4 #include "object-store.h"
6 #include "xdiff-interface.h"
12 static int grep_source_load(struct grep_source
*gs
);
13 static int grep_source_is_binary(struct grep_source
*gs
);
15 static struct grep_opt grep_defaults
;
17 static void std_output(struct grep_opt
*opt
, const void *buf
, size_t size
)
19 fwrite(buf
, size
, 1, stdout
);
22 static void color_set(char *dst
, const char *color_bytes
)
24 xsnprintf(dst
, COLOR_MAXLEN
, "%s", color_bytes
);
28 * Initialize the grep_defaults template with hardcoded defaults.
29 * We could let the compiler do this, but without C99 initializers
30 * the code gets unwieldy and unreadable, so...
32 void init_grep_defaults(void)
34 struct grep_opt
*opt
= &grep_defaults
;
41 memset(opt
, 0, sizeof(*opt
));
45 opt
->pattern_type_option
= GREP_PATTERN_TYPE_UNSPECIFIED
;
46 color_set(opt
->color_context
, "");
47 color_set(opt
->color_filename
, "");
48 color_set(opt
->color_function
, "");
49 color_set(opt
->color_lineno
, "");
50 color_set(opt
->color_match_context
, GIT_COLOR_BOLD_RED
);
51 color_set(opt
->color_match_selected
, GIT_COLOR_BOLD_RED
);
52 color_set(opt
->color_selected
, "");
53 color_set(opt
->color_sep
, GIT_COLOR_CYAN
);
55 opt
->output
= std_output
;
58 static int parse_pattern_type_arg(const char *opt
, const char *arg
)
60 if (!strcmp(arg
, "default"))
61 return GREP_PATTERN_TYPE_UNSPECIFIED
;
62 else if (!strcmp(arg
, "basic"))
63 return GREP_PATTERN_TYPE_BRE
;
64 else if (!strcmp(arg
, "extended"))
65 return GREP_PATTERN_TYPE_ERE
;
66 else if (!strcmp(arg
, "fixed"))
67 return GREP_PATTERN_TYPE_FIXED
;
68 else if (!strcmp(arg
, "perl"))
69 return GREP_PATTERN_TYPE_PCRE
;
70 die("bad %s argument: %s", opt
, arg
);
74 * Read the configuration file once and store it in
75 * the grep_defaults template.
77 int grep_config(const char *var
, const char *value
, void *cb
)
79 struct grep_opt
*opt
= &grep_defaults
;
82 if (userdiff_config(var
, value
) < 0)
85 if (!strcmp(var
, "grep.extendedregexp")) {
86 opt
->extended_regexp_option
= git_config_bool(var
, value
);
90 if (!strcmp(var
, "grep.patterntype")) {
91 opt
->pattern_type_option
= parse_pattern_type_arg(var
, value
);
95 if (!strcmp(var
, "grep.linenumber")) {
96 opt
->linenum
= git_config_bool(var
, value
);
100 if (!strcmp(var
, "grep.fullname")) {
101 opt
->relative
= !git_config_bool(var
, value
);
105 if (!strcmp(var
, "color.grep"))
106 opt
->color
= git_config_colorbool(var
, value
);
107 else if (!strcmp(var
, "color.grep.context"))
108 color
= opt
->color_context
;
109 else if (!strcmp(var
, "color.grep.filename"))
110 color
= opt
->color_filename
;
111 else if (!strcmp(var
, "color.grep.function"))
112 color
= opt
->color_function
;
113 else if (!strcmp(var
, "color.grep.linenumber"))
114 color
= opt
->color_lineno
;
115 else if (!strcmp(var
, "color.grep.matchcontext"))
116 color
= opt
->color_match_context
;
117 else if (!strcmp(var
, "color.grep.matchselected"))
118 color
= opt
->color_match_selected
;
119 else if (!strcmp(var
, "color.grep.selected"))
120 color
= opt
->color_selected
;
121 else if (!strcmp(var
, "color.grep.separator"))
122 color
= opt
->color_sep
;
123 else if (!strcmp(var
, "color.grep.match")) {
126 return config_error_nonbool(var
);
127 rc
|= color_parse(value
, opt
->color_match_context
);
128 rc
|= color_parse(value
, opt
->color_match_selected
);
134 return config_error_nonbool(var
);
135 return color_parse(value
, color
);
141 * Initialize one instance of grep_opt and copy the
142 * default values from the template we read the configuration
143 * information in an earlier call to git_config(grep_config).
145 void grep_init(struct grep_opt
*opt
, const char *prefix
)
147 struct grep_opt
*def
= &grep_defaults
;
149 memset(opt
, 0, sizeof(*opt
));
150 opt
->prefix
= prefix
;
151 opt
->prefix_length
= (prefix
&& *prefix
) ? strlen(prefix
) : 0;
152 opt
->pattern_tail
= &opt
->pattern_list
;
153 opt
->header_tail
= &opt
->header_list
;
155 opt
->color
= def
->color
;
156 opt
->extended_regexp_option
= def
->extended_regexp_option
;
157 opt
->pattern_type_option
= def
->pattern_type_option
;
158 opt
->linenum
= def
->linenum
;
159 opt
->max_depth
= def
->max_depth
;
160 opt
->pathname
= def
->pathname
;
161 opt
->relative
= def
->relative
;
162 opt
->output
= def
->output
;
164 color_set(opt
->color_context
, def
->color_context
);
165 color_set(opt
->color_filename
, def
->color_filename
);
166 color_set(opt
->color_function
, def
->color_function
);
167 color_set(opt
->color_lineno
, def
->color_lineno
);
168 color_set(opt
->color_match_context
, def
->color_match_context
);
169 color_set(opt
->color_match_selected
, def
->color_match_selected
);
170 color_set(opt
->color_selected
, def
->color_selected
);
171 color_set(opt
->color_sep
, def
->color_sep
);
174 static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
177 * When committing to the pattern type by setting the relevant
178 * fields in grep_opt it's generally not necessary to zero out
179 * the fields we're not choosing, since they won't have been
180 * set by anything. The extended_regexp_option field is the
181 * only exception to this.
183 * This is because in the process of parsing grep.patternType
184 * & grep.extendedRegexp we set opt->pattern_type_option and
185 * opt->extended_regexp_option, respectively. We then
186 * internally use opt->extended_regexp_option to see if we're
187 * compiling an ERE. It must be unset if that's not actually
190 if (pattern_type
!= GREP_PATTERN_TYPE_ERE
&&
191 opt
->extended_regexp_option
)
192 opt
->extended_regexp_option
= 0;
194 switch (pattern_type
) {
195 case GREP_PATTERN_TYPE_UNSPECIFIED
:
198 case GREP_PATTERN_TYPE_BRE
:
201 case GREP_PATTERN_TYPE_ERE
:
202 opt
->extended_regexp_option
= 1;
205 case GREP_PATTERN_TYPE_FIXED
:
209 case GREP_PATTERN_TYPE_PCRE
:
214 * It's important that pcre1 always be assigned to
215 * even when there's no USE_LIBPCRE* defined. We still
216 * call the PCRE stub function, it just dies with
217 * "cannot use Perl-compatible regexes[...]".
