary.left->hit = 0;
593 x = x->u.binary.right;
597 static int chk_hit_marker(struct grep_expr *x)
599 /* Top level nodes have hit markers. See if they all are hits */
601 if (x->node != GREP_NODE_OR)
603 if (!x->u.binary.left->hit)
605 x = x->u.binary.right;
609 int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size)
612 * we do not have to do the two-pass grep when we do not check
613 * buffer-wide "all-match".
616 return grep_buffer_1(opt, name, buf, size, 0);
618 /* Otherwise the toplevel "or" terms hit a bit differently.
619 * We first clear hit markers from them.
621 clr_hit_marker(opt->pattern_expression);
622 grep_buffer_1(opt, name, buf, size, 1);
624 if (!chk_hit_marker(opt->pattern_expression))
627 return grep_buffer_1(opt, name, buf, size, 0);