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
,
15 struct index_state
*istate
);
17 static struct grep_opt grep_defaults
;
19 static const char *color_grep_slots
[] = {
20 [GREP_COLOR_CONTEXT
] = "context",
21 [GREP_COLOR_FILENAME
] = "filename",
22 [GREP_COLOR_FUNCTION
] = "function",
23 [GREP_COLOR_LINENO
] = "lineNumber",
24 [GREP_COLOR_COLUMNNO
] = "column",
25 [GREP_COLOR_MATCH_CONTEXT
] = "matchContext",
26 [GREP_COLOR_MATCH_SELECTED
] = "matchSelected",
27 [GREP_COLOR_SELECTED
] = "selected",
28 [GREP_COLOR_SEP
] = "separator",
31 static void std_output(struct grep_opt
*opt
, const void *buf
, size_t size
)
33 fwrite(buf
, size
, 1, stdout
);
36 static void color_set(char *dst
, const char *color_bytes
)
38 xsnprintf(dst
, COLOR_MAXLEN
, "%s", color_bytes
);
42 * Initialize the grep_defaults template with hardcoded defaults.
43 * We could let the compiler do this, but without C99 initializers
44 * the code gets unwieldy and unreadable, so...
46 void init_grep_defaults(struct repository
*repo
)
48 struct grep_opt
*opt
= &grep_defaults
;
55 memset(opt
, 0, sizeof(*opt
));
60 opt
->pattern_type_option
= GREP_PATTERN_TYPE_UNSPECIFIED
;
61 color_set(opt
->colors
[GREP_COLOR_CONTEXT
], "");
62 color_set(opt
->colors
[GREP_COLOR_FILENAME
], "");
63 color_set(opt
->colors
[GREP_COLOR_FUNCTION
], "");
64 color_set(opt
->colors
[GREP_COLOR_LINENO
], "");
65 color_set(opt
->colors
[GREP_COLOR_COLUMNNO
], "");
66 color_set(opt
->colors
[GREP_COLOR_MATCH_CONTEXT
], GIT_COLOR_BOLD_RED
);
67 color_set(opt
->colors
[GREP_COLOR_MATCH_SELECTED
], GIT_COLOR_BOLD_RED
);
68 color_set(opt
->colors
[GREP_COLOR_SELECTED
], "");
69 color_set(opt
->colors
[GREP_COLOR_SEP
], GIT_COLOR_CYAN
);
70 opt
->only_matching
= 0;
72 opt
->output
= std_output
;
75 static int parse_pattern_type_arg(const char *opt
, const char *arg
)
77 if (!strcmp(arg
, "default"))
78 return GREP_PATTERN_TYPE_UNSPECIFIED
;
79 else if (!strcmp(arg
, "basic"))
80 return GREP_PATTERN_TYPE_BRE
;
81 else if (!strcmp(arg
, "extended"))
82 return GREP_PATTERN_TYPE_ERE
;
83 else if (!strcmp(arg
, "fixed"))
84 return GREP_PATTERN_TYPE_FIXED
;
85 else if (!strcmp(arg
, "perl"))
86 return GREP_PATTERN_TYPE_PCRE
;
87 die("bad %s argument: %s", opt
, arg
);
90 define_list_config_array_extra(color_grep_slots
, {"match"});
93 * Read the configuration file once and store it in
94 * the grep_defaults template.
96 int grep_config(const char *var
, const char *value
, void *cb
)
98 struct grep_opt
*opt
= &grep_defaults
;
101 if (userdiff_config(var
, value
) < 0)
104 if (!strcmp(var
, "grep.extendedregexp")) {
105 opt
->extended_regexp_option
= git_config_bool(var
, value
);
109 if (!strcmp(var
, "grep.patterntype")) {
110 opt
->pattern_type_option
= parse_pattern_type_arg(var
, value
);
114 if (!strcmp(var
, "grep.linenumber")) {
115 opt
->linenum
= git_config_bool(var
, value
);
118 if (!strcmp(var
, "grep.column")) {
119 opt
->columnnum
= git_config_bool(var
, value
);
123 if (!strcmp(var
, "grep.fullname")) {
124 opt
->relative
= !git_config_bool(var
, value
);
128 if (!strcmp(var
, "color.grep"))
129 opt
->color
= git_config_colorbool(var
, value
);
130 if (!strcmp(var
, "color.grep.match")) {
131 if (grep_config("color.grep.matchcontext", value
, cb
) < 0)
133 if (grep_config("color.grep.matchselected", value
, cb
) < 0)
135 } else if (skip_prefix(var
, "color.grep.", &slot
)) {
136 int i
= LOOKUP_CONFIG(color_grep_slots
, slot
);
141 color
= opt
->colors
[i
];
143 return config_error_nonbool(var
);
144 return color_parse(value
, color
);
150 * Initialize one instance of grep_opt and copy the
151 * default values from the template we read the configuration
152 * information in an earlier call to git_config(grep_config).
154 void grep_init(struct grep_opt
*opt
, struct repository
*repo
, const char *prefix
)
156 struct grep_opt
*def
= &grep_defaults
;
159 memset(opt
, 0, sizeof(*opt
));
161 opt
->prefix
= prefix
;
162 opt
->prefix_length
= (prefix
&& *prefix
) ? strlen(prefix
) : 0;
163 opt
->pattern_tail
= &opt
->pattern_list
;
164 opt
->header_tail
= &opt
->header_list
;
166 opt
->only_matching
= def
->only_matching
;
167 opt
->color
= def
->color
;
168 opt
->extended_regexp_option
= def
->extended_regexp_option
;
169 opt
->pattern_type_option
= def
->pattern_type_option
;
170 opt
->linenum
= def
->linenum
;
171 opt
->columnnum
= def
->columnnum
;
172 opt
->max_depth
= def
->max_depth
;
173 opt
->pathname
= def
->pathname
;
174 opt
->relative
= def
->relative
;
175 opt
->output
= def
->output
;
177 for (i
= 0; i
< NR_GREP_COLORS
; i
++)
178 color_set(opt
->colors
[i
], def
->colors
[i
]);
181 static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
184 * When committing to the pattern type by setting the relevant
185 * fields in grep_opt it's generally not necessary to zero out
186 * the fields we're not choosing, since they won't have been
187 * set by anything. The extended_regexp_option field is the
188 * only exception to this.
190 * This is because in the process of parsing grep.patternType
191 * & grep.extendedRegexp we set opt->pattern_type_option and
192 * opt->extended_regexp_option, respectively. We then
193 * internally use opt->extended_regexp_option to see if we're
194 * compiling an ERE. It must be unset if that's not actually
197 if (pattern_type
!= GREP_PATTERN_TYPE_ERE
&&
198 opt
->extended_regexp_option
)
199 opt
->extended_regexp_option
= 0;
201 switch (pattern_type
) {
202 case GREP_PATTERN_TYPE_UNSPECIFIED
:
205 case GREP_PATTERN_TYPE_BRE
:
208 case GREP_PATTERN_TYPE_ERE
:
209 opt
->extended_regexp_option
= 1;
212 case GREP_PATTERN_TYPE_FIXED
:
216 case GREP_PATTERN_TYPE_PCRE
:
221 * It's important that pcre1 always be assigned to
222 * even when there's no USE_LIBPCRE* defined. We still
223 * call the PCRE stub function, it just dies with
224 * "cannot use Perl-compatible regexes[...]".