225 void grep_commit_pattern_type(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
227 if (pattern_type
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
228 grep_set_pattern_type_option(pattern_type
, opt
);
229 else if (opt
->pattern_type_option
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
230 grep_set_pattern_type_option(opt
->pattern_type_option
, opt
);
231 else if (opt
->extended_regexp_option
)
233 * This branch *must* happen after setting from the
234 * opt->pattern_type_option above, we don't want
235 * grep.extendedRegexp to override grep.patternType!
237 grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE
, opt
);
240 static struct grep_pat
*create_grep_pat(const char *pat
, size_t patlen
,
241 const char *origin
, int no
,
242 enum grep_pat_token t
,
243 enum grep_header_field field
)
245 struct grep_pat
*p
= xcalloc(1, sizeof(*p
));
246 p
->pattern
= xmemdupz(pat
, patlen
);
247 p
->patternlen
= patlen
;
255 static void do_append_grep_pat(struct grep_pat
***tail
, struct grep_pat
*p
)
262 case GREP_PATTERN
: /* atom */
263 case GREP_PATTERN_HEAD
:
264 case GREP_PATTERN_BODY
:
266 struct grep_pat
*new_pat
;
268 char *cp
= p
->pattern
+ p
->patternlen
, *nl
= NULL
;
269 while (++len
<= p
->patternlen
) {
270 if (*(--cp
) == '\n') {
277 new_pat
= create_grep_pat(nl
+ 1, len
- 1, p
->origin
,
278 p
->no
, p
->token
, p
->field
);
279 new_pat
->next
= p
->next
;
281 *tail
= &new_pat
->next
;
284 p
->patternlen
-= len
;
292 void append_header_grep_pattern(struct grep_opt
*opt
,
293 enum grep_header_field field
, const char *pat
)
295 struct grep_pat
*p
= create_grep_pat(pat
, strlen(pat
), "header", 0,
296 GREP_PATTERN_HEAD
, field
);
297 if (field
== GREP_HEADER_REFLOG
)
298 opt
->use_reflog_filter
= 1;
299 do_append_grep_pat(&opt
->header_tail
, p
);
302 void append_grep_pattern(struct grep_opt
*opt
, const char *pat
,
303 const char *origin
, int no
, enum grep_pat_token t
)
305 append_grep_pat(opt
, pat
, strlen(pat
), origin
, no
, t
);
308 void append_grep_pat(struct grep_opt
*opt
, const char *pat
, size_t patlen
,
309 const char *origin
, int no
, enum grep_pat_token t
)
311 struct grep_pat
*p
= create_grep_pat(pat
, patlen
, origin
, no
, t
, 0);
312 do_append_grep_pat(&opt
->pattern_tail
, p
);
315 struct grep_opt
*grep_opt_dup(const struct grep_opt
*opt
)
317 struct grep_pat
*pat
;
318 struct grep_opt
*ret
= xmalloc(sizeof(struct grep_opt
));
321 ret
->pattern_list
= NULL
;
322 ret
->pattern_tail
= &ret
->pattern_list
;
324 for(pat
= opt
->pattern_list
; pat
!= NULL
; pat
= pat
->next
)
326 if(pat
->token
== GREP_PATTERN_HEAD
)
327 append_header_grep_pattern(ret
, pat
->field
,
330 append_grep_pat(ret
, pat
->pattern
, pat
->patternlen
,
331 pat
->origin
, pat
->no
, pat
->token
);
337 static NORETURN
void compile_regexp_failed(const struct grep_pat
*p
,
343 xsnprintf(where
, sizeof(where
), "In '%s' at %d, ", p
->origin
, p
->no
);
345 xsnprintf(where
, sizeof(where
), "%s, ", p
->origin
);
349 die("%s'%s': %s", where
, p
->pattern
, error
);
352 static int is_fixed(const char *s
, size_t len
)
356 for (i
= 0; i
< len
; i
++) {
357 if (is_regex_special(s
[i
]))
364 static int has_null(const char *s
, size_t len
)
367 * regcomp cannot accept patterns with NULs so when using it
368 * we consider any pattern containing a NUL fixed.
370 if (memchr(s
, 0, len
))
377 static void compile_pcre1_regexp(struct grep_pat
*p
, const struct grep_opt
*opt
)
381 int options
= PCRE_MULTILINE
;
383 if (opt
->ignore_case
) {
384 if (has_non_ascii(p
->pattern
))
385 p
->pcre1_tables
= pcre_maketables();
386 options
|= PCRE_CASELESS
;
388 if (is_utf8_locale() && has_non_ascii(p
->pattern
))
389 options
|= PCRE_UTF8
;
391 p
->pcre1_regexp
= pcre_compile(p
->pattern
, options
, &error
, &erroffset
,
393 if (!p
->pcre1_regexp
)
394 compile_regexp_failed(p
, error
);
396 p
->pcre1_extra_info
= pcre_study(p
->pcre1_regexp
, GIT_PCRE_STUDY_JIT_COMPILE
, &error
);
397 if (!p
->pcre1_extra_info
&& error
)
400 #ifdef GIT_PCRE1_USE_JIT
401 pcre_config(PCRE_CONFIG_JIT
, &p
->pcre1_jit_on
);
402 if (p
->pcre1_jit_on
== 1) {
403 p
->pcre1_jit_stack
= pcre_jit_stack_alloc(1, 1024 * 1024);
404 if (!p
->pcre1_jit_stack
)
405 die("Couldn't allocate PCRE JIT stack");
406 pcre_assign_jit_stack(p
->pcre1_extra_info
, NULL
, p
->pcre1_jit_stack
);
407 } else if (p
->pcre1_jit_on
!= 0) {
408 die("BUG: The pcre1_jit_on variable should be 0 or 1, not %d",
414 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
415 regmatch_t
*match
, int eflags
)
417 int ovector
[30], ret
, flags
= 0;
419 if (eflags
& REG_NOTBOL
)
420 flags
|= PCRE_NOTBOL
;
422 #ifdef GIT_PCRE1_USE_JIT
423 if (p
->pcre1_jit_on
) {
424 ret
= pcre_jit_exec(p
->pcre1_regexp
, p
->pcre1_extra_info
, line
,
425 eol
- line
, 0, flags
, ovector
,
426 ARRAY_SIZE(ovector
), p
->pcre1_jit_stack
);
430 ret
= pcre_exec(p
