git/grep.c

1872 lines
44 KiB
C
Raw Normal View History

#include "cache.h"
#include "grep.h"
#include "userdiff.h"
#include "xdiff-interface.h"
#include "diff.h"
#include "diffcore.h"
#include "commit.h"
#include "quote.h"
static int grep_source_load(struct grep_source *gs);
static int grep_source_is_binary(struct grep_source *gs);
static struct grep_opt grep_defaults;
/*
* Initialize the grep_defaults template with hardcoded defaults.
* We could let the compiler do this, but without C99 initializers
* the code gets unwieldy and unreadable, so...
*/
void init_grep_defaults(void)
{
struct grep_opt *opt = &grep_defaults;
static int run_once;
if (run_once)
return;
run_once++;
memset(opt, 0, sizeof(*opt));
opt->relative = 1;
opt->pathname = 1;
opt->regflags = REG_NEWLINE;
opt->max_depth = -1;
opt->pattern_type_option = GREP_PATTERN_TYPE_UNSPECIFIED;
opt->extended_regexp_option = 0;
color_set(opt->color_context, "");
color_set(opt->color_filename, "");
color_set(opt->color_function, "");
color_set(opt->color_lineno, "");
color_set(opt->color_match_context, GIT_COLOR_BOLD_RED);
color_set(opt->color_match_selected, GIT_COLOR_BOLD_RED);
color_set(opt->color_selected, "");
color_set(opt->color_sep, GIT_COLOR_CYAN);
opt->color = -1;
}
static int parse_pattern_type_arg(const char *opt, const char *arg)
{
if (!strcmp(arg, "default"))
return GREP_PATTERN_TYPE_UNSPECIFIED;
else if (!strcmp(arg, "basic"))
return GREP_PATTERN_TYPE_BRE;
else if (!strcmp(arg, "extended"))
return GREP_PATTERN_TYPE_ERE;
else if (!strcmp(arg, "fixed"))
return GREP_PATTERN_TYPE_FIXED;
else if (!strcmp(arg, "perl"))
return GREP_PATTERN_TYPE_PCRE;
die("bad %s argument: %s", opt, arg);
}
/*
* Read the configuration file once and store it in
* the grep_defaults template.
*/
int grep_config(const char *var, const char *value, void *cb)
{
struct grep_opt *opt = &grep_defaults;
char *color = NULL;
if (userdiff_config(var, value) < 0)
return -1;
if (!strcmp(var, "grep.extendedregexp")) {
if (git_config_bool(var, value))
opt->extended_regexp_option = 1;
else
opt->extended_regexp_option = 0;
return 0;
}
if (!strcmp(var, "grep.patterntype")) {
opt->pattern_type_option = parse_pattern_type_arg(var, value);
return 0;
}
if (!strcmp(var, "grep.linenumber")) {
opt->linenum = git_config_bool(var, value);
return 0;
}
if (!strcmp(var, "grep.fullname")) {
opt->relative = !git_config_bool(var, value);
return 0;
}
if (!strcmp(var, "color.grep"))
opt->color = git_config_colorbool(var, value);
else if (!strcmp(var, "color.grep.context"))
color = opt->color_context;
else if (!strcmp(var, "color.grep.filename"))
color = opt->color_filename;
else if (!strcmp(var, "color.grep.function"))
color = opt->color_function;
else if (!strcmp(var, "color.grep.linenumber"))
color = opt->color_lineno;
else if (!strcmp(var, "color.grep.matchcontext"))
color = opt->color_match_context;
else if (!strcmp(var, "color.grep.matchselected"))
color = opt->color_match_selected;
else if (!strcmp(var, "color.grep.selected"))
color = opt->color_selected;
else if (!strcmp(var, "color.grep.separator"))
color = opt->color_sep;
else if (!strcmp(var, "color.grep.match")) {
int rc = 0;
if (!value)
return config_error_nonbool(var);
rc |= color_parse(value, opt->color_match_context);
rc |= color_parse(value, opt->color_match_selected);
return rc;
}
if (color) {
if (!value)
return config_error_nonbool(var);
return color_parse(value, color);
}
return 0;
}
/*
* Initialize one instance of grep_opt and copy the
* default values from the template we read the configuration
* information in an earlier call to git_config(grep_config).
*/
void grep_init(struct grep_opt *opt, const char *prefix)
{
struct grep_opt *def = &grep_defaults;
memset(opt, 0, sizeof(*opt));
opt->prefix = prefix;
opt->prefix_length = (prefix && *prefix) ? strlen(prefix) : 0;
opt->pattern_tail = &opt->pattern_list;
opt->header_tail = &opt->header_list;
opt->color = def->color;
opt->extended_regexp_option = def->extended_regexp_option;
opt->pattern_type_option = def->pattern_type_option;
opt->linenum = def->linenum;
opt->max_depth = def->max_depth;
opt->pathname = def->pathname;
opt->regflags = def->regflags;
opt->relative = def->relative;
color_set(opt->color_context, def->color_context);
color_set(opt->color_filename, def->color_filename);
color_set(opt->color_function, def->color_function);
color_set(opt->color_lineno, def->color_lineno);
color_set(opt->color_match_context, def->color_match_context);
color_set(opt->color_match_selected, def->color_match_selected);
color_set(opt->color_selected, def->color_selected);
color_set(opt->color_sep, def->color_sep);
}
static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type, struct grep_opt *opt)
{
switch (pattern_type) {
case GREP_PATTERN_TYPE_UNSPECIFIED:
/* fall through */
case GREP_PATTERN_TYPE_BRE:
opt->fixed = 0;
opt->pcre = 0;
opt->regflags &= ~REG_EXTENDED;
break;
case GREP_PATTERN_TYPE_ERE:
opt->fixed = 0;
opt->pcre = 0;
opt->regflags |= REG_EXTENDED;
break;
case GREP_PATTERN_TYPE_FIXED:
opt->fixed = 1;
opt->pcre = 0;
opt->regflags &= ~REG_EXTENDED;
break;
case GREP_PATTERN_TYPE_PCRE:
opt->fixed = 0;
opt->pcre = 1;
opt->regflags &= ~REG_EXTENDED;
break;
}
}
void grep_commit_pattern_type(enum grep_pattern_type pattern_type, struct grep_opt *opt)
{
if (pattern_type != GREP_PATTERN_TYPE_UNSPECIFIED)
grep_set_pattern_type_option(pattern_type, opt);
else if (opt->pattern_type_option != GREP_PATTERN_TYPE_UNSPECIFIED)
grep_set_pattern_type_option(opt->pattern_type_option, opt);
else if (opt->extended_regexp_option)
grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE, opt);
}
static struct grep_pat *create_grep_pat(const char *pat, size_t patlen,
const char *origin, int no,
enum grep_pat_token t,
enum grep_header_field field)
log --author/--committer: really match only with name part When we tried to find commits done by AUTHOR, the first implementation tried to pattern match a line with "^author .*AUTHOR", which later was enhanced to strip leading caret and look for "^author AUTHOR" when the search pattern was anchored at the left end (i.e. --author="^AUTHOR"). This had a few problems: * When looking for fixed strings (e.g. "git log -F --author=x --grep=y"), the regexp internally used "^author .*x" would never match anything; * To match at the end (e.g. "git log --author='google.com>$'"), the generated regexp has to also match the trailing timestamp part the commit header lines have. Also, in order to determine if the '$' at the end means "match at the end of the line" or just a literal dollar sign (probably backslash-quoted), we would need to parse the regexp ourselves. An earlier alternative tried to make sure that a line matches "^author " (to limit by field name) and the user supplied pattern at the same time. While it solved the -F problem by introducing a special override for matching the "^author ", it did not solve the trailing timestamp nor tail match problem. It also would have matched every commit if --author=author was asked for, not because the author's email part had this string, but because every commit header line that talks about the author begins with that field name, regardleses of who wrote it. Instead of piling more hacks on top of hacks, this rethinks the grep machinery that is used to look for strings in the commit header, and makes sure that (1) field name matches literally at the beginning of the line, followed by a SP, and (2) the user supplied pattern is matched against the remainder of the line, excluding the trailing timestamp data. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2008-09-05 05:15:02 +00:00
{
struct grep_pat *p = xcalloc(1, sizeof(*p));
p->pattern = xmemdupz(pat, patlen);
p->patternlen = patlen;
p->origin = origin;
p->no = no;
p->token = t;
log --author/--committer: really match only with name part When we tried to find commits done by AUTHOR, the first implementation tried to pattern match a line with "^author .*AUTHOR", which later was enhanced to strip leading caret and look for "^author AUTHOR" when the search pattern was anchored at the left end (i.e. --author="^AUTHOR"). This had a few problems: * When looking for fixed strings (e.g. "git log -F --author=x --grep=y"), the regexp internally used "^author .*x" would never match anything; * To match at the end (e.g. "git log --author='google.com>$'"), the generated regexp has to also match the trailing timestamp part the commit header lines have. Also, in order to determine if the '$' at the end means "match at the end of the line" or just a literal dollar sign (probably backslash-quoted), we would need to parse the regexp ourselves. An earlier alternative tried to make sure that a line matches "^author " (to limit by field name) and the user supplied pattern at the same time. While it solved the -F problem by introducing a special override for matching the "^author ", it did not solve the trailing timestamp nor tail match problem. It also would have matched every commit if --author=author was asked for, not because the author's email part had this string, but because every commit header line that talks about the author begins with that field name, regardleses of who wrote it. Instead of piling more hacks on top of hacks, this rethinks the grep machinery that is used to look for strings in the commit header, and makes sure that (1) field name matches literally at the beginning of the line, followed by a SP, and (2) the user supplied pattern is matched against the remainder of the line, excluding the trailing timestamp data. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2008-09-05 05:15:02 +00:00
p->field = field;
return p;
}
static void do_append_grep_pat(struct grep_pat ***tail, struct grep_pat *p)
{
**tail = p;
*tail = &p->next;