232 void grep_commit_pattern_type(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
234 if (pattern_type
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
235 grep_set_pattern_type_option(pattern_type
, opt
);
236 else if (opt
->pattern_type_option
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
237 grep_set_pattern_type_option(opt
->pattern_type_option
, opt
);
238 else if (opt
->extended_regexp_option
)
240 * This branch *must* happen after setting from the
241 * opt->pattern_type_option above, we don't want
242 * grep.extendedRegexp to override grep.patternType!
244 grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE
, opt
);
247 static struct grep_pat
*create_grep_pat(const char *pat
, size_t patlen
,
248 const char *origin
, int no
,
249 enum grep_pat_token t
,
250 enum grep_header_field field
)
252 struct grep_pat
*p
= xcalloc(1, sizeof(*p
));
253 p
->pattern
= xmemdupz(pat
, patlen
);
254 p
->patternlen
= patlen
;
262 static void do_append_grep_pat(struct grep_pat
***tail
, struct grep_pat
*p
)
269 case GREP_PATTERN
: /* atom */
270 case GREP_PATTERN_HEAD
:
271 case GREP_PATTERN_BODY
:
273 struct grep_pat
*new_pat
;
275 char *cp
= p
->pattern
+ p
->patternlen
, *nl
= NULL
;
276 while (++len
<= p
->patternlen
) {
277 if (*(--cp
) == '\n') {
284 new_pat
= create_grep_pat(nl
+ 1, len
- 1, p
->origin
,
285 p
->no
, p
->token
, p
->field
);
286 new_pat
->next
= p
->next
;
288 *tail
= &new_pat
->next
;
291 p
->patternlen
-= len
;
299 void append_header_grep_pattern(struct grep_opt
*opt
,
300 enum grep_header_field field
, const char *pat
)
302 struct grep_pat
*p
= create_grep_pat(pat
, strlen(pat
), "header", 0,
303 GREP_PATTERN_HEAD
, field
);
304 if (field
== GREP_HEADER_REFLOG
)
305 opt
->use_reflog_filter
= 1;
306 do_append_grep_pat(&opt
->header_tail
, p
);
309 void append_grep_pattern(struct grep_opt
*opt
, const char *pat
,
310 const char *origin
, int no
, enum grep_pat_token t
)
312 append_grep_pat(opt
, pat
, strlen(pat
), origin
, no
, t
);
315 void append_grep_pat(struct grep_opt
*opt
, const char *pat
, size_t patlen
,
316 const char *origin
, int no
, enum grep_pat_token t
)
318 struct grep_pat
*p
= create_grep_pat(pat
, patlen
, origin
, no
, t
, 0);
319 do_append_grep_pat(&opt
->pattern_tail
, p
);
322 struct grep_opt
*grep_opt_dup(const struct grep_opt
*opt
)
324 struct grep_pat
*pat
;
325 struct grep_opt
*ret
= xmalloc(sizeof(struct grep_opt
));
328 ret
->pattern_list
= NULL
;
329 ret
->pattern_tail
= &ret
->pattern_list
;
331 for(pat
= opt
->pattern_list
; pat
!= NULL
; pat
= pat
->next
)
333 if(pat
->token
== GREP_PATTERN_HEAD
)
334 append_header_grep_pattern(ret
, pat
->field
,
337 append_grep_pat(ret
, pat
->pattern
, pat
->patternlen
,
338 pat
->origin
, pat
->no
, pat
->token
);
344 static NORETURN
void compile_regexp_failed(const struct grep_pat
*p
,
350 xsnprintf(where
, sizeof(where
), "In '%s' at %d, ", p
->origin
, p
->no
);
352 xsnprintf(where
, sizeof(where
), "%s, ", p
->origin
);
356 die("%s'%s': %s", where
, p
->pattern
, error
);
359 static int is_fixed(const char *s
, size_t len
)
363 for (i
= 0; i
< len
; i
++) {
364 if (is_regex_special(s
[i
]))
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 (!opt
->ignore_locale
&& has_non_ascii(p
->pattern
))
380 p
->pcre1_tables
= pcre_maketables();
381 options
|= PCRE_CASELESS
;
383 if (!opt
->ignore_locale
&& 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
);
400 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
401 regmatch_t
*match
, int eflags
)
403 int ovector
[30], ret
, flags
= 0;
405 if (eflags
& REG_NOTBOL
)
406 flags
|= PCRE_NOTBOL
;
408 ret
= pcre_exec(p
->pcre1_regexp
, p
->pcre1_extra_info
, line
,
409 eol
- line
, 0, flags
, ovector
,
410 ARRAY_SIZE(ovector
));
412 if (ret
< 0 && ret
!= PCRE_ERROR_NOMATCH
)
413 die("pcre_exec failed with error code %d", ret
);
416 match
->rm_so
= ovector
[0];
417 match
->rm_eo
= ovector
[1];
423 static void free_pcre1_regexp(struct grep_pat
*p
)
425 pcre_free(p
->pcre1_regexp
);
426 #ifdef GIT_PCRE1_USE_JIT
428 pcre_free_study(p
->pcre1_extra_info
);
431 pcre_free(p
->pcre1_extra_info
);
432 pcre_free((void *)p
->pcre1_tables
);
434 #else /* !USE_LIBPCRE1 */
435 static void compile_pcre1_regexp(struct grep_pat
*p
, const struct grep_opt
*opt
)
437 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
440 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
441 regmatch_t
*match
, int eflags
)
446 static void free_pcre1_regexp(struct grep_pat
*p
)
449 #endif /* !USE_LIBPCRE1 */
452 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
455 PCRE2_UCHAR errbuf
[256];
456 PCRE2_SIZE erroffset
;
457 int options
= PCRE2_MULTILINE
;
458 const uint8_t *character_tables
= NULL
;
465 p
->pcre2_compile_context
= NULL
;
467 if (opt
->ignore_case
) {
468 if (!opt
->ignore_locale
&& has_non_ascii(p
->pattern
)) {
469 character_tables
= pcre2_maketables(NULL
);
470 p
->pcre2_compile_context
= pcre2_compile_context_create(NULL
);
471 pcre2_set_character_tables(p
->pcre2_compile_context
, character_tables
);
473 options
|= PCRE2_CASELESS
;
475 if (!opt
->ignore_locale
&& is_utf8_locale() && has_non_ascii(p
->pattern
))
476 options
|= PCRE2_UTF
;
478 p
->pcre2_pattern
= pcre2_compile((PCRE2_SPTR
)p
->pattern
,
479 p
->patternlen
, options
, &error
, &erroffset
,
480 p
->pcre2_compile_context
);
482 if (p
->pcre2_pattern
) {
483 p
->pcre2_match_data
= pcre2_match_data_create_from_pattern(p
->pcre2_pattern
, NULL
);
484 if (!p
->pcre2_match_data
)
485 die("Couldn't allocate PCRE2 match data");
487 pcre2_get_error_message(error
, errbuf
, sizeof(errbuf
));
488 compile_regexp_failed(p
, (const char *)&errbuf
);
491 pcre2_config(PCRE2_CONFIG_JIT
, &p
->pcre2_jit_on
);
492 if (p
->pcre2_jit_on
) {
493 jitret
= pcre2_jit_compile(p
->pcre2_pattern
, PCRE2_JIT_COMPLETE
);
495 die("Couldn't JIT the PCRE2 pattern '%s', got '%d'\n", p
->pattern
, jitret
);
498 * The pcre2_config(PCRE2_CONFIG_JIT, ...) call just
499 * tells us whether the library itself supports JIT,
500 * but to see whether we're going to be actually using
501 * JIT we need to extract PCRE2_INFO_JITSIZE from the
502 * pattern *after* we do pcre2_jit_compile() above.