->pcre1_regexp
, p
->pcre1_extra_info
, line
,
431 eol
- line
, 0, flags
, ovector
,
432 ARRAY_SIZE(ovector
));
435 if (ret
< 0 && ret
!= PCRE_ERROR_NOMATCH
)
436 die("pcre_exec failed with error code %d", ret
);
439 match
->rm_so
= ovector
[0];
440 match
->rm_eo
= ovector
[1];
446 static void free_pcre1_regexp(struct grep_pat
*p
)
448 pcre_free(p
->pcre1_regexp
);
449 #ifdef GIT_PCRE1_USE_JIT
450 if (p
->pcre1_jit_on
) {
451 pcre_free_study(p
->pcre1_extra_info
);
452 pcre_jit_stack_free(p
->pcre1_jit_stack
);
456 pcre_free(p
->pcre1_extra_info
);
458 pcre_free((void *)p
->pcre1_tables
);
460 #else /* !USE_LIBPCRE1 */
461 static void compile_pcre1_regexp(struct grep_pat
*p
, const struct grep_opt
*opt
)
463 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
466 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
467 regmatch_t
*match
, int eflags
)
472 static void free_pcre1_regexp(struct grep_pat
*p
)
475 #endif /* !USE_LIBPCRE1 */
478 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
481 PCRE2_UCHAR errbuf
[256];
482 PCRE2_SIZE erroffset
;
483 int options
= PCRE2_MULTILINE
;
484 const uint8_t *character_tables
= NULL
;
491 p
->pcre2_compile_context
= NULL
;
493 if (opt
->ignore_case
) {
494 if (has_non_ascii(p
->pattern
)) {
495 character_tables
= pcre2_maketables(NULL
);
496 p
->pcre2_compile_context
= pcre2_compile_context_create(NULL
);
497 pcre2_set_character_tables(p
->pcre2_compile_context
, character_tables
);
499 options
|= PCRE2_CASELESS
;
501 if (is_utf8_locale() && has_non_ascii(p
->pattern
))
502 options
|= PCRE2_UTF
;
504 p
->pcre2_pattern
= pcre2_compile((PCRE2_SPTR
)p
->pattern
,
505 p
->patternlen
, options
, &error
, &erroffset
,
506 p
->pcre2_compile_context
);
508 if (p
->pcre2_pattern
) {
509 p
->pcre2_match_data
= pcre2_match_data_create_from_pattern(p
->pcre2_pattern
, NULL
);
510 if (!p
->pcre2_match_data
)
511 die("Couldn't allocate PCRE2 match data");
513 pcre2_get_error_message(error
, errbuf
, sizeof(errbuf
));
514 compile_regexp_failed(p
, (const char *)&errbuf
);
517 pcre2_config(PCRE2_CONFIG_JIT
, &p
->pcre2_jit_on
);
518 if (p
->pcre2_jit_on
== 1) {
519 jitret
= pcre2_jit_compile(p
->pcre2_pattern
, PCRE2_JIT_COMPLETE
);
521 die("Couldn't JIT the PCRE2 pattern '%s', got '%d'\n", p
->pattern
, jitret
);
524 * The pcre2_config(PCRE2_CONFIG_JIT, ...) call just
525 * tells us whether the library itself supports JIT,
526 * but to see whether we're going to be actually using
527 * JIT we need to extract PCRE2_INFO_JITSIZE from the
528 * pattern *after* we do pcre2_jit_compile() above.
530 * This is because if the pattern contains the
531 * (*NO_JIT) verb (see pcre2syntax(3))
532 * pcre2_jit_compile() will exit early with 0. If we
533 * then proceed to call pcre2_jit_match() further down
534 * the line instead of pcre2_match() we'll either
535 * segfault (pre PCRE 10.31) or run into a fatal error
538 patinforet
= pcre2_pattern_info(p
->pcre2_pattern
, PCRE2_INFO_JITSIZE
, &jitsizearg
);
540 BUG("pcre2_pattern_info() failed: %d", patinforet
);
541 if (jitsizearg
== 0) {
546 p
->pcre2_jit_stack
= pcre2_jit_stack_create(1, 1024 * 1024, NULL
);
547 if (!p
->pcre2_jit_stack
)
548 die("Couldn't allocate PCRE2 JIT stack");
549 p
->pcre2_match_context
= pcre2_match_context_create(NULL
);
550 if (!p
->pcre2_match_context
)
551 die("Couldn't allocate PCRE2 match context");
552 pcre2_jit_stack_assign(p
->pcre2_match_context
, NULL
, p
->pcre2_jit_stack
);
553 } else if (p
->pcre2_jit_on
!= 0) {
554 die("BUG: The pcre2_jit_on variable should be 0 or 1, not %d",
559 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
560 regmatch_t
*match
, int eflags
)
564 PCRE2_UCHAR errbuf
[256];
566 if (eflags
& REG_NOTBOL
)
567 flags
|= PCRE2_NOTBOL
;
570 ret
= pcre2_jit_match(p
->pcre2_pattern
, (unsigned char *)line
,
571 eol
- line
, 0, flags
, p
->pcre2_match_data
,
574 ret
= pcre2_match(p
->pcre2_pattern
, (unsigned char *)line
,
575 eol
- line
, 0, flags
, p
->pcre2_match_data
,
578 if (ret
< 0 && ret
!= PCRE2_ERROR_NOMATCH
) {
579 pcre2_get_error_message(ret
, errbuf
, sizeof(errbuf
));
580 die("%s failed with error code %d: %s",
581 (p
->pcre2_jit_on
? "pcre2_jit_match" : "pcre2_match"), ret
,
585 ovector
= pcre2_get_ovector_pointer(p
->pcre2_match_data
);
587 match
->rm_so
= (int)ovector
[0];
588 match
->rm_eo
= (int)ovector
[1];
594 static void free_pcre2_pattern(struct grep_pat
*p
)
596 pcre2_compile_context_free(p
->pcre2_compile_context
);
597 pcre2_code_free(p
->pcre2_pattern
);
598 pcre2_match_data_free(p
->pcre2_match_data
);
599 pcre2_jit_stack_free(p
->pcre2_jit_stack
);
600 pcre2_match_context_free(p
->pcre2_match_context
);
602 #else /* !USE_LIBPCRE2 */
603 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
606 * Unreachable until USE_LIBPCRE2 becomes synonymous with
607 * USE_LIBPCRE. See the sibling comment in
608 * grep_set_pattern_type_option().