log --author/--committer: really match only with name part When we tried to find commits done by AUTHOR, the first implementation tried to pattern match a line with "^author .*AUTHOR", which later was enhanced to strip leading caret and look for "^author AUTHOR" when the search pattern was anchored at the left end (i.e. --author="^AUTHOR"). This had a few problems: * When looking for fixed strings (e.g. "git log -F --author=x --grep=y"), the regexp internally used "^author .*x" would never match anything; * To match at the end (e.g. "git log --author='google.com>$'"), the generated regexp has to also match the trailing timestamp part the commit header lines have. Also, in order to determine if the '$' at the end means "match at the end of the line" or just a literal dollar sign (probably backslash-quoted), we would need to parse the regexp ourselves. An earlier alternative tried to make sure that a line matches "^author " (to limit by field name) and the user supplied pattern at the same time. While it solved the -F problem by introducing a special override for matching the "^author ", it did not solve the trailing timestamp nor tail match problem. It also would have matched every commit if --author=author was asked for, not because the author's email part had this string, but because every commit header line that talks about the author begins with that field name, regardleses of who wrote it. Instead of piling more hacks on top of hacks, this rethinks the grep machinery that is used to look for strings in the commit header, and makes sure that (1) field name matches literally at the beginning of the line, followed by a SP, and (2) the user supplied pattern is matched against the remainder of the line, excluding the trailing timestamp data. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2008-09-05 05:15:02 +00:00
p->next = NULL;
switch (p->token) {
case GREP_PATTERN: /* atom */
case GREP_PATTERN_HEAD:
case GREP_PATTERN_BODY:
for (;;) {
struct grep_pat *new_pat;
size_t len = 0;
char *cp = p->pattern + p->patternlen, *nl = NULL;
while (++len <= p->patternlen) {
if (*(--cp) == '\n') {
nl = cp;
break;
}
}
if (!nl)
break;
new_pat = create_grep_pat(nl + 1, len - 1, p->origin,
p->no, p->token, p->field);
new_pat->next = p->next;
if (!p->next)
*tail = &new_pat->next;
p->next = new_pat;
*nl = '\0';
p->patternlen -= len;
}
break;
default:
break;
}
}
void append_header_grep_pattern(struct grep_opt *opt,
enum grep_header_field field, const char *pat)
{
struct grep_pat *p = create_grep_pat(pat, strlen(pat), "header", 0,
GREP_PATTERN_HEAD, field);
if (field == GREP_HEADER_REFLOG)
opt->use_reflog_filter = 1;
do_append_grep_pat(&opt->header_tail, p);
log --author/--committer: really match only with name part When we tried to find commits done by AUTHOR, the first implementation tried to pattern match a line with "^author .*AUTHOR", which later was enhanced to strip leading caret and look for "^author AUTHOR" when the search pattern was anchored at the left end (i.e. --author="^AUTHOR"). This had a few problems: * When looking for fixed strings (e.g. "git log -F --author=x --grep=y"), the regexp internally used "^author .*x" would never match anything; * To match at the end (e.g. "git log --author='google.com>$'"), the generated regexp has to also match the trailing timestamp part the commit header lines have. Also, in order to determine if the '$' at the end means "match at the end of the line" or just a literal dollar sign (probably backslash-quoted), we would need to parse the regexp ourselves. An earlier alternative tried to make sure that a line matches "^author " (to limit by field name) and the user supplied pattern at the same time. While it solved the -F problem by introducing a special override for matching the "^author ", it did not solve the trailing timestamp nor tail match problem. It also would have matched every commit if --author=author was asked for, not because the author's email part had this string, but because every commit header line that talks about the author begins with that field name, regardleses of who wrote it. Instead of piling more hacks on top of hacks, this rethinks the grep machinery that is used to look for strings in the commit header, and makes sure that (1) field name matches literally at the beginning of the line, followed by a SP, and (2) the user supplied pattern is matched against the remainder of the line, excluding the trailing timestamp data. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2008-09-05 05:15:02 +00:00
}
void append_grep_pattern(struct grep_opt *opt, const char *pat,
const char *origin, int no, enum grep_pat_token t)
{
append_grep_pat(opt, pat, strlen(pat), origin, no, t);
}
void append_grep_pat(struct grep_opt *opt, const char *pat, size_t patlen,
const char *origin, int no, enum grep_pat_token t)
{
struct grep_pat *p = create_grep_pat(pat, patlen, origin, no, t, 0);
do_append_grep_pat(&opt->pattern_tail, p);
}
struct grep_opt *grep_opt_dup(const struct grep_opt *opt)
{
struct grep_pat *pat;
struct grep_opt *ret = xmalloc(sizeof(struct grep_opt));
*ret = *opt;
ret->pattern_list = NULL;
ret->pattern_tail = &ret->pattern_list;
for(pat = opt->pattern_list; pat != NULL; pat = pat->next)
{
if(pat->token == GREP_PATTERN_HEAD)
append_header_grep_pattern(ret, pat->field,
pat->pattern);
else
append_grep_pat(ret, pat->pattern, pat->patternlen,
pat->origin, pat->no, pat->token);
}
return ret;
}
static NORETURN void compile_regexp_failed(const struct grep_pat *p,
const char *error)
{
char where[1024];
if (p->no)
xsnprintf(where, sizeof(where), "In '%s' at %d, ", p->origin, p->no);
else if (p->origin)
xsnprintf(where, sizeof(where), "%s, ", p->origin);
else
where[0] = 0;
die("%s'%s': %s", where, p->pattern, error);
}
#ifdef USE_LIBPCRE
static void compile_pcre_regexp(struct grep_pat *p, const struct grep_opt *opt)
{
const char *error;
int erroffset;
int options = PCRE_MULTILINE;
if (opt->ignore_case) {
if (has_non_ascii(p->pattern))
p->pcre_tables = pcre_maketables();
options |= PCRE_CASELESS;
}
if (is_utf8_locale() && has_non_ascii(p->pattern))
options |= PCRE_UTF8;
p->pcre_regexp = pcre_compile(p->pattern, options, &error, &erroffset,
p->pcre_tables);
if (!p->pcre_regexp)
compile_regexp_failed(p, error);
p->pcre_extra_info = pcre_study(p->pcre_regexp, 0, &error);
if (!p->pcre_extra_info && error)
die("%s", error);
}
static int pcrematch(struct grep_pat *p, const char *line, const char *eol,
regmatch_t *match, int eflags)
{
int ovector[30], ret, flags = 0;
if (eflags & REG_NOTBOL)
flags |= PCRE_NOTBOL;
ret = pcre_exec(p->pcre_regexp, p->pcre_extra_info, line, eol - line,
0, flags, ovector, ARRAY_SIZE(ovector));
if (ret < 0 && ret != PCRE_ERROR_NOMATCH)
die("pcre_exec failed with error code %d", ret);
if (ret > 0) {
ret = 0;
match->rm_so = ovector[0];
match->rm_eo = ovector[1];
}
return ret;
}
static void free_pcre_regexp(struct grep_pat *p)
{
pcre_free(p->pcre_regexp);
pcre_free(p->pcre_extra_info);
pcre_free((void *)p->pcre_tables);
}
#else /* !USE_LIBPCRE */
static void compile_pcre_regexp(struct grep_pat *p, const struct grep_opt *opt)
{
die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
}
static int pcrematch(struct grep_pat *p, const char *line, const char *eol,
regmatch_t *match, int eflags)
{
return 1;
}
static void free_pcre_regexp(struct grep_pat *p)
{
}
#endif /* !USE_LIBPCRE */
Use kwset in grep Benchmarks for the hot cache case: before: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 3,478,085 cache-misses # 2.322 M/sec ( +- 2.690% ) 11,356,177 cache-references # 7.582 M/sec ( +- 2.598% ) 3,872,184 branch-misses # 0.363 % ( +- 0.258% ) 1,067,367,848 branches # 712.673 M/sec ( +- 2.622% ) 3,828,370,782 instructions # 0.947 IPC ( +- 0.033% ) 4,043,832,831 cycles # 2700.037 M/sec ( +- 0.167% ) 8,518 page-faults # 0.006 M/sec ( +- 3.648% ) 847 CPU-migrations # 0.001 M/sec ( +- 3.262% ) 6,546 context-switches # 0.004 M/sec ( +- 2.292% ) 1497.695495 task-clock-msecs # 3.303 CPUs ( +- 2.550% ) 0.453394396 seconds time elapsed ( +- 0.912% ) after: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 2,989,918 cache-misses # 3.166 M/sec ( +- 5.013% ) 10,986,041 cache-references # 11.633 M/sec ( +- 4.899% ) (scaled from 95.06%) 3,511,993 branch-misses # 1.422 % ( +- 0.785% ) 246,893,561 branches # 261.433 M/sec ( +- 3.967% ) 1,392,727,757 instructions # 0.564 IPC ( +- 0.040% ) 2,468,142,397 cycles # 2613.494 M/sec ( +- 0.110% ) 7,747 page-faults # 0.008 M/sec ( +- 3.995% ) 897 CPU-migrations # 0.001 M/sec ( +- 2.383% ) 6,535 context-switches # 0.007 M/sec ( +- 1.993% ) 944.384228 task-clock-msecs # 3.177 CPUs ( +- 0.268% ) 0.297257643 seconds time elapsed ( +- 0.450% ) So we gain about 35% by using the kwset code. As a side effect of using kwset two grep tests are fixed by this patch. The first is fixed because kwset can deal with case-insensitive search containing NULs, something strcasestr cannot do. The second one is fixed because we consider patterns containing NULs as fixed strings (regcomp cannot accept patterns with NULs). Signed-off-by: Fredrik Kuivinen <frekui@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-20 22:42:18 +00:00
static int is_fixed(const char *s, size_t len)
{
size_t i;
/* regcomp cannot accept patterns with NULs so we
* consider any pattern containing a NUL fixed.
*/
if (memchr(s, 0, len))
return 1;
for (i = 0; i < len; i++) {
if (is_regex_special(s[i]))
return 0;
}
return 1;
}
static void compile_fixed_regexp(struct grep_pat *p, struct grep_opt *opt)
{
struct strbuf sb = STRBUF_INIT;
int err;
int regflags;
basic_regex_quote_buf(&sb, p->pattern);
regflags = opt->regflags & ~REG_EXTENDED;
if (opt->ignore_case)
regflags |= REG_ICASE;
err = regcomp(&p->regexp, sb.buf, regflags);
if (opt->debug)
fprintf(stderr, "fixed %s\n", sb.buf);
strbuf_release(&sb);
if (err) {
char errbuf[1024];
regerror(err, &p->regexp, errbuf, sizeof(errbuf));
regfree(&p->regexp);
compile_regexp_failed(p, errbuf);
}
}
static void compile_regexp(struct grep_pat *p, struct grep_opt *opt)
{
int icase, ascii_only;
int err;
p->word_regexp = opt->word_regexp;
p->ignore_case = opt->ignore_case;
icase = opt->regflags & REG_ICASE || p->ignore_case;
ascii_only = !has_non_ascii(p->pattern);
/*
* Even when -F (fixed) asks us to do a non-regexp search, we
* may not be able to correctly case-fold when -i
* (ignore-case) is asked (in which case, we'll synthesize a
* regexp to match the pattern that matches regexp special
* characters literally, while ignoring case differences). On
* the other hand, even without -F, if the pattern does not
* have any regexp special characters and there is no need for
* case-folding search, we can internally turn it into a
* simple string match using kws. p->fixed tells us if we
* want to use kws.