504 * This is because if the pattern contains the
505 * (*NO_JIT) verb (see pcre2syntax(3))
506 * pcre2_jit_compile() will exit early with 0. If we
507 * then proceed to call pcre2_jit_match() further down
508 * the line instead of pcre2_match() we'll either
509 * segfault (pre PCRE 10.31) or run into a fatal error
512 patinforet
= pcre2_pattern_info(p
->pcre2_pattern
, PCRE2_INFO_JITSIZE
, &jitsizearg
);
514 BUG("pcre2_pattern_info() failed: %d", patinforet
);
515 if (jitsizearg
== 0) {
522 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
523 regmatch_t
*match
, int eflags
)
527 PCRE2_UCHAR errbuf
[256];
529 if (eflags
& REG_NOTBOL
)
530 flags
|= PCRE2_NOTBOL
;
533 ret
= pcre2_jit_match(p
->pcre2_pattern
, (unsigned char *)line
,
534 eol
- line
, 0, flags
, p
->pcre2_match_data
,
537 ret
= pcre2_match(p
->pcre2_pattern
, (unsigned char *)line
,
538 eol
- line
, 0, flags
, p
->pcre2_match_data
,
541 if (ret
< 0 && ret
!= PCRE2_ERROR_NOMATCH
) {
542 pcre2_get_error_message(ret
, errbuf
, sizeof(errbuf
));
543 die("%s failed with error code %d: %s",
544 (p
->pcre2_jit_on
? "pcre2_jit_match" : "pcre2_match"), ret
,
548 ovector
= pcre2_get_ovector_pointer(p
->pcre2_match_data
);
550 match
->rm_so
= (int)ovector
[0];
551 match
->rm_eo
= (int)ovector
[1];
557 static void free_pcre2_pattern(struct grep_pat
*p
)
559 pcre2_compile_context_free(p
->pcre2_compile_context
);
560 pcre2_code_free(p
->pcre2_pattern
);
561 pcre2_match_data_free(p
->pcre2_match_data
);
563 #else /* !USE_LIBPCRE2 */
564 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
567 * Unreachable until USE_LIBPCRE2 becomes synonymous with
568 * USE_LIBPCRE. See the sibling comment in
569 * grep_set_pattern_type_option().
571 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
574 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
575 regmatch_t
*match
, int eflags
)
580 static void free_pcre2_pattern(struct grep_pat
*p
)
584 static void compile_fixed_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
586 struct strbuf sb
= STRBUF_INIT
;
590 basic_regex_quote_buf(&sb
, p
->pattern
);
591 if (opt
->ignore_case
)
592 regflags
|= REG_ICASE
;
593 err
= regcomp(&p
->regexp
, sb
.buf
, regflags
);
595 fprintf(stderr
, "fixed %s\n", sb
.buf
);
599 regerror(err
, &p
->regexp
, errbuf
, sizeof(errbuf
));
600 compile_regexp_failed(p
, errbuf
);
603 #endif /* !USE_LIBPCRE2 */
605 static void compile_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
608 int regflags
= REG_NEWLINE
;
610 p
->word_regexp
= opt
->word_regexp
;
611 p
->ignore_case
= opt
->ignore_case
;
612 p
->fixed
= opt
->fixed
;
614 if (memchr(p
->pattern
, 0, p
->patternlen
) && !opt
->pcre2
)
615 die(_("given pattern contains NULL byte (via -f <file>). This is only supported with -P under PCRE v2"));
617 p
->is_fixed
= is_fixed(p
->pattern
, p
->patternlen
);
619 if (!p
->fixed
&& !p
->is_fixed
) {
620 const char *no_jit
= "(*NO_JIT)";
621 const int no_jit_len
= strlen(no_jit
);
622 if (starts_with(p
->pattern
, no_jit
) &&
623 is_fixed(p
->pattern
+ no_jit_len
,
624 p
->patternlen
- no_jit_len
))
628 if (p
->fixed
|| p
->is_fixed
) {
632 compile_pcre2_pattern(p
, opt
);
635 * E.g. t7811-grep-open.sh relies on the
636 * pattern being restored.
638 char *old_pattern
= p
->pattern
;
639 size_t old_patternlen
= p
->patternlen
;
640 struct strbuf sb
= STRBUF_INIT
;
643 * There is the PCRE2_LITERAL flag, but it's
644 * only in PCRE v2 10.30 and later. Needing to
645 * ifdef our way around that and dealing with
646 * it + PCRE2_MULTILINE being an error is more
647 * complex than just quoting this ourselves.
649 strbuf_add(&sb
, "\\Q", 2);
650 strbuf_add(&sb
, p
->pattern
, p
->patternlen
);
651 strbuf_add(&sb
, "\\E", 2);
654 p
->patternlen
= sb
.len
;
655 compile_pcre2_pattern(p
, opt
);
656 p
->pattern
= old_pattern
;
657 p
->patternlen
= old_patternlen
;
660 #else /* !USE_LIBPCRE2 */
661 compile_fixed_regexp(p
, opt
);
662 #endif /* !USE_LIBPCRE2 */
667 compile_pcre2_pattern(p
, opt
);
672 compile_pcre1_regexp(p
, opt
);
677 regflags
|= REG_ICASE
;
678 if (opt
->extended_regexp_option
)
679 regflags
|= REG_EXTENDED
;
680 err
= regcomp(&p
->regexp
, p
->pattern
, regflags
);
683 regerror(err
, &p
->regexp
, errbuf
, 1024);
684 compile_regexp_failed(p
, errbuf
);
688 static struct grep_expr
*compile_pattern_or(struct grep_pat
**);
689 static struct grep_expr
*compile_pattern_atom(struct grep_pat
**list
)
698 case GREP_PATTERN
: /* atom */
699 case GREP_PATTERN_HEAD
:
700 case GREP_PATTERN_BODY
:
701 x
= xcalloc(1, sizeof (struct grep_expr
));
702 x
->node
= GREP_NODE_ATOM
;
706 case GREP_OPEN_PAREN
:
708 x
= compile_pattern_or(list
);
709 if (!*list
|| (*list
)->token
!= GREP_CLOSE_PAREN
)
710 die("unmatched parenthesis");
711 *list
= (*list
)->next
;
718 static struct grep_expr
*compile_pattern_not(struct grep_pat
**list
)
729 die("--not not followed by pattern expression");
731 x
= xcalloc(1, sizeof (struct grep_expr
));
732 x
->node
= GREP_NODE_NOT
;
733 x
->u
.unary
= compile_pattern_not(list
);
735 die("--not followed by non pattern expression");
738 return compile_pattern_atom(list
);
742 static struct grep_expr
*compile_pattern_and(struct grep_pat
**list
)
745 struct grep_expr
*x
, *y
, *z
;
747 x
= compile_pattern_not(list
);
749 if (p
&& p
->token
== GREP_AND
) {
751 die("--and not followed by pattern expression");
753 y
= compile_pattern_and(list
);
755 die("--and not followed by pattern expression");
756 z
= xcalloc(1, sizeof (struct grep_expr
));
757 z
->node
= GREP_NODE_AND
;
758 z
->u
.binary
.left
= x
;
759 z
->u
.binary
.right
= y
;
765 static struct grep_expr
*compile_pattern_or(struct grep_pat
**list
)
768 struct grep_expr
*x
, *y
, *z
;
770 x
= compile_pattern_and(list
