4 #include "object-store.h"
6 #include "xdiff-interface.h"
13 static int grep_source_load(struct grep_source
*gs
);
14 static int grep_source_is_binary(struct grep_source
*gs
);
16 static struct grep_opt grep_defaults
;
18 static const char *color_grep_slots
[] = {
19 [GREP_COLOR_CONTEXT
] = "context",
20 [GREP_COLOR_FILENAME
] = "filename",
21 [GREP_COLOR_FUNCTION
] = "function",
22 [GREP_COLOR_LINENO
] = "lineNumber",
23 [GREP_COLOR_MATCH_CONTEXT
] = "matchContext",
24 [GREP_COLOR_MATCH_SELECTED
] = "matchSelected",
25 [GREP_COLOR_SELECTED
] = "selected",
26 [GREP_COLOR_SEP
] = "separator",
29 static void std_output(struct grep_opt
*opt
, const void *buf
, size_t size
)
31 fwrite(buf
, size
, 1, stdout
);
34 static void color_set(char *dst
, const char *color_bytes
)
36 xsnprintf(dst
, COLOR_MAXLEN
, "%s", color_bytes
);
40 * Initialize the grep_defaults template with hardcoded defaults.
41 * We could let the compiler do this, but without C99 initializers
42 * the code gets unwieldy and unreadable, so...
44 void init_grep_defaults(void)
46 struct grep_opt
*opt
= &grep_defaults
;
53 memset(opt
, 0, sizeof(*opt
));
57 opt
->pattern_type_option
= GREP_PATTERN_TYPE_UNSPECIFIED
;
58 color_set(opt
->colors
[GREP_COLOR_CONTEXT
], "");
59 color_set(opt
->colors
[GREP_COLOR_FILENAME
], "");
60 color_set(opt
->colors
[GREP_COLOR_FUNCTION
], "");
61 color_set(opt
->colors
[GREP_COLOR_LINENO
], "");
62 color_set(opt
->colors
[GREP_COLOR_MATCH_CONTEXT
], GIT_COLOR_BOLD_RED
);
63 color_set(opt
->colors
[GREP_COLOR_MATCH_SELECTED
], GIT_COLOR_BOLD_RED
);
64 color_set(opt
->colors
[GREP_COLOR_SELECTED
], "");
65 color_set(opt
->colors
[GREP_COLOR_SEP
], GIT_COLOR_CYAN
);
67 opt
->output
= std_output
;
70 static int parse_pattern_type_arg(const char *opt
, const char *arg
)
72 if (!strcmp(arg
, "default"))
73 return GREP_PATTERN_TYPE_UNSPECIFIED
;
74 else if (!strcmp(arg
, "basic"))
75 return GREP_PATTERN_TYPE_BRE
;
76 else if (!strcmp(arg
, "extended"))
77 return GREP_PATTERN_TYPE_ERE
;
78 else if (!strcmp(arg
, "fixed"))
79 return GREP_PATTERN_TYPE_FIXED
;
80 else if (!strcmp(arg
, "perl"))
81 return GREP_PATTERN_TYPE_PCRE
;
82 die("bad %s argument: %s", opt
, arg
);
85 define_list_config_array_extra(color_grep_slots
, {"match"});
88 * Read the configuration file once and store it in
89 * the grep_defaults template.
91 int grep_config(const char *var
, const char *value
, void *cb
)
93 struct grep_opt
*opt
= &grep_defaults
;
96 if (userdiff_config(var
, value
) < 0)
99 if (!strcmp(var
, "grep.extendedregexp")) {
100 opt
->extended_regexp_option
= git_config_bool(var
, value
);
104 if (!strcmp(var
, "grep.patterntype")) {
105 opt
->pattern_type_option
= parse_pattern_type_arg(var
, value
);
109 if (!strcmp(var
, "grep.linenumber")) {
110 opt
->linenum
= git_config_bool(var
, value
);
114 if (!strcmp(var
, "grep.fullname")) {
115 opt
->relative
= !git_config_bool(var
, value
);
119 if (!strcmp(var
, "color.grep"))
120 opt
->color
= git_config_colorbool(var
, value
);
121 if (!strcmp(var
, "color.grep.match")) {
122 if (grep_config("color.grep.matchcontext", value
, cb
) < 0)
124 if (grep_config("color.grep.matchselected", value
, cb
) < 0)
126 } else if (skip_prefix(var
, "color.grep.", &slot
)) {
127 int i
= LOOKUP_CONFIG(color_grep_slots
, slot
);
132 color
= opt
->colors
[i
];
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
;
150 memset(opt
, 0, sizeof(*opt
));
151 opt
->prefix
= prefix
;
152 opt
->prefix_length
= (prefix
&& *prefix
) ? strlen(prefix
) : 0;
153 opt
->pattern_tail
= &opt
->pattern_list
;
154 opt
->header_tail
= &opt
->header_list
;
156 opt
->color
= def
->color
;
157 opt
->extended_regexp_option
= def
->extended_regexp_option
;
158 opt
->pattern_type_option
= def
->pattern_type_option
;
159 opt
->linenum
= def
->linenum
;
160 opt
->max_depth
= def
->max_depth
;
161 opt
->pathname
= def
->pathname
;
162 opt
->relative
= def
->relative
;
163 opt
->output
= def
->output
;
165 for (i
= 0; i
< NR_GREP_COLORS
; i
++)
166 color_set(opt
->colors
[i
], def
->colors
[i
]);
169 static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
172 * When committing to the pattern type by setting the relevant
173 * fields in grep_opt it's generally not necessary to zero out
174 * the fields we're not choosing, since they won't have been
175 * set by anything. The extended_regexp_option field is the
176 * only exception to this.
178 * This is because in the process of parsing grep.patternType
179 * & grep.extendedRegexp we set opt->pattern_type_option and
180 * opt->extended_regexp_option, respectively. We then
181 * internally use opt->extended_regexp_option to see if we're
182 * compiling an ERE. It must be unset if that's not actually
185 if (pattern_type
!= GREP_PATTERN_TYPE_ERE
&&
186 opt
->extended_regexp_option
)
187 opt
->extended_regexp_option
= 0;
189 switch (pattern_type
) {
190 case GREP_PATTERN_TYPE_UNSPECIFIED
:
193 case GREP_PATTERN_TYPE_BRE
:
196 case GREP_PATTERN_TYPE_ERE
:
197 opt
->extended_regexp_option
= 1;
200 case GREP_PATTERN_TYPE_FIXED
:
204 case GREP_PATTERN_TYPE_PCRE
:
209 * It's important that pcre1 always be assigned to
210 * even when there's no USE_LIBPCRE* defined. We still
211 * call the PCRE stub function, it just dies with
212 * "cannot use Perl-compatible regexes[...]".
220 void grep_commit_pattern_type(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
222 if (pattern_type
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
223 grep_set_pattern_type_option(pattern_type
, opt
);
224 else if (opt
->pattern_type_option
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
225 grep_set_pattern_type_option(opt
->pattern_type_option
, opt
);
226 else if (opt
->extended_regexp_option
)
228 * This branch *must* happen after setting from the
229 * opt->pattern_type_option above, we don't want
230 * grep.extendedRegexp to override grep.patternType!