*/
Use kwset in grep Benchmarks for the hot cache case: before: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 3,478,085 cache-misses # 2.322 M/sec ( +- 2.690% ) 11,356,177 cache-references # 7.582 M/sec ( +- 2.598% ) 3,872,184 branch-misses # 0.363 % ( +- 0.258% ) 1,067,367,848 branches # 712.673 M/sec ( +- 2.622% ) 3,828,370,782 instructions # 0.947 IPC ( +- 0.033% ) 4,043,832,831 cycles # 2700.037 M/sec ( +- 0.167% ) 8,518 page-faults # 0.006 M/sec ( +- 3.648% ) 847 CPU-migrations # 0.001 M/sec ( +- 3.262% ) 6,546 context-switches # 0.004 M/sec ( +- 2.292% ) 1497.695495 task-clock-msecs # 3.303 CPUs ( +- 2.550% ) 0.453394396 seconds time elapsed ( +- 0.912% ) after: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 2,989,918 cache-misses # 3.166 M/sec ( +- 5.013% ) 10,986,041 cache-references # 11.633 M/sec ( +- 4.899% ) (scaled from 95.06%) 3,511,993 branch-misses # 1.422 % ( +- 0.785% ) 246,893,561 branches # 261.433 M/sec ( +- 3.967% ) 1,392,727,757 instructions # 0.564 IPC ( +- 0.040% ) 2,468,142,397 cycles # 2613.494 M/sec ( +- 0.110% ) 7,747 page-faults # 0.008 M/sec ( +- 3.995% ) 897 CPU-migrations # 0.001 M/sec ( +- 2.383% ) 6,535 context-switches # 0.007 M/sec ( +- 1.993% ) 944.384228 task-clock-msecs # 3.177 CPUs ( +- 0.268% ) 0.297257643 seconds time elapsed ( +- 0.450% ) So we gain about 35% by using the kwset code. As a side effect of using kwset two grep tests are fixed by this patch. The first is fixed because kwset can deal with case-insensitive search containing NULs, something strcasestr cannot do. The second one is fixed because we consider patterns containing NULs as fixed strings (regcomp cannot accept patterns with NULs). Signed-off-by: Fredrik Kuivinen <frekui@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-20 22:42:18 +00:00
if (opt->fixed || is_fixed(p->pattern, p->patternlen))
p->fixed = !icase || ascii_only;
Use kwset in grep Benchmarks for the hot cache case: before: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 3,478,085 cache-misses # 2.322 M/sec ( +- 2.690% ) 11,356,177 cache-references # 7.582 M/sec ( +- 2.598% ) 3,872,184 branch-misses # 0.363 % ( +- 0.258% ) 1,067,367,848 branches # 712.673 M/sec ( +- 2.622% ) 3,828,370,782 instructions # 0.947 IPC ( +- 0.033% ) 4,043,832,831 cycles # 2700.037 M/sec ( +- 0.167% ) 8,518 page-faults # 0.006 M/sec ( +- 3.648% ) 847 CPU-migrations # 0.001 M/sec ( +- 3.262% ) 6,546 context-switches # 0.004 M/sec ( +- 2.292% ) 1497.695495 task-clock-msecs # 3.303 CPUs ( +- 2.550% ) 0.453394396 seconds time elapsed ( +- 0.912% ) after: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 2,989,918 cache-misses # 3.166 M/sec ( +- 5.013% ) 10,986,041 cache-references # 11.633 M/sec ( +- 4.899% ) (scaled from 95.06%) 3,511,993 branch-misses # 1.422 % ( +- 0.785% ) 246,893,561 branches # 261.433 M/sec ( +- 3.967% ) 1,392,727,757 instructions # 0.564 IPC ( +- 0.040% ) 2,468,142,397 cycles # 2613.494 M/sec ( +- 0.110% ) 7,747 page-faults # 0.008 M/sec ( +- 3.995% ) 897 CPU-migrations # 0.001 M/sec ( +- 2.383% ) 6,535 context-switches # 0.007 M/sec ( +- 1.993% ) 944.384228 task-clock-msecs # 3.177 CPUs ( +- 0.268% ) 0.297257643 seconds time elapsed ( +- 0.450% ) So we gain about 35% by using the kwset code. As a side effect of using kwset two grep tests are fixed by this patch. The first is fixed because kwset can deal with case-insensitive search containing NULs, something strcasestr cannot do. The second one is fixed because we consider patterns containing NULs as fixed strings (regcomp cannot accept patterns with NULs). Signed-off-by: Fredrik Kuivinen <frekui@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-20 22:42:18 +00:00
else
p->fixed = 0;
if (p->fixed) {
p->kws = kwsalloc(icase ? tolower_trans_tbl : NULL);
Use kwset in grep Benchmarks for the hot cache case: before: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 3,478,085 cache-misses # 2.322 M/sec ( +- 2.690% ) 11,356,177 cache-references # 7.582 M/sec ( +- 2.598% ) 3,872,184 branch-misses # 0.363 % ( +- 0.258% ) 1,067,367,848 branches # 712.673 M/sec ( +- 2.622% ) 3,828,370,782 instructions # 0.947 IPC ( +- 0.033% ) 4,043,832,831 cycles # 2700.037 M/sec ( +- 0.167% ) 8,518 page-faults # 0.006 M/sec ( +- 3.648% ) 847 CPU-migrations # 0.001 M/sec ( +- 3.262% ) 6,546 context-switches # 0.004 M/sec ( +- 2.292% ) 1497.695495 task-clock-msecs # 3.303 CPUs ( +- 2.550% ) 0.453394396 seconds time elapsed ( +- 0.912% ) after: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 2,989,918 cache-misses # 3.166 M/sec ( +- 5.013% ) 10,986,041 cache-references # 11.633 M/sec ( +- 4.899% ) (scaled from 95.06%) 3,511,993 branch-misses # 1.422 % ( +- 0.785% ) 246,893,561 branches # 261.433 M/sec ( +- 3.967% ) 1,392,727,757 instructions # 0.564 IPC ( +- 0.040% ) 2,468,142,397 cycles # 2613.494 M/sec ( +- 0.110% ) 7,747 page-faults # 0.008 M/sec ( +- 3.995% ) 897 CPU-migrations # 0.001 M/sec ( +- 2.383% ) 6,535 context-switches # 0.007 M/sec ( +- 1.993% ) 944.384228 task-clock-msecs # 3.177 CPUs ( +- 0.268% ) 0.297257643 seconds time elapsed ( +- 0.450% ) So we gain about 35% by using the kwset code. As a side effect of using kwset two grep tests are fixed by this patch. The first is fixed because kwset can deal with case-insensitive search containing NULs, something strcasestr cannot do. The second one is fixed because we consider patterns containing NULs as fixed strings (regcomp cannot accept patterns with NULs). Signed-off-by: Fredrik Kuivinen <frekui@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-20 22:42:18 +00:00
kwsincr(p->kws, p->pattern, p->patternlen);
kwsprep(p->kws);
return;
} else if (opt->fixed) {
/*
* We come here when the pattern has the non-ascii
* characters we cannot case-fold, and asked to
* ignore-case.
*/
compile_fixed_regexp(p, opt);
return;
Use kwset in grep Benchmarks for the hot cache case: before: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 3,478,085 cache-misses # 2.322 M/sec ( +- 2.690% ) 11,356,177 cache-references # 7.582 M/sec ( +- 2.598% ) 3,872,184 branch-misses # 0.363 % ( +- 0.258% ) 1,067,367,848 branches # 712.673 M/sec ( +- 2.622% ) 3,828,370,782 instructions # 0.947 IPC ( +- 0.033% ) 4,043,832,831 cycles # 2700.037 M/sec ( +- 0.167% ) 8,518 page-faults # 0.006 M/sec ( +- 3.648% ) 847 CPU-migrations # 0.001 M/sec ( +- 3.262% ) 6,546 context-switches # 0.004 M/sec ( +- 2.292% ) 1497.695495 task-clock-msecs # 3.303 CPUs ( +- 2.550% ) 0.453394396 seconds time elapsed ( +- 0.912% ) after: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 2,989,918 cache-misses # 3.166 M/sec ( +- 5.013% ) 10,986,041 cache-references # 11.633 M/sec ( +- 4.899% ) (scaled from 95.06%) 3,511,993 branch-misses # 1.422 % ( +- 0.785% ) 246,893,561 branches # 261.433 M/sec ( +- 3.967% ) 1,392,727,757 instructions # 0.564 IPC ( +- 0.040% ) 2,468,142,397 cycles # 2613.494 M/sec ( +- 0.110% ) 7,747 page-faults # 0.008 M/sec ( +- 3.995% ) 897 CPU-migrations # 0.001 M/sec ( +- 2.383% ) 6,535 context-switches # 0.007 M/sec ( +- 1.993% ) 944.384228 task-clock-msecs # 3.177 CPUs ( +- 0.268% ) 0.297257643 seconds time elapsed ( +- 0.450% ) So we gain about 35% by using the kwset code. As a side effect of using kwset two grep tests are fixed by this patch. The first is fixed because kwset can deal with case-insensitive search containing NULs, something strcasestr cannot do. The second one is fixed because we consider patterns containing NULs as fixed strings (regcomp cannot accept patterns with NULs). Signed-off-by: Fredrik Kuivinen <frekui@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-20 22:42:18 +00:00
}
if (opt->pcre) {
compile_pcre_regexp(p, opt);
return;
}
err = regcomp(&p->regexp, p->pattern, opt->regflags);
if (err) {
char errbuf[1024];
regerror(err, &p->regexp, errbuf, 1024);
regfree(&p->regexp);
compile_regexp_failed(p, errbuf);
}
}
static struct grep_expr *compile_pattern_or(struct grep_pat **);
static struct grep_expr *compile_pattern_atom(struct grep_pat **list)
{
struct grep_pat *p;
struct grep_expr *x;
p = *list;
if (!p)
return NULL;
switch (p->token) {
case GREP_PATTERN: /* atom */
case GREP_PATTERN_HEAD:
case GREP_PATTERN_BODY:
x = xcalloc(1, sizeof (struct grep_expr));
x->node = GREP_NODE_ATOM;
x->u.atom = p;
*list = p->next;
return x;
case GREP_OPEN_PAREN:
*list = p->next;
x = compile_pattern_or(list);
if (!*list || (*list)->token != GREP_CLOSE_PAREN)
die("unmatched parenthesis");
*list = (*list)->next;
return x;
default:
return NULL;
}
}
static struct grep_expr *compile_pattern_not(struct grep_pat **list)
{
struct grep_pat *p;
struct grep_expr *x;
p = *list;
if (!p)
return NULL;
switch (p->token) {
case GREP_NOT:
if (!p->next)
die("--not not followed by pattern expression");
*list = p->next;
x = xcalloc(1, sizeof (struct grep_expr));
x->node = GREP_NODE_NOT;
x->u.unary = compile_pattern_not(list);
if (!x->u.unary)
die("--not followed by non pattern expression");
return x;
default:
return compile_pattern_atom(list);
}
}
static struct grep_expr *compile_pattern_and(struct grep_pat **list)
{
struct grep_pat *p;
struct grep_expr *x, *y, *z;
x = compile_pattern_not(list);
p = *list;
if (p && p->token == GREP_AND) {
if (!p->next)
die("--and not followed by pattern expression");
*list = p->next;
y = compile_pattern_and(list);
if (!y)
die("--and not followed by pattern expression");
z = xcalloc(1, sizeof (struct grep_expr));
z->node = GREP_NODE_AND;
z->u.binary.left = x;
z->u.binary.right = y;
return z;
}
return x;
}
static struct grep_expr *compile_pattern_or(struct grep_pat **list)
{
struct grep_pat *p;
struct grep_expr *x, *y, *z;
x = compile_pattern_and(list);
p = *list;
if (x && p && p->token != GREP_CLOSE_PAREN) {
y = compile_pattern_or(list);
if (!y)
die("not a pattern expression %s", p->pattern);
z = xcalloc(1, sizeof (struct grep_expr));
z->node = GREP_NODE_OR;
z->u.binary.left = x;
z->u.binary.right = y;
return z;
}
return x;
}
static struct grep_expr *compile_pattern_expr(struct grep_pat **list)
{
return compile_pattern_or(list);
}
static void indent(int in)
{
while (in-- > 0)
fputc(' ', stderr);
}
static void dump_grep_pat(struct grep_pat *p)
{
switch (p->token) {
case GREP_AND: fprintf(stderr, "*and*"); break;
case GREP_OPEN_PAREN: fprintf(stderr, "*(*"); break;
case GREP_CLOSE_PAREN: fprintf(stderr, "*)*"); break;
case GREP_NOT: fprintf(stderr, "*not*"); break;
case GREP_OR: fprintf(stderr, "*or*"); break;
case GREP_PATTERN: fprintf(stderr, "pattern"); break;
case GREP_PATTERN_HEAD: fprintf(stderr, "pattern_head"); break;
case GREP_PATTERN_BODY: fprintf(stderr, "pattern_body"); break;
}
switch (p->token) {
default: break;
case GREP_PATTERN_HEAD:
fprintf(stderr, "<head %d>", p->field); break;
case GREP_PATTERN_BODY:
fprintf(stderr, "<body>"); break;
}
switch (p->token) {
default: break;
case GREP_PATTERN_HEAD:
case GREP_PATTERN_BODY:
case GREP_PATTERN:
fprintf(stderr, "%.*s", (int)p->patternlen, p->pattern);
break;
}
fputc('\n', stderr);
}
static void dump_grep_expression_1(struct grep_expr *x, int in)
{
indent(in);
switch (x->node) {
case GREP_NODE_TRUE:
fprintf(stderr, "true\n");
break;
case GREP_NODE_ATOM:
dump_grep_pat(x->u.atom);
break;
case GREP_NODE_NOT:
fprintf(stderr, "(not\n");
dump_grep_expression_1(x->u.unary, in+1);
indent(in);
fprintf(stderr, ")\n");
break;
case GREP_NODE_AND:
fprintf(stderr, "(and\n");
dump_grep_expression_1(x->u.binary.left, in+1);
dump_grep_expression_1(x->u.binary.right, in+1);
indent(in);
fprintf(stderr, ")\n");
break;
case GREP_NODE_OR:
fprintf(stderr, "(or\n");
dump_grep_expression_1(x->u.binary.left, in+1);
dump_grep_expression_1(x->u.binary.right, in+1);
indent(in);
fprintf(stderr, ")\n");
break;
}
}
static void dump_grep_expression(struct grep_opt *opt)
{
struct grep_expr *x = opt->pattern_expression;
if (opt->all_match)
fprintf(stderr, "[all-match]\n");
dump_grep_expression_1(x, 0);
fflush(NULL);
}
static struct grep_expr *grep_true_expr(void)
{
struct grep_expr *z = xcalloc(1, sizeof(*z));
z->node = GREP_NODE_TRUE;
return z;
}
static struct grep_expr *grep_or_expr(struct grep_expr *left, struct grep_expr *right)
{
struct grep_expr *z = xcalloc(1, sizeof(*z));
z->node = GREP_NODE_OR;
z->u.binary.left = left;
z->u.binary.right = right;
return z;
}
static struct grep_expr *prep_header_patterns(struct grep_opt *opt)
{
struct grep_pat *p;
struct grep_expr *header_expr;
struct grep_expr *(header_group[GREP_HEADER_FIELD_MAX]);
enum grep_header_field fld;
if (!opt->header_list)
return NULL;
for (p = opt->header_list; p; p = p->next) {
if (p->token != GREP_PATTERN_HEAD)
die("bug: a non-header pattern in grep header list.");
if (p->field < GREP_HEADER_FIELD_MIN ||
GREP_HEADER_FIELD_MAX <= p->field)
die("bug: unknown header field %d", p->field);
compile_regexp(p, opt);
}
for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++)
header_group[fld] = NULL;
