1
0
Fork 0
mirror of https://github.com/git/git.git synced 2024-11-17 06:25:13 +01:00
git/builtin-grep.c

895 lines
20 KiB
C
Raw Normal View History

/*
* Builtin "git grep"
*
* Copyright (c) 2006 Junio C Hamano
*/
#include "cache.h"
#include "blob.h"
#include "tree.h"
#include "commit.h"
#include "tag.h"
#include "tree-walk.h"
#include "builtin.h"
#include <regex.h>
#include <fnmatch.h>
#include <sys/wait.h>
/*
* git grep pathspecs are somewhat different from diff-tree pathspecs;
* pathname wildcards are allowed.
*/
static int pathspec_matches(const char **paths, const char *name)
{
int namelen, i;
if (!paths || !*paths)
return 1;
namelen = strlen(name);
for (i = 0; paths[i]; i++) {
const char *match = paths[i];
int matchlen = strlen(match);
const char *cp, *meta;
if (!matchlen ||
((matchlen <= namelen) &&
!strncmp(name, match, matchlen) &&
(match[matchlen-1] == '/' ||
name[matchlen] == '\0' || name[matchlen] == '/')))
return 1;
if (!fnmatch(match, name, 0))
return 1;
if (name[namelen-1] != '/')
continue;
/* We are being asked if the directory ("name") is worth
* descending into.
*
* Find the longest leading directory name that does
* not have metacharacter in the pathspec; the name
* we are looking at must overlap with that directory.
*/
for (cp = match, meta = NULL; cp - match < matchlen; cp++) {
char ch = *cp;
if (ch == '*' || ch == '[' || ch == '?') {
meta = cp;
break;
}
}
if (!meta)
meta = cp; /* fully literal */
if (namelen <= meta - match) {
/* Looking at "Documentation/" and
* the pattern says "Documentation/howto/", or
* "Documentation/diff*.txt". The name we
* have should match prefix.
*/
if (!memcmp(match, name, namelen))
return 1;
continue;
}
if (meta - match < namelen) {
/* Looking at "Documentation/howto/" and
* the pattern says "Documentation/h*";
* match up to "Do.../h"; this avoids descending
* into "Documentation/technical/".
*/
if (!memcmp(match, name, meta - match))
return 1;
continue;
}
}
return 0;
}
struct grep_pat {
struct grep_pat *next;
const char *origin;
int no;
const char *pattern;
regex_t regexp;
};
struct grep_opt {
struct grep_pat *pattern_list;
struct grep_pat **pattern_tail;
regex_t regexp;
unsigned linenum:1;
unsigned invert:1;
unsigned name_only:1;
unsigned unmatch_name_only:1;
unsigned count:1;
unsigned word_regexp:1;
unsigned fixed:1;
#define GREP_BINARY_DEFAULT 0
#define GREP_BINARY_NOMATCH 1
#define GREP_BINARY_TEXT 2
unsigned binary:2;
int regflags;
unsigned pre_context;
unsigned post_context;
};
static void add_pattern(struct grep_opt *opt, const char *pat,
const char *origin, int no)
{
struct grep_pat *p = xcalloc(1, sizeof(*p));
p->pattern = pat;
p->origin = origin;
p->no = no;
*opt->pattern_tail = p;
opt->pattern_tail = &p->next;
p->next = NULL;
}
static void compile_patterns(struct grep_opt *opt)
{
struct grep_pat *p;
for (p = opt->pattern_list; p; p = p->next) {
int err = regcomp(&p->regexp, p->pattern, opt->regflags);
if (err) {
char errbuf[1024];
char where[1024];
if (p->no)
sprintf(where, "In '%s' at %d, ",
p->origin, p->no);
else if (p->origin)
sprintf(where, "%s, ", p->origin);
else
where[0] = 0;
regerror(err, &p->regexp, errbuf, 1024);
regfree(&p->regexp);
die("%s'%s': %s", where, p->pattern, errbuf);
}
}
}