);
772 if (x
&& p
&& p
->token
!= GREP_CLOSE_PAREN
) {
773 y
= compile_pattern_or(list
);
775 die("not a pattern expression %s", p
->pattern
);
776 z
= xcalloc(1, sizeof (struct grep_expr
));
777 z
->node
= GREP_NODE_OR
;
778 z
->u
.binary
.left
= x
;
779 z
->u
.binary
.right
= y
;
785 static struct grep_expr
*compile_pattern_expr(struct grep_pat
**list
)
787 return compile_pattern_or(list
);
790 static void indent(int in
)
796 static void dump_grep_pat(struct grep_pat
*p
)
799 case GREP_AND
: fprintf(stderr
, "*and*"); break;
800 case GREP_OPEN_PAREN
: fprintf(stderr
, "*(*"); break;
801 case GREP_CLOSE_PAREN
: fprintf(stderr
, "*)*"); break;
802 case GREP_NOT
: fprintf(stderr
, "*not*"); break;
803 case GREP_OR
: fprintf(stderr
, "*or*"); break;
805 case GREP_PATTERN
: fprintf(stderr
, "pattern"); break;
806 case GREP_PATTERN_HEAD
: fprintf(stderr
, "pattern_head"); break;
807 case GREP_PATTERN_BODY
: fprintf(stderr
, "pattern_body"); break;
812 case GREP_PATTERN_HEAD
:
813 fprintf(stderr
, "<head %d>", p
->field
); break;
814 case GREP_PATTERN_BODY
:
815 fprintf(stderr
, "<body>"); break;
819 case GREP_PATTERN_HEAD
:
820 case GREP_PATTERN_BODY
:
822 fprintf(stderr
, "%.*s", (int)p
->patternlen
, p
->pattern
);
828 static void dump_grep_expression_1(struct grep_expr
*x
, int in
)
833 fprintf(stderr
, "true\n");
836 dump_grep_pat(x
->u
.atom
);
839 fprintf(stderr
, "(not\n");
840 dump_grep_expression_1(x
->u
.unary
, in
+1);
842 fprintf(stderr
, ")\n");
845 fprintf(stderr
, "(and\n");
846 dump_grep_expression_1(x
->u
.binary
.left
, in
+1);
847 dump_grep_expression_1(x
->u
.binary
.right
, in
+1);
849 fprintf(stderr
, ")\n");
852 fprintf(stderr
, "(or\n");
853 dump_grep_expression_1(x
->u
.binary
.left
, in
+1);
854 dump_grep_expression_1(x
->u
.binary
.right
, in
+1);
856 fprintf(stderr
, ")\n");
861 static void dump_grep_expression(struct grep_opt
*opt
)
863 struct grep_expr
*x
= opt
->pattern_expression
;
866 fprintf(stderr
, "[all-match]\n");
867 dump_grep_expression_1(x
, 0);
871 static struct grep_expr
*grep_true_expr(void)
873 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
874 z
->node
= GREP_NODE_TRUE
;
878 static struct grep_expr
*grep_or_expr(struct grep_expr
*left
, struct grep_expr
*right
)
880 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
881 z
->node
= GREP_NODE_OR
;
882 z
->u
.binary
.left
= left
;
883 z
->u
.binary
.right
= right
;
887 static struct grep_expr
*prep_header_patterns(struct grep_opt
*opt
)
890 struct grep_expr
*header_expr
;
891 struct grep_expr
*(header_group
[GREP_HEADER_FIELD_MAX
]);
892 enum grep_header_field fld
;
894 if (!opt
->header_list
)
897 for (p
= opt
->header_list
; p
; p
= p
->next
) {
898 if (p
->token
!= GREP_PATTERN_HEAD
)
899 BUG("a non-header pattern in grep header list.");
900 if (p
->field
< GREP_HEADER_FIELD_MIN
||
901 GREP_HEADER_FIELD_MAX
<= p
->field
)
902 BUG("unknown header field %d", p
->field
);
903 compile_regexp(p
, opt
);
906 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++)
907 header_group
[fld
] = NULL
;
909 for (p
= opt
->header_list
; p
; p
= p
->next
) {
911 struct grep_pat
*pp
= p
;
913 h
= compile_pattern_atom(&pp
);
914 if (!h
|| pp
!= p
->next
)
915 BUG("malformed header expr");
916 if (!header_group
[p
->field
]) {
917 header_group
[p
->field
] = h
;
920 header_group
[p
->field
] = grep_or_expr(h
, header_group
[p
->field
]);
925 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++) {
926 if (!header_group
[fld
])
929 header_expr
= grep_true_expr();
930 header_expr
= grep_or_expr(header_group
[fld
], header_expr
);
935 static struct grep_expr
*grep_splice_or(struct grep_expr
*x
, struct grep_expr
*y
)
937 struct grep_expr
*z
= x
;
940 assert(x
->node
== GREP_NODE_OR
);
941 if (x
->u
.binary
.right
&&
942 x
->u
.binary
.right
->node
== GREP_NODE_TRUE
) {
943 x
->u
.binary
.right
= y
;
946 x
= x
->u
.binary
.right
;
951 static void compile_grep_patterns_real(struct grep_opt
*opt
)
954 struct grep_expr
*header_expr
= prep_header_patterns(opt
);
956 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
958 case GREP_PATTERN
: /* atom */
959 case GREP_PATTERN_HEAD
:
960 case GREP_PATTERN_BODY
:
961 compile_regexp(p
, opt
);
969 if (opt
->all_match
|| header_expr
)
971 else if (!opt
->extended
&& !opt
->debug
)
974 p
= opt
->pattern_list
;
976 opt
->pattern_expression
= compile_pattern_expr(&p
);
978 die("incomplete pattern expression: %s", p
->pattern
);
983 if (!opt
->pattern_expression
)
984 opt
->pattern_expression
= header_expr
;
985 else if (opt
->all_match
)
986 opt
->pattern_expression
= grep_splice_or(header_expr
,
987 opt
->pattern_expression
);
989 opt
->pattern_expression
= grep_or_expr(opt
->pattern_expression
,
994 void compile_grep_patterns(struct grep_opt
*opt
)
996 compile_grep_patterns_real(opt
);
998 dump_grep_expression(opt
);
1001 static void free_pattern_expr(struct grep_expr
*x
)
1004 case GREP_NODE_TRUE
:
1005 case GREP_NODE_ATOM
:
1008 free_pattern_expr(x
->u
.unary
);
1012 free_pattern_expr(x
->u
.binary
.left
);
1013 free_pattern_expr(x
->u
.binary
.right
);
1019 void free_grep_patterns(struct grep_opt
*opt
)
1021 struct grep_pat
*p
, *n
;
1023 for (p
= opt
->pattern_list
; p
; p
= n
) {
1026 case GREP_PATTERN
: /* atom */
1027 case GREP_PATTERN_HEAD
:
1028 case GREP_PATTERN_BODY
:
1029 if (p
->pcre1_regexp
)
1030 free_pcre1_regexp(p
);
1031 else if (p
->pcre2_pattern
)
1032 free_pcre2_pattern(p
);
1034 regfree(&p
->regexp
);
1045 free_pattern_expr(opt
->pattern_expression
);
1048 static char *end_of_line(char *cp
, unsigned long *left
)
1050 unsigned long l
= *left
;
1051 while (l
&& *cp
!= '\n') {
1059 static int word_char(char ch
)
1061 return isalnum(ch
) || ch
== '_';