610 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
613 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
614 regmatch_t
*match
, int eflags
)
619 static void free_pcre2_pattern(struct grep_pat
*p
)
622 #endif /* !USE_LIBPCRE2 */
624 static void compile_fixed_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
626 struct strbuf sb
= STRBUF_INIT
;
630 basic_regex_quote_buf(&sb
, p
->pattern
);
631 if (opt
->ignore_case
)
632 regflags
|= REG_ICASE
;
633 err
= regcomp(&p
->regexp
, sb
.buf
, regflags
);
635 fprintf(stderr
, "fixed %s\n", sb
.buf
);
639 regerror(err
, &p
->regexp
, errbuf
, sizeof(errbuf
));
641 compile_regexp_failed(p
, errbuf
);
645 static void compile_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
649 int regflags
= REG_NEWLINE
;
651 p
->word_regexp
= opt
->word_regexp
;
652 p
->ignore_case
= opt
->ignore_case
;
653 ascii_only
= !has_non_ascii(p
->pattern
);
656 * Even when -F (fixed) asks us to do a non-regexp search, we
657 * may not be able to correctly case-fold when -i
658 * (ignore-case) is asked (in which case, we'll synthesize a
659 * regexp to match the pattern that matches regexp special
660 * characters literally, while ignoring case differences). On
661 * the other hand, even without -F, if the pattern does not
662 * have any regexp special characters and there is no need for
663 * case-folding search, we can internally turn it into a
664 * simple string match using kws. p->fixed tells us if we
668 has_null(p
->pattern
, p
->patternlen
) ||
669 is_fixed(p
->pattern
, p
->patternlen
))
670 p
->fixed
= !p
->ignore_case
|| ascii_only
;
673 p
->kws
= kwsalloc(p
->ignore_case
? tolower_trans_tbl
: NULL
);
674 kwsincr(p
->kws
, p
->pattern
, p
->patternlen
);
677 } else if (opt
->fixed
) {
679 * We come here when the pattern has the non-ascii
680 * characters we cannot case-fold, and asked to
683 compile_fixed_regexp(p
, opt
);
688 compile_pcre2_pattern(p
, opt
);
693 compile_pcre1_regexp(p
, opt
);
698 regflags
|= REG_ICASE
;
699 if (opt
->extended_regexp_option
)
700 regflags
|= REG_EXTENDED
;
701 err
= regcomp(&p
->regexp
, p
->pattern
, regflags
);
704 regerror(err
, &p
->regexp
, errbuf
, 1024);
706 compile_regexp_failed(p
, errbuf
);
710 static struct grep_expr
*compile_pattern_or(struct grep_pat
**);
711 static struct grep_expr
*compile_pattern_atom(struct grep_pat
**list
)
720 case GREP_PATTERN
: /* atom */
721 case GREP_PATTERN_HEAD
:
722 case GREP_PATTERN_BODY
:
723 x
= xcalloc(1, sizeof (struct grep_expr
));
724 x
->node
= GREP_NODE_ATOM
;
728 case GREP_OPEN_PAREN
:
730 x
= compile_pattern_or(list
);
731 if (!*list
|| (*list
)->token
!= GREP_CLOSE_PAREN
)
732 die("unmatched parenthesis");
733 *list
= (*list
)->next
;
740 static struct grep_expr
*compile_pattern_not(struct grep_pat
**list
)
751 die("--not not followed by pattern expression");
753 x
= xcalloc(1, sizeof (struct grep_expr
));
754 x
->node
= GREP_NODE_NOT
;
755 x
->u
.unary
= compile_pattern_not(list
);
757 die("--not followed by non pattern expression");
760 return compile_pattern_atom(list
);
764 static struct grep_expr
*compile_pattern_and(struct grep_pat
**list
)
767 struct grep_expr
*x
, *y
, *z
;
769 x
= compile_pattern_not(list
);
771 if (p
&& p
->token
== GREP_AND
) {
773 die("--and not followed by pattern expression");
775 y
= compile_pattern_and(list
);
777 die("--and not followed by pattern expression");
778 z
= xcalloc(1, sizeof (struct grep_expr
));
779 z
->node
= GREP_NODE_AND
;
780 z
->u
.binary
.left
= x
;
781 z
->u
.binary
.right
= y
;
787 static struct grep_expr
*compile_pattern_or(struct grep_pat
**list
)
790 struct grep_expr
*x
, *y
, *z
;
792 x
= compile_pattern_and(list
);
794 if (x
&& p
&& p
->token
!= GREP_CLOSE_PAREN
) {
795 y
= compile_pattern_or(list
);
797 die("not a pattern expression %s", p
->pattern
);
798 z
= xcalloc(1, sizeof (struct grep_expr
));
799 z
->node
= GREP_NODE_OR
;
800 z
->u
.binary
.left
= x
;
801 z
->u
.binary
.right
= y
;
807 static struct grep_expr
*compile_pattern_expr(struct grep_pat
**list
)
809 return compile_pattern_or(list
);
812 static void indent(int in
)
818 static void dump_grep_pat(struct grep_pat
*p
)
821 case GREP_AND
: fprintf(stderr
, "*and*"); break;
822 case GREP_OPEN_PAREN
: fprintf(stderr
, "*(*"); break;
823 case GREP_CLOSE_PAREN
: fprintf(stderr
, "*)*"); break;
824 case GREP_NOT
: fprintf(stderr
, "*not*"); break;
825 case GREP_OR
: fprintf(stderr
, "*or*"); break;
827 case GREP_PATTERN
: fprintf(stderr
, "pattern"); break;
828 case GREP_PATTERN_HEAD
: fprintf(stderr
, "pattern_head"); break;
829 case GREP_PATTERN_BODY
: fprintf(stderr
, "pattern_body"); break;
834 case GREP_PATTERN_HEAD
:
835 fprintf(stderr
, "<head %d>", p
->field
); break;
836 case GREP_PATTERN_BODY
:
837 fprintf(stderr
, "<body>"); break;
841 case GREP_PATTERN_HEAD
:
842 case GREP_PATTERN_BODY
:
844 fprintf(stderr
, "%.*s", (int)p
->patternlen
, p
->pattern
);
850 static void dump_grep_expression_1(struct grep_expr
*x
, int in
)
855 fprintf(stderr
, "true\n");
858 dump_grep_pat(x
->u
.atom
);
861 fprintf(stderr
, "(not\n");
862 dump_grep_expression_1(x
->u
.unary
, in
+1);
864 fprintf(stderr
, ")\n");
867 fprintf(stderr
, "(and\n");
868 dump_grep_expression_1(x
->u
.binary
.left
, in
+1);
869 dump_grep_expression_1(x
->u
.binary
.right
, in
+1);
871 fprintf(stderr
, ")\n");
874 fprintf(stderr
, "(or\n");
875 dump_grep_expression_1(x
->u
.binary
.left
, in
+1);
876 dump_grep_expression_1(x
->u
.binary
.right
, in
+1);
878 fprintf(stderr
, ")\n");
883 static void dump_grep_expression(struct grep_opt
*opt
)
885 struct grep_expr
*x
= opt
->pattern_expression
;
888 fprintf(stderr
, "[all-match]\n");
889 dump_grep_expression_1(x
, 0);
893 static struct grep_expr
*grep_true_expr(void)
895 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
896 z
->node
= GREP_NODE_TRUE
;
900 static struct grep_expr
*grep_or_expr(struct grep_expr
*left
, struct grep_expr
*right
)
902 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
903 z
->node
= GREP_NODE_OR
;
904 z
->u
.binary
.left
= left
;
905 z
->u
.binary
.right
= right
;
909 static struct grep_expr
*prep_header_patterns(struct grep_opt
*opt
)
912 struct grep_expr
*header_expr
;
913 struct grep_expr
*(header_group
[GREP_HEADER_FIELD_MAX
]);
914 enum grep_header_field fld
;
916 if (!opt
->header_list
)
919 for (p
= opt
->header_list
; p
; p
= p
->next