for (p = opt->header_list; p; p = p->next) {
struct grep_expr *h;
struct grep_pat *pp = p;
h = compile_pattern_atom(&pp);
if (!h || pp != p->next)
die("bug: malformed header expr");
if (!header_group[p->field]) {
header_group[p->field] = h;
continue;
}
header_group[p->field] = grep_or_expr(h, header_group[p->field]);
}
header_expr = NULL;
for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) {
if (!header_group[fld])
continue;
if (!header_expr)
header_expr = grep_true_expr();
header_expr = grep_or_expr(header_group[fld], header_expr);
}
return header_expr;
}
log --grep/--author: honor --all-match honored for multiple --grep patterns When we have both header expression (which has to be an OR node by construction) and a pattern expression (which could be anything), we create a new top-level OR node to bind them together, and the resulting expression structure looks like this: OR / \ / \ pattern OR / \ / \ ..... committer OR / \ author TRUE The three elements on the top-level backbone that are inspected by the "all-match" logic are "pattern", "committer" and "author". When there are more than one elements in the "pattern", the top-level node of the "pattern" part of the subtree is an OR, and that node is inspected by "all-match". The result ends up ignoring the "--all-match" given from the command line. A match on either side of the pattern is considered a match, hence: git log --grep=A --grep=B --author=C --all-match shows the same "authored by C and has either A or B" that is correct only when run without "--all-match". Fix this by turning the resulting expression around when "--all-match" is in effect, like this: OR / \ / \ / OR committer / \ author \ pattern The set of nodes on the top-level backbone in the resulting expression becomes "committer", "author", and the nodes that are on the top-level backbone of the "pattern" subexpression. This makes the "all-match" logic inspect the same nodes in "pattern" as the case without the author and/or the committer restriction, and makes the earlier "log" example to show "authored by C and has A and has B", which is what the command line expects. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2012-09-13 23:26:57 +00:00
static struct grep_expr *grep_splice_or(struct grep_expr *x, struct grep_expr *y)
{
struct grep_expr *z = x;
while (x) {
assert(x->node == GREP_NODE_OR);
if (x->u.binary.right &&
x->u.binary.right->node == GREP_NODE_TRUE) {
x->u.binary.right = y;
break;
}
x = x->u.binary.right;
}
return z;
}
static void compile_grep_patterns_real(struct grep_opt *opt)
{
struct grep_pat *p;
struct grep_expr *header_expr = prep_header_patterns(opt);
for (p = opt->pattern_list; p; p = p->next) {
switch (p->token) {
case GREP_PATTERN: /* atom */
case GREP_PATTERN_HEAD:
case GREP_PATTERN_BODY:
compile_regexp(p, opt);
break;
default:
opt->extended = 1;
break;
}
}
if (opt->all_match || header_expr)
opt->extended = 1;
else if (!opt->extended && !opt->debug)
return;
p = opt->pattern_list;
if (p)
opt->pattern_expression = compile_pattern_expr(&p);
if (p)
die("incomplete pattern expression: %s", p->pattern);
if (!header_expr)
return;
if (!opt->pattern_expression)
opt->pattern_expression = header_expr;
log --grep/--author: honor --all-match honored for multiple --grep patterns When we have both header expression (which has to be an OR node by construction) and a pattern expression (which could be anything), we create a new top-level OR node to bind them together, and the resulting expression structure looks like this: OR / \ / \ pattern OR / \ / \ ..... committer OR / \ author TRUE The three elements on the top-level backbone that are inspected by the "all-match" logic are "pattern", "committer" and "author". When there are more than one elements in the "pattern", the top-level node of the "pattern" part of the subtree is an OR, and that node is inspected by "all-match". The result ends up ignoring the "--all-match" given from the command line. A match on either side of the pattern is considered a match, hence: git log --grep=A --grep=B --author=C --all-match shows the same "authored by C and has either A or B" that is correct only when run without "--all-match". Fix this by turning the resulting expression around when "--all-match" is in effect, like this: OR / \ / \ / OR committer / \ author \ pattern The set of nodes on the top-level backbone in the resulting expression becomes "committer", "author", and the nodes that are on the top-level backbone of the "pattern" subexpression. This makes the "all-match" logic inspect the same nodes in "pattern" as the case without the author and/or the committer restriction, and makes the earlier "log" example to show "authored by C and has A and has B", which is what the command line expects. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2012-09-13 23:26:57 +00:00
else if (opt->all_match)
opt->pattern_expression = grep_splice_or(header_expr,
opt->pattern_expression);
else
opt->pattern_expression = grep_or_expr(opt->pattern_expression,
header_expr);
opt->all_match = 1;
}
void compile_grep_patterns(struct grep_opt *opt)
{
compile_grep_patterns_real(opt);
if (opt->debug)
dump_grep_expression(opt);
}
static void free_pattern_expr(struct grep_expr *x)
{
switch (x->node) {
case GREP_NODE_TRUE:
case GREP_NODE_ATOM:
break;
case GREP_NODE_NOT:
free_pattern_expr(x->u.unary);
break;
case GREP_NODE_AND:
case GREP_NODE_OR:
free_pattern_expr(x->u.binary.left);
free_pattern_expr(x->u.binary.right);
break;
}
free(x);
}
void free_grep_patterns(struct grep_opt *opt)
{
struct grep_pat *p, *n;
for (p = opt->pattern_list; p; p = n) {
n = p->next;
switch (p->token) {
case GREP_PATTERN: /* atom */
case GREP_PATTERN_HEAD:
case GREP_PATTERN_BODY:
Use kwset in grep Benchmarks for the hot cache case: before: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 3,478,085 cache-misses # 2.322 M/sec ( +- 2.690% ) 11,356,177 cache-references # 7.582 M/sec ( +- 2.598% ) 3,872,184 branch-misses # 0.363 % ( +- 0.258% ) 1,067,367,848 branches # 712.673 M/sec ( +- 2.622% ) 3,828,370,782 instructions # 0.947 IPC ( +- 0.033% ) 4,043,832,831 cycles # 2700.037 M/sec ( +- 0.167% ) 8,518 page-faults # 0.006 M/sec ( +- 3.648% ) 847 CPU-migrations # 0.001 M/sec ( +- 3.262% ) 6,546 context-switches # 0.004 M/sec ( +- 2.292% ) 1497.695495 task-clock-msecs # 3.303 CPUs ( +- 2.550% ) 0.453394396 seconds time elapsed ( +- 0.912% ) after: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 2,989,918 cache-misses # 3.166 M/sec ( +- 5.013% ) 10,986,041 cache-references # 11.633 M/sec ( +- 4.899% ) (scaled from 95.06%) 3,511,993 branch-misses # 1.422 % ( +- 0.785% ) 246,893,561 branches # 261.433 M/sec ( +- 3.967% ) 1,392,727,757 instructions # 0.564 IPC ( +- 0.040% ) 2,468,142,397 cycles # 2613.494 M/sec ( +- 0.110% ) 7,747 page-faults # 0.008 M/sec ( +- 3.995% ) 897 CPU-migrations # 0.001 M/sec ( +- 2.383% ) 6,535 context-switches # 0.007 M/sec ( +- 1.993% ) 944.384228 task-clock-msecs # 3.177 CPUs ( +- 0.268% ) 0.297257643 seconds time elapsed ( +- 0.450% ) So we gain about 35% by using the kwset code. As a side effect of using kwset two grep tests are fixed by this patch. The first is fixed because kwset can deal with case-insensitive search containing NULs, something strcasestr cannot do. The second one is fixed because we consider patterns containing NULs as fixed strings (regcomp cannot accept patterns with NULs). Signed-off-by: Fredrik Kuivinen <frekui@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-20 22:42:18 +00:00
if (p->kws)
kwsfree(p->kws);
else if (p->pcre_regexp)
free_pcre_regexp(p);
else
regfree(&p->regexp);
free(p->pattern);
break;
default:
break;
}
free(p);
}
if (!opt->extended)
return;
free_pattern_expr(opt->pattern_expression);
}
static char *end_of_line(char *cp, unsigned long *left)
{
unsigned long l = *left;
while (l && *cp != '\n') {
l--;
cp++;
}
*left = l;
return cp;
}
static int word_char(char ch)
{
return isalnum(ch) || ch == '_';
}
static void output_color(struct grep_opt *opt, const void *data, size_t size,
const char *color)
{
color: delay auto-color decision until point of use When we read a color value either from a config file or from the command line, we use git_config_colorbool to convert it from the tristate always/never/auto into a single yes/no boolean value. This has some timing implications with respect to starting a pager. If we start (or decide not to start) the pager before checking the colorbool, everything is fine. Either isatty(1) will give us the right information, or we will properly check for pager_in_use(). However, if we decide to start a pager after we have checked the colorbool, things are not so simple. If stdout is a tty, then we will have already decided to use color. However, the user may also have configured color.pager not to use color with the pager. In this case, we need to actually turn off color. Unfortunately, the pager code has no idea which color variables were turned on (and there are many of them throughout the code, and they may even have been manipulated after the colorbool selection by something like "--color" on the command line). This bug can be seen any time a pager is started after config and command line options are checked. This has affected "git diff" since 89d07f7 (diff: don't run pager if user asked for a diff style exit code, 2007-08-12). It has also affect the log family since 1fda91b (Fix 'git log' early pager startup error case, 2010-08-24). This patch splits the notion of parsing a colorbool and actually checking the configuration. The "use_color" variables now have an additional possible value, GIT_COLOR_AUTO. Users of the variable should use the new "want_color()" wrapper, which will lazily determine and cache the auto-color decision. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-18 05:04:23 +00:00
if (want_color(opt->color) && color && color[0]) {
opt->output(opt, color, strlen(color));
opt->output(opt, data, size);
opt->output(opt, GIT_COLOR_RESET, strlen(GIT_COLOR_RESET));
} else
opt->output(opt, data, size);
}
static void output_sep(struct grep_opt *opt, char sign)
{
if (opt->null_following_name)
opt->output(opt, "\0", 1);
else
output_color(opt, &sign, 1, opt->color_sep);
}
static void show_name(struct grep_opt *opt, const char *name)
{
output_color(opt, name, strlen(name), opt->color_filename);
opt->output(opt, opt->null_following_name ? "\0" : "\n", 1);
}
static int fixmatch(struct grep_pat *p, char *line, char *eol,
regmatch_t *match)
{
Use kwset in grep Benchmarks for the hot cache case: before: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 3,478,085 cache-misses # 2.322 M/sec ( +- 2.690% ) 11,356,177 cache-references # 7.582 M/sec ( +- 2.598% ) 3,872,184 branch-misses # 0.363 % ( +- 0.258% ) 1,067,367,848 branches # 712.673 M/sec ( +- 2.622% ) 3,828,370,782 instructions # 0.947 IPC ( +- 0.033% ) 4,043,832,831 cycles # 2700.037 M/sec ( +- 0.167% ) 8,518 page-faults # 0.006 M/sec ( +- 3.648% ) 847 CPU-migrations # 0.001 M/sec ( +- 3.262% ) 6,546 context-switches # 0.004 M/sec ( +- 2.292% ) 1497.695495 task-clock-msecs # 3.303 CPUs ( +- 2.550% ) 0.453394396 seconds time elapsed ( +- 0.912% ) after: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 2,989,918 cache-misses # 3.166 M/sec ( +- 5.013% ) 10,986,041 cache-references # 11.633 M/sec ( +- 4.899% ) (scaled from 95.06%) 3,511,993 branch-misses # 1.422 % ( +- 0.785% ) 246,893,561 branches # 261.433 M/sec ( +- 3.967% ) 1,392,727,757 instructions # 0.564 IPC ( +- 0.040% ) 2,468,142,397 cycles # 2613.494 M/sec ( +- 0.110% ) 7,747 page-faults # 0.008 M/sec ( +- 3.995% ) 897 CPU-migrations # 0.001 M/sec ( +- 2.383% ) 6,535 context-switches # 0.007 M/sec ( +- 1.993% ) 944.384228 task-clock-msecs # 3.177 CPUs ( +- 0.268% ) 0.297257643 seconds time elapsed ( +- 0.450% ) So we gain about 35% by using the kwset code. As a side effect of using kwset two grep tests are fixed by this patch. The first is fixed because kwset can deal with case-insensitive search containing NULs, something strcasestr cannot do. The second one is fixed because we consider patterns containing NULs as fixed strings (regcomp cannot accept patterns with NULs). Signed-off-by: Fredrik Kuivinen <frekui@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-20 22:42:18 +00:00
struct kwsmatch kwsm;
size_t offset = kwsexec(p->kws, line, eol - line, &kwsm);
if (offset == -1) {
match->rm_so = match->rm_eo = -1;
return REG_NOMATCH;