1064 static void output_color(struct grep_opt
*opt
, const void *data
, size_t size
,
1067 if (want_color(opt
->color
) && color
&& color
[0]) {
1068 opt
->output(opt
, color
, strlen(color
));
1069 opt
->output(opt
, data
, size
);
1070 opt
->output(opt
, GIT_COLOR_RESET
, strlen(GIT_COLOR_RESET
));
1072 opt
->output(opt
, data
, size
);
1075 static void output_sep(struct grep_opt
*opt
, char sign
)
1077 if (opt
->null_following_name
)
1078 opt
->output(opt
, "\0", 1);
1080 output_color(opt
, &sign
, 1, opt
->colors
[GREP_COLOR_SEP
]);
1083 static void show_name(struct grep_opt
*opt
, const char *name
)
1085 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1086 opt
->output(opt
, opt
->null_following_name
? "\0" : "\n", 1);
1089 static int patmatch(struct grep_pat
*p
, char *line
, char *eol
,
1090 regmatch_t
*match
, int eflags
)
1094 if (p
->pcre1_regexp
)
1095 hit
= !pcre1match(p
, line
, eol
, match
, eflags
);
1096 else if (p
->pcre2_pattern
)
1097 hit
= !pcre2match(p
, line
, eol
, match
, eflags
);
1099 hit
= !regexec_buf(&p
->regexp
, line
, eol
- line
, 1, match
,
1105 static int strip_timestamp(char *bol
, char **eol_p
)
1110 while (bol
< --eol
) {
1124 } header_field
[] = {
1126 { "committer ", 10 },
1130 static int match_one_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1131 enum grep_context ctx
,
1132 regmatch_t
*pmatch
, int eflags
)
1136 const char *start
= bol
;
1138 if ((p
->token
!= GREP_PATTERN
) &&
1139 ((p
->token
== GREP_PATTERN_HEAD
) != (ctx
== GREP_CONTEXT_HEAD
)))
1142 if (p
->token
== GREP_PATTERN_HEAD
) {
1145 assert(p
->field
< ARRAY_SIZE(header_field
));
1146 field
= header_field
[p
->field
].field
;
1147 len
= header_field
[p
->field
].len
;
1148 if (strncmp(bol
, field
, len
))
1152 case GREP_HEADER_AUTHOR
:
1153 case GREP_HEADER_COMMITTER
:
1154 saved_ch
= strip_timestamp(bol
, &eol
);
1162 hit
= patmatch(p
, bol
, eol
, pmatch
, eflags
);
1164 if (hit
&& p
->word_regexp
) {
1165 if ((pmatch
[0].rm_so
< 0) ||
1166 (eol
- bol
) < pmatch
[0].rm_so
||
1167 (pmatch
[0].rm_eo
< 0) ||
1168 (eol
- bol
) < pmatch
[0].rm_eo
)
1169 die("regexp returned nonsense");
1171 /* Match beginning must be either beginning of the
1172 * line, or at word boundary (i.e. the last char must
1173 * not be a word char). Similarly, match end must be
1174 * either end of the line, or at word boundary
1175 * (i.e. the next char must not be a word char).
1177 if ( ((pmatch
[0].rm_so
== 0) ||
1178 !word_char(bol
[pmatch
[0].rm_so
-1])) &&
1179 ((pmatch
[0].rm_eo
== (eol
-bol
)) ||
1180 !word_char(bol
[pmatch
[0].rm_eo
])) )
1185 /* Words consist of at least one character. */
1186 if (pmatch
->rm_so
== pmatch
->rm_eo
)
1189 if (!hit
&& pmatch
[0].rm_so
+ bol
+ 1 < eol
) {
1190 /* There could be more than one match on the
1191 * line, and the first match might not be
1192 * strict word match. But later ones could be!
1193 * Forward to the next possible start, i.e. the
1194 * next position following a non-word char.
1196 bol
= pmatch
[0].rm_so
+ bol
+ 1;
1197 while (word_char(bol
[-1]) && bol
< eol
)
1199 eflags
|= REG_NOTBOL
;
1204 if (p
->token
== GREP_PATTERN_HEAD
&& saved_ch
)
1207 pmatch
[0].rm_so
+= bol
- start
;
1208 pmatch
[0].rm_eo
+= bol
- start
;
1213 static int match_expr_eval(struct grep_opt
*opt
, struct grep_expr
*x
, char *bol
,
1214 char *eol
, enum grep_context ctx
, ssize_t
*col
,
1215 ssize_t
*icol
, int collect_hits
)
1220 die("Not a valid grep expression");
1222 case GREP_NODE_TRUE
:
1225 case GREP_NODE_ATOM
:
1228 h
= match_one_pattern(x
->u
.atom
, bol
, eol
, ctx
,
1230 if (h
&& (*col
< 0 || tmp
.rm_so
< *col
))
1236 * Upon visiting a GREP_NODE_NOT, col and icol become swapped.
1238 h
= !match_expr_eval(opt
, x
->u
.unary
, bol
, eol
, ctx
, icol
, col
,
1242 h
= match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
, ctx
, col
,
1244 if (h
|| opt
->columnnum
) {
1246 * Don't short-circuit AND when given --column, since a
1247 * NOT earlier in the tree may turn this into an OR. In
1248 * this case, see the below comment.
1250 h
&= match_expr_eval(opt
, x
->u
.binary
.right
, bol
, eol
,
1255 if (!(collect_hits
|| opt
->columnnum
)) {
1257 * Don't short-circuit OR when given --column (or
1258 * collecting hits) to ensure we don't skip a later
1259 * child that would produce an earlier match.
1261 return (match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
,
1262 ctx
, col
, icol
, 0) ||
1263 match_expr_eval(opt
, x
->u
.binary
.right
, bol
,
1264 eol
, ctx
, col
, icol
, 0));
1266 h
= match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
, ctx
, col
,
1269 x
->u
.binary
.left
->hit
|= h
;
1270 h
|= match_expr_eval(opt
, x
->u
.binary
.right
, bol
, eol
, ctx
, col
,
1271 icol
, collect_hits
);
1274 die("Unexpected node type (internal error) %d", x
->node
);
1281 static int match_expr(struct grep_opt
*opt
, char *bol
, char *eol
,
1282 enum grep_context ctx
, ssize_t
*col
,
1283 ssize_t
*icol
, int collect_hits
)
1285 struct grep_expr
*x
= opt
->pattern_expression
;
1286 return match_expr_eval(opt
, x
, bol
, eol
, ctx
, col
, icol
, collect_hits
);
1289 static int match_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1290 ssize_t
*col
, ssize_t
*icol
,
1291 enum grep_context ctx
, int collect_hits
)
1297 return match_expr(opt
, bol
, eol
, ctx
, col
, icol
,
1300 /* we do not call with collect_hits without being extended */
1301 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1303 if (match_one_pattern(p
, bol
, eol
, ctx
, &tmp
, 0)) {
1305 if (!opt
->columnnum
) {
1307 * Without --column, any single match on a line
1308 * is enough to know that it needs to be
1309 * printed. With --column, scan _all_ patterns
1310 * to find the earliest.