) {
920 if (p
->token
!= GREP_PATTERN_HEAD
)
921 die("BUG: a non-header pattern in grep header list.");
922 if (p
->field
< GREP_HEADER_FIELD_MIN
||
923 GREP_HEADER_FIELD_MAX
<= p
->field
)
924 die("BUG: unknown header field %d", p
->field
);
925 compile_regexp(p
, opt
);
928 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++)
929 header_group
[fld
] = NULL
;
931 for (p
= opt
->header_list
; p
; p
= p
->next
) {
933 struct grep_pat
*pp
= p
;
935 h
= compile_pattern_atom(&pp
);
936 if (!h
|| pp
!= p
->next
)
937 die("BUG: malformed header expr");
938 if (!header_group
[p
->field
]) {
939 header_group
[p
->field
] = h
;
942 header_group
[p
->field
] = grep_or_expr(h
, header_group
[p
->field
]);
947 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++) {
948 if (!header_group
[fld
])
951 header_expr
= grep_true_expr();
952 header_expr
= grep_or_expr(header_group
[fld
], header_expr
);
957 static struct grep_expr
*grep_splice_or(struct grep_expr
*x
, struct grep_expr
*y
)
959 struct grep_expr
*z
= x
;
962 assert(x
->node
== GREP_NODE_OR
);
963 if (x
->u
.binary
.right
&&
964 x
->u
.binary
.right
->node
== GREP_NODE_TRUE
) {
965 x
->u
.binary
.right
= y
;
968 x
= x
->u
.binary
.right
;
973 static void compile_grep_patterns_real(struct grep_opt
*opt
)
976 struct grep_expr
*header_expr
= prep_header_patterns(opt
);
978 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
980 case GREP_PATTERN
: /* atom */
981 case GREP_PATTERN_HEAD
:
982 case GREP_PATTERN_BODY
:
983 compile_regexp(p
, opt
);
991 if (opt
->all_match
|| header_expr
)
993 else if (!opt
->extended
&& !opt
->debug
)
996 p
= opt
->pattern_list
;
998 opt
->pattern_expression
= compile_pattern_expr(&p
);
1000 die("incomplete pattern expression: %s", p
->pattern
);
1005 if (!opt
->pattern_expression
)
1006 opt
->pattern_expression
= header_expr
;
1007 else if (opt
->all_match
)
1008 opt
->pattern_expression
= grep_splice_or(header_expr
,
1009 opt
->pattern_expression
);
1011 opt
->pattern_expression
= grep_or_expr(opt
->pattern_expression
,
1016 void compile_grep_patterns(struct grep_opt
*opt
)
1018 compile_grep_patterns_real(opt
);
1020 dump_grep_expression(opt
);
1023 static void free_pattern_expr(struct grep_expr
*x
)
1026 case GREP_NODE_TRUE
:
1027 case GREP_NODE_ATOM
:
1030 free_pattern_expr(x
->u
.unary
);
1034 free_pattern_expr(x
->u
.binary
.left
);
1035 free_pattern_expr(x
->u
.binary
.right
);
1041 void free_grep_patterns(struct grep_opt
*opt
)
1043 struct grep_pat
*p
, *n
;
1045 for (p
= opt
->pattern_list
; p
; p
= n
) {
1048 case GREP_PATTERN
: /* atom */
1049 case GREP_PATTERN_HEAD
:
1050 case GREP_PATTERN_BODY
:
1053 else if (p
->pcre1_regexp
)
1054 free_pcre1_regexp(p
);
1055 else if (p
->pcre2_pattern
)
1056 free_pcre2_pattern(p
);
1058 regfree(&p
->regexp
);
1069 free_pattern_expr(opt
->pattern_expression
);
1072 static char *end_of_line(char *cp
, unsigned long *left
)
1074 unsigned long l
= *left
;
1075 while (l
&& *cp
!= '\n') {
1083 static int word_char(char ch
)
1085 return isalnum(ch
) || ch
== '_';
1088 static void output_color(struct grep_opt
*opt
, const void *data
, size_t size
,
1091 if (want_color(opt
->color
) && color
&& color
[0]) {
1092 opt
->output(opt
, color
, strlen(color
));
1093 opt
->output(opt
, data
, size
);
1094 opt
->output(opt
, GIT_COLOR_RESET
, strlen(GIT_COLOR_RESET
));
1096 opt
->output(opt
, data
, size
);
1099 static void output_sep(struct grep_opt
*opt
, char sign
)
1101 if (opt
->null_following_name
)
1102 opt
->output(opt
, "\0", 1);
1104 output_color(opt
, &sign
, 1, opt
->color_sep
);
1107 static void show_name(struct grep_opt
*opt
, const char *name
)
1109 output_color(opt
, name
, strlen(name
), opt
->color_filename
);
1110 opt
->output(opt
, opt
->null_following_name
? "\0" : "\n", 1);
1113 static int fixmatch(struct grep_pat
*p
, char *line
, char *eol
,
1116 struct kwsmatch kwsm
;
1117 size_t offset
= kwsexec(p
->kws
, line
, eol
- line
, &kwsm
);
1119 match
->rm_so
= match
->rm_eo
= -1;
1122 match
->rm_so
= offset
;
1123 match
->rm_eo
= match
->rm_so
+ kwsm
.size
[0];
1128 static int patmatch(struct grep_pat
*p
, char *line
, char *eol
,
1129 regmatch_t
*match
, int eflags
)
1134 hit
= !fixmatch(p
, line
, eol
, match
);
1135 else if (p
->pcre1_regexp
)
1136 hit
= !pcre1match(p
, line
, eol
, match
, eflags
);
1137 else if (p
->pcre2_pattern
)
1138 hit
= !pcre2match(p
, line
, eol
, match
, eflags
);
1140 hit
= !regexec_buf(&p
->regexp
, line
, eol
- line
, 1, match
,
1146 static int strip_timestamp(char *bol
, char **eol_p
)
1151 while (bol
< --eol
) {
1165 } header_field
[] = {
1167 { "committer ", 10 },
1171 static int match_one_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1172 enum grep_context ctx
,
1173 regmatch_t
*pmatch
, int eflags
)
1177 const char *start
= bol
;
1179 if ((p
->token
!= GREP_PATTERN
) &&
1180 ((p
->token
== GREP_PATTERN_HEAD
) != (ctx
== GREP_CONTEXT_HEAD
)))
1183 if (p
->token
== GREP_PATTERN_HEAD
) {
1186 assert(p
->field
< ARRAY_SIZE(header_field
));
1187 field
= header_field
[p
->field
].field
;
1188 len
= header_field
[p
->field
].len
;
1189 if (strncmp(bol
, field
, len
))
1193 case GREP_HEADER_AUTHOR
:
1194 case GREP_HEADER_COMMITTER
:
1195 saved_ch
= strip_timestamp(bol
, &eol
);
1203 hit
= patmatch(p
, bol
, eol
, pmatch
, eflags
);
1205 if (hit
&& p
->word_regexp
) {
1206 if ((pmatch
[0].rm_so
< 0) ||
1207 (eol
- bol
) < pmatch
[0].rm_so
||
1208 (pmatch
[0].rm_eo
< 0) ||
1209 (eol
- bol
) < pmatch
[0].rm_eo
)
1210 die("regexp returned nonsense");
1212 /* Match beginning must be either beginning of the
1213 * line, or at word boundary (i.e. the last char must
1214 * not be a word char). Similarly, match end must be
1215 * either end of the line, or at word boundary
1216 * (i.e. the next char must not be a word char).
1218 if ( ((pmatch
[0].rm_so
== 0) ||
1219 !word_char(bol
[pmatch
[0].rm_so
-1])) &&
1220 ((pmatch
[0].rm_eo
== (eol
-bol
)) ||
1221 !word_char(bol
[pmatch
[0].rm_eo
])) )
1226 /* Words consist of at least one character. */
1227 if (pmatch
->rm_so
== pmatch
->rm_eo
)
1230 if (!hit
&& pmatch
[0].rm_so
+ bol
+ 1 < eol
) {
1231 /* There could be more than one match on the
1232 * line, and the first match might not be
1233 * strict word match. But later ones could be!
1234 * Forward to the next possible start, i.e. the
1235 * next position following a non-word char.
