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
->header_tail
= p
;
15 opt
->header_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 struct grep_opt
*grep_opt_dup(const struct grep_opt
*opt
)
35 struct grep_opt
*ret
= xmalloc(sizeof(struct grep_opt
));
38 ret
->pattern_list
= NULL
;
39 ret
->pattern_tail
= &ret
->pattern_list
;
41 for(pat
= opt
->pattern_list
; pat
!= NULL
; pat
= pat
->next
)
43 if(pat
->token
== GREP_PATTERN_HEAD
)
44 append_header_grep_pattern(ret
, pat
->field
,
47 append_grep_pattern(ret
, pat
->pattern
, pat
->origin
,
54 static void compile_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
58 p
->word_regexp
= opt
->word_regexp
;
59 p
->ignore_case
= opt
->ignore_case
;
60 p
->fixed
= opt
->fixed
;
65 err
= regcomp(&p
->regexp
, p
->pattern
, opt
->regflags
);
70 sprintf(where
, "In '%s' at %d, ",
73 sprintf(where
, "%s, ", p
->origin
);
76 regerror(err
, &p
->regexp
, errbuf
, 1024);
78 die("%s'%s': %s", where
, p
->pattern
, errbuf
);
82 static struct grep_expr
*compile_pattern_or(struct grep_pat
**);
83 static struct grep_expr
*compile_pattern_atom(struct grep_pat
**list
)
92 case GREP_PATTERN
: /* atom */
93 case GREP_PATTERN_HEAD
:
94 case GREP_PATTERN_BODY
:
95 x
= xcalloc(1, sizeof (struct grep_expr
));
96 x
->node
= GREP_NODE_ATOM
;
100 case GREP_OPEN_PAREN
:
102 x
= compile_pattern_or(list
);
103 if (!*list
|| (*list
)->token
!= GREP_CLOSE_PAREN
)
104 die("unmatched parenthesis");
105 *list
= (*list
)->next
;
112 static struct grep_expr
*compile_pattern_not(struct grep_pat
**list
)
123 die("--not not followed by pattern expression");
125 x
= xcalloc(1, sizeof (struct grep_expr
));
126 x
->node
= GREP_NODE_NOT
;
127 x
->u
.unary
= compile_pattern_not(list
);
129 die("--not followed by non pattern expression");
132 return compile_pattern_atom(list
);
136 static struct grep_expr
*compile_pattern_and(struct grep_pat
**list
)
139 struct grep_expr
*x
, *y
, *z
;
141 x
= compile_pattern_not(list
);
143 if (p
&& p
->token
== GREP_AND
) {
145 die("--and not followed by pattern expression");
147 y
= compile_pattern_and(list
);
149 die("--and not followed by pattern expression");
150 z
= xcalloc(1, sizeof (struct grep_expr
));
151 z
->node
= GREP_NODE_AND
;
152 z
->u
.binary
.left
= x
;
153 z
->u
.binary
.right
= y
;
159 static struct grep_expr
*compile_pattern_or(struct grep_pat
**list
)
162 struct grep_expr
*x
, *y
, *z
;
164 x
= compile_pattern_and(list
);
166 if (x
&& p
&& p
->token
!= GREP_CLOSE_PAREN
) {
167 y
= compile_pattern_or(list
);
169 die("not a pattern expression %s", p
->pattern
);
170 z
= xcalloc(1, sizeof (struct grep_expr
));
171 z
->node
= GREP_NODE_OR
;
172 z
->u
.binary
.left
= x
;
173 z
->u
.binary
.right
= y
;
179 static struct grep_expr
*compile_pattern_expr(struct grep_pat
**list
)
181 return compile_pattern_or(list
);
184 void compile_grep_patterns(struct grep_opt
*opt
)
187 struct grep_expr
*header_expr
= NULL
;
189 if (opt
->header_list
) {
190 p
= opt
->header_list
;
191 header_expr
= compile_pattern_expr(&p
);
193 die("incomplete pattern expression: %s", p
->pattern
);
194 for (p
= opt
->header_list
; p
; p
= p
->next
) {
196 case GREP_PATTERN
: /* atom */
197 case GREP_PATTERN_HEAD
:
198 case GREP_PATTERN_BODY
:
199 compile_regexp(p
, opt
);
208 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
210 case GREP_PATTERN
: /* atom */
211 case GREP_PATTERN_HEAD
:
212 case GREP_PATTERN_BODY
:
213 compile_regexp(p
, opt
);
221 if (opt
->all_match
|| header_expr
)
223 else if (!opt
->extended
)
226 /* Then bundle them up in an expression.