1314 if (*col
< 0 || tmp
.rm_so
< *col
)
1321 static int match_next_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1322 enum grep_context ctx
,
1323 regmatch_t
*pmatch
, int eflags
)
1327 if (!match_one_pattern(p
, bol
, eol
, ctx
, &match
, eflags
))
1329 if (match
.rm_so
< 0 || match
.rm_eo
< 0)
1331 if (pmatch
->rm_so
>= 0 && pmatch
->rm_eo
>= 0) {
1332 if (match
.rm_so
> pmatch
->rm_so
)
1334 if (match
.rm_so
== pmatch
->rm_so
&& match
.rm_eo
< pmatch
->rm_eo
)
1337 pmatch
->rm_so
= match
.rm_so
;
1338 pmatch
->rm_eo
= match
.rm_eo
;
1342 static int next_match(struct grep_opt
*opt
, char *bol
, char *eol
,
1343 enum grep_context ctx
, regmatch_t
*pmatch
, int eflags
)
1348 pmatch
->rm_so
= pmatch
->rm_eo
= -1;
1350 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1352 case GREP_PATTERN
: /* atom */
1353 case GREP_PATTERN_HEAD
:
1354 case GREP_PATTERN_BODY
:
1355 hit
|= match_next_pattern(p
, bol
, eol
, ctx
,
1366 static void show_line_header(struct grep_opt
*opt
, const char *name
,
1367 unsigned lno
, ssize_t cno
, char sign
)
1369 if (opt
->heading
&& opt
->last_shown
== 0) {
1370 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1371 opt
->output(opt
, "\n", 1);
1373 opt
->last_shown
= lno
;
1375 if (!opt
->heading
&& opt
->pathname
) {
1376 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1377 output_sep(opt
, sign
);
1381 xsnprintf(buf
, sizeof(buf
), "%d", lno
);
1382 output_color(opt
, buf
, strlen(buf
), opt
->colors
[GREP_COLOR_LINENO
]);
1383 output_sep(opt
, sign
);
1386 * Treat 'cno' as the 1-indexed offset from the start of a non-context
1387 * line to its first match. Otherwise, 'cno' is 0 indicating that we are
1388 * being called with a context line.
1390 if (opt
->columnnum
&& cno
) {
1392 xsnprintf(buf
, sizeof(buf
), "%"PRIuMAX
, (uintmax_t)cno
);
1393 output_color(opt
, buf
, strlen(buf
), opt
->colors
[GREP_COLOR_COLUMNNO
]);
1394 output_sep(opt
, sign
);
1398 static void show_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1399 const char *name
, unsigned lno
, ssize_t cno
, char sign
)
1401 int rest
= eol
- bol
;
1402 const char *match_color
= NULL
;
1403 const char *line_color
= NULL
;
1405 if (opt
->file_break
&& opt
->last_shown
== 0) {
1406 if (opt
->show_hunk_mark
)
1407 opt
->output(opt
, "\n", 1);
1408 } else if (opt
->pre_context
|| opt
->post_context
|| opt
->funcbody
) {
1409 if (opt
->last_shown
== 0) {
1410 if (opt
->show_hunk_mark
) {
1411 output_color(opt
, "--", 2, opt
->colors
[GREP_COLOR_SEP
]);
1412 opt
->output(opt
, "\n", 1);
1414 } else if (lno
> opt
->last_shown
+ 1) {
1415 output_color(opt
, "--", 2, opt
->colors
[GREP_COLOR_SEP
]);
1416 opt
->output(opt
, "\n", 1);
1419 if (!opt
->only_matching
) {
1421 * In case the line we're being called with contains more than
1422 * one match, leave printing each header to the loop below.
1424 show_line_header(opt
, name
, lno
, cno
, sign
);
1426 if (opt
->color
|| opt
->only_matching
) {
1428 enum grep_context ctx
= GREP_CONTEXT_BODY
;
1434 match_color
= opt
->colors
[GREP_COLOR_MATCH_SELECTED
];
1436 match_color
= opt
->colors
[GREP_COLOR_MATCH_CONTEXT
];
1438 line_color
= opt
->colors
[GREP_COLOR_SELECTED
];
1439 else if (sign
== '-')
1440 line_color
= opt
->colors
[GREP_COLOR_CONTEXT
];
1441 else if (sign
== '=')
1442 line_color
= opt
->colors
[GREP_COLOR_FUNCTION
];
1445 while (next_match(opt
, bol
, eol
, ctx
, &match
, eflags
)) {
1446 if (match
.rm_so
== match
.rm_eo
)
1449 if (opt
->only_matching
)
1450 show_line_header(opt
, name
, lno
, cno
, sign
);
1452 output_color(opt
, bol
, match
.rm_so
, line_color
);
1453 output_color(opt
, bol
+ match
.rm_so
,
1454 match
.rm_eo
- match
.rm_so
, match_color
);
1455 if (opt
->only_matching
)
1456 opt
->output(opt
, "\n", 1);
1459 rest
-= match
.rm_eo
;
1460 eflags
= REG_NOTBOL
;
1464 if (!opt
->only_matching
) {
1465 output_color(opt
, bol
, rest
, line_color
);
1466 opt
->output(opt
, "\n", 1);
1473 * This lock protects access to the gitattributes machinery, which is
1476 pthread_mutex_t grep_attr_mutex
;
1478 static inline void grep_attr_lock(void)
1481 pthread_mutex_lock(&grep_attr_mutex
);
1484 static inline void grep_attr_unlock(void)
1487 pthread_mutex_unlock(&grep_attr_mutex
);
1491 * Same as git_attr_mutex, but protecting the thread-unsafe object db access.