1237 bol
= pmatch
[0].rm_so
+ bol
+ 1;
1238 while (word_char(bol
[-1]) && bol
< eol
)
1240 eflags
|= REG_NOTBOL
;
1245 if (p
->token
== GREP_PATTERN_HEAD
&& saved_ch
)
1248 pmatch
[0].rm_so
+= bol
- start
;
1249 pmatch
[0].rm_eo
+= bol
- start
;
1254 static int match_expr_eval(struct grep_expr
*x
, char *bol
, char *eol
,
1255 enum grep_context ctx
, int collect_hits
)
1261 die("Not a valid grep expression");
1263 case GREP_NODE_TRUE
:
1266 case GREP_NODE_ATOM
:
1267 h
= match_one_pattern(x
->u
.atom
, bol
, eol
, ctx
, &match
, 0);
1270 h
= !match_expr_eval(x
->u
.unary
, bol
, eol
, ctx
, 0);
1273 if (!match_expr_eval(x
->u
.binary
.left
, bol
, eol
, ctx
, 0))
1275 h
= match_expr_eval(x
->u
.binary
.right
, bol
, eol
, ctx
, 0);
1279 return (match_expr_eval(x
->u
.binary
.left
,
1280 bol
, eol
, ctx
, 0) ||
1281 match_expr_eval(x
->u
.binary
.right
,
1283 h
= match_expr_eval(x
->u
.binary
.left
, bol
, eol
, ctx
, 0);
1284 x
->u
.binary
.left
->hit
|= h
;
1285 h
|= match_expr_eval(x
->u
.binary
.right
, bol
, eol
, ctx
, 1);
1288 die("Unexpected node type (internal error) %d", x
->node
);
1295 static int match_expr(struct grep_opt
*opt
, char *bol
, char *eol
,
1296 enum grep_context ctx
, int collect_hits
)
1298 struct grep_expr
*x
= opt
->pattern_expression
;
1299 return match_expr_eval(x
, bol
, eol
, ctx
, collect_hits
);
1302 static int match_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1303 enum grep_context ctx
, int collect_hits
)
1309 return match_expr(opt
, bol
, eol
, ctx
, collect_hits
);
1311 /* we do not call with collect_hits without being extended */
1312 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1313 if (match_one_pattern(p
, bol
, eol
, ctx
, &match
, 0))
1319 static int match_next_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1320 enum grep_context ctx
,
1321 regmatch_t
*pmatch
, int eflags
)
1325 if (!match_one_pattern(p
, bol
, eol
, ctx
, &match
, eflags
))
1327 if (match
.rm_so
< 0 || match
.rm_eo
< 0)
1329 if (pmatch
->rm_so
>= 0 && pmatch
->rm_eo
>= 0) {
1330 if (match
.rm_so
> pmatch
->rm_so
)
1332 if (match
.rm_so
== pmatch
->rm_so
&& match
.rm_eo
< pmatch
->rm_eo
)
1335 pmatch
->rm_so
= match
.rm_so
;
1336 pmatch
->rm_eo
= match
.rm_eo
;
1340 static int next_match(struct grep_opt
*opt
, char *bol
, char *eol
,
1341 enum grep_context ctx
, regmatch_t
*pmatch
, int eflags
)
1346 pmatch
->rm_so
= pmatch
->rm_eo
= -1;
1348 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1350 case GREP_PATTERN
: /* atom */
1351 case GREP_PATTERN_HEAD
:
1352 case GREP_PATTERN_BODY
:
1353 hit
|= match_next_pattern(p
, bol
, eol
, ctx
,
1364 static void show_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1365 const char *name
, unsigned lno
, char sign
)
1367 int rest
= eol
- bol
;
1368 const char *match_color
, *line_color
= NULL
;
1370 if (opt
->file_break
&& opt
->last_shown
== 0) {
1371 if (opt
->show_hunk_mark
)
1372 opt
->output(opt
, "\n", 1);
1373 } else if (opt
->pre_context
|| opt
->post_context
|| opt
->funcbody
) {
1374 if (opt
->last_shown
== 0) {
1375 if (opt
->show_hunk_mark
) {
1376 output_color(opt
, "--", 2, opt
->color_sep
);
1377 opt
->output(opt
, "\n", 1);
1379 } else if (lno
> opt
->last_shown
+ 1) {
1380 output_color(opt
, "--", 2, opt
->color_sep
);
1381 opt
->output(opt
, "\n", 1);
1384 if (opt
->heading
&& opt
->last_shown
== 0) {
1385 output_color(opt
, name
, strlen(name
), opt
->color_filename
);
1386 opt
->output(opt
, "\n", 1);
1388 opt
->last_shown
= lno
;
1390 if (!opt
->heading
&& opt
->pathname
) {
1391 output_color(opt
, name
, strlen(name
), opt
->color_filename
);
1392 output_sep(opt
, sign
);
1396 xsnprintf(buf
, sizeof(buf
), "%d", lno
);
1397 output_color(opt
, buf
, strlen(buf
), opt
->color_lineno
);
1398 output_sep(opt
, sign
);
1402 enum grep_context ctx
= GREP_CONTEXT_BODY
;
1407 match_color
= opt
->color_match_selected
;
1409 match_color
= opt
->color_match_context
;
1411 line_color
= opt
->color_selected
;
1412 else if (sign
== '-')
1413 line_color
= opt
->color_context
;
1414 else if (sign
== '=')
1415 line_color
= opt
->color_function
;
1417 while (next_match(opt
, bol
, eol
, ctx
, &match
, eflags
)) {
1418 if (match
.rm_so
== match
.rm_eo
)
1421 output_color(opt
, bol
, match
.rm_so
, line_color
);
1422 output_color(opt
, bol
+ match
.rm_so
,
1423 match
.rm_eo
- match
.rm_so
, match_color
);
1425 rest
-= match
.rm_eo
;
1426 eflags
= REG_NOTBOL
;
1430 output_color(opt
, bol
, rest
, line_color
);
1431 opt
->output(opt
, "\n", 1);
1438 * This lock protects access to the gitattributes machinery, which is
1441 pthread_mutex_t grep_attr_mutex
;
1443 static inline void grep_attr_lock(void)
1446 pthread_mutex_lock(&grep_attr_mutex
);
1449 static inline void grep_attr_unlock(void)
1452 pthread_mutex_unlock(&grep_attr_mutex
);
1456 * Same as git_attr_mutex, but protecting the thread-unsafe object db access.