227 * A classic recursive descent parser would do.
229 p
= opt
->pattern_list
;
231 opt
->pattern_expression
= compile_pattern_expr(&p
);
233 die("incomplete pattern expression: %s", p
->pattern
);
238 if (opt
->pattern_expression
) {
240 z
= xcalloc(1, sizeof(*z
));
241 z
->node
= GREP_NODE_OR
;
242 z
->u
.binary
.left
= opt
->pattern_expression
;
243 z
->u
.binary
.right
= header_expr
;
244 opt
->pattern_expression
= z
;
246 opt
->pattern_expression
= header_expr
;
251 static void free_pattern_expr(struct grep_expr
*x
)
257 free_pattern_expr(x
->u
.unary
);
261 free_pattern_expr(x
->u
.binary
.left
);
262 free_pattern_expr(x
->u
.binary
.right
);
268 void free_grep_patterns(struct grep_opt
*opt
)
270 struct grep_pat
*p
, *n
;
272 for (p
= opt
->pattern_list
; p
; p
= n
) {
275 case GREP_PATTERN
: /* atom */
276 case GREP_PATTERN_HEAD
:
277 case GREP_PATTERN_BODY
:
288 free_pattern_expr(opt
->pattern_expression
);
291 static char *end_of_line(char *cp
, unsigned long *left
)
293 unsigned long l
= *left
;
294 while (l
&& *cp
!= '\n') {
302 static int word_char(char ch
)
304 return isalnum(ch
) || ch
== '_';
307 static void show_name(struct grep_opt
*opt
, const char *name
)
309 opt
->output(opt
, name
, strlen(name
));
310 opt
->output(opt
, opt
->null_following_name
? "\0" : "\n", 1);
314 static int fixmatch(const char *pattern
, char *line
, int ignore_case
, regmatch_t
*match
)
318 hit
= strcasestr(line
, pattern
);
320 hit
= strstr(line
, pattern
);
323 match
->rm_so
= match
->rm_eo
= -1;
327 match
->rm_so
= hit
- line
;
328 match
->rm_eo
= match
->rm_so
+ strlen(pattern
);
333 static int strip_timestamp(char *bol
, char **eol_p
)
338 while (bol
< --eol
) {
354 { "committer ", 10 },
357 static int match_one_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
358 enum grep_context ctx
,
359 regmatch_t
*pmatch
, int eflags
)
363 const char *start
= bol
;
365 if ((p
->token
!= GREP_PATTERN
) &&
366 ((p
->token
== GREP_PATTERN_HEAD
) != (ctx
== GREP_CONTEXT_HEAD
)))
369 if (p
->token
== GREP_PATTERN_HEAD
) {
372 assert(p
->field
< ARRAY_SIZE(header_field
));
373 field
= header_field
[p
->field
].field
;
374 len
= header_field
[p
->field
].len
;
375 if (strncmp(bol
, field
, len
))
378 saved_ch
= strip_timestamp(bol
, &eol
);
383 hit
= !fixmatch(p
->pattern
, bol
, p
->ignore_case
, pmatch
);
385 hit
= !regexec(&p
->regexp
, bol
, 1, pmatch
, eflags
);
387 if (hit
&& p
->word_regexp
) {
388 if ((pmatch
[0].rm_so
< 0) ||
389 (eol
- bol
) < pmatch
[0].rm_so
||
390 (pmatch
[0].rm_eo
< 0) ||
391 (eol
- bol
) < pmatch
[0].rm_eo
)
392 die("regexp returned nonsense");
394 /* Match beginning must be either beginning of the
395 * line, or at word boundary (i.e. the last char must
396 * not be a word char). Similarly, match end must be
397 * either end of the line, or at word boundary
398 * (i.e. the next char must not be a word char).
400 if ( ((pmatch
[0].rm_so
== 0) ||
401 !word_char(bol
[pmatch
[0].rm_so
-1])) &&
402 ((pmatch
[0].rm_eo
== (eol
-bol
)) ||
403 !word_char(bol
[pmatch
[0].rm_eo
])) )
408 /* Words consist of at least one character. */
409 if (pmatch
->rm_so
== pmatch
->rm_eo
)
412 if (!hit
&& pmatch
[0].rm_so
+ bol
+ 1 < eol
) {
413 /* There could be more than one match on the
414 * line, and the first match might not be
415 * strict word match. But later ones could be!
416 * Forward to the next possible start, i.e. the
417 * next position following a non-word char.