1493 pthread_mutex_t grep_read_mutex
;
1495 static int match_funcname(struct grep_opt
*opt
, struct grep_source
*gs
, char *bol
, char *eol
)
1497 xdemitconf_t
*xecfg
= opt
->priv
;
1498 if (xecfg
&& !xecfg
->find_func
) {
1499 grep_source_load_driver(gs
, opt
->repo
->index
);
1500 if (gs
->driver
->funcname
.pattern
) {
1501 const struct userdiff_funcname
*pe
= &gs
->driver
->funcname
;
1502 xdiff_set_find_func(xecfg
, pe
->pattern
, pe
->cflags
);
1504 xecfg
= opt
->priv
= NULL
;
1510 return xecfg
->find_func(bol
, eol
- bol
, buf
, 1,
1511 xecfg
->find_func_priv
) >= 0;
1516 if (isalpha(*bol
) || *bol
== '_' || *bol
== '$')
1521 static void show_funcname_line(struct grep_opt
*opt
, struct grep_source
*gs
,
1522 char *bol
, unsigned lno
)
1524 while (bol
> gs
->buf
) {
1527 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1531 if (lno
<= opt
->last_shown
)
1534 if (match_funcname(opt
, gs
, bol
, eol
)) {
1535 show_line(opt
, bol
, eol
, gs
->name
, lno
, 0, '=');
1541 static int is_empty_line(const char *bol
, const char *eol
);
1543 static void show_pre_context(struct grep_opt
*opt
, struct grep_source
*gs
,
1544 char *bol
, char *end
, unsigned lno
)
1546 unsigned cur
= lno
, from
= 1, funcname_lno
= 0, orig_from
;
1547 int funcname_needed
= !!opt
->funcname
, comment_needed
= 0;
1549 if (opt
->pre_context
< lno
)
1550 from
= lno
- opt
->pre_context
;
1551 if (from
<= opt
->last_shown
)
1552 from
= opt
->last_shown
+ 1;
1554 if (opt
->funcbody
) {
1555 if (match_funcname(opt
, gs
, bol
, end
))
1558 funcname_needed
= 1;
1559 from
= opt
->last_shown
+ 1;
1563 while (bol
> gs
->buf
&& cur
> from
) {
1564 char *next_bol
= bol
;
1567 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1570 if (comment_needed
&& (is_empty_line(bol
, eol
) ||
1571 match_funcname(opt
, gs
, bol
, eol
))) {
1580 if (funcname_needed
&& match_funcname(opt
, gs
, bol
, eol
)) {
1582 funcname_needed
= 0;
1590 /* We need to look even further back to find a function signature. */
1591 if (opt
->funcname
&& funcname_needed
)
1592 show_funcname_line(opt
, gs
, bol
, cur
);
1596 char *eol
= bol
, sign
= (cur
== funcname_lno
) ? '=' : '-';
1598 while (*eol
!= '\n')
1600 show_line(opt
, bol
, eol
, gs
->name
, cur
, 0, sign
);
1606 static int should_lookahead(struct grep_opt
*opt
)
1611 return 0; /* punt for too complex stuff */
1614 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1615 if (p
->token
!= GREP_PATTERN
)
1616 return 0; /* punt for "header only" and stuff */
1621 static int look_ahead(struct grep_opt
*opt
,
1622 unsigned long *left_p
,
1626 unsigned lno
= *lno_p
;
1629 char *sp
, *last_bol
;
1630 regoff_t earliest
= -1;
1632 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1636 hit
= patmatch(p
, bol
, bol
+ *left_p
, &m
, 0);
1637 if (!hit
|| m
.rm_so
< 0 || m
.rm_eo
< 0)
1639 if (earliest
< 0 || m
.rm_so
< earliest
)
1644 *bol_p
= bol
+ *left_p
;
1648 for (sp
= bol
+ earliest
; bol
< sp
&& sp
[-1] != '\n'; sp
--)
1649 ; /* find the beginning of the line */
1652 for (sp
= bol
; sp
< last_bol
; sp
++) {
1656 *left_p
-= last_bol
- bol
;
1662 static int fill_textconv_grep(struct repository
*r
,
1663 struct userdiff_driver
*driver
,
1664 struct grep_source
*gs
)
1666 struct diff_filespec
*df
;
1670 if (!driver
|| !driver
->textconv
)
1671 return grep_source_load(gs
);
1674 * The textconv interface is intimately tied to diff_filespecs, so we
1675 * have to pretend to be one. If we could unify the grep_source
1676 * and diff_filespec structs, this mess could just go away.
1678 df
= alloc_filespec(gs
->path
);
1680 case GREP_SOURCE_OID
:
1681 fill_filespec(df
, gs
->identifier
, 1, 0100644);
1683 case GREP_SOURCE_FILE
:
1684 fill_filespec(df
, &null_oid
, 0, 0100644);
1687 BUG("attempt to textconv something without a path?");
1691 * fill_textconv is not remotely thread-safe; it may load objects
1692 * behind the scenes, and it modifies the global diff tempfile
1696 size
= fill_textconv(r
, driver
, df
, &buf
);
1701 * The normal fill_textconv usage by the diff machinery would just keep
1702 * the textconv'd buf separate from the diff_filespec. But much of the
1703 * grep code passes around a grep_source and assumes that its "buf"
1704 * pointer is the beginning of the thing we are searching. So let's
1705 * install our textconv'd version into the grep_source, taking care not
1706 * to leak any existing buffer.
1708 grep_source_clear_data(gs
);
1715 static int is_empty_line(const char *bol
, const char *eol
)
1717 while (bol
< eol
&& isspace(*bol
))
1722 static int grep_source_1(struct grep_opt
*opt
, struct grep_source
*gs
, int collect_hits
)
1725 char *peek_bol
= NULL
;
1728 unsigned last_hit
= 0;
1729 int binary_match_only
= 0;
1731 int try_lookahead
= 0;
1732 int show_function
= 0;
1733 struct userdiff_driver
*textconv
= NULL
;
1734 enum grep_context ctx
= GREP_CONTEXT_HEAD
;
1737 if (!opt
->status_only
&& gs
->name
== NULL
)
1738 BUG("grep call which could print a name requires "
1739 "grep_source.name be non-NULL");
1742 opt
->output
= std_output
;
1744 if (opt
->pre_context
|| opt
->post_context
|| opt
->file_break
||
1746 /* Show hunk marks, except for the first file. */
1747 if (opt
->last_shown
)
1748 opt
->show_hunk_mark
= 1;
1750 * If we're using threads then we can't easily identify
1751 * the first file. Always put hunk marks in that case
1752 * and skip the very first one later in work_done().
1754 if (opt
->output
!= std_output
)
1755 opt
->show_hunk_mark
= 1;
1757 opt
->last_shown
= 0;
1759 if (opt
->allow_textconv
) {
1760 grep_source_load_driver(gs
, opt
->repo
->index
);
1762 * We might set up the shared textconv cache data here, which
1763 * is not thread-safe.
1766 textconv
= userdiff_get_textconv(opt
->repo
, gs
->driver
);
1771 * We know the result of a textconv is text, so we only have to care
1772 * about binary handling if we are not using it.
1775 switch (opt
->binary
) {
1776 case GREP_BINARY_DEFAULT
:
1777 if (grep_source_is_binary(gs
, opt
->repo
->index
))
1778 binary_match_only
= 1;
1780 case GREP_BINARY_NOMATCH
:
1781 if (grep_source_is_binary(gs
, opt
->repo
->index
))
1782 return 0; /* Assume unmatch */
1784 case GREP_BINARY_TEXT
:
1787 BUG("unknown binary handling mode");
1791 memset(&xecfg
, 0, sizeof(xecfg
));
1794 try_lookahead
= should_lookahead(opt
);
1796 if (fill_textconv_grep(opt
->repo
, textconv
, gs
) < 0)
1805 ssize_t col
= -1, icol
= -1;
1808 * look_ahead() skips quickly to the line that possibly
1809 * has the next hit; don't call it if we need to do
1810 * something more than just skipping the current line
1811 * in response to an unmatch for the current line. E.g.
1812 * inside a post-context window, we will show the current
1813 * line as a context around the previous hit when it
1818 && (show_function
||
1819 lno
<= last_hit
+ opt
->post_context
))
1820 && look_ahead(opt
, &left
, &lno
, &bol
))
1822 eol
= end_of_line(bol
, &left
);
1826 if ((ctx
== GREP_CONTEXT_HEAD
) && (eol
== bol
))
1827 ctx
= GREP_CONTEXT_BODY
;
1829 hit
= match_line(opt
, bol
, eol
, &col
, &icol
, ctx
, collect_hits
);
1835 /* "grep -v -e foo -e bla" should list lines
1836 * that do not have either, so inversion should
1841 if (opt
->unmatch_name_only
) {
1848 if (opt
->status_only
)
1850 if (opt
->name_only
) {
1851 show_name(opt
, gs
->name
);
1856 if (binary_match_only
) {
1857 opt
->output(opt
, "Binary file ", 12);
1858 output_color(opt
, gs
->name
, strlen(gs
->name
),
1859 opt
->colors
[GREP_COLOR_FILENAME
]);
1860 opt
->output(opt
, " matches\n", 9);
1863 /* Hit at this line. If we haven't shown the
1864 * pre-context lines, we would need to show them.