1458 pthread_mutex_t grep_read_mutex
;
1461 #define grep_attr_lock()
1462 #define grep_attr_unlock()
1465 static int match_funcname(struct grep_opt
*opt
, struct grep_source
*gs
, char *bol
, char *eol
)
1467 xdemitconf_t
*xecfg
= opt
->priv
;
1468 if (xecfg
&& !xecfg
->find_func
) {
1469 grep_source_load_driver(gs
);
1470 if (gs
->driver
->funcname
.pattern
) {
1471 const struct userdiff_funcname
*pe
= &gs
->driver
->funcname
;
1472 xdiff_set_find_func(xecfg
, pe
->pattern
, pe
->cflags
);
1474 xecfg
= opt
->priv
= NULL
;
1480 return xecfg
->find_func(bol
, eol
- bol
, buf
, 1,
1481 xecfg
->find_func_priv
) >= 0;
1486 if (isalpha(*bol
) || *bol
== '_' || *bol
== '$')
1491 static void show_funcname_line(struct grep_opt
*opt
, struct grep_source
*gs
,
1492 char *bol
, unsigned lno
)
1494 while (bol
> gs
->buf
) {
1497 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1501 if (lno
<= opt
->last_shown
)
1504 if (match_funcname(opt
, gs
, bol
, eol
)) {
1505 show_line(opt
, bol
, eol
, gs
->name
, lno
, '=');
1511 static int is_empty_line(const char *bol
, const char *eol
);
1513 static void show_pre_context(struct grep_opt
*opt
, struct grep_source
*gs
,
1514 char *bol
, char *end
, unsigned lno
)
1516 unsigned cur
= lno
, from
= 1, funcname_lno
= 0, orig_from
;
1517 int funcname_needed
= !!opt
->funcname
, comment_needed
= 0;
1519 if (opt
->pre_context
< lno
)
1520 from
= lno
- opt
->pre_context
;
1521 if (from
<= opt
->last_shown
)
1522 from
= opt
->last_shown
+ 1;
1524 if (opt
->funcbody
) {
1525 if (match_funcname(opt
, gs
, bol
, end
))
1528 funcname_needed
= 1;
1529 from
= opt
->last_shown
+ 1;
1533 while (bol
> gs
->buf
&& cur
> from
) {
1534 char *next_bol
= bol
;
1537 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1540 if (comment_needed
&& (is_empty_line(bol
, eol
) ||
1541 match_funcname(opt
, gs
, bol
, eol
))) {
1550 if (funcname_needed
&& match_funcname(opt
, gs
, bol
, eol
)) {
1552 funcname_needed
= 0;
1560 /* We need to look even further back to find a function signature. */
1561 if (opt
->funcname
&& funcname_needed
)
1562 show_funcname_line(opt
, gs
, bol
, cur
);
1566 char *eol
= bol
, sign
= (cur
== funcname_lno
) ? '=' : '-';
1568 while (*eol
!= '\n')
1570 show_line(opt
, bol
, eol
, gs
->name
, cur
, sign
);
1576 static int should_lookahead(struct grep_opt
*opt
)
1581 return 0; /* punt for too complex stuff */
1584 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1585 if (p
->token
!= GREP_PATTERN
)
1586 return 0; /* punt for "header only" and stuff */
1591 static int look_ahead(struct grep_opt
*opt
,
1592 unsigned long *left_p
,
1596 unsigned lno
= *lno_p
;
1599 char *sp
, *last_bol
;
1600 regoff_t earliest
= -1;
1602 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1606 hit
= patmatch(p
, bol
, bol
+ *left_p
, &m
, 0);
1607 if (!hit
|| m
.rm_so
< 0 || m
.rm_eo
< 0)
1609 if (earliest
< 0 || m
.rm_so
< earliest
)
1614 *bol_p
= bol
+ *left_p
;
1618 for (sp
= bol
+ earliest
; bol
< sp
&& sp
[-1] != '\n'; sp
--)
1619 ; /* find the beginning of the line */
1622 for (sp
= bol
; sp
< last_bol
; sp
++) {
1626 *left_p
-= last_bol
- bol
;
1632 static int fill_textconv_grep(struct userdiff_driver
*driver
,
1633 struct grep_source
*gs
)
1635 struct diff_filespec
*df
;
1639 if (!driver
|| !driver
->textconv
)
1640 return grep_source_load(gs
);
1643 * The textconv interface is intimately tied to diff_filespecs, so we
1644 * have to pretend to be one. If we could unify the grep_source
1645 * and diff_filespec structs, this mess could just go away.
1647 df
= alloc_filespec(gs
->path
);
1649 case GREP_SOURCE_OID
:
1650 fill_filespec(df
, gs
->identifier
, 1, 0100644);
1652 case GREP_SOURCE_FILE
:
1653 fill_filespec(df
, &null_oid
, 0, 0100644);
1656 die("BUG: attempt to textconv something without a path?");
1660 * fill_textconv is not remotely thread-safe; it may load objects
1661 * behind the scenes, and it modifies the global diff tempfile
1665 size
= fill_textconv(driver
, df
, &buf
);
1670 * The normal fill_textconv usage by the diff machinery would just keep
1671 * the textconv'd buf separate from the diff_filespec. But much of the
1672 * grep code passes around a grep_source and assumes that its "buf"
1673 * pointer is the beginning of the thing we are searching. So let's
1674 * install our textconv'd version into the grep_source, taking care not
1675 * to leak any existing buffer.
1677 grep_source_clear_data(gs
);
1684 static int is_empty_line(const char *bol
, const char *eol
)
1686 while (bol
< eol
&& isspace(*bol
))
1691 static int grep_source_1(struct grep_opt
*opt
, struct grep_source
*gs
, int collect_hits
)
1694 char *peek_bol
= NULL
;
1697 unsigned last_hit
= 0;
1698 int binary_match_only
= 0;
1700 int try_lookahead
= 0;
1701 int show_function
= 0;
1702 struct userdiff_driver
*textconv
= NULL
;
1703 enum grep_context ctx
= GREP_CONTEXT_HEAD
;
1707 opt
->output
= std_output
;
1709 if (opt
->pre_context
|| opt
->post_context
|| opt
->file_break
||
1711 /* Show hunk marks, except for the first file. */
1712 if (opt
->last_shown
)
1713 opt
->show_hunk_mark
= 1;
1715 * If we're using threads then we can't easily identify
1716 * the first file. Always put hunk marks in that case
1717 * and skip the very first one later in work_done().
1719 if (opt
->output
!= std_output
)
1720 opt
->show_hunk_mark
= 1;
1722 opt
->last_shown
= 0;
1724 if (opt
->allow_textconv
) {
1725 grep_source_load_driver(gs
);
1727 * We might set up the shared textconv cache data here, which
1728 * is not thread-safe.
1731 textconv
= userdiff_get_textconv(gs
->driver
);
1736 * We know the result of a textconv is text, so we only have to care
1737 * about binary handling if we are not using it.
1740 switch (opt
->binary
) {
1741 case GREP_BINARY_DEFAULT
:
1742 if (grep_source_is_binary(gs
))
1743 binary_match_only
= 1;
1745 case GREP_BINARY_NOMATCH
:
1746 if (grep_source_is_binary(gs
))
1747 return 0; /* Assume unmatch */
1749 case GREP_BINARY_TEXT
:
1752 die("BUG: unknown binary handling mode");
1756 memset(&xecfg
, 0, sizeof(xecfg
));
1759 try_lookahead
= should_lookahead(opt
);
1761 if (fill_textconv_grep(textconv
, gs
) < 0)
1771 * look_ahead() skips quickly to the line that possibly
1772 * has the next hit; don't call it if we need to do
1773 * something more than just skipping the current line
1774 * in response to an unmatch for the current line. E.g.
