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
->pattern_tail
= p
;
15 opt
->pattern_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 static int is_fixed(const char *s
)
34 while (*s
&& !is_regex_special(*s
))
39 static void compile_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
43 p
->word_regexp
= opt
->word_regexp
;
45 if (opt
->fixed
|| is_fixed(p
->pattern
))
47 if (opt
->regflags
& REG_ICASE
)
52 err
= regcomp(&p
->regexp
, p
->pattern
, opt
->regflags
);
57 sprintf(where
, "In '%s' at %d, ",
60 sprintf(where
, "%s, ", p
->origin
);
63 regerror(err
, &p
->regexp
, errbuf
, 1024);
65 die("%s'%s': %s", where
, p
->pattern
, errbuf
);
69 static struct grep_expr
*compile_pattern_or(struct grep_pat
**);
70 static struct grep_expr
*compile_pattern_atom(struct grep_pat
**list
)
79 case GREP_PATTERN
: /* atom */
80 case GREP_PATTERN_HEAD
:
81 case GREP_PATTERN_BODY
:
82 x
= xcalloc(1, sizeof (struct grep_expr
));
83 x
->node
= GREP_NODE_ATOM
;
89 x
= compile_pattern_or(list
);
90 if (!*list
|| (*list
)->token
!= GREP_CLOSE_PAREN
)
91 die("unmatched parenthesis");
92 *list
= (*list
)->next
;
99 static struct grep_expr
*compile_pattern_not(struct grep_pat
**list
)
110 die("--not not followed by pattern expression");
112 x
= xcalloc(1, sizeof (struct grep_expr
));
113 x
->node
= GREP_NODE_NOT
;
114 x
->u
.unary
= compile_pattern_not(list
);
116 die("--not followed by non pattern expression");
119 return compile_pattern_atom(list
);
123 static struct grep_expr
*compile_pattern_and(struct grep_pat
**list
)
126 struct grep_expr
*x
, *y
, *z
;
128 x
= compile_pattern_not(list
);
130 if (p
&& p
->token
== GREP_AND
) {
132 die("--and not followed by pattern expression");
134 y
= compile_pattern_and(list
);
136 die("--and not followed by pattern expression");
137 z
= xcalloc(1, sizeof (struct grep_expr
));
138 z
->node
= GREP_NODE_AND
;
139 z
->u
.binary
.left
= x
;
140 z
->u
.binary
.right
= y
;
146 static struct grep_expr
*compile_pattern_or(struct grep_pat
**list
)
149 struct grep_expr
*x
, *y
, *z
;
151 x
= compile_pattern_and(list
);
153 if (x
&& p
&& p
->token
!= GREP_CLOSE_PAREN
) {
154 y
= compile_pattern_or(list
);
156 die("not a pattern expression %s", p
->pattern
);
157 z
= xcalloc(1, sizeof (struct grep_expr
));
158 z
->node
= GREP_NODE_OR
;
159 z
->u
.binary
.left
= x
;
160 z
->u
.binary
.right
= y
;
166 static struct grep_expr
*compile_pattern_expr(struct grep_pat
**list
)
168 return compile_pattern_or(list
);
171 void compile_grep_patterns(struct grep_opt
*opt
)
178 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
180 case GREP_PATTERN
: /* atom */
181 case GREP_PATTERN_HEAD
:
182 case GREP_PATTERN_BODY
:
183 compile_regexp(p
, opt
);
194 /* Then bundle them up in an expression.
195 * A classic recursive descent parser would do.