1866 if (opt
->pre_context
|| opt
->funcbody
)
1867 show_pre_context(opt
, gs
, bol
, eol
, lno
);
1868 else if (opt
->funcname
)
1869 show_funcname_line(opt
, gs
, bol
, lno
);
1870 cno
= opt
->invert
? icol
: col
;
1873 * A negative cno indicates that there was no
1874 * match on the line. We are thus inverted and
1875 * being asked to show all lines that _don't_
1876 * match a given expression. Therefore, set cno
1877 * to 0 to suggest the whole line matches.
1881 show_line(opt
, bol
, eol
, gs
->name
, lno
, cno
+ 1, ':');
1887 if (show_function
&& (!peek_bol
|| peek_bol
< bol
)) {
1888 unsigned long peek_left
= left
;
1889 char *peek_eol
= eol
;
1892 * Trailing empty lines are not interesting.
1893 * Peek past them to see if they belong to the
1894 * body of the current function.
1897 while (is_empty_line(peek_bol
, peek_eol
)) {
1898 peek_bol
= peek_eol
+ 1;
1899 peek_eol
= end_of_line(peek_bol
, &peek_left
);
1902 if (match_funcname(opt
, gs
, peek_bol
, peek_eol
))
1905 if (show_function
||
1906 (last_hit
&& lno
<= last_hit
+ opt
->post_context
)) {
1907 /* If the last hit is within the post context,
1908 * we need to show this line.
1910 show_line(opt
, bol
, eol
, gs
->name
, lno
, col
+ 1, '-');
1924 if (opt
->status_only
)
1925 return opt
->unmatch_name_only
;
1926 if (opt
->unmatch_name_only
) {
1927 /* We did not see any hit, so we want to show this */
1928 show_name(opt
, gs
->name
);
1932 xdiff_clear_find_func(&xecfg
);
1936 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1937 * which feels mostly useless but sometimes useful. Maybe
1938 * make it another option? For now suppress them.
1940 if (opt
->count
&& count
) {
1942 if (opt
->pathname
) {
1943 output_color(opt
, gs
->name
, strlen(gs
->name
),
1944 opt
->colors
[GREP_COLOR_FILENAME
]);
1945 output_sep(opt
, ':');
1947 xsnprintf(buf
, sizeof(buf
), "%u\n", count
);
1948 opt
->output(opt
, buf
, strlen(buf
));
1954 static void clr_hit_marker(struct grep_expr
*x
)
1956 /* All-hit markers are meaningful only at the very top level
1961 if (x
->node
!= GREP_NODE_OR
)
1963 x
->u
.binary
.left
->hit
= 0;
1964 x
= x
->u
.binary
.right
;
1968 static int chk_hit_marker(struct grep_expr
*x
)
1970 /* Top level nodes have hit markers. See if they all are hits */
1972 if (x
->node
!= GREP_NODE_OR
)
1974 if (!x
->u
.binary
.left
->hit
)
1976 x
= x
->u
.binary
.right
;
1980 int grep_source(struct grep_opt
*opt
, struct grep_source
*gs
)
1983 * we do not have to do the two-pass grep when we do not check
1984 * buffer-wide "all-match".
1986 if (!opt
->all_match
)
1987 return grep_source_1(opt
, gs
, 0);
1989 /* Otherwise the toplevel "or" terms hit a bit differently.
1990 * We first clear hit markers from them.
1992 clr_hit_marker(opt
->pattern_expression
);
1993 grep_source_1(opt
, gs
, 1);
1995 if (!chk_hit_marker(opt
->pattern_expression
))
1998 return grep_source_1(opt
, gs
, 0);
2001 int grep_buffer(struct grep_opt
*opt
, char *buf
, unsigned long size
)
2003 struct grep_source gs
;
2006 grep_source_init(&gs
, GREP_SOURCE_BUF
, NULL
, NULL
, NULL
);
2010 r
= grep_source(opt
, &gs
);
2012 grep_source_clear(&gs
);
2016 void grep_source_init(struct grep_source
*gs
, enum grep_source_type type
,
2017 const char *name
, const char *path
,
2018 const void *identifier
)
2021 gs
->name
= xstrdup_or_null(name
);
2022 gs
->path
= xstrdup_or_null(path
);
2028 case GREP_SOURCE_FILE
:
2029 gs
->identifier
= xstrdup(identifier
);
2031 case GREP_SOURCE_OID
:
2032 gs
->identifier
= oiddup(identifier
);
2034 case GREP_SOURCE_BUF
:
2035 gs
->identifier
= NULL
;
2040 void grep_source_clear(struct grep_source
*gs
)
2042 FREE_AND_NULL(gs
->name
);
2043 FREE_AND_NULL(gs
->path
);
2044 FREE_AND_NULL(gs
->identifier
);
2045 grep_source_clear_data(gs
);
2048 void grep_source_clear_data(struct grep_source
*gs
)
2051 case GREP_SOURCE_FILE
:
2052 case GREP_SOURCE_OID
:
2053 FREE_AND_NULL(gs
->buf
);
2056 case GREP_SOURCE_BUF
:
2057 /* leave user-provided buf intact */
2062 static int grep_source_load_oid(struct grep_source
*gs
)
2064 enum object_type type
;
2067 gs
->buf
= read_object_file(gs
->identifier
, &type
, &gs
->size
);
2071 return error(_("'%s': unable to read %s"),
2073 oid_to_hex(gs
->identifier
));
2077 static int grep_source_load_file(struct grep_source
*gs
)
2079 const char *filename
= gs
->identifier
;
2085 if (lstat(filename
, &st
) < 0) {
2087 if (errno
!= ENOENT
)
2088 error_errno(_("failed to stat '%s'"), filename
);
2091 if (!S_ISREG(st
.st_mode
))
2093 size
= xsize_t(st
.st_size
);
2094 i
= open(filename
, O_RDONLY
);
2097 data
= xmallocz(size
);
2098 if (st
.st_size
!= read_in_full(i
, data
, size
)) {
2099 error_errno(_("'%s': short read"), filename
);
2111 static int grep_source_load(struct grep_source
*gs
)
2117 case GREP_SOURCE_FILE
:
2118 return grep_source_load_file(gs
);
2119 case GREP_SOURCE_OID
:
2120 return grep_source_load_oid(gs
);
2121 case GREP_SOURCE_BUF
:
2122 return gs
->buf
? 0 : -1;
2124 BUG("invalid grep_source type to load");
2127 void grep_source_load_driver(struct grep_source
*gs
,
2128 struct index_state
*istate
)
2135 gs
->driver
= userdiff_find_by_path(istate
, gs
->path
);
2137 gs
->driver
= userdiff_find_by_name("default");
2141 static int grep_source_is_binary(struct grep_source
*gs
,
2142 struct index_state
*istate
)
2144 grep_source_load_driver(gs
, istate
);
2145 if (gs
->driver
->binary
!= -1)
2146 return gs
->driver
->binary
;
2148 if (!grep_source_load(gs
))
2149 return buffer_is_binary(gs
->buf
, gs
->size
);