Use kwset in grep Benchmarks for the hot cache case: before: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 3,478,085 cache-misses # 2.322 M/sec ( +- 2.690% ) 11,356,177 cache-references # 7.582 M/sec ( +- 2.598% ) 3,872,184 branch-misses # 0.363 % ( +- 0.258% ) 1,067,367,848 branches # 712.673 M/sec ( +- 2.622% ) 3,828,370,782 instructions # 0.947 IPC ( +- 0.033% ) 4,043,832,831 cycles # 2700.037 M/sec ( +- 0.167% ) 8,518 page-faults # 0.006 M/sec ( +- 3.648% ) 847 CPU-migrations # 0.001 M/sec ( +- 3.262% ) 6,546 context-switches # 0.004 M/sec ( +- 2.292% ) 1497.695495 task-clock-msecs # 3.303 CPUs ( +- 2.550% ) 0.453394396 seconds time elapsed ( +- 0.912% ) after: $ perf stat --repeat=5 git grep qwerty > /dev/null Performance counter stats for 'git grep qwerty' (5 runs): 2,989,918 cache-misses # 3.166 M/sec ( +- 5.013% ) 10,986,041 cache-references # 11.633 M/sec ( +- 4.899% ) (scaled from 95.06%) 3,511,993 branch-misses # 1.422 % ( +- 0.785% ) 246,893,561 branches # 261.433 M/sec ( +- 3.967% ) 1,392,727,757 instructions # 0.564 IPC ( +- 0.040% ) 2,468,142,397 cycles # 2613.494 M/sec ( +- 0.110% ) 7,747 page-faults # 0.008 M/sec ( +- 3.995% ) 897 CPU-migrations # 0.001 M/sec ( +- 2.383% ) 6,535 context-switches # 0.007 M/sec ( +- 1.993% ) 944.384228 task-clock-msecs # 3.177 CPUs ( +- 0.268% ) 0.297257643 seconds time elapsed ( +- 0.450% ) So we gain about 35% by using the kwset code. As a side effect of using kwset two grep tests are fixed by this patch. The first is fixed because kwset can deal with case-insensitive search containing NULs, something strcasestr cannot do. The second one is fixed because we consider patterns containing NULs as fixed strings (regcomp cannot accept patterns with NULs). Signed-off-by: Fredrik Kuivinen <frekui@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2011-08-20 22:42:18 +00:00
} else {
match->rm_so = offset;
match->rm_eo = match->rm_so + kwsm.size[0];
return 0;
}
}
static int regmatch(const regex_t *preg, char *line, char *eol,
regmatch_t *match, int eflags)
{
#ifdef REG_STARTEND
match->rm_so = 0;
match->rm_eo = eol - line;
eflags |= REG_STARTEND;
#endif
return regexec(preg, line, 1, match, eflags);
}
static int patmatch(struct grep_pat *p, char *line, char *eol,
regmatch_t *match, int eflags)
{
int hit;
if (p->fixed)
hit = !fixmatch(p, line, eol, match);
else if (p->pcre_regexp)
hit = !pcrematch(p, line, eol, match, eflags);
else
hit = !regmatch(&p->regexp, line, eol, match, eflags);
return hit;
}
log --author/--committer: really match only with name part When we tried to find commits done by AUTHOR, the first implementation tried to pattern match a line with "^author .*AUTHOR", which later was enhanced to strip leading caret and look for "^author AUTHOR" when the search pattern was anchored at the left end (i.e. --author="^AUTHOR"). This had a few problems: * When looking for fixed strings (e.g. "git log -F --author=x --grep=y"), the regexp internally used "^author .*x" would never match anything; * To match at the end (e.g. "git log --author='google.com>$'"), the generated regexp has to also match the trailing timestamp part the commit header lines have. Also, in order to determine if the '$' at the end means "match at the end of the line" or just a literal dollar sign (probably backslash-quoted), we would need to parse the regexp ourselves. An earlier alternative tried to make sure that a line matches "^author " (to limit by field name) and the user supplied pattern at the same time. While it solved the -F problem by introducing a special override for matching the "^author ", it did not solve the trailing timestamp nor tail match problem. It also would have matched every commit if --author=author was asked for, not because the author's email part had this string, but because every commit header line that talks about the author begins with that field name, regardleses of who wrote it. Instead of piling more hacks on top of hacks, this rethinks the grep machinery that is used to look for strings in the commit header, and makes sure that (1) field name matches literally at the beginning of the line, followed by a SP, and (2) the user supplied pattern is matched against the remainder of the line, excluding the trailing timestamp data. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2008-09-05 05:15:02 +00:00
static int strip_timestamp(char *bol, char **eol_p)
{
char *eol = *eol_p;
int ch;
while (bol < --eol) {
if (*eol != '>')
continue;
*eol_p = ++eol;
ch = *eol;
*eol = '\0';
return ch;
}
return 0;
}
static struct {
const char *field;
size_t len;
} header_field[] = {
{ "author ", 7 },
{ "committer ", 10 },
{ "reflog ", 7 },
log --author/--committer: really match only with name part When we tried to find commits done by AUTHOR, the first implementation tried to pattern match a line with "^author .*AUTHOR", which later was enhanced to strip leading caret and look for "^author AUTHOR" when the search pattern was anchored at the left end (i.e. --author="^AUTHOR"). This had a few problems: * When looking for fixed strings (e.g. "git log -F --author=x --grep=y"), the regexp internally used "^author .*x" would never match anything; * To match at the end (e.g. "git log --author='google.com>$'"), the generated regexp has to also match the trailing timestamp part the commit header lines have. Also, in order to determine if the '$' at the end means "match at the end of the line" or just a literal dollar sign (probably backslash-quoted), we would need to parse the regexp ourselves. An earlier alternative tried to make sure that a line matches "^author " (to limit by field name) and the user supplied pattern at the same time. While it solved the -F problem by introducing a special override for matching the "^author ", it did not solve the trailing timestamp nor tail match problem. It also would have matched every commit if --author=author was asked for, not because the author's email part had this string, but because every commit header line that talks about the author begins with that field name, regardleses of who wrote it. Instead of piling more hacks on top of hacks, this rethinks the grep machinery that is used to look for strings in the commit header, and makes sure that (1) field name matches literally at the beginning of the line, followed by a SP, and (2) the user supplied pattern is matched against the remainder of the line, excluding the trailing timestamp data. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2008-09-05 05:15:02 +00:00
};
static int match_one_pattern(struct grep_pat *p, char *bol, char *eol,
enum grep_context ctx,
regmatch_t *pmatch, int eflags)
{
int hit = 0;
log --author/--committer: really match only with name part When we tried to find commits done by AUTHOR, the first implementation tried to pattern match a line with "^author .*AUTHOR", which later was enhanced to strip leading caret and look for "^author AUTHOR" when the search pattern was anchored at the left end (i.e. --author="^AUTHOR"). This had a few problems: * When looking for fixed strings (e.g. "git log -F --author=x --grep=y"), the regexp internally used "^author .*x" would never match anything; * To match at the end (e.g. "git log --author='google.com>$'"), the generated regexp has to also match the trailing timestamp part the commit header lines have. Also, in order to determine if the '$' at the end means "match at the end of the line" or just a literal dollar sign (probably backslash-quoted), we would need to parse the regexp ourselves. An earlier alternative tried to make sure that a line matches "^author " (to limit by field name) and the user supplied pattern at the same time. While it solved the -F problem by introducing a special override for matching the "^author ", it did not solve the trailing timestamp nor tail match problem. It also would have matched every commit if --author=author was asked for, not because the author's email part had this string, but because every commit header line that talks about the author begins with that field name, regardleses of who wrote it. Instead of piling more hacks on top of hacks, this rethinks the grep machinery that is used to look for strings in the commit header, and makes sure that (1) field name matches literally at the beginning of the line, followed by a SP, and (2) the user supplied pattern is matched against the remainder of the line, excluding the trailing timestamp data. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2008-09-05 05:15:02 +00:00
int saved_ch = 0;
const char *start = bol;
if ((p->token != GREP_PATTERN) &&
((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
return 0;
log --author/--committer: really match only with name part When we tried to find commits done by AUTHOR, the first implementation tried to pattern match a line with "^author .*AUTHOR", which later was enhanced to strip leading caret and look for "^author AUTHOR" when the search pattern was anchored at the left end (i.e. --author="^AUTHOR"). This had a few problems: * When looking for fixed strings (e.g. "git log -F --author=x --grep=y"), the regexp internally used "^author .*x" would never match anything; * To match at the end (e.g. "git log --author='google.com>$'"), the generated regexp has to also match the trailing timestamp part the commit header lines have. Also, in order to determine if the '$' at the end means "match at the end of the line" or just a literal dollar sign (probably backslash-quoted), we would need to parse the regexp ourselves. An earlier alternative tried to make sure that a line matches "^author " (to limit by field name) and the user supplied pattern at the same time. While it solved the -F problem by introducing a special override for matching the "^author ", it did not solve the trailing timestamp nor tail match problem. It also would have matched every commit if --author=author was asked for, not because the author's email part had this string, but because every commit header line that talks about the author begins with that field name, regardleses of who wrote it. Instead of piling more hacks on top of hacks, this rethinks the grep machinery that is used to look for strings in the commit header, and makes sure that (1) field name matches literally at the beginning of the line, followed by a SP, and (2) the user supplied pattern is matched against the remainder of the line, excluding the trailing timestamp data. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2008-09-05 05:15:02 +00:00
if (p->token == GREP_PATTERN_HEAD) {
const char *field;
size_t len;
assert(p->field < ARRAY_SIZE(header_field));
field = header_field[p->field].field;
len = header_field[p->field].len;
if (strncmp(bol, field, len))
return 0;
bol += len;
switch (p->field) {
case GREP_HEADER_AUTHOR:
case GREP_HEADER_COMMITTER:
saved_ch = strip_timestamp(bol, &eol);
break;
default:
break;
}
log --author/--committer: really match only with name part When we tried to find commits done by AUTHOR, the first implementation tried to pattern match a line with "^author .*AUTHOR", which later was enhanced to strip leading caret and look for "^author AUTHOR" when the search pattern was anchored at the left end (i.e. --author="^AUTHOR"). This had a few problems: * When looking for fixed strings (e.g. "git log -F --author=x --grep=y"), the regexp internally used "^author .*x" would never match anything; * To match at the end (e.g. "git log --author='google.com>$'"), the generated regexp has to also match the trailing timestamp part the commit header lines have. Also, in order to determine if the '$' at the end means "match at the end of the line" or just a literal dollar sign (probably backslash-quoted), we would need to parse the regexp ourselves. An earlier alternative tried to make sure that a line matches "^author " (to limit by field name) and the user supplied pattern at the same time. While it solved the -F problem by introducing a special override for matching the "^author ", it did not solve the trailing timestamp nor tail match problem. It also would have matched every commit if --author=author was asked for, not because the author's email part had this string, but because every commit header line that talks about the author begins with that field name, regardleses of who wrote it. Instead of piling more hacks on top of hacks, this rethinks the grep machinery that is used to look for strings in the commit header, and makes sure that (1) field name matches literally at the beginning of the line, followed by a SP, and (2) the user supplied pattern is matched against the remainder of the line, excluding the trailing timestamp data. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2008-09-05 05:15:02 +00:00
}
again:
hit = patmatch(p, bol, eol, pmatch, eflags);
if (hit && p->word_regexp) {
if ((pmatch[0].rm_so < 0) ||
(eol - bol) < pmatch[0].rm_so ||
(pmatch[0].rm_eo < 0) ||
(eol - bol) < pmatch[0].rm_eo)
die("regexp returned nonsense");
/* Match beginning must be either beginning of the
* line, or at word boundary (i.e. the last char must
* not be a word char). Similarly, match end must be
* either end of the line, or at word boundary
* (i.e. the next char must not be a word char).