232 grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE
, opt
);
235 static struct grep_pat
*create_grep_pat(const char *pat
, size_t patlen
,
236 const char *origin
, int no
,
237 enum grep_pat_token t
,
238 enum grep_header_field field
)
240 struct grep_pat
*p
= xcalloc(1, sizeof(*p
));
241 p
->pattern
= xmemdupz(pat
, patlen
);
242 p
->patternlen
= patlen
;
250 static void do_append_grep_pat(struct grep_pat
***tail
, struct grep_pat
*p
)
257 case GREP_PATTERN
: /* atom */
258 case GREP_PATTERN_HEAD
:
259 case GREP_PATTERN_BODY
:
261 struct grep_pat
*new_pat
;
263 char *cp
= p
->pattern
+ p
->patternlen
, *nl
= NULL
;
264 while (++len
<= p
->patternlen
) {
265 if (*(--cp
) == '\n') {
272 new_pat
= create_grep_pat(nl
+ 1, len
- 1, p
->origin
,
273 p
->no
, p
->token
, p
->field
);
274 new_pat
->next
= p
->next
;
276 *tail
= &new_pat
->next
;
279 p
->patternlen
-= len
;
287 void append_header_grep_pattern(struct grep_opt
*opt
,
288 enum grep_header_field field
, const char *pat
)
290 struct grep_pat
*p
= create_grep_pat(pat
, strlen(pat
), "header", 0,
291 GREP_PATTERN_HEAD
, field
);
292 if (field
== GREP_HEADER_REFLOG
)
293 opt
->use_reflog_filter
= 1;
294 do_append_grep_pat(&opt
->header_tail
, p
);
297 void append_grep_pattern(struct grep_opt
*opt
, const char *pat
,
298 const char *origin
, int no
, enum grep_pat_token t
)
300 append_grep_pat(opt
, pat
, strlen(pat
), origin
, no
, t
);
303 void append_grep_pat(struct grep_opt
*opt
, const char *pat
, size_t patlen
,
304 const char *origin
, int no
, enum grep_pat_token t
)
306 struct grep_pat
*p
= create_grep_pat(pat
, patlen
, origin
, no
, t
, 0);
307 do_append_grep_pat(&opt
->pattern_tail
, p
);
310 struct grep_opt
*grep_opt_dup(const struct grep_opt
*opt
)
312 struct grep_pat
*pat
;
313 struct grep_opt
*ret
= xmalloc(sizeof(struct grep_opt
));
316 ret
->pattern_list
= NULL
;
317 ret
->pattern_tail
= &ret
->pattern_list
;
319 for(pat
= opt
->pattern_list
; pat
!= NULL
; pat
= pat
->next
)
321 if(pat
->token
== GREP_PATTERN_HEAD
)
322 append_header_grep_pattern(ret
, pat
->field
,
325 append_grep_pat(ret
, pat
->pattern
, pat
->patternlen
,
326 pat
->origin
, pat
->no
, pat
->token
);
332 static NORETURN
void compile_regexp_failed(const struct grep_pat
*p
,
338 xsnprintf(where
, sizeof(where
), "In '%s' at %d, ", p
->origin
, p
->no
);
340 xsnprintf(where
, sizeof(where
), "%s, ", p
->origin
);
344 die("%s'%s': %s", where
, p
->pattern
, error
);
347 static int is_fixed(const char *s
, size_t len
)
351 for (i
= 0; i
< len
; i
++) {
352 if (is_regex_special(s
[i
]))
359 static int has_null(const char *s
, size_t len
)
362 * regcomp cannot accept patterns with NULs so when using it
363 * we consider any pattern containing a NUL fixed.
365 if (memchr(s
, 0, len
))
372 static void compile_pcre1_regexp(struct grep_pat
*p
, const struct grep_opt
*opt
)
376 int options
= PCRE_MULTILINE
;
378 if (opt
->ignore_case
) {
379 if (has_non_ascii(p
->pattern
))
380 p
->pcre1_tables
= pcre_maketables();
381 options
|= PCRE_CASELESS
;
383 if (is_utf8_locale() && has_non_ascii(p
->pattern
))
384 options
|= PCRE_UTF8
;
386 p
->pcre1_regexp
= pcre_compile(p
->pattern
, options
, &error
, &erroffset
,
388 if (!p
->pcre1_regexp
)
389 compile_regexp_failed(p
, error
);
391 p
->pcre1_extra_info
= pcre_study(p
->pcre1_regexp
, GIT_PCRE_STUDY_JIT_COMPILE
, &error
);
392 if (!p
->pcre1_extra_info
&& error
)
395 #ifdef GIT_PCRE1_USE_JIT
396 pcre_config(PCRE_CONFIG_JIT
, &p
->pcre1_jit_on
);
397 if (p
->pcre1_jit_on
== 1) {
398 p
->pcre1_jit_stack
= pcre_jit_stack_alloc(1, 1024 * 1024);
399 if (!p
->pcre1_jit_stack
)
400 die("Couldn't allocate PCRE JIT stack");
401 pcre_assign_jit_stack(p
->pcre1_extra_info
, NULL
, p
->pcre1_jit_stack
);
402 } else if (p
->pcre1_jit_on
!= 0) {
403 BUG("The pcre1_jit_on variable should be 0 or 1, not %d",
409 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
410 regmatch_t
*match
, int eflags
)
412 int ovector
[30], ret
, flags
= 0;
414 if (eflags
& REG_NOTBOL
)
415 flags
|= PCRE_NOTBOL
;
417 #ifdef GIT_PCRE1_USE_JIT
418 if (p
->pcre1_jit_on
) {
419 ret
= pcre_jit_exec(p
->pcre1_regexp
, p
->pcre1_extra_info
, line
,
420 eol
- line
, 0, flags
, ovector
,
421 ARRAY_SIZE(ovector
), p
->pcre1_jit_stack
);
425 ret
= pcre_exec(p
->pcre1_regexp
, p
->pcre1_extra_info
, line
,
426 eol
- line
, 0, flags
, ovector
,
427 ARRAY_SIZE(ovector
));
430 if (ret
< 0 && ret
!= PCRE_ERROR_NOMATCH
)
431 die("pcre_exec failed with error code %d", ret
);
434 match
->rm_so
= ovector
[0];
435 match
->rm_eo
= ovector
[1];
441 static void free_pcre1_regexp(struct grep_pat
*p
)
443 pcre_free(p
->pcre1_regexp
);
444 #ifdef GIT_PCRE1_USE_JIT
445 if (p
->pcre1_jit_on
) {
446 pcre_free_study(p
->pcre1_extra_info
);
447 pcre_jit_stack_free(p
->pcre1_jit_stack
);
451 pcre_free(p
->pcre1_extra_info
);
453 pcre_free((void *)p
->pcre1_tables
);
455 #else /* !USE_LIBPCRE1 */
456 static void compile_pcre1_regexp(struct grep_pat
*p
, const struct grep_opt
*opt
)
458 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
461 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
462 regmatch_t
*match
, int eflags
)
467 static void free_pcre1_regexp(struct grep_pat
*p
)
470 #endif /* !USE_LIBPCRE1 */
473 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
476 PCRE2_UCHAR errbuf
[256];
477 PCRE2_SIZE erroffset
;
478 int options
= PCRE2_MULTILINE
;
479 const uint8_t *character_tables
= NULL
;
486 p
->pcre2_compile_context
= NULL
;
488 if (opt
->ignore_case
) {
489 if (has_non_ascii(p
->pattern
)) {
490 character_tables
= pcre2_maketables(NULL
);
491 p
->pcre2_compile_context
= pcre2_compile_context_create(NULL
);
492 pcre2_set_character_tables(p
->pcre2_compile_context
, character_tables
);
494 options
|= PCRE2_CASELESS
;
496 if (is_utf8_locale() && has_non_ascii(p
->pattern
))
497 options
|= PCRE2_UTF
;
499 p
->pcre2_pattern
= pcre2_compile((PCRE2_SPTR
)p
->pattern
,
500 p
->patternlen
, options
, &error
, &erroffset
,
501 p
->pcre2_compile_context
);
503 if (p
->pcre2_pattern
) {
504 p
->pcre2_match_data
= pcre2_match_data_create_from_pattern(p
->pcre2_pattern
, NULL
);
505 if (!p
->pcre2_match_data
)
506 die("Couldn't allocate PCRE2 match data");
508 pcre2_get_error_message(error
, errbuf
, sizeof(errbuf
));
509 compile_regexp_failed(p
, (const char *)&errbuf
);
512 pcre2_config(PCRE2_CONFIG_JIT
, &p
->pcre2_jit_on
);
513 if (p
->pcre2_jit_on
== 1) {
514 jitret
= pcre2_jit_compile(p
->pcre2_pattern
, PCRE2_JIT_COMPLETE
);
516 die("Couldn't JIT the PCRE2 pattern '%s', got '%d'\n", p
->pattern
, jitret
);
519 * The pcre2_config(PCRE2_CONFIG_JIT, ...) call just
520 * tells us whether the library itself supports JIT,
521 * but to see whether we're going to be actually using
522 * JIT we need to extract PCRE2_INFO_JITSIZE from the
523 * pattern *after* we do pcre2_jit_compile() above.