197 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_name(struct grep_opt
*opt
, const char *name
)
262 printf("%s%c", name
, opt
->null_following_name
? '\0' : '\n');
265 static int fixmatch(const char *pattern
, char *line
, regmatch_t
*match
)
267 char *hit
= strstr(line
, pattern
);
269 match
->rm_so
= match
->rm_eo
= -1;
273 match
->rm_so
= hit
- line
;
274 match
->rm_eo
= match
->rm_so
+ strlen(pattern
);
279 static int strip_timestamp(char *bol
, char **eol_p
)
284 while (bol
< --eol
) {
300 { "committer ", 10 },
303 static int match_one_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
304 enum grep_context ctx
,
305 regmatch_t
*pmatch
, int eflags
)
309 const char *start
= bol
;
311 if ((p
->token
!= GREP_PATTERN
) &&
312 ((p
->token
== GREP_PATTERN_HEAD
) != (ctx
== GREP_CONTEXT_HEAD
)))
315 if (p
->token
== GREP_PATTERN_HEAD
) {
318 assert(p
->field
< ARRAY_SIZE(header_field
));
319 field
= header_field
[p
->field
].field
;
320 len
= header_field
[p
->field
].len
;
321 if (strncmp(bol
, field
, len
))
324 saved_ch
= strip_timestamp(bol
, &eol
);
329 hit
= !fixmatch(p
->pattern
, bol
, pmatch
);
331 hit
= !regexec(&p
->regexp
, bol
, 1, pmatch
, eflags
);
333 if (hit
&& p
->word_regexp
) {
334 if ((pmatch
[0].rm_so
< 0) ||
335 (eol
- bol
) < pmatch
[0].rm_so
||
336 (pmatch
[0].rm_eo
< 0) ||
337 (eol
- bol
) < pmatch
[0].rm_eo
)
338 die("regexp returned nonsense");
340 /* Match beginning must be either beginning of the
341 * line, or at word boundary (i.e. the last char must
342 * not be a word char). Similarly, match end must be
343 * either end of the line, or at word boundary
344 * (i.e. the next char must not be a word char).
346 if ( ((pmatch
[0].rm_so
== 0) ||
347 !word_char(bol
[pmatch
[0].rm_so
-1])) &&
348 ((pmatch
[0].rm_eo
== (eol
-bol
)) ||
349 !word_char(bol
[pmatch
[0].rm_eo
])) )
354 /* Words consist of at least one character. */
355 if (pmatch
->rm_so
== pmatch
->rm_eo
)
358 if (!hit
&& pmatch
[0].rm_so
+ bol
+ 1 < eol
) {
359 /* There could be more than one match on the
360 * line, and the first match might not be
361 * strict word match. But later ones could be!
362 * Forward to the next possible start, i.e. the
363 * next position following a non-word char.
365 bol
= pmatch
[0].rm_so
+ bol
+ 1;
366 while (word_char(bol
[-1]) && bol
< eol
)
368 eflags
|= REG_NOTBOL
;
373 if (p
->token
== GREP_PATTERN_HEAD
&& saved_ch
)
376 pmatch
[0].rm_so
+= bol
- start
;
377 pmatch
[0].rm_eo
+= bol
- start
;
382 static int match_expr_eval(struct grep_expr
*x
, char *bol
, char *eol
,
383 enum grep_context ctx
, int collect_hits
)
389 die("Not a valid grep expression");
392 h
= match_one_pattern(x
->u
.atom
, bol
, eol
, ctx
, &match
, 0);
395 h
= !match_expr_eval(x
->u
.unary
, bol
, eol
, ctx
, 0);
398 if (!match_expr_eval(x
->u
.binary
.left
, bol
, eol
, ctx
, 0))
400 h
= match_expr_eval(x
->u
.binary
.right
, bol
, eol
, ctx
, 0);
404 return (match_expr_eval(x
->u
.binary
.left
,
406 match_expr_eval(x
->u
.binary
.right
,
408 h
= match_expr_eval(x
->u
.binary
.left
, bol
, eol
, ctx
, 0);
409 x
->u
.binary
.left
->hit
|= h
;
410 h
|= match_expr_eval(x
->u
.binary
.right
, bol
, eol
, ctx
, 1);
413 die("Unexpected node type (internal error) %d", x
->node
);
420 static int match_expr(struct grep_opt
*opt
, char *bol
, char *eol
,
421 enum grep_context ctx
, int collect_hits
)
423 struct grep_expr
*x
= opt
->pattern_expression
;
424 return match_expr_eval(x
, bol
, eol
, ctx