*/
if ( ((pmatch[0].rm_so == 0) ||
!word_char(bol[pmatch[0].rm_so-1])) &&
((pmatch[0].rm_eo == (eol-bol)) ||
!word_char(bol[pmatch[0].rm_eo])) )
;
else
hit = 0;
/* Words consist of at least one character. */
if (pmatch->rm_so == pmatch->rm_eo)
hit = 0;
if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
/* There could be more than one match on the
* line, and the first match might not be
* strict word match. But later ones could be!
* Forward to the next possible start, i.e. the
* next position following a non-word char.
*/
bol = pmatch[0].rm_so + bol + 1;
while (word_char(bol[-1]) && bol < eol)
bol++;
eflags |= REG_NOTBOL;
if (bol < eol)
goto again;
}
}
log --author/--committer: really match only with name part When we tried to find commits done by AUTHOR, the first implementation tried to pattern match a line with "^author .*AUTHOR", which later was enhanced to strip leading caret and look for "^author AUTHOR" when the search pattern was anchored at the left end (i.e. --author="^AUTHOR"). This had a few problems: * When looking for fixed strings (e.g. "git log -F --author=x --grep=y"), the regexp internally used "^author .*x" would never match anything; * To match at the end (e.g. "git log --author='google.com>$'"), the generated regexp has to also match the trailing timestamp part the commit header lines have. Also, in order to determine if the '$' at the end means "match at the end of the line" or just a literal dollar sign (probably backslash-quoted), we would need to parse the regexp ourselves. An earlier alternative tried to make sure that a line matches "^author " (to limit by field name) and the user supplied pattern at the same time. While it solved the -F problem by introducing a special override for matching the "^author ", it did not solve the trailing timestamp nor tail match problem. It also would have matched every commit if --author=author was asked for, not because the author's email part had this string, but because every commit header line that talks about the author begins with that field name, regardleses of who wrote it. Instead of piling more hacks on top of hacks, this rethinks the grep machinery that is used to look for strings in the commit header, and makes sure that (1) field name matches literally at the beginning of the line, followed by a SP, and (2) the user supplied pattern is matched against the remainder of the line, excluding the trailing timestamp data. Signed-off-by: Junio C Hamano <gitster@pobox.com>
2008-09-05 05:15:02 +00:00
if (p->token == GREP_PATTERN_HEAD && saved_ch)
*eol = saved_ch;
if (hit) {
pmatch[0].rm_so += bol - start;
pmatch[0].rm_eo += bol - start;
}
return hit;
}
static int match_expr_eval(struct grep_expr *x, char *bol, char *eol,
enum grep_context ctx, int collect_hits)
{
int h = 0;
regmatch_t match;
if (!x)
die("Not a valid grep expression");
switch (x->node) {
case GREP_NODE_TRUE:
h = 1;
break;
case GREP_NODE_ATOM:
h = match_one_pattern(x->u.atom, bol, eol, ctx, &match, 0);
break;
case GREP_NODE_NOT:
h = !match_expr_eval(x->u.unary, bol, eol, ctx, 0);
break;
case GREP_NODE_AND:
if (!match_expr_eval(x->u.binary.left, bol, eol, ctx, 0))
return 0;
h = match_expr_eval(x->u.binary.right, bol, eol, ctx, 0);
break;
case GREP_NODE_OR:
if (!collect_hits)
return (match_expr_eval(x->u.binary.left,
bol, eol, ctx, 0) ||
match_expr_eval(x->u.binary.right,
bol, eol, ctx, 0));
h = match_expr_eval(x->u.binary.left, bol, eol, ctx, 0);
x->u.binary.left->hit |= h;
h |= match_expr_eval(x->u.binary.right, bol, eol, ctx, 1);
break;
default:
die("Unexpected node type (internal error) %d", x->node);
}
if (collect_hits)
x->hit |= h;
return h;
}
static int match_expr(struct grep_opt *opt, char *bol, char *eol,
enum grep_context ctx, int collect_hits)
{
struct grep_expr *x = opt->pattern_expression;
return match_expr_eval(x, bol, eol, ctx, collect_hits);
}
static int match_line(struct grep_opt *opt, char *bol, char *eol,
enum grep_context ctx, int collect_hits)
{
struct grep_pat *p;
regmatch_t match;
if (opt->extended)
return match_expr(opt, bol, eol, ctx, collect_hits);
/* we do not call with collect_hits without being extended */
for (p = opt->pattern_list; p; p = p->next) {
if (match_one_pattern(p, bol, eol, ctx, &match, 0))
return 1;
}
return 0;
}
static int match_next_pattern(struct grep_pat *p, char *bol, char *eol,
enum grep_context ctx,
regmatch_t *pmatch, int eflags)
{
regmatch_t match;
if (!match_one_pattern(p, bol, eol, ctx, &match, eflags))
return 0;
if (match.rm_so < 0 || match.rm_eo < 0)
return 0;
if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) {
if (match.rm_so > pmatch->rm_so)
return 1;
if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo)
return 1;
}
pmatch->rm_so = match.rm_so;
pmatch->rm_eo = match.rm_eo;
return 1;
}
static int next_match(struct grep_opt *opt, char *bol, char *eol,
enum grep_context ctx, regmatch_t *pmatch, int eflags)
{
struct grep_pat *p;
int hit = 0;
pmatch->rm_so = pmatch->rm_eo = -1;
if (bol < eol) {
for (p = opt->pattern_list; p; p = p->next) {
switch (p->token) {
case GREP_PATTERN: /* atom */
case GREP_PATTERN_HEAD:
case GREP_PATTERN_BODY:
hit |= match_next_pattern(p, bol, eol, ctx,
pmatch, eflags);
break;
default:
break;
}
}
}
return hit;
}
static void show_line(struct grep_opt *opt, char *bol, char *eol,
const char *name, unsigned lno, char sign)
{
int rest = eol - bol;
const char *match_color, *line_color = NULL;
if (opt->file_break && opt->last_shown == 0) {
if (opt->show_hunk_mark)
opt->output(opt, "\n", 1);
} else if (opt->pre_context || opt->post_context || opt->funcbody) {
if (opt->last_shown == 0) {
if (opt->show_hunk_mark) {
output_color(opt, "--", 2, opt->color_sep);
opt->output(opt, "\n", 1);
}
} else if (lno > opt->last_shown + 1) {
output_color(opt, "--", 2, opt->color_sep);
opt->output(opt, "\n", 1);
}
}
if (opt->heading && opt->last_shown == 0) {
output_color(opt, name, strlen(name), opt->color_filename);
opt->output(opt, "\n", 1);
}
opt->last_shown = lno;
if (!opt->heading && opt->pathname) {
output_color(opt, name, strlen(name), opt->color_filename);
output_sep(opt, sign);
}
if (opt->linenum) {
char buf[32];
snprintf(buf, sizeof(buf), "%d", lno);
output_color(opt, buf, strlen(buf), opt->color_lineno);
output_sep(opt, sign);
}
if (opt->color) {
regmatch_t match;
enum grep_context ctx = GREP_CONTEXT_BODY;
int ch = *eol;
int eflags = 0;
if (sign == ':')
match_color = opt->color_match_selected;
else
match_color = opt->color_match_context;
if (sign == ':')
line_color = opt->color_selected;
else if (sign == '-')
line_color = opt->color_context;
else if (sign == '=')
line_color = opt->color_function;
*eol = '\0';
while (next_match(opt, bol, eol, ctx, &match, eflags)) {
if (match.rm_so == match.rm_eo)
break;
output_color(opt, bol, match.rm_so, line_color);
output_color(opt, bol + match.rm_so,
match.rm_eo - match.rm_so, match_color);
bol += match.rm_eo;
rest -= match.rm_eo;
eflags = REG_NOTBOL;
}
*eol = ch;
}
output_color(opt, bol, rest, line_color);
opt->output(opt, "\n", 1);
}
#ifndef NO_PTHREADS
grep: make locking flag global The low-level grep code traditionally didn't care about threading, as it doesn't do any threading itself and didn't call out to other non-thread-safe code. That changed with 0579f91 (grep: enable threading with -p and -W using lazy attribute lookup, 2011-12-12), which pushed the lookup of funcname attributes (which is not thread-safe) into the low-level grep code. As a result, the low-level code learned about a new global "grep_attr_mutex" to serialize access to the attribute code. A multi-threaded caller (e.g., builtin/grep.c) is expected to initialize the mutex and set "use_threads" in the grep_opt structure. The low-level code only uses the lock if use_threads is set. However, putting the use_threads flag into the grep_opt struct is not the most logical place. Whether threading is in use is not something that matters for each call to grep_buffer, but is instead global to the whole program (i.e., if any thread is doing multi-threaded grep, every other thread, even if it thinks it is doing its own single-threaded grep, would need to use the locking). In practice, this distinction isn't a problem for us, because the only user of multi-threaded grep is "git-grep", which does nothing except call grep. This patch turns the opt->use_threads flag into a global flag. More important than the nit-picking semantic argument above is that this means that the locking functions don't need to actually have access to a grep_opt to know whether to lock. Which in turn can make adding new locks simpler, as we don't need to pass around a grep_opt. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2012-02-02 08:18:29 +00:00
int grep_use_locks;
/*
* This lock protects access to the gitattributes machinery, which is
* not thread-safe.