525 * This is because if the pattern contains the
526 * (*NO_JIT) verb (see pcre2syntax(3))
527 * pcre2_jit_compile() will exit early with 0. If we
528 * then proceed to call pcre2_jit_match() further down
529 * the line instead of pcre2_match() we'll either
530 * segfault (pre PCRE 10.31) or run into a fatal error
533 patinforet
= pcre2_pattern_info(p
->pcre2_pattern
, PCRE2_INFO_JITSIZE
, &jitsizearg
);
535 BUG("pcre2_pattern_info() failed: %d", patinforet
);
536 if (jitsizearg
== 0) {
541 p
->pcre2_jit_stack
= pcre2_jit_stack_create(1, 1024 * 1024, NULL
);
542 if (!p
->pcre2_jit_stack
)
543 die("Couldn't allocate PCRE2 JIT stack");
544 p
->pcre2_match_context
= pcre2_match_context_create(NULL
);
545 if (!p
->pcre2_match_context
)
546 die("Couldn't allocate PCRE2 match context");
547 pcre2_jit_stack_assign(p
->pcre2_match_context
, NULL
, p
->pcre2_jit_stack
);
548 } else if (p
->pcre2_jit_on
!= 0) {
549 BUG("The pcre2_jit_on variable should be 0 or 1, not %d",
554 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
555 regmatch_t
*match
, int eflags
)
559 PCRE2_UCHAR errbuf
[256];
561 if (eflags
& REG_NOTBOL
)
562 flags
|= PCRE2_NOTBOL
;
565 ret
= pcre2_jit_match(p
->pcre2_pattern
, (unsigned char *)line
,
566 eol
- line
, 0, flags
, p
->pcre2_match_data
,
569 ret
= pcre2_match(p
->pcre2_pattern
, (unsigned char *)line
,
570 eol
- line
, 0, flags
, p
->pcre2_match_data
,
573 if (ret
< 0 && ret
!= PCRE2_ERROR_NOMATCH
) {
574 pcre2_get_error_message(ret
, errbuf
, sizeof(errbuf
));
575 die("%s failed with error code %d: %s",
576 (p
->pcre2_jit_on
? "pcre2_jit_match" : "pcre2_match"), ret
,
580 ovector
= pcre2_get_ovector_pointer(p
->pcre2_match_data
);
582 match
->rm_so
= (int)ovector
[0];
583 match
->rm_eo
= (int)ovector
[1];
589 static void free_pcre2_pattern(struct grep_pat
*p
)
591 pcre2_compile_context_free(p
->pcre2_compile_context
);
592 pcre2_code_free(p
->pcre2_pattern
);
593 pcre2_match_data_free(p
->pcre2_match_data
);
594 pcre2_jit_stack_free(p
->pcre2_jit_stack
);
595 pcre2_match_context_free(p
->pcre2_match_context
);
597 #else /* !USE_LIBPCRE2 */
598 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
601 * Unreachable until USE_LIBPCRE2 becomes synonymous with
602 * USE_LIBPCRE. See the sibling comment in
603 * grep_set_pattern_type_option().