419 bol
= pmatch
[0].rm_so
+ bol
+ 1;
420 while (word_char(bol
[-1]) && bol
< eol
)
422 eflags
|= REG_NOTBOL
;
427 if (p
->token
== GREP_PATTERN_HEAD
&& saved_ch
)
430 pmatch
[0].rm_so
+= bol
- start
;
431 pmatch
[0].rm_eo
+= bol
- start
;
436 static int match_expr_eval(struct grep_expr
*x
, char *bol
, char *eol
,
437 enum grep_context ctx
, int collect_hits
)
443 die("Not a valid grep expression");
446 h
= match_one_pattern(x
->u
.atom
, bol
, eol
, ctx
, &match
, 0);
449 h
= !match_expr_eval(x
->u
.unary
, bol
, eol
, ctx
, 0);
452 if (!match_expr_eval(x
->u
.binary
.left
, bol
, eol
, ctx
, 0))
454 h
= match_expr_eval(x
->u
.binary
.right
, bol
, eol
, ctx
, 0);
458 return (match_expr_eval(x
->u
.binary
.left
,
460 match_expr_eval(x
->u
.binary
.right
,
462 h
= match_expr_eval(x
->u
.binary
.left
, bol
, eol
, ctx
, 0);
463 x
->u
.binary
.left
->hit
|= h
;
464 h
|= match_expr_eval(x
->u
.binary
.right
, bol
, eol
, ctx
, 1);
467 die("Unexpected node type (internal error) %d", x
->node
);
474 static int match_expr(struct grep_opt
*opt
, char *bol
, char *eol
,
475 enum grep_context ctx
, int collect_hits
)
477 struct grep_expr
*x
= opt
->pattern_expression
;
478 return match_expr_eval(x
, bol
, eol
, ctx
, collect_hits
);
481 static int match_line(struct grep_opt
*opt
, char *bol
, char *eol
,
482 enum grep_context ctx
, int collect_hits
)
488 return match_expr(opt
, bol
, eol
, ctx
, collect_hits
);
490 /* we do not call with collect_hits without being extended */
491 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
492 if (match_one_pattern(p
, bol
, eol
, ctx
, &match
, 0))
498 static int match_next_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
499 enum grep_context ctx
,
500 regmatch_t
*pmatch
, int eflags
)
504 if (!match_one_pattern(p
, bol
, eol
, ctx
, &match
, eflags
))
506 if (match
.rm_so
< 0 || match
.rm_eo
< 0)
508 if (pmatch
->rm_so
>= 0 && pmatch
->rm_eo
>= 0) {
509 if (match
.rm_so
> pmatch
->rm_so
)
511 if (match
.rm_so
== pmatch
->rm_so
&& match
.rm_eo
< pmatch
->rm_eo
)
514 pmatch
->rm_so
= match
.rm_so
;
515 pmatch
->rm_eo
= match
.rm_eo
;
519 static int next_match(struct grep_opt
*opt
, char *bol
, char *eol
,
520 enum grep_context ctx
, regmatch_t
*pmatch
, int eflags
)
525 pmatch
->rm_so
= pmatch
->rm_eo
= -1;
527 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
529 case GREP_PATTERN
: /* atom */
530 case GREP_PATTERN_HEAD
:
531 case GREP_PATTERN_BODY
:
532 hit
|= match_next_pattern(p
, bol
, eol
, ctx
,
543 static void show_line(struct grep_opt
*opt
, char *bol
, char *eol
,
544 const char *name
, unsigned lno
, char sign
)
546 int rest
= eol
- bol
;
550 if (opt
->pre_context
|| opt
->post_context
) {
551 if (opt
->last_shown
== 0) {
552 if (opt
->show_hunk_mark
)
553 opt
->output(opt
, "--\n", 3);
554 } else if (lno
> opt
->last_shown
+ 1)
555 opt
->output(opt
, "--\n", 3);
557 opt
->last_shown
= lno
;
559 if (opt
->null_following_name
)
562 opt
->output(opt
, name
, strlen(name
));
563 opt
->output(opt
, sign_str
, 1);
567 snprintf(buf
, sizeof(buf
), "%d", lno
);
568 opt
->output(opt
, buf