*/
pthread_mutex_t grep_attr_mutex;
grep: make locking flag global The low-level grep code traditionally didn't care about threading, as it doesn't do any threading itself and didn't call out to other non-thread-safe code. That changed with 0579f91 (grep: enable threading with -p and -W using lazy attribute lookup, 2011-12-12), which pushed the lookup of funcname attributes (which is not thread-safe) into the low-level grep code. As a result, the low-level code learned about a new global "grep_attr_mutex" to serialize access to the attribute code. A multi-threaded caller (e.g., builtin/grep.c) is expected to initialize the mutex and set "use_threads" in the grep_opt structure. The low-level code only uses the lock if use_threads is set. However, putting the use_threads flag into the grep_opt struct is not the most logical place. Whether threading is in use is not something that matters for each call to grep_buffer, but is instead global to the whole program (i.e., if any thread is doing multi-threaded grep, every other thread, even if it thinks it is doing its own single-threaded grep, would need to use the locking). In practice, this distinction isn't a problem for us, because the only user of multi-threaded grep is "git-grep", which does nothing except call grep. This patch turns the opt->use_threads flag into a global flag. More important than the nit-picking semantic argument above is that this means that the locking functions don't need to actually have access to a grep_opt to know whether to lock. Which in turn can make adding new locks simpler, as we don't need to pass around a grep_opt. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2012-02-02 08:18:29 +00:00
static inline void grep_attr_lock(void)
{
grep: make locking flag global The low-level grep code traditionally didn't care about threading, as it doesn't do any threading itself and didn't call out to other non-thread-safe code. That changed with 0579f91 (grep: enable threading with -p and -W using lazy attribute lookup, 2011-12-12), which pushed the lookup of funcname attributes (which is not thread-safe) into the low-level grep code. As a result, the low-level code learned about a new global "grep_attr_mutex" to serialize access to the attribute code. A multi-threaded caller (e.g., builtin/grep.c) is expected to initialize the mutex and set "use_threads" in the grep_opt structure. The low-level code only uses the lock if use_threads is set. However, putting the use_threads flag into the grep_opt struct is not the most logical place. Whether threading is in use is not something that matters for each call to grep_buffer, but is instead global to the whole program (i.e., if any thread is doing multi-threaded grep, every other thread, even if it thinks it is doing its own single-threaded grep, would need to use the locking). In practice, this distinction isn't a problem for us, because the only user of multi-threaded grep is "git-grep", which does nothing except call grep. This patch turns the opt->use_threads flag into a global flag. More important than the nit-picking semantic argument above is that this means that the locking functions don't need to actually have access to a grep_opt to know whether to lock. Which in turn can make adding new locks simpler, as we don't need to pass around a grep_opt. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2012-02-02 08:18:29 +00:00
if (grep_use_locks)
pthread_mutex_lock(&grep_attr_mutex);
}
grep: make locking flag global The low-level grep code traditionally didn't care about threading, as it doesn't do any threading itself and didn't call out to other non-thread-safe code. That changed with 0579f91 (grep: enable threading with -p and -W using lazy attribute lookup, 2011-12-12), which pushed the lookup of funcname attributes (which is not thread-safe) into the low-level grep code. As a result, the low-level code learned about a new global "grep_attr_mutex" to serialize access to the attribute code. A multi-threaded caller (e.g., builtin/grep.c) is expected to initialize the mutex and set "use_threads" in the grep_opt structure. The low-level code only uses the lock if use_threads is set. However, putting the use_threads flag into the grep_opt struct is not the most logical place. Whether threading is in use is not something that matters for each call to grep_buffer, but is instead global to the whole program (i.e., if any thread is doing multi-threaded grep, every other thread, even if it thinks it is doing its own single-threaded grep, would need to use the locking). In practice, this distinction isn't a problem for us, because the only user of multi-threaded grep is "git-grep", which does nothing except call grep. This patch turns the opt->use_threads flag into a global flag. More important than the nit-picking semantic argument above is that this means that the locking functions don't need to actually have access to a grep_opt to know whether to lock. Which in turn can make adding new locks simpler, as we don't need to pass around a grep_opt. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2012-02-02 08:18:29 +00:00
static inline void grep_attr_unlock(void)
{
grep: make locking flag global The low-level grep code traditionally didn't care about threading, as it doesn't do any threading itself and didn't call out to other non-thread-safe code. That changed with 0579f91 (grep: enable threading with -p and -W using lazy attribute lookup, 2011-12-12), which pushed the lookup of funcname attributes (which is not thread-safe) into the low-level grep code. As a result, the low-level code learned about a new global "grep_attr_mutex" to serialize access to the attribute code. A multi-threaded caller (e.g., builtin/grep.c) is expected to initialize the mutex and set "use_threads" in the grep_opt structure. The low-level code only uses the lock if use_threads is set. However, putting the use_threads flag into the grep_opt struct is not the most logical place. Whether threading is in use is not something that matters for each call to grep_buffer, but is instead global to the whole program (i.e., if any thread is doing multi-threaded grep, every other thread, even if it thinks it is doing its own single-threaded grep, would need to use the locking). In practice, this distinction isn't a problem for us, because the only user of multi-threaded grep is "git-grep", which does nothing except call grep. This patch turns the opt->use_threads flag into a global flag. More important than the nit-picking semantic argument above is that this means that the locking functions don't need to actually have access to a grep_opt to know whether to lock. Which in turn can make adding new locks simpler, as we don't need to pass around a grep_opt. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2012-02-02 08:18:29 +00:00
if (grep_use_locks)
pthread_mutex_unlock(&grep_attr_mutex);
}
/*
* Same as git_attr_mutex, but protecting the thread-unsafe object db access.
*/
pthread_mutex_t grep_read_mutex;
#else
grep: make locking flag global The low-level grep code traditionally didn't care about threading, as it doesn't do any threading itself and didn't call out to other non-thread-safe code. That changed with 0579f91 (grep: enable threading with -p and -W using lazy attribute lookup, 2011-12-12), which pushed the lookup of funcname attributes (which is not thread-safe) into the low-level grep code. As a result, the low-level code learned about a new global "grep_attr_mutex" to serialize access to the attribute code. A multi-threaded caller (e.g., builtin/grep.c) is expected to initialize the mutex and set "use_threads" in the grep_opt structure. The low-level code only uses the lock if use_threads is set. However, putting the use_threads flag into the grep_opt struct is not the most logical place. Whether threading is in use is not something that matters for each call to grep_buffer, but is instead global to the whole program (i.e., if any thread is doing multi-threaded grep, every other thread, even if it thinks it is doing its own single-threaded grep, would need to use the locking). In practice, this distinction isn't a problem for us, because the only user of multi-threaded grep is "git-grep", which does nothing except call grep. This patch turns the opt->use_threads flag into a global flag. More important than the nit-picking semantic argument above is that this means that the locking functions don't need to actually have access to a grep_opt to know whether to lock. Which in turn can make adding new locks simpler, as we don't need to pass around a grep_opt. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2012-02-02 08:18:29 +00:00
#define grep_attr_lock()
#define grep_attr_unlock()
#endif
static int match_funcname(struct grep_opt *opt, struct grep_source *gs, char *bol, char *eol)
{
xdemitconf_t *xecfg = opt->priv;
if (xecfg && !xecfg->find_func) {
grep_source_load_driver(gs);
if (gs->driver->funcname.pattern) {
const struct userdiff_funcname *pe = &gs->driver->funcname;
xdiff_set_find_func(xecfg, pe->pattern, pe->cflags);
} else {
xecfg = opt->priv = NULL;
}
}
if (xecfg) {
char buf[1];
return xecfg->find_func(bol, eol - bol, buf, 1,
xecfg->find_func_priv) >= 0;
}
if (bol == eol)
return 0;
if (isalpha(*bol) || *bol == '_' || *bol == '$')
return 1;
return 0;
}
static void show_funcname_line(struct grep_opt *opt, struct grep_source *gs,
char *bol, unsigned lno)
{
while (bol > gs->buf) {
char *eol = --bol;
while (bol > gs->buf && bol[-1] != '\n')
bol--;
lno--;
if (lno <= opt->last_shown)
break;
if (match_funcname(opt, gs, bol, eol)) {
show_line(opt, bol, eol, gs->name, lno, '=');
break;
}
}
}
static void show_pre_context(struct grep_opt *opt, struct grep_source *gs,
char *bol, char *end, unsigned lno)
{
unsigned cur = lno, from = 1, funcname_lno = 0;
int funcname_needed = !!opt->funcname;
if (opt->funcbody && !match_funcname(opt, gs, bol, end))
funcname_needed = 2;
if (opt->pre_context < lno)
from = lno - opt->pre_context;
if (from <= opt->last_shown)
from = opt->last_shown + 1;
/* Rewind. */
while (bol > gs->buf &&
cur > (funcname_needed == 2 ? opt->last_shown + 1 : from)) {
char *eol = --bol;
while (bol > gs->buf && bol[-1] != '\n')
bol--;
cur--;
if (funcname_needed && match_funcname(opt, gs, bol, eol)) {
funcname_lno = cur;
funcname_needed = 0;
}
}
/* We need to look even further back to find a function signature. */
if (opt->funcname && funcname_needed)
show_funcname_line(opt, gs, bol, cur);
/* Back forward. */
while (cur < lno) {
char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-';
while (*eol != '\n')
eol++;
show_line(opt, bol, eol, gs->name, cur, sign);
bol = eol + 1;
cur++;
}
}
static int should_lookahead(struct grep_opt *opt)
{
struct grep_pat *p;
if (opt->extended)
return 0; /* punt for too complex stuff */
if (opt->invert)
return 0;
for (p = opt->pattern_list; p; p = p->next) {
if (p->token != GREP_PATTERN)
return 0; /* punt for "header only" and stuff */
}
return 1;
}
static int look_ahead(struct grep_opt *opt,
unsigned long *left_p,
unsigned *lno_p,
char **bol_p)
{
unsigned lno = *lno_p;
char *bol = *bol_p;
struct grep_pat *p;
char *sp, *last_bol;
regoff_t earliest = -1;
for (p = opt->pattern_list; p; p = p->next) {
int hit;
regmatch_t m;
hit = patmatch(p, bol, bol + *left_p, &m, 0);
if (!hit || m.rm_so < 0 || m.rm_eo < 0)
continue;
if (earliest < 0 || m.rm_so < earliest)
earliest = m.rm_so;
}
if (earliest < 0) {
*bol_p = bol + *left_p;
*left_p = 0;
return 1;
}
for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--)
; /* find the beginning of the line */
last_bol = sp;
for (sp = bol; sp < last_bol; sp++) {
if (*sp == '\n')
lno++;
}
*left_p -= last_bol - bol;
*bol_p = last_bol;
*lno_p = lno;
return 0;
}
static void std_output(struct grep_opt *opt, const void *buf, size_t size)
{
fwrite(buf, size, 1, stdout);
}
static int fill_textconv_grep(struct userdiff_driver *driver,
struct grep_source *gs)
{
struct diff_filespec *df;
char *buf;
size_t size;
if (!driver || !driver->textconv)
return grep_source_load(gs);
/*
* The textconv interface is intimately tied to diff_filespecs, so we
* have to pretend to be one. If we could unify the grep_source
* and diff_filespec structs, this mess could just go away.