605 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
608 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
609 regmatch_t
*match
, int eflags
)
614 static void free_pcre2_pattern(struct grep_pat
*p
)
617 #endif /* !USE_LIBPCRE2 */
619 static void compile_fixed_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
621 struct strbuf sb
= STRBUF_INIT
;
625 basic_regex_quote_buf(&sb
, p
->pattern
);
626 if (opt
->ignore_case
)
627 regflags
|= REG_ICASE
;
628 err
= regcomp(&p
->regexp
, sb
.buf
, regflags
);
630 fprintf(stderr
, "fixed %s\n", sb
.buf
);
634 regerror(err
, &p
->regexp
, errbuf
, sizeof(errbuf
));
635 compile_regexp_failed(p
, errbuf
);
639 static void compile_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
643 int regflags
= REG_NEWLINE
;
645 p
->word_regexp
= opt
->word_regexp
;
646 p
->ignore_case
= opt
->ignore_case
;
647 ascii_only
= !has_non_ascii(p
->pattern
);
650 * Even when -F (fixed) asks us to do a non-regexp search, we
651 * may not be able to correctly case-fold when -i
652 * (ignore-case) is asked (in which case, we'll synthesize a
653 * regexp to match the pattern that matches regexp special
654 * characters literally, while ignoring case differences). On
655 * the other hand, even without -F, if the pattern does not
656 * have any regexp special characters and there is no need for
657 * case-folding search, we can internally turn it into a
658 * simple string match using kws. p->fixed tells us if we
662 has_null(p
->pattern
, p
->patternlen
) ||
663 is_fixed(p
->pattern
, p
->patternlen
))
664 p
->fixed
= !p
->ignore_case
|| ascii_only
;
667 p
->kws
= kwsalloc(p
->ignore_case
? tolower_trans_tbl
: NULL
);
668 kwsincr(p
->kws
, p
->pattern
, p
->patternlen
);
671 } else if (opt
->fixed
) {
673 * We come here when the pattern has the non-ascii
674 * characters we cannot case-fold, and asked to
677 compile_fixed_regexp(p
, opt
);
682 compile_pcre2_pattern(p
, opt
);
687 compile_pcre1_regexp(p
, opt
);
692 regflags
|= REG_ICASE
;
693 if (opt
->extended_regexp_option
)
694 regflags
|= REG_EXTENDED
;
695 err
= regcomp(&p
->regexp
, p
->pattern
, regflags
);
698 regerror(err
, &p
->regexp
, errbuf
, 1024);
699 compile_regexp_failed(p
, errbuf
);
703 static struct grep_expr
*compile_pattern_or(struct grep_pat
**);
704 static struct grep_expr
*compile_pattern_atom(struct grep_pat
**list
)
713 case GREP_PATTERN
: /* atom */
714 case GREP_PATTERN_HEAD
:
715 case GREP_PATTERN_BODY
:
716 x
= xcalloc(1, sizeof (struct grep_expr
));
717 x
->node
= GREP_NODE_ATOM
;
721 case GREP_OPEN_PAREN
:
723 x
= compile_pattern_or(list
);
724 if (!*list
|| (*list
)->token
!= GREP_CLOSE_PAREN
)
725 die("unmatched parenthesis");
726 *list
= (*list
)->next
;
733 static struct grep_expr
*compile_pattern_not(struct grep_pat
**list
)
744 die("--not not followed by pattern expression");
746 x
= xcalloc(1, sizeof (struct grep_expr
));
747 x
->node
= GREP_NODE_NOT
;
748 x
->u
.unary
= compile_pattern_not(list
);
750 die("--not followed by non pattern expression");
753 return compile_pattern_atom(list
);
757 static struct grep_expr
*compile_pattern_and(struct grep_pat
**list
)
760 struct grep_expr
*x
, *y
, *z
;
762 x
= compile_pattern_not(list
);
764 if (p
&& p
->token
== GREP_AND
) {
766 die("--and not followed by pattern expression");
768 y
= compile_pattern_and(list
);
770 die("--and not followed by pattern expression");
771 z
= xcalloc(1, sizeof (struct grep_expr
));
772 z
->node
= GREP_NODE_AND
;
773 z
->u
.binary
.left
= x
;
774 z
->u
.binary
.right
= y
;
780 static struct grep_expr
*compile_pattern_or(struct grep_pat
**list
)
783 struct grep_expr
*x
, *y
, *z
;
785 x
= compile_pattern_and(list
);
787 if (x
&& p
&& p
->token
!= GREP_CLOSE_PAREN
) {
788 y
= compile_pattern_or(list
);
790 die("not a pattern expression %s", p
->pattern
);
791 z
= xcalloc(1, sizeof (struct grep_expr
));
792 z
->node
= GREP_NODE_OR
;
793 z
->u
.binary
.left
= x
;
794 z
->u
.binary
.right
= y
;
800 static struct grep_expr
*compile_pattern_expr(struct grep_pat
**list
)
802 return compile_pattern_or(list
);
805 static void indent(int in
)
811 static void dump_grep_pat(struct grep_pat
*p
)
814 case GREP_AND
: fprintf(stderr
, "*and*"); break;
815 case GREP_OPEN_PAREN
: fprintf(stderr
, "*(*"); break;
816 case GREP_CLOSE_PAREN
: fprintf(stderr
, "*)*"); break;
817 case GREP_NOT
: fprintf(stderr
, "*not*"); break;
818 case GREP_OR
: fprintf(stderr
, "*or*"); break;
820 case GREP_PATTERN
: fprintf(stderr
, "pattern"); break;
821 case GREP_PATTERN_HEAD
: fprintf(stderr
, "pattern_head"); break;
822 case GREP_PATTERN_BODY
: fprintf(stderr
, "pattern_body"); break;
827 case GREP_PATTERN_HEAD
:
828 fprintf(stderr
, "<head %d>", p
->field
); break;
829 case GREP_PATTERN_BODY
:
830 fprintf(stderr
, "<body>"); break;
834 case GREP_PATTERN_HEAD
:
835 case GREP_PATTERN_BODY
:
837 fprintf(stderr
, "%.*s", (int)p
->patternlen
, p
->pattern
);
843 static void dump_grep_expression_1(struct grep_expr
*x
, int in
)
848 fprintf(stderr
, "true\n");
851 dump_grep_pat(x
->u
.atom
);
854 fprintf(stderr
, "(not\n");
855 dump_grep_expression_1(x
->u
.unary
, in
+1);
857 fprintf(stderr
, ")\n");
860 fprintf(stderr
, "(and\n");
861 dump_grep_expression_1(x
->u
.binary
.left
, in
+1);
862 dump_grep_expression_1(x
->u
.binary
.right
, in
+1);
864 fprintf(stderr
, ")\n");
867 fprintf(stderr
, "(or\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");
876 static void dump_grep_expression(struct grep_opt
*opt
)
878 struct grep_expr
*x
= opt
->pattern_expression
;
881 fprintf(stderr
, "[all-match]\n");
882 dump_grep_expression_1(x
, 0);
886 static struct grep_expr
*grep_true_expr(void)
888 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
889 z
->node
= GREP_NODE_TRUE
;
893 static struct grep_expr
*grep_or_expr(struct grep_expr
*left
, struct grep_expr
*right
)
895 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
896 z
->node
= GREP_NODE_OR
;
897 z
->u
.binary
.left
= left
;
898 z
->u
.binary
.right
= right
;
902 static struct grep_expr
*prep_header_patterns(struct grep_opt
*opt
)
905 struct grep_expr
*header_expr
;
906 struct grep_expr
*(header_group
[GREP_HEADER_FIELD_MAX
]);
907 enum grep_header_field fld
;
909 if (!opt
->header_list
)
912 for (p
= opt
->header_list
; p
; p
= p
->next
) {
913 if (p
->token
!= GREP_PATTERN_HEAD