1775 * inside a post-context window, we will show the current
1776 * line as a context around the previous hit when it
1781 && (show_function
||
1782 lno
<= last_hit
+ opt
->post_context
))
1783 && look_ahead(opt
, &left
, &lno
, &bol
))
1785 eol
= end_of_line(bol
, &left
);
1789 if ((ctx
== GREP_CONTEXT_HEAD
) && (eol
== bol
))
1790 ctx
= GREP_CONTEXT_BODY
;
1792 hit
= match_line(opt
, bol
, eol
, ctx
, collect_hits
);
1798 /* "grep -v -e foo -e bla" should list lines
1799 * that do not have either, so inversion should
1804 if (opt
->unmatch_name_only
) {
1811 if (opt
->status_only
)
1813 if (opt
->name_only
) {
1814 show_name(opt
, gs
->name
);
1819 if (binary_match_only
) {
1820 opt
->output(opt
, "Binary file ", 12);
1821 output_color(opt
, gs
->name
, strlen(gs
->name
),
1822 opt
->color_filename
);
1823 opt
->output(opt
, " matches\n", 9);
1826 /* Hit at this line. If we haven't shown the
1827 * pre-context lines, we would need to show them.
1829 if (opt
->pre_context
|| opt
->funcbody
)
1830 show_pre_context(opt
, gs
, bol
, eol
, lno
);
1831 else if (opt
->funcname
)
1832 show_funcname_line(opt
, gs
, bol
, lno
);
1833 show_line(opt
, bol
, eol
, gs
->name
, lno
, ':');
1839 if (show_function
&& (!peek_bol
|| peek_bol
< bol
)) {
1840 unsigned long peek_left
= left
;
1841 char *peek_eol
= eol
;
1844 * Trailing empty lines are not interesting.
1845 * Peek past them to see if they belong to the
1846 * body of the current function.
1849 while (is_empty_line(peek_bol
, peek_eol
)) {
1850 peek_bol
= peek_eol
+ 1;
1851 peek_eol
= end_of_line(peek_bol
, &peek_left
);
1854 if (match_funcname(opt
, gs
, peek_bol
, peek_eol
))
1857 if (show_function
||
1858 (last_hit
&& lno
<= last_hit
+ opt
->post_context
)) {
1859 /* If the last hit is within the post context,
1860 * we need to show this line.
1862 show_line(opt
, bol
, eol
, gs
->name
, lno
, '-');
1876 if (opt
->status_only
)
1877 return opt
->unmatch_name_only
;
1878 if (opt
->unmatch_name_only
) {
1879 /* We did not see any hit, so we want to show this */
1880 show_name(opt
, gs
->name
);
1884 xdiff_clear_find_func(&xecfg
);
1888 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1889 * which feels mostly useless but sometimes useful. Maybe
1890 * make it another option? For now suppress them.
1892 if (opt
->count
&& count
) {
1894 if (opt
->pathname
) {
1895 output_color(opt
, gs
->name
, strlen(gs
->name
),
1896 opt
->color_filename
);
1897 output_sep(opt
, ':');
1899 xsnprintf(buf
, sizeof(buf
), "%u\n", count
);
1900 opt
->output(opt
, buf
, strlen(buf
));
1906 static void clr_hit_marker(struct grep_expr
*x
)
1908 /* All-hit markers are meaningful only at the very top level
1913 if (x
->node
!= GREP_NODE_OR
)
1915 x
->u
.binary
.left
->hit
= 0;
1916 x
= x
->u
.binary
.right
;
1920 static int chk_hit_marker(struct grep_expr
*x
)
1922 /* Top level nodes have hit markers. See if they all are hits */
1924 if (x
->node
!= GREP_NODE_OR
)
1926 if (!x
->u
.binary
.left
->hit
)
1928 x
= x
->u
.binary
.right
;
1932 int grep_source(struct grep_opt
*opt
, struct grep_source
*gs
)
1935 * we do not have to do the two-pass grep when we do not check
1936 * buffer-wide "all-match".
1938 if (!opt
->all_match
)
1939 return grep_source_1(opt
, gs
, 0);
1941 /* Otherwise the toplevel "or" terms hit a bit differently.
1942 * We first clear hit markers from them.
1944 clr_hit_marker(opt
->pattern_expression
);
1945 grep_source_1(opt
, gs
, 1);
1947 if (!chk_hit_marker(opt
->pattern_expression
))
1950 return grep_source_1(opt
, gs
, 0);
1953 int grep_buffer(struct grep_opt
*opt
, char *buf
, unsigned long size
)
1955 struct grep_source gs
;
1958 grep_source_init(&gs
, GREP_SOURCE_BUF
, NULL
, NULL
, NULL
);
1962 r
= grep_source(opt
, &gs
);
1964 grep_source_clear(&gs
);
1968 void grep_source_init(struct grep_source
*gs
, enum grep_source_type type
,
1969 const char *name
, const char *path
,
1970 const void *identifier
)
1973 gs
->name
= xstrdup_or_null(name
);
1974 gs
->path
= xstrdup_or_null(path
);
1980 case GREP_SOURCE_FILE
:
1981 gs
->identifier
= xstrdup(identifier
);
1983 case GREP_SOURCE_OID
:
1984 gs
->identifier
= oiddup(identifier
);
1986 case GREP_SOURCE_BUF
:
1987 gs
->identifier
= NULL
;
1992 void grep_source_clear(struct grep_source
*gs
)
1994 FREE_AND_NULL(gs
->name
);
1995 FREE_AND_NULL(gs
->path
);
1996 FREE_AND_NULL(gs
->identifier
);
1997 grep_source_clear_data(gs
);
2000 void grep_source_clear_data(struct grep_source
*gs
)
2003 case GREP_SOURCE_FILE
:
2004 case GREP_SOURCE_OID
:
2005 FREE_AND_NULL(gs
->buf
);
2008 case GREP_SOURCE_BUF
:
2009 /* leave user-provided buf intact */
2014 static int grep_source_load_oid(struct grep_source
*gs
)
2016 enum object_type type
;
2019 gs
->buf
= read_object_file(gs
->identifier
, &type
, &gs
->size
);
2023 return error(_("'%s': unable to read %s"),
2025 oid_to_hex(gs
->identifier
));
2029 static int grep_source_load_file(struct grep_source
*gs
)
2031 const char *filename
= gs
->identifier
;
2037 if (lstat(filename
, &st
) < 0) {
2039 if (errno
!= ENOENT
)
2040 error_errno(_("failed to stat '%s'"), filename
);
2043 if (!S_ISREG(st
.st_mode
))
2045 size
= xsize_t(st
.st_size
);
2046 i
= open(filename
, O_RDONLY
);
2049 data
= xmallocz(size
);
2050 if (st
.st_size
!= read_in_full(i
, data
, size
)) {
2051 error_errno(_("'%s': short read"), filename
);
2063 static int grep_source_load(struct grep_source
*gs
)
2069 case GREP_SOURCE_FILE
:
2070 return grep_source_load_file(gs
);
2071 case GREP_SOURCE_OID
:
2072 return grep_source_load_oid(gs
);
2073 case GREP_SOURCE_BUF
:
2074 return gs
->buf
? 0 : -1;
2076 die("BUG: invalid grep_source type to load");
2079 void grep_source_load_driver(struct grep_source
*gs
)
2086 gs
->driver
= userdiff_find_by_path(gs
->path
);
2088 gs
->driver
= userdiff_find_by_name("default");
2092 static int grep_source_is_binary(struct grep_source
*gs
)
2094 grep_source_load_driver(gs
);
2095 if (gs
->driver
->binary
!= -1)
2096 return gs
->driver
->binary
;
2098 if (!grep_source_load(gs
))
2099 return buffer_is_binary(gs
->buf
, gs
->size
);