, collect_hits
);
427 static int match_line(struct grep_opt
*opt
, char *bol
, char *eol
,
428 enum grep_context ctx
, int collect_hits
)
434 return match_expr(opt
, bol
, eol
, ctx
, collect_hits
);
436 /* we do not call with collect_hits without being extended */
437 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
438 if (match_one_pattern(p
, bol
, eol
, ctx
, &match
, 0))
444 static int match_next_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
445 enum grep_context ctx
,
446 regmatch_t
*pmatch
, int eflags
)
450 if (!match_one_pattern(p
, bol
, eol
, ctx
, &match
, eflags
))
452 if (match
.rm_so
< 0 || match
.rm_eo
< 0)
454 if (pmatch
->rm_so
>= 0 && pmatch
->rm_eo
>= 0) {
455 if (match
.rm_so
> pmatch
->rm_so
)
457 if (match
.rm_so
== pmatch
->rm_so
&& match
.rm_eo
< pmatch
->rm_eo
)
460 pmatch
->rm_so
= match
.rm_so
;
461 pmatch
->rm_eo
= match
.rm_eo
;
465 static int next_match(struct grep_opt
*opt
, char *bol
, char *eol
,
466 enum grep_context ctx
, regmatch_t
*pmatch
, int eflags
)
471 pmatch
->rm_so
= pmatch
->rm_eo
= -1;
473 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
475 case GREP_PATTERN
: /* atom */
476 case GREP_PATTERN_HEAD
:
477 case GREP_PATTERN_BODY
:
478 hit
|= match_next_pattern(p
, bol
, eol
, ctx
,
489 static void show_line(struct grep_opt
*opt
, char *bol
, char *eol
,
490 const char *name
, unsigned lno
, char sign
)
492 int rest
= eol
- bol
;
494 if (opt
->pre_context
|| opt
->post_context
) {
495 if (opt
->last_shown
== 0) {
496 if (opt
->show_hunk_mark
)
497 fputs("--\n", stdout
);
499 opt
->show_hunk_mark
= 1;
500 } else if (lno
> opt
->last_shown
+ 1)
501 fputs("--\n", stdout
);
503 opt
->last_shown
= lno
;
505 if (opt
->null_following_name
)
508 printf("%s%c", name
, sign
);
510 printf("%d%c", lno
, sign
);
513 enum grep_context ctx
= GREP_CONTEXT_BODY
;
518 while (next_match(opt
, bol
, eol
, ctx
, &match
, eflags
)) {
519 if (match
.rm_so
== match
.rm_eo
)
521 printf("%.*s%s%.*s%s",
522 (int)match
.rm_so
, bol
,
524 (int)(match
.rm_eo
- match
.rm_so
), bol
+ match
.rm_so
,
532 printf("%.*s\n", rest
, bol
);
535 static int match_funcname(struct grep_opt
*opt
, char *bol
, char *eol
)
537 xdemitconf_t
*xecfg
= opt
->priv
;
538 if (xecfg
&& xecfg
->find_func
) {
540 return xecfg
->find_func(bol
, eol
- bol
, buf
, 1,
541 xecfg
->find_func_priv
) >= 0;
546 if (isalpha(*bol
) || *bol
== '_' || *bol
== '$')
551 static void show_funcname_line(struct grep_opt
*opt
, const char *name
,
552 char *buf
, char *bol
, unsigned lno
)
557 while (bol
> buf
&& bol
[-1] != '\n')
561 if (lno
<= opt
->last_shown
)
564 if (match_funcname(opt
, bol
, eol
)) {
565 show_line(opt
, bol
, eol
, name
, lno
, '=');
571 static void show_pre_context(struct grep_opt
*opt
, const char *name
, char *buf
,
572 char *bol
, unsigned lno
)
574 unsigned cur
= lno
, from
= 1, funcname_lno
= 0;
575 int funcname_needed
= opt
->funcname
;
577 if (opt
->pre_context
< lno
)
578 from
= lno
- opt
->pre_context
;
579 if (from
<= opt
->last_shown
)
580 from
= opt
->last_shown
+ 1;
583 while (bol
> buf
&& cur
> from
) {
586 while (bol
> buf
&& bol
[-1] != '\n')
589 if (funcname_needed
&& match_funcname(opt
, bol
, eol
)) {
595 /* We need to look even further back to find a function signature. */
596 if (opt
->funcname
&& funcname_needed
)
597 show_funcname_line(opt
, name
, buf
, bol
, cur
);
601 char *eol
= bol
, sign
= (cur
== funcname_lno
) ? '=' : '-';
605 show_line(opt
, bol
, eol
, name