)
914 BUG("a non-header pattern in grep header list.");
915 if (p
->field
< GREP_HEADER_FIELD_MIN
||
916 GREP_HEADER_FIELD_MAX
<= p
->field
)
917 BUG("unknown header field %d", p
->field
);
918 compile_regexp(p
, opt
);
921 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++)
922 header_group
[fld
] = NULL
;
924 for (p
= opt
->header_list
; p
; p
= p
->next
) {
926 struct grep_pat
*pp
= p
;
928 h
= compile_pattern_atom(&pp
);
929 if (!h
|| pp
!= p
->next
)
930 BUG("malformed header expr");
931 if (!header_group
[p
->field
]) {
932 header_group
[p
->field
] = h
;
935 header_group
[p
->field
] = grep_or_expr(h
, header_group
[p
->field
]);
940 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++) {
941 if (!header_group
[fld
])
944 header_expr
= grep_true_expr();
945 header_expr
= grep_or_expr(header_group
[fld
], header_expr
);
950 static struct grep_expr
*grep_splice_or(struct grep_expr
*x
, struct grep_expr
*y
)
952 struct grep_expr
*z
= x
;
955 assert(x
->node
== GREP_NODE_OR
);
956 if (x
->u
.binary
.right
&&
957 x
->u
.binary
.right
->node
== GREP_NODE_TRUE
) {
958 x
->u
.binary
.right
= y
;
961 x
= x
->u
.binary
.right
;
966 static void compile_grep_patterns_real(struct grep_opt
*opt
)
969 struct grep_expr
*header_expr
= prep_header_patterns(opt
);
971 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
973 case GREP_PATTERN
: /* atom */
974 case GREP_PATTERN_HEAD
:
975 case GREP_PATTERN_BODY
:
976 compile_regexp(p
, opt
);
984 if (opt
->all_match
|| header_expr
)
986 else if (!opt
->extended
&& !opt
->debug
)
989 p
= opt
->pattern_list
;
991 opt
->pattern_expression
= compile_pattern_expr(&p
);
993 die("incomplete pattern expression: %s", p
->pattern
);
998 if (!opt
->pattern_expression
)
999 opt
->pattern_expression
= header_expr
;
1000 else if (opt
->all_match
)
1001 opt
->pattern_expression
= grep_splice_or(header_expr
,
1002 opt
->pattern_expression
);
1004 opt
->pattern_expression
= grep_or_expr(opt
->pattern_expression
,
1009 void compile_grep_patterns(struct grep_opt
*opt
)
1011 compile_grep_patterns_real(opt
);
1013 dump_grep_expression(opt
);
1016 static void free_pattern_expr(struct grep_expr
*x
)
1019 case GREP_NODE_TRUE
:
1020 case GREP_NODE_ATOM
:
1023 free_pattern_expr(x
->u
.unary
);
1027 free_pattern_expr(x
->u
.binary
.left
);
1028 free_pattern_expr(x
->u
.binary
.right
);
1034 void free_grep_patterns(struct grep_opt
*opt
)
1036 struct grep_pat
*p
, *n
;
1038 for (p
= opt
->pattern_list
; p
; p
= n
) {
1041 case GREP_PATTERN
: /* atom */
1042 case GREP_PATTERN_HEAD
:
1043 case GREP_PATTERN_BODY
:
1046 else if (p
->pcre1_regexp
)
1047 free_pcre1_regexp(p
);
1048 else if (p
->pcre2_pattern
)
1049 free_pcre2_pattern(p
);
1051 regfree(&p
->regexp
);
1062 free_pattern_expr(opt
->pattern_expression
);
1065 static char *end_of_line(char *cp
, unsigned long *left
)
1067 unsigned long l
= *left
;
1068 while (l
&& *cp
!= '\n') {
1076 static int word_char(char ch
)
1078 return isalnum(ch
) || ch
== '_';
1081 static void output_color(struct grep_opt
*opt
, const void *data
, size_t size
,
1084 if (want_color(opt
->color
) && color
&& color
[0]) {
1085 opt
->output(opt
, color
, strlen(color
));
1086 opt
->output(opt
, data
, size
);
1087 opt
->output(opt
, GIT_COLOR_RESET
, strlen(GIT_COLOR_RESET
));
1089 opt
->output(opt
, data
, size
);
1092 static void output_sep(struct grep_opt
*opt
, char sign
)
1094 if (opt
->null_following_name
)
1095 opt
->output(opt
, "\0", 1);
1097 output_color(opt
, &sign
, 1, opt
->colors
[GREP_COLOR_SEP
]);
1100 static void show_name(struct grep_opt
*opt
, const char *name
)
1102 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1103 opt
->output(opt
, opt
->null_following_name
? "\0" : "\n", 1);
1106 static int fixmatch(struct grep_pat
*p
, char *line
, char *eol
,
1109 struct kwsmatch kwsm
;
1110 size_t offset
= kwsexec(p
->kws
, line
, eol
- line
, &kwsm
);
1112 match
->rm_so
= match
->rm_eo
= -1;
1115 match
->rm_so
= offset
;
1116 match
->rm_eo
= match
->rm_so
+ kwsm
.size
[0];
1121 static int patmatch(struct grep_pat
*p
, char *line
, char *eol
,
1122 regmatch_t
*match
, int eflags
)
1127 hit
= !fixmatch(p
, line
, eol
, match
);
1128 else if (p
->pcre1_regexp
)
1129 hit
= !pcre1match(p
, line
, eol
, match
, eflags
);
1130 else if (p
->pcre2_pattern
)
1131 hit
= !pcre2match(p
, line
, eol
, match
, eflags
);
1133 hit
= !regexec_buf(&p
->regexp
, line
, eol
- line
, 1, match
,
1139 static int strip_timestamp(char *bol
, char **eol_p
)
1144 while (bol
< --eol
) {
1158 } header_field
[] = {
1160 { "committer ", 10 },
1164 static int match_one_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1165 enum grep_context ctx
,
1166 regmatch_t
*pmatch
, int eflags
)
1170 const char *start
= bol
;
1172 if ((p
->token
!= GREP_PATTERN
) &&
1173 ((p
->token
== GREP_PATTERN_HEAD
) != (ctx
== GREP_CONTEXT_HEAD
)))
1176 if (p
->token
== GREP_PATTERN_HEAD
) {
1179 assert(p
->field
< ARRAY_SIZE(header_field
));
1180 field
= header_field
[p
->field
].field
;
1181 len
= header_field
[p
->field
].len
;
1182 if (strncmp(bol
, field
, len
))
1186 case GREP_HEADER_AUTHOR
:
1187 case GREP_HEADER_COMMITTER
:
1188 saved_ch
= strip_timestamp(bol
, &eol
);
1196 hit
= patmatch(p
, bol
, eol
, pmatch
, eflags
);
1198 if (hit
&& p
->word_regexp
) {
1199 if ((pmatch
[0].rm_so
< 0) ||
1200 (eol
- bol
) < pmatch
[0].rm_so
||
1201 (pmatch
[0].rm_eo
< 0) ||
1202 (eol
- bol
) < pmatch
[0].rm_eo
)
1203 die("regexp returned nonsense");
1205 /* Match beginning must be either beginning of the
1206 * line, or at word boundary (i.e. the last char must
1207 * not be a word char). Similarly, match end must be
1208 * either end of the line, or at word boundary
1209 * (i.e. the next char must not be a word char).
1211 if ( ((pmatch
[0].rm_so
== 0) ||
1212 !word_char(bol
[pmatch
[0].rm_so
-1])) &&
1213 ((pmatch
[0].rm_eo
== (eol
-bol
)) ||
1214 !word_char(bol
[pmatch
[0].rm_eo
])) )
1219 /* Words consist of at least one character. */
1220 if (pmatch
->rm_so
== pmatch
->rm_eo
)
1223 if (!hit
&& pmatch
[0].rm_so
+ bol
+ 1 < eol
) {
1224 /* There could be more than one match on the
1225 * line, and the first match might not be
1226 * strict word match. But later ones could be!
1227 * Forward to the next possible start, i.e. the
1228 * next position following a non-word char.