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 show_line(struct grep_opt *opt, const char *bol, const char *eol,
const char *name, unsigned lno, char sign)
{
printf("%s%c", name, sign);
if (opt->linenum)
printf("%d%c", lno, sign);
printf("%.*s\n", (int)(eol-bol), bol);
}
/*
* NEEDSWORK: share code with diff.c
*/
#define FIRST_FEW_BYTES 8000
static int buffer_is_binary(const char *ptr, unsigned long size)
{
if (FIRST_FEW_BYTES < size)
size = FIRST_FEW_BYTES;
if (memchr(ptr, 0, size))
return 1;
return 0;
}
static int fixmatch(const char *pattern, char *line, regmatch_t *match)
{
char *hit = strstr(line, pattern);
if (!hit) {
match->rm_so = match->rm_eo = -1;
return REG_NOMATCH;
}
else {
match->rm_so = hit - line;
match->rm_eo = match->rm_so + strlen(pattern);
return 0;
}
}
static int grep_buffer(struct grep_opt *opt, const char *name,
char *buf, unsigned long size)
{
char *bol = buf;
unsigned long left = size;
unsigned lno = 1;
struct pre_context_line {
char *bol;
char *eol;
} *prev = NULL, *pcl;
unsigned last_hit = 0;
unsigned last_shown = 0;
int binary_match_only = 0;
const char *hunk_mark = "";
unsigned count = 0;
if (buffer_is_binary(buf, size)) {
switch (opt->binary) {
case GREP_BINARY_DEFAULT:
binary_match_only = 1;
break;
case GREP_BINARY_NOMATCH:
return 0; /* Assume unmatch */
break;
default:
break;
}
}
if (opt->pre_context)
prev = xcalloc(opt->pre_context, sizeof(*prev));
if (opt->pre_context || opt->post_context)
hunk_mark = "--\n";
while (left) {
regmatch_t pmatch[10];
char *eol, ch;
int hit = 0;
struct grep_pat *p;
eol = end_of_line(bol, &left);
ch = *eol;
*eol = 0;
for (p = opt->pattern_list; p; p = p->next) {
if (!opt->fixed) {
regex_t *exp = &p->regexp;
hit = !regexec(exp, bol, ARRAY_SIZE(pmatch),
pmatch, 0);
}
else {
hit = !fixmatch(p->pattern, bol, pmatch);
}
if (hit && opt->word_regexp) {
/* Match beginning must be either
* beginning of the line, or at word
* boundary (i.e. the last char must
* not be alnum or underscore).
*/
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");
if (pmatch[0].rm_so != 0 &&
word_char(bol[pmatch[0].rm_so-1]))
hit = 0;
if (pmatch[0].rm_eo != (eol-bol) &&
word_char(bol[pmatch[0].rm_eo]))
hit = 0;
}
if (hit)
break;
}
/* "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 (binary_match_only) {
printf("Binary file %s matches\n", name);
return 1;
}
if (opt->name_only) {
printf("%s\n", name);
return 1;
}
/* Hit at this line. If we haven't shown the
* pre-context lines, we would need to show them.
* When asked to do "count", this still show
* the context which is nonsense, but the user
* deserves to get that ;-).
*/
if (opt->pre_context) {
unsigned from;
if (opt->pre_context < lno)
from = lno - opt->pre_context;
else
from = 1;
if (from <= last_shown)
from = last_shown + 1;
if (last_shown && from != last_shown + 1)
printf(hunk_mark);
while (from < lno) {
pcl = &prev[lno-from-1];
show_line(opt, pcl->bol, pcl->eol,
name, from, '-');
from++;
}
last_shown = lno-1;
}
if (last_shown && lno != last_shown + 1)
printf(hunk_mark);
if (!opt->count)
show_line(opt, bol, eol, name, lno, ':');
last_shown = last_hit = lno;
}
else if (last_hit &&
lno <= last_hit + opt->post_context) {
/* If the last hit is within the post context,
* we need to show this line.