, cur
, sign
);
611 static int grep_buffer_1(struct grep_opt
*opt
, const char *name
,
612 char *buf
, unsigned long size
, int collect_hits
)
615 unsigned long left
= size
;
617 unsigned last_hit
= 0;
618 int binary_match_only
= 0;
620 enum grep_context ctx
= GREP_CONTEXT_HEAD
;
625 if (buffer_is_binary(buf
, size
)) {
626 switch (opt
->binary
) {
627 case GREP_BINARY_DEFAULT
:
628 binary_match_only
= 1;
630 case GREP_BINARY_NOMATCH
:
631 return 0; /* Assume unmatch */
638 memset(&xecfg
, 0, sizeof(xecfg
));
639 if (opt
->funcname
&& !opt
->unmatch_name_only
&& !opt
->status_only
&&
640 !opt
->name_only
&& !binary_match_only
&& !collect_hits
) {
641 struct userdiff_driver
*drv
= userdiff_find_by_path(name
);
642 if (drv
&& drv
->funcname
.pattern
) {
643 const struct userdiff_funcname
*pe
= &drv
->funcname
;
644 xdiff_set_find_func(&xecfg
, pe
->pattern
, pe
->cflags
);
653 eol
= end_of_line(bol
, &left
);
657 if ((ctx
== GREP_CONTEXT_HEAD
) && (eol
== bol
))
658 ctx
= GREP_CONTEXT_BODY
;
660 hit
= match_line(opt
, bol
, eol
, ctx
, collect_hits
);
666 /* "grep -v -e foo -e bla" should list lines
667 * that do not have either, so inversion should
672 if (opt
->unmatch_name_only
) {
679 if (opt
->status_only
)
681 if (binary_match_only
) {
682 printf("Binary file %s matches\n", name
);
685 if (opt
->name_only
) {
686 show_name(opt
, name
);
689 /* Hit at this line. If we haven't shown the
690 * pre-context lines, we would need to show them.
691 * When asked to do "count", this still show
692 * the context which is nonsense, but the user
693 * deserves to get that ;-).
695 if (opt
->pre_context
)
696 show_pre_context(opt
, name
, buf
, bol
, lno
);
697 else if (opt
->funcname
)
698 show_funcname_line(opt
, name
, buf
, bol
, lno
);
700 show_line(opt
, bol
, eol
, name
, lno
, ':');
704 lno
<= last_hit
+ opt
->post_context
) {
705 /* If the last hit is within the post context,
706 * we need to show this line.
708 show_line(opt
, bol
, eol
, name
, lno
, '-');
722 if (opt
->status_only
)
724 if (opt
->unmatch_name_only
) {
725 /* We did not see any hit, so we want to show this */
726 show_name(opt
, name
);
730 xdiff_clear_find_func(&xecfg
);
734 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
735 * which feels mostly useless but sometimes useful. Maybe
736 * make it another option? For now suppress them.
738 if (opt
->count
&& count
)
739 printf("%s%c%u\n", name
,
740 opt
->null_following_name
? '\0' : ':', count
);
744 static void clr_hit_marker(struct grep_expr
*x
)
746 /* All-hit markers are meaningful only at the very top level
751 if (x
->node
!= GREP_NODE_OR
)
753 x
->u
.binary
.left
->hit
= 0;
754 x
= x
->u
.binary
.right
;
758 static int chk_hit_marker(struct grep_expr
*x
)
760 /* Top level nodes have hit markers. See if they all are hits */
762 if (x
->node
!= GREP_NODE_OR
)
764 if (!x
->u
.binary
.left
->hit
)
766 x
= x
->u
.binary
.right
;
770 int grep_buffer(struct grep_opt
*opt
, const char *name
, char *buf
, unsigned long size
)
773 * we do not have to do the two-pass grep when we do not check
774 * buffer-wide "all-match".
777 return grep_buffer_1(opt
, name
, buf
, size
, 0);
779 /* Otherwise the toplevel "or" terms hit a bit differently.
780 * We first clear hit markers from them.
782 clr_hit_marker(opt
->pattern_expression
);
783 grep_buffer_1(opt
, name
, buf
, size
, 1);
785 if (!chk_hit_marker(opt
->pattern_expression
))
788 return grep_buffer_1(opt
, name
, buf
, size
, 0);