1230 bol
= pmatch
[0].rm_so
+ bol
+ 1;
1231 while (word_char(bol
[-1]) && bol
< eol
)
1233 eflags
|= REG_NOTBOL
;
1238 if (p
->token
== GREP_PATTERN_HEAD
&& saved_ch
)
1241 pmatch
[0].rm_so
+= bol
- start
;
1242 pmatch
[0].rm_eo
+= bol
- start
;
1247 static int match_expr_eval(struct grep_expr
*x
, char *bol
, char *eol
,
1248 enum grep_context ctx
, int collect_hits
)
1254 die("Not a valid grep expression");
1256 case GREP_NODE_TRUE
:
1259 case GREP_NODE_ATOM
:
1260 h
= match_one_pattern(x
->u
.atom
, bol
, eol
, ctx
, &match
, 0);
1263 h
= !match_expr_eval(x
->u
.unary
, bol
, eol
, ctx
, 0);
1266 if (!match_expr_eval(x
->u
.binary
.left
, bol
, eol
, ctx
, 0))
1268 h
= match_expr_eval(x
->u
.binary
.right
, bol
, eol
, ctx
, 0);
1272 return (match_expr_eval(x
->u
.binary
.left
,
1273 bol
, eol
, ctx
, 0) ||
1274 match_expr_eval(x
->u
.binary
.right
,
1276 h
= match_expr_eval(x
->u
.binary
.left
, bol
, eol
, ctx
, 0);
1277 x
->u
.binary
.left
->hit
|= h
;
1278 h
|= match_expr_eval(x
->u
.binary
.right
, bol
, eol
, ctx
, 1);
1281 die("Unexpected node type (internal error) %d", x
->node
);
1288 static int match_expr(struct grep_opt
*opt
, char *bol
, char *eol
,
1289 enum grep_context ctx
, int collect_hits
)
1291 struct grep_expr
*x
= opt
->pattern_expression
;
1292 return match_expr_eval(x
, bol
, eol
, ctx
, collect_hits
);
1295 static int match_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1296 enum grep_context ctx
, int collect_hits
)
1302 return match_expr(opt
, bol
, eol
, ctx
, collect_hits
);
1304 /* we do not call with collect_hits without being extended */
1305 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1306 if (match_one_pattern(p
, bol
, eol
, ctx
, &match
, 0))
1312 static int match_next_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1313 enum grep_context ctx
,
1314 regmatch_t
*pmatch
, int eflags
)
1318 if (!match_one_pattern(p
, bol
, eol
, ctx
, &match
, eflags
))
1320 if (match
.rm_so
< 0 || match
.rm_eo
< 0)
1322 if (pmatch
->rm_so
>= 0 && pmatch
->rm_eo
>= 0) {
1323 if (match
.rm_so
> pmatch
->rm_so
)
1325 if (match
.rm_so
== pmatch
->rm_so
&& match
.rm_eo
< pmatch
->rm_eo
)
1328 pmatch
->rm_so
= match
.rm_so
;
1329 pmatch
->rm_eo
= match
.rm_eo
;
1333 static int next_match(struct grep_opt
*opt
, char *bol
, char *eol
,
1334 enum grep_context ctx
, regmatch_t
*pmatch
, int eflags
)
1339 pmatch
->rm_so
= pmatch
->rm_eo
= -1;
1341 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1343 case GREP_PATTERN
: /* atom */
1344 case GREP_PATTERN_HEAD
:
1345 case GREP_PATTERN_BODY
:
1346 hit
|= match_next_pattern(p
, bol
, eol
, ctx
,
1357 static void show_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1358 const char *name
, unsigned lno
, char sign
)
1360 int rest
= eol
- bol
;
1361 const char *match_color
, *line_color
= NULL
;
1363 if (opt
->file_break
&& opt
->last_shown
== 0) {
1364 if (opt
->show_hunk_mark
)
1365 opt
->output(opt
, "\n", 1);
1366 } else if (opt
->pre_context
|| opt
->post_context
|| opt
->funcbody
) {
1367 if (opt
->last_shown
== 0) {
1368 if (opt
->show_hunk_mark
) {
1369 output_color(opt
, "--", 2, opt
->colors
[GREP_COLOR_SEP
]);
1370 opt
->output(opt
, "\n", 1);
1372 } else if (lno
> opt
->last_shown
+ 1) {
1373 output_color(opt
, "--", 2, opt
->colors
[GREP_COLOR_SEP
]);
1374 opt
->output(opt
, "\n", 1);
1377 if (opt
->heading
&& opt
->last_shown
== 0) {
1378 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1379 opt
->output(opt
, "\n", 1);
1381 opt
->last_shown
= lno
;
1383 if (!opt
->heading
&& opt
->pathname
) {
1384 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1385 output_sep(opt
, sign
);
1389 xsnprintf(buf
, sizeof(buf
), "%d", lno
);
1390 output_color(opt
, buf
, strlen(buf
), opt
->colors
[GREP_COLOR_LINENO
]);
1391 output_sep(opt
, sign
);
1395 enum grep_context ctx
= GREP_CONTEXT_BODY
;
1400 match_color
= opt
->colors
[GREP_COLOR_MATCH_SELECTED
];
1402 match_color
= opt
->colors
[GREP_COLOR_MATCH_CONTEXT
];
1404 line_color
= opt
->colors
[GREP_COLOR_SELECTED
];
1405 else if (sign
== '-')
1406 line_color
= opt
->colors
[GREP_COLOR_CONTEXT
];
1407 else if (sign
== '=')
1408 line_color
= opt
->colors
[GREP_COLOR_FUNCTION
];
1410 while (next_match(opt
, bol
, eol
, ctx
, &match
, eflags
)) {
1411 if (match
.rm_so
== match
.rm_eo
)
1414 output_color(opt
, bol
, match
.rm_so
, line_color
);
1415 output_color(opt
, bol
+ match
.rm_so
,
1416 match
.rm_eo
- match
.rm_so
, match_color
);
1418 rest
-= match
.rm_eo
;
1419 eflags
= REG_NOTBOL
;
1423 output_color(opt
, bol
, rest
, line_color
);
1424 opt
->output(opt
, "\n", 1);
1431 * This lock protects access to the gitattributes machinery, which is
1434 pthread_mutex_t grep_attr_mutex
;
1436 static inline void grep_attr_lock(void)
1439 pthread_mutex_lock(&grep_attr_mutex
);
1442 static inline void grep_attr_unlock(void)
1445 pthread_mutex_unlock(&grep_attr_mutex
);
1449 * Same as git_attr_mutex, but protecting the thread-unsafe object db access.