*/
if (last_shown && lno != last_shown + 1)
printf(hunk_mark);
show_line(opt, bol, eol, name, lno, '-');
last_shown = lno;
}
if (opt->pre_context) {
memmove(prev+1, prev,
(opt->pre_context-1) * sizeof(*prev));
prev->bol = bol;
prev->eol = eol;
}
next_line:
*eol = ch;
bol = eol + 1;
if (!left)
break;
left--;
lno++;
}
if (opt->unmatch_name_only) {
/* We did not see any hit, so we want to show this */
printf("%s\n", name);
return 1;
}
/* 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)
printf("%s:%u\n", name, count);
return !!last_hit;
}
static int grep_sha1(struct grep_opt *opt, const unsigned char *sha1, const char *name)
{
unsigned long size;
char *data;
char type[20];
int hit;
data = read_sha1_file(sha1, type, &size);
if (!data) {
error("'%s': unable to read %s", name, sha1_to_hex(sha1));
return 0;
}
hit = grep_buffer(opt, name, data, size);
free(data);
return hit;
}
static int grep_file(struct grep_opt *opt, const char *filename)
{
struct stat st;
int i;
char *data;
if (lstat(filename, &st) < 0) {
err_ret:
if (errno != ENOENT)
error("'%s': %s", filename, strerror(errno));
return 0;
}
if (!st.st_size)
return 0; /* empty file -- no grep hit */
if (!S_ISREG(st.st_mode))
return 0;
i = open(filename, O_RDONLY);
if (i < 0)
goto err_ret;
data = xmalloc(st.st_size + 1);
if (st.st_size != xread(i, data, st.st_size)) {
error("'%s': short read %s", filename, strerror(errno));
close(i);
free(data);
return 0;
}
close(i);
i = grep_buffer(opt, filename, data, st.st_size);
free(data);
return i;
}
static int exec_grep(int argc, const char **argv)
{
pid_t pid;
int status;
argv[argc] = NULL;
pid = fork();
if (pid < 0)
return pid;
if (!pid) {
execvp("grep", (char **) argv);
exit(255);
}
while (waitpid(pid, &status, 0) < 0) {
if (errno == EINTR)
continue;
return -1;
}
if (WIFEXITED(status)) {
if (!WEXITSTATUS(status))
return 1;
return 0;
}
return -1;
}
#define MAXARGS 1000
#define ARGBUF 4096
#define push_arg(a) do { \
if (nr < MAXARGS) argv[nr++] = (a); \
else die("maximum number of args exceeded"); \
} while (0)
static int external_grep(struct grep_opt *opt, const char **paths, int cached)
{
int i, nr, argc, hit, len;
const char *argv[MAXARGS+1];
char randarg[ARGBUF];
char *argptr = randarg;
struct grep_pat *p;
len = nr = 0;
push_arg("grep");
if (opt->fixed)
push_arg("-F");
if (opt->linenum)
push_arg("-n");
if (opt->regflags & REG_EXTENDED)
push_arg("-E");
if (opt->regflags & REG_ICASE)
push_arg("-i");
if (opt->word_regexp)
push_arg("-w");
if (opt->name_only)
push_arg("-l");
if (opt->unmatch_name_only)
push_arg("-L");
if (opt->count)
push_arg("-c");
if (opt->post_context || opt->pre_context) {
if (opt->post_context != opt->pre_context) {
if (opt->pre_context) {
push_arg("-B");
len += snprintf(argptr, sizeof(randarg)-len,
"%u", opt->pre_context);
if (sizeof(randarg) <= len)
die("maximum length of args exceeded");
push_arg(argptr);
argptr += len;
}
if (opt->post_context) {
push_arg("-A");
len += snprintf(argptr, sizeof(randarg)-len,
"%u", opt->post_context);
if (sizeof(randarg) <= len)
die("maximum length of args exceeded");
push_arg(argptr);
argptr += len;
}
}
else {
push_arg("-C");
len += snprintf(argptr, sizeof(randarg)-len,
"%u", opt->post_context);
if (sizeof(randarg) <= len)
die("maximum length of args exceeded");
push_arg(argptr);
argptr += len;
}
}
for (p = opt->pattern_list; p; p = p->next) {
push_arg("-e");
push_arg(p->pattern);
}
/*
* To make sure we get the header printed out when we want it,
* add /dev/null to the paths to grep. This is unnecessary
* (and wrong) with "-l" or "-L", which always print out the
* name anyway.