, strlen(buf
));
569 opt
->output(opt
, sign_str
, 1);
573 enum grep_context ctx
= GREP_CONTEXT_BODY
;
578 while (next_match(opt
, bol
, eol
, ctx
, &match
, eflags
)) {
579 if (match
.rm_so
== match
.rm_eo
)
582 opt
->output(opt
, bol
, match
.rm_so
);
583 opt
->output(opt
, opt
->color_match
,
584 strlen(opt
->color_match
));
585 opt
->output(opt
, bol
+ match
.rm_so
,
586 (int)(match
.rm_eo
- match
.rm_so
));
587 opt
->output(opt
, GIT_COLOR_RESET
,
588 strlen(GIT_COLOR_RESET
));
595 opt
->output(opt
, bol
, rest
);
596 opt
->output(opt
, "\n", 1);
599 static int match_funcname(struct grep_opt
*opt
, char *bol
, char *eol
)
601 xdemitconf_t
*xecfg
= opt
->priv
;
602 if (xecfg
&& xecfg
->find_func
) {
604 return xecfg
->find_func(bol
, eol
- bol
, buf
, 1,
605 xecfg
->find_func_priv
) >= 0;
610 if (isalpha(*bol
) || *bol
== '_' || *bol
== '$')
615 static void show_funcname_line(struct grep_opt
*opt
, const char *name
,
616 char *buf
, char *bol
, unsigned lno
)
621 while (bol
> buf
&& bol
[-1] != '\n')
625 if (lno
<= opt
->last_shown
)
628 if (match_funcname(opt
, bol
, eol
)) {
629 show_line(opt
, bol
, eol
, name
, lno
, '=');
635 static void show_pre_context(struct grep_opt
*opt
, const char *name
, char *buf
,
636 char *bol
, unsigned lno
)
638 unsigned cur
= lno
, from
= 1, funcname_lno
= 0;
639 int funcname_needed
= opt
->funcname
;
641 if (opt
->pre_context
< lno
)
642 from
= lno
- opt
->pre_context
;
643 if (from
<= opt
->last_shown
)
644 from
= opt
->last_shown
+ 1;
647 while (bol
> buf
&& cur
> from
) {
650 while (bol
> buf
&& bol
[-1] != '\n')
653 if (funcname_needed
&& match_funcname(opt
, bol
, eol
)) {
659 /* We need to look even further back to find a function signature. */
660 if (opt
->funcname
&& funcname_needed
)
661 show_funcname_line(opt
, name
, buf
, bol
, cur
);
665 char *eol
= bol
, sign
= (cur
== funcname_lno
) ? '=' : '-';
669 show_line(opt
, bol
, eol
, name
, cur
, sign
);
675 static int should_lookahead(struct grep_opt
*opt
)
680 return 0; /* punt for too complex stuff */
683 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
684 if (p
->token
!= GREP_PATTERN
)
685 return 0; /* punt for "header only" and stuff */
690 static int look_ahead(struct grep_opt
*opt
,
691 unsigned long *left_p
,
695 unsigned lno
= *lno_p
;
699 regoff_t earliest
= -1;
701 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
706 hit
= !fixmatch(p
->pattern
, bol
, p
->ignore_case
, &m
);
711 hit
= !regexec(&p
->regexp
, bol
, 1, &m
, REG_STARTEND
);
713 hit
= !regexec(&p
->regexp
, bol
, 1, &m
, 0);
716 if (!hit
|| m
.rm_so
< 0 || m
.rm_eo
< 0)
718 if (earliest
< 0 || m
.rm_so
< earliest
)
723 *bol_p
= bol
+ *left_p
;
727 for (sp
= bol
+ earliest
; bol
< sp
&& sp
[-1] != '\n'; sp
--)
728 ; /* find the beginning of the line */
731 for (sp
= bol
; sp
< last_bol
; sp
++) {
735 *left_p
-= last_bol
- bol
;
741 int grep_threads_ok(const struct grep_opt
*opt
)
743 /* If this condition is true, then we may use the attribute
744 * machinery in grep_buffer_1. The attribute code is not
745 * thread safe, so we disable the use of threads.