*/
df = alloc_filespec(gs->path);
switch (gs->type) {
case GREP_SOURCE_SHA1:
fill_filespec(df, gs->identifier, 1, 0100644);
break;
case GREP_SOURCE_FILE:
fill_filespec(df, null_sha1, 0, 0100644);
break;
default:
die("BUG: attempt to textconv something without a path?");
}
/*
* fill_textconv is not remotely thread-safe; it may load objects
* behind the scenes, and it modifies the global diff tempfile
* structure.
*/
grep_read_lock();
size = fill_textconv(driver, df, &buf);
grep_read_unlock();
free_filespec(df);
/*
* The normal fill_textconv usage by the diff machinery would just keep
* the textconv'd buf separate from the diff_filespec. But much of the
* grep code passes around a grep_source and assumes that its "buf"
* pointer is the beginning of the thing we are searching. So let's
* install our textconv'd version into the grep_source, taking care not
* to leak any existing buffer.
*/
grep_source_clear_data(gs);
gs->buf = buf;
gs->size = size;
return 0;
}
static int is_empty_line(const char *bol, const char *eol)
{
while (bol < eol && isspace(*bol))
bol++;
return bol == eol;
}
static int grep_source_1(struct grep_opt *opt, struct grep_source *gs, int collect_hits)
{
char *bol;
char *peek_bol = NULL;
unsigned long left;
unsigned lno = 1;
unsigned last_hit = 0;
int binary_match_only = 0;
unsigned count = 0;
int try_lookahead = 0;
int show_function = 0;
struct userdiff_driver *textconv = NULL;
enum grep_context ctx = GREP_CONTEXT_HEAD;
xdemitconf_t xecfg;
if (!opt->output)
opt->output = std_output;
if (opt->pre_context || opt->post_context || opt->file_break ||
opt->funcbody) {
/* Show hunk marks, except for the first file. */
if (opt->last_shown)
opt->show_hunk_mark = 1;
/*
* If we're using threads then we can't easily identify
* the first file. Always put hunk marks in that case
* and skip the very first one later in work_done().
*/
if (opt->output != std_output)
opt->show_hunk_mark = 1;
}
opt->last_shown = 0;
if (opt->allow_textconv) {
grep_source_load_driver(gs);
/*
* We might set up the shared textconv cache data here, which
* is not thread-safe.
*/
grep_attr_lock();
textconv = userdiff_get_textconv(gs->driver);
grep_attr_unlock();
}
/*
* We know the result of a textconv is text, so we only have to care
* about binary handling if we are not using it.
*/
if (!textconv) {
switch (opt->binary) {
case GREP_BINARY_DEFAULT:
if (grep_source_is_binary(gs))
binary_match_only = 1;
break;
case GREP_BINARY_NOMATCH:
if (grep_source_is_binary(gs))
return 0; /* Assume unmatch */
break;
case GREP_BINARY_TEXT:
break;
default:
die("bug: unknown binary handling mode");
}
}
memset(&xecfg, 0, sizeof(xecfg));
opt->priv = &xecfg;
try_lookahead = should_lookahead(opt);
if (fill_textconv_grep(textconv, gs) < 0)
return 0;
bol = gs->buf;
left = gs->size;
while (left) {
char *eol, ch;
int hit;
/*
* look_ahead() skips quickly to the line that possibly
* has the next hit; don't call it if we need to do
* something more than just skipping the current line
* in response to an unmatch for the current line. E.g.
* inside a post-context window, we will show the current
* line as a context around the previous hit when it
* doesn't hit.
*/
if (try_lookahead
&& !(last_hit
&& (show_function ||
lno <= last_hit + opt->post_context))
&& look_ahead(opt, &left, &lno, &bol))
break;
eol = end_of_line(bol, &left);
ch = *eol;
*eol = 0;
if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
ctx = GREP_CONTEXT_BODY;
hit = match_line(opt, bol, eol, ctx, collect_hits);
*eol = ch;
if (collect_hits)
goto next_line;
/* "grep -v -e foo -e bla" should list lines
* that do not have either, so inversion should
* be done outside.
*/
if (opt->invert)
hit = !hit;
if (opt->unmatch_name_only) {
if (hit)
return 0;
goto next_line;
}
if (hit) {
count++;
if (opt->status_only)
return 1;
if (opt->name_only) {
show_name(opt, gs->name);
return 1;
}
if (opt->count)
goto next_line;
if (binary_match_only) {
opt->output(opt, "Binary file ", 12);
output_color(opt, gs->name, strlen(gs->name),
opt->color_filename);
opt->output(opt, " matches\n", 9);
return 1;
}
/* Hit at this line. If we haven't shown the
* pre-context lines, we would need to show them.
*/
if (opt->pre_context || opt->funcbody)
show_pre_context(opt, gs, bol, eol, lno);
else if (opt->funcname)
show_funcname_line(opt, gs, bol, lno);
show_line(opt, bol, eol, gs->name, lno, ':');
last_hit = lno;
if (opt->funcbody)
show_function = 1;
goto next_line;
}
if (show_function && (!peek_bol || peek_bol < bol)) {
unsigned long peek_left = left;
char *peek_eol = eol;
/*
* Trailing empty lines are not interesting.
* Peek past them to see if they belong to the
* body of the current function.
*/
peek_bol = bol;
while (is_empty_line(peek_bol, peek_eol)) {
peek_bol = peek_eol + 1;
peek_eol = end_of_line(peek_bol, &peek_left);
}
if (match_funcname(opt, gs, peek_bol, peek_eol))
show_function = 0;
}
if (show_function ||
(last_hit && lno <= last_hit + opt->post_context)) {
/* If the last hit is within the post context,
* we need to show this line.
*/
show_line(opt, bol, eol, gs->name, lno, '-');
}
next_line:
bol = eol + 1;
if (!left)
break;
left--;
lno++;
}
if (collect_hits)
return 0;
if (opt->status_only)
return 0;
if (opt->unmatch_name_only) {
/* We did not see any hit, so we want to show this */
show_name(opt, gs->name);
return 1;
}
xdiff_clear_find_func(&xecfg);
opt->priv = NULL;
/* NEEDSWORK:
* The real "grep -c foo *.c" gives many "bar.c:0" lines,
* which feels mostly useless but sometimes useful. Maybe
* make it another option? For now suppress them.
*/
if (opt->count && count) {
char buf[32];
if (opt->pathname) {
output_color(opt, gs->name, strlen(gs->name),
opt->color_filename);
output_sep(opt, ':');
}
snprintf(buf, sizeof(buf), "%u\n", count);
opt->output(opt, buf, strlen(buf));
return 1;
}
return !!last_hit;
}
static void clr_hit_marker(struct grep_expr *x)
{
/* All-hit markers are meaningful only at the very top level
* OR node.
*/
while (1) {
x->hit = 0;
if (x->node != GREP_NODE_OR)
return;
x->u.binary.left->hit = 0;
x = x->u.binary.right;
}
}
static int chk_hit_marker(struct grep_expr *x)
{
/* Top level nodes have hit markers. See if they all are hits */
while (1) {
if (x->node != GREP_NODE_OR)
return x->hit;
if (!x->u.binary.left->hit)
return 0;
x = x->u.binary.right;
}
}
int grep_source(struct grep_opt *opt, struct grep_source *gs)
{
/*
* we do not have to do the two-pass grep when we do not check
* buffer-wide "all-match".
*/
if (!opt->all_match)
return grep_source_1(opt, gs, 0);
/* Otherwise the toplevel "or" terms hit a bit differently.
* We first clear hit markers from them.
*/
clr_hit_marker(opt->pattern_expression);
grep_source_1(opt, gs, 1);
if (!chk_hit_marker(opt->pattern_expression))
return 0;
return grep_source_1(opt, gs, 0);
}
int grep_buffer(struct grep_opt *opt, char *buf, unsigned long size)
{
struct grep_source gs;
int r;
grep_source_init(&gs, GREP_SOURCE_BUF, NULL, NULL, NULL);
gs.buf = buf;
gs.size = size;
r = grep_source(opt, &gs);
grep_source_clear(&gs);
return r;
}
void grep_source_init(struct grep_source *gs, enum grep_source_type type,
const char *name, const char *path,
const void *identifier)
{
gs->type = type;
gs->name = xstrdup_or_null(name);
gs->path = xstrdup_or_null(path);
gs->buf = NULL;
gs->size = 0;
gs->driver = NULL;
switch (type) {
case GREP_SOURCE_FILE:
gs->identifier = xstrdup(identifier);
break;
case GREP_SOURCE_SHA1:
gs->identifier = xmalloc(20);
hashcpy(gs->identifier, identifier);
break;
case GREP_SOURCE_BUF:
gs->identifier = NULL;
}
}
void grep_source_clear(struct grep_source *gs)
{
free(gs->name);
gs->name = NULL;
free(gs->path);
gs->path = NULL;
free(gs->identifier);
gs->identifier = NULL;
grep_source_clear_data(gs);
}
void grep_source_clear_data(struct grep_source *gs)
{
switch (gs->type) {
case GREP_SOURCE_FILE:
case GREP_SOURCE_SHA1:
free(gs->buf);
gs->buf = NULL;
gs->size = 0;
break;
case GREP_SOURCE_BUF:
/* leave user-provided buf intact */
break;
}
}
static int grep_source_load_sha1(struct grep_source *gs)
{
enum object_type type;
grep_read_lock();
gs->buf = read_sha1_file(gs->identifier, &type, &gs->size);
grep_read_unlock();
if (!gs->buf)
return error(_("'%s': unable to read %s"),
gs->name,
sha1_to_hex(gs->identifier));
return 0;
}
static int grep_source_load_file(struct grep_source *gs)
{
const char *filename = gs->identifier;
struct stat st;
char *data;
size_t size;
int i;
if (lstat(filename, &st) < 0) {
err_ret:
if (errno != ENOENT)
error_errno(_("failed to stat '%s'"), filename);
return -1;
}
if (!S_ISREG(st.st_mode))
return -1;
size = xsize_t(st.st_size);
i = open(filename, O_RDONLY);
if (i < 0)
goto err_ret;
data = xmallocz(size);
if (st.st_size != read_in_full(i, data, size)) {
error_errno(_("'%s': short read"), filename);
close(i);
free(data);
return -1;
}
close(i);
gs->buf = data;
gs->size = size;
return 0;
}
static int grep_source_load(struct grep_source *gs)
{
if (gs->buf)
return 0;
switch (gs->type) {
case GREP_SOURCE_FILE:
return grep_source_load_file(gs);
case GREP_SOURCE_SHA1:
return grep_source_load_sha1(gs);
case GREP_SOURCE_BUF:
return gs->buf ? 0 : -1;
}
die("BUG: invalid grep_source type");
}
void grep_source_load_driver(struct grep_source *gs)
{
if (gs->driver)
return;
grep_attr_lock();
if (gs->path)
gs->driver = userdiff_find_by_path(gs->path);
if (!gs->driver)
gs->driver = userdiff_find_by_name("default");
grep_attr_unlock();
}
static int grep_source_is_binary(struct grep_source *gs)
{
grep_source_load_driver(gs);
if (gs->driver->binary != -1)
return gs->driver->binary;
if (!grep_source_load(gs))
return buffer_is_binary(gs->buf, gs->size);
return 0;
}