1451 pthread_mutex_t grep_read_mutex
;
1454 #define grep_attr_lock()
1455 #define grep_attr_unlock()
1458 static int match_funcname(struct grep_opt
*opt
, struct grep_source
*gs
, char *bol
, char *eol
)
1460 xdemitconf_t
*xecfg
= opt
->priv
;
1461 if (xecfg
&& !xecfg
->find_func
) {
1462 grep_source_load_driver(gs
);
1463 if (gs
->driver
->funcname
.pattern
) {
1464 const struct userdiff_funcname
*pe
= &gs
->driver
->funcname
;
1465 xdiff_set_find_func(xecfg
, pe
->pattern
, pe
->cflags
);
1467 xecfg
= opt
->priv
= NULL
;
1473 return xecfg
->find_func(bol
, eol
- bol
, buf
, 1,
1474 xecfg
->find_func_priv
) >= 0;
1479 if (isalpha(*bol
) || *bol
== '_' || *bol
== '$')
1484 static void show_funcname_line(struct grep_opt
*opt
, struct grep_source
*gs
,
1485 char *bol
, unsigned lno
)
1487 while (bol
> gs
->buf
) {
1490 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1494 if (lno
<= opt
->last_shown
)
1497 if (match_funcname(opt
, gs
, bol
, eol
)) {
1498 show_line(opt
, bol
, eol
, gs
->name
, lno
, '=');
1504 static int is_empty_line(const char *bol
, const char *eol
);
1506 static void show_pre_context(struct grep_opt
*opt
, struct grep_source
*gs
,
1507 char *bol
, char *end
, unsigned lno
)
1509 unsigned cur
= lno
, from
= 1, funcname_lno
= 0, orig_from
;
1510 int funcname_needed
= !!opt
->funcname
, comment_needed
= 0;
1512 if (opt
->pre_context
< lno
)
1513 from
= lno
- opt
->pre_context
;
1514 if (from
<= opt
->last_shown
)
1515 from
= opt
->last_shown
+ 1;
1517 if (opt
->funcbody
) {
1518 if (match_funcname(opt
, gs
, bol
, end
))
1521 funcname_needed
= 1;
1522 from
= opt
->last_shown
+ 1;
1526 while (bol
> gs
->buf
&& cur
> from
) {
1527 char *next_bol
= bol
;
1530 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1533 if (comment_needed
&& (is_empty_line(bol
, eol
) ||
1534 match_funcname(opt
, gs
, bol
, eol
))) {
1543 if (funcname_needed
&& match_funcname(opt
, gs
, bol
, eol
)) {
1545 funcname_needed
= 0;
1553 /* We need to look even further back to find a function signature. */
1554 if (opt
->funcname
&& funcname_needed
)
1555 show_funcname_line(opt
, gs
, bol
, cur
);
1559 char *eol
= bol
, sign
= (cur
== funcname_lno
) ? '=' : '-';
1561 while (*eol
!= '\n')
1563 show_line(opt
, bol
, eol
, gs
->name
, cur
, sign
);
1569 static int should_lookahead(struct grep_opt
*opt
)
1574 return 0; /* punt for too complex stuff */
1577 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1578 if (p
->token
!= GREP_PATTERN
)
1579 return 0; /* punt for "header only" and stuff */
1584 static int look_ahead(struct grep_opt
*opt
,
1585 unsigned long *left_p
,
1589 unsigned lno
= *lno_p
;
1592 char *sp
, *last_bol
;
1593 regoff_t earliest
= -1;
1595 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1599 hit
= patmatch(p
, bol
, bol
+ *left_p
, &m
, 0);
1600 if (!hit
|| m
.rm_so
< 0 || m
.rm_eo
< 0)
1602 if (earliest
< 0 || m
.rm_so
< earliest
)
1607 *bol_p
= bol
+ *left_p
;
1611 for (sp
= bol
+ earliest
; bol
< sp
&& sp
[-1] != '\n'; sp
--)
1612 ; /* find the beginning of the line */
1615 for (sp
= bol
; sp
< last_bol
; sp
++) {
1619 *left_p
-= last_bol
- bol
;
1625 static int fill_textconv_grep(struct userdiff_driver
*driver
,
1626 struct grep_source
*gs
)
1628 struct diff_filespec
*df
;
1632 if (!driver
|| !driver
->textconv
)
1633 return grep_source_load(gs
);
1636 * The textconv interface is intimately tied to diff_filespecs, so we
1637 * have to pretend to be one. If we could unify the grep_source
1638 * and diff_filespec structs, this mess could just go away.
1640 df
= alloc_filespec(gs
->path
);
1642 case GREP_SOURCE_OID
:
1643 fill_filespec(df
, gs
->identifier
, 1, 0100644);
1645 case GREP_SOURCE_FILE
:
1646 fill_filespec(df
, &null_oid
, 0, 0100644);
1649 BUG("attempt to textconv something without a path?");
1653 * fill_textconv is not remotely thread-safe; it may load objects
1654 * behind the scenes, and it modifies the global diff tempfile
1658 size
= fill_textconv(driver
, df
, &buf
);
1663 * The normal fill_textconv usage by the diff machinery would just keep
1664 * the textconv'd buf separate from the diff_filespec. But much of the
1665 * grep code passes around a grep_source and assumes that its "buf"
1666 * pointer is the beginning of the thing we are searching. So let's
1667 * install our textconv'd version into the grep_source, taking care not
1668 * to leak any existing buffer.
1670 grep_source_clear_data(gs
);
1677 static int is_empty_line(const char *bol
, const char *eol
)
1679 while (bol
< eol
&& isspace(*bol
))
1684 static int grep_source_1(struct grep_opt
*opt
, struct grep_source
*gs
, int collect_hits
)
1687 char *peek_bol
= NULL
;
1690 unsigned last_hit
= 0;
1691 int binary_match_only
= 0;
1693 int try_lookahead
= 0;
1694 int show_function
= 0;
1695 struct userdiff_driver
*textconv
= NULL
;
1696 enum grep_context ctx
= GREP_CONTEXT_HEAD
;
1700 opt
->output
= std_output
;
1702 if (opt
->pre_context
|| opt
->post_context
|| opt
->file_break
||
1704 /* Show hunk marks, except for the first file. */
1705 if (opt
->last_shown
)
1706 opt
->show_hunk_mark
= 1;
1708 * If we're using threads then we can't easily identify
1709 * the first file. Always put hunk marks in that case
1710 * and skip the very first one later in work_done().
1712 if (opt
->output
!= std_output
)
1713 opt
->show_hunk_mark
= 1;
1715 opt
->last_shown
= 0;
1717 if (opt
->allow_textconv
) {
1718 grep_source_load_driver(gs
);
1720 * We might set up the shared textconv cache data here, which
1721 * is not thread-safe.
1724 textconv
= userdiff_get_textconv(gs
->driver
);
1729 * We know the result of a textconv is text, so we only have to care
1730 * about binary handling if we are not using it.
1733 switch (opt
->binary
) {
1734 case GREP_BINARY_DEFAULT
:
1735 if (grep_source_is_binary(gs
))
1736 binary_match_only
= 1;
1738 case GREP_BINARY_NOMATCH
:
1739 if (grep_source_is_binary(gs
))
1740 return 0; /* Assume unmatch */
1742 case GREP_BINARY_TEXT
:
1745 BUG("unknown binary handling mode");
1749 memset(&xecfg
, 0, sizeof(xecfg
));
1752 try_lookahead
= should_lookahead(opt
);
1754 if (fill_textconv_grep(textconv
, gs
) < 0)
1764 * look_ahead() skips quickly to the line that possibly
1765 * has the next hit; don't call it if we need to do
1766 * something more than just skipping the current line
1767 * in response to an unmatch for the current line. E.g.
1768 * inside a post-context window, we will show the current
1769 * line as a context around the previous hit when it
1774 && (show_function
||
1775 lno
<= last_hit
+ opt
->post_context
))
1776 && look_ahead(opt
, &left
, &lno
, &bol
))
1778 eol
= end_of_line(bol
, &left
);
1782 if ((ctx
== GREP_CONTEXT_HEAD
) && (eol
== bol
))
1783 ctx
= GREP_CONTEXT_BODY
;
1785 hit
= match_line(opt
, bol
, eol
, ctx
, collect_hits
);
1791 /* "grep -v -e foo -e bla" should list lines
1792 * that do not have either, so inversion should
1797 if (opt
->unmatch_name_only
) {
1804 if (opt
->status_only
)
1806 if (opt
->name_only
) {
1807 show_name(opt
, gs
->name
);
1812 if (binary_match_only
) {
1813 opt
->output(opt
, "Binary file ", 12);
1814 output_color(opt
, gs
->name
, strlen(gs
->name
),
1815 opt
->colors
[GREP_COLOR_FILENAME
]);
1816 opt
->output(opt
, " matches\n", 9);
1819 /* Hit at this line. If we haven't shown the
1820 * pre-context lines, we would need to show them.