747 if (opt
->funcname
&& !opt
->unmatch_name_only
&& !opt
->status_only
&&
754 static void std_output(struct grep_opt
*opt
, const void *buf
, size_t size
)
756 fwrite(buf
, size
, 1, stdout
);
759 static int grep_buffer_1(struct grep_opt
*opt
, const char *name
,
760 char *buf
, unsigned long size
, int collect_hits
)
763 unsigned long left
= size
;
765 unsigned last_hit
= 0;
766 int binary_match_only
= 0;
768 int try_lookahead
= 0;
769 enum grep_context ctx
= GREP_CONTEXT_HEAD
;
773 opt
->output
= std_output
;
775 if (opt
->last_shown
&& (opt
->pre_context
|| opt
->post_context
) &&
776 opt
->output
== std_output
)
777 opt
->show_hunk_mark
= 1;
780 if (buffer_is_binary(buf
, size
)) {
781 switch (opt
->binary
) {
782 case GREP_BINARY_DEFAULT
:
783 binary_match_only
= 1;
785 case GREP_BINARY_NOMATCH
:
786 return 0; /* Assume unmatch */
793 memset(&xecfg
, 0, sizeof(xecfg
));
794 if (opt
->funcname
&& !opt
->unmatch_name_only
&& !opt
->status_only
&&
795 !opt
->name_only
&& !binary_match_only
&& !collect_hits
) {
796 struct userdiff_driver
*drv
= userdiff_find_by_path(name
);
797 if (drv
&& drv
->funcname
.pattern
) {
798 const struct userdiff_funcname
*pe
= &drv
->funcname
;
799 xdiff_set_find_func(&xecfg
, pe
->pattern
, pe
->cflags
);
803 try_lookahead
= should_lookahead(opt
);
810 * look_ahead() skips quicly to the line that possibly
811 * has the next hit; don't call it if we need to do
812 * something more than just skipping the current line
813 * in response to an unmatch for the current line. E.g.
814 * inside a post-context window, we will show the current
815 * line as a context around the previous hit when it
820 && lno
<= last_hit
+ opt
->post_context
)
821 && look_ahead(opt
, &left
, &lno
, &bol
))
823 eol
= end_of_line(bol
, &left
);
827 if ((ctx
== GREP_CONTEXT_HEAD
) && (eol
== bol
))
828 ctx
= GREP_CONTEXT_BODY
;
830 hit
= match_line(opt
, bol
, eol
, ctx
, collect_hits
);
836 /* "grep -v -e foo -e bla" should list lines
837 * that do not have either, so inversion should
842 if (opt
->unmatch_name_only
) {
849 if (opt
->status_only
)
851 if (binary_match_only
) {
852 opt
->output(opt
, "Binary file ", 12);
853 opt
->output(opt
, name
, strlen(name
));
854 opt
->output(opt
, " matches\n", 9);
857 if (opt
->name_only
) {
858 show_name(opt
, name
);
861 /* Hit at this line. If we haven't shown the
862 * pre-context lines, we would need to show them.
863 * When asked to do "count", this still show
864 * the context which is nonsense, but the user
865 * deserves to get that ;-).
867 if (opt
->pre_context
)
868 show_pre_context(opt
, name
, buf
, bol
, lno
);
869 else if (opt
->funcname
)
870 show_funcname_line(opt
, name
, buf
, bol
, lno
);
872 show_line(opt
, bol
, eol
, name
, lno
, ':');
876 lno
<= last_hit
+ opt
->post_context
) {
877 /* If the last hit is within the post context,
878 * we need to show this line.
880 show_line(opt
, bol
, eol
, name
, lno
, '-');
894 if (opt
->status_only
)
896 if (opt
->unmatch_name_only
) {
897 /* We did not see any hit, so we want to show this */
898 show_name(opt
, name
);
902 xdiff_clear_find_func(&xecfg
);
906 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
907 * which feels mostly useless but sometimes useful. Maybe
908 * make it another option? For now suppress them.
910 if (opt
->count
&& count
) {
912 opt
->output(opt
, name
, strlen(name
));
913 snprintf(buf
, sizeof(buf
), "%c%u\n",
914 opt
->null_following_name
? '\0' : ':', count
);
915 opt
->output(opt
, buf
, strlen(buf
));
920 static void clr_hit_marker(struct grep_expr
*x
)
922 /* All-hit markers are meaningful only at the very top level
927 if (x
->node
!= GREP_NODE_OR
)
929 x
->u
.binary
.left
->hit
= 0;
930 x
= x
->u
.binary
.right
;
934 static int chk_hit_marker(struct grep_expr
*x
)
936 /* Top level nodes have hit markers. See if they all are hits */
938 if (x
->node
!= GREP_NODE_OR
)
940 if (!x
->u
.binary
.left
->hit
)
942 x
= x
->u
.binary
.right
;
946 int grep_buffer(struct grep_opt
*opt
, const char *name
, char *buf
, unsigned long size
)
949 * we do not have to do the two-pass grep when we do not check
950 * buffer-wide "all-match".
953 return grep_buffer_1(opt
, name
, buf
, size
, 0);
955 /* Otherwise the toplevel "or" terms hit a bit differently.
956 * We first clear hit markers from them.
958 clr_hit_marker(opt
->pattern_expression
);
959 grep_buffer_1(opt
, name
, buf
, size
, 1);
961 if (!chk_hit_marker(opt
->pattern_expression
))
964 return grep_buffer_1(opt
, name
, buf
, size
, 0);