1822 if (opt
->pre_context
|| opt
->funcbody
)
1823 show_pre_context(opt
, gs
, bol
, eol
, lno
);
1824 else if (opt
->funcname
)
1825 show_funcname_line(opt
, gs
, bol
, lno
);
1826 show_line(opt
, bol
, eol
, gs
->name
, lno
, ':');
1832 if (show_function
&& (!peek_bol
|| peek_bol
< bol
)) {
1833 unsigned long peek_left
= left
;
1834 char *peek_eol
= eol
;
1837 * Trailing empty lines are not interesting.
1838 * Peek past them to see if they belong to the
1839 * body of the current function.
1842 while (is_empty_line(peek_bol
, peek_eol
)) {
1843 peek_bol
= peek_eol
+ 1;
1844 peek_eol
= end_of_line(peek_bol
, &peek_left
);
1847 if (match_funcname(opt
, gs
, peek_bol
, peek_eol
))
1850 if (show_function
||
1851 (last_hit
&& lno
<= last_hit
+ opt
->post_context
)) {
1852 /* If the last hit is within the post context,
1853 * we need to show this line.
1855 show_line(opt
, bol
, eol
, gs
->name
, lno
, '-');
1869 if (opt
->status_only
)
1870 return opt
->unmatch_name_only
;
1871 if (opt
->unmatch_name_only
) {
1872 /* We did not see any hit, so we want to show this */
1873 show_name(opt
, gs
->name
);
1877 xdiff_clear_find_func(&xecfg
);
1881 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1882 * which feels mostly useless but sometimes useful. Maybe
1883 * make it another option? For now suppress them.
1885 if (opt
->count
&& count
) {
1887 if (opt
->pathname
) {
1888 output_color(opt
, gs
->name
, strlen(gs
->name
),
1889 opt
->colors
[GREP_COLOR_FILENAME
]);
1890 output_sep(opt
, ':');
1892 xsnprintf(buf
, sizeof(buf
), "%u\n", count
);
1893 opt
->output(opt
, buf
, strlen(buf
));
1899 static void clr_hit_marker(struct grep_expr
*x
)
1901 /* All-hit markers are meaningful only at the very top level
1906 if (x
->node
!= GREP_NODE_OR
)
1908 x
->u
.binary
.left
->hit
= 0;
1909 x
= x
->u
.binary
.right
;
1913 static int chk_hit_marker(struct grep_expr
*x
)
1915 /* Top level nodes have hit markers. See if they all are hits */
1917 if (x
->node
!= GREP_NODE_OR
)
1919 if (!x
->u
.binary
.left
->hit
)
1921 x
= x
->u
.binary
.right
;
1925 int grep_source(struct grep_opt
*opt
, struct grep_source
*gs
)
1928 * we do not have to do the two-pass grep when we do not check
1929 * buffer-wide "all-match".
1931 if (!opt
->all_match
)
1932 return grep_source_1(opt
, gs
, 0);
1934 /* Otherwise the toplevel "or" terms hit a bit differently.
1935 * We first clear hit markers from them.
1937 clr_hit_marker(opt
->pattern_expression
);
1938 grep_source_1(opt
, gs
, 1);
1940 if (!chk_hit_marker(opt
->pattern_expression
))
1943 return grep_source_1(opt
, gs
, 0);
1946 int grep_buffer(struct grep_opt
*opt
, char *buf
, unsigned long size
)
1948 struct grep_source gs
;
1951 grep_source_init(&gs
, GREP_SOURCE_BUF
, NULL
, NULL
, NULL
);
1955 r
= grep_source(opt
, &gs
);
1957 grep_source_clear(&gs
);
1961 void grep_source_init(struct grep_source
*gs
, enum grep_source_type type
,
1962 const char *name
, const char *path
,
1963 const void *identifier
)
1966 gs
->name
= xstrdup_or_null(name
);
1967 gs
->path
= xstrdup_or_null(path
);
1973 case GREP_SOURCE_FILE
:
1974 gs
->identifier
= xstrdup(identifier
);
1976 case GREP_SOURCE_OID
:
1977 gs
->identifier
= oiddup(identifier
);
1979 case GREP_SOURCE_BUF
:
1980 gs
->identifier
= NULL
;
1985 void grep_source_clear(struct grep_source
*gs
)
1987 FREE_AND_NULL(gs
->name
);
1988 FREE_AND_NULL(gs
->path
);
1989 FREE_AND_NULL(gs
->identifier
);
1990 grep_source_clear_data(gs
);
1993 void grep_source_clear_data(struct grep_source
*gs
)
1996 case GREP_SOURCE_FILE
:
1997 case GREP_SOURCE_OID
:
1998 FREE_AND_NULL(gs
->buf
);
2001 case GREP_SOURCE_BUF
:
2002 /* leave user-provided buf intact */
2007 static int grep_source_load_oid(struct grep_source
*gs
)
2009 enum object_type type
;
2012 gs
->buf
= read_object_file(gs
->identifier
, &type
, &gs
->size
);
2016 return error(_("'%s': unable to read %s"),
2018 oid_to_hex(gs
->identifier
));
2022 static int grep_source_load_file(struct grep_source
*gs
)
2024 const char *filename
= gs
->identifier
;
2030 if (lstat(filename
, &st
) < 0) {
2032 if (errno
!= ENOENT
)
2033 error_errno(_("failed to stat '%s'"), filename
);
2036 if (!S_ISREG(st
.st_mode
))
2038 size
= xsize_t(st
.st_size
);
2039 i
= open(filename
, O_RDONLY
);
2042 data
= xmallocz(size
);
2043 if (st
.st_size
!= read_in_full(i
, data
, size
)) {
2044 error_errno(_("'%s': short read"), filename
);
2056 static int grep_source_load(struct grep_source
*gs
)
2062 case GREP_SOURCE_FILE
:
2063 return grep_source_load_file(gs
);
2064 case GREP_SOURCE_OID
:
2065 return grep_source_load_oid(gs
);
2066 case GREP_SOURCE_BUF
:
2067 return gs
->buf
? 0 : -1;
2069 BUG("invalid grep_source type to load");
2072 void grep_source_load_driver(struct grep_source
*gs
)
2079 gs
->driver
= userdiff_find_by_path(gs
->path
);
2081 gs
->driver
= userdiff_find_by_name("default");
2085 static int grep_source_is_binary(struct grep_source
*gs
)
2087 grep_source_load_driver(gs
);
2088 if (gs
->driver
->binary
!= -1)
2089 return gs
->driver
->binary
;
2091 if (!grep_source_load(gs
))
2092 return buffer_is_binary(gs
->buf
, gs
->size
);