*
* GNU grep has "-H", but this is portable.
*/
if (!opt->name_only && !opt->unmatch_name_only)
push_arg("/dev/null");
hit = 0;
argc = nr;
for (i = 0; i < active_nr; i++) {
struct cache_entry *ce = active_cache[i];
char *name;
if (ce_stage(ce) || !S_ISREG(ntohl(ce->ce_mode)))
continue;
if (!pathspec_matches(paths, ce->name))
continue;
name = ce->name;
if (name[0] == '-') {
int len = ce_namelen(ce);
name = xmalloc(len + 3);
memcpy(name, "./", 2);
memcpy(name + 2, ce->name, len + 1);
}
argv[argc++] = name;
if (argc < MAXARGS)
continue;
hit += exec_grep(argc, argv);
argc = nr;
}
if (argc > nr)
hit += exec_grep(argc, argv);
return 0;
}
static int grep_cache(struct grep_opt *opt, const char **paths, int cached)
{
int hit = 0;
int nr;
read_cache();
#ifdef __unix__
/*
* Use the external "grep" command for the case where
* we grep through the checked-out files. It tends to
* be a lot more optimized
*/
if (!cached) {
hit = external_grep(opt, paths, cached);
if (hit >= 0)
return hit;
}
#endif
for (nr = 0; nr < active_nr; nr++) {
struct cache_entry *ce = active_cache[nr];
if (ce_stage(ce) || !S_ISREG(ntohl(ce->ce_mode)))
continue;
if (!pathspec_matches(paths, ce->name))
continue;
if (cached)
hit |= grep_sha1(opt, ce->sha1, ce->name);
else
hit |= grep_file(opt, ce->name);
}
return hit;
}
static int grep_tree(struct grep_opt *opt, const char **paths,
struct tree_desc *tree,
const char *tree_name, const char *base)
{
int len;
int hit = 0;
tree_entry(): new tree-walking helper function This adds a "tree_entry()" function that combines the common operation of doing a "tree_entry_extract()" + "update_tree_entry()". It also has a simplified calling convention, designed for simple loops that traverse over a whole tree: the arguments are pointers to the tree descriptor and a name_entry structure to fill in, and it returns a boolean "true" if there was an entry left to be gotten in the tree. This allows tree traversal with struct tree_desc desc; struct name_entry entry; desc.buf = tree->buffer; desc.size = tree->size; while (tree_entry(&desc, &entry) { ... use "entry.{path, sha1, mode, pathlen}" ... } which is not only shorter than writing it out in full, it's hopefully less error prone too. [ It's actually a tad faster too - we don't need to recalculate the entry pathlength in both extract and update, but need to do it only once. Also, some callers can avoid doing a "strlen()" on the result, since it's returned as part of the name_entry structure. However, by now we're talking just 1% speedup on "git-rev-list --objects --all", and we're definitely at the point where tree walking is no longer the issue any more. ] NOTE! Not everybody wants to use this new helper function, since some of the tree walkers very much on purpose do the descriptor update separately from the entry extraction. So the "extract + update" sequence still remains as the core sequence, this is just a simplified interface. We should probably add a silly two-line inline helper function for initializing the descriptor from the "struct tree" too, just to cut down on the noise from that common "desc" initializer. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-05-30 18:45:45 +02:00
struct name_entry entry;
char *down;
char *path_buf = xmalloc(PATH_MAX + strlen(tree_name) + 100);
if (tree_name[0]) {
int offset = sprintf(path_buf, "%s:", tree_name);
down = path_buf + offset;
strcat(down, base);
}
else {
down = path_buf;
strcpy(down, base);
}
len = strlen(path_buf);
tree_entry(): new tree-walking helper function This adds a "tree_entry()" function that combines the common operation of doing a "tree_entry_extract()" + "update_tree_entry()". It also has a simplified calling convention, designed for simple loops that traverse over a whole tree: the arguments are pointers to the tree descriptor and a name_entry structure to fill in, and it returns a boolean "true" if there was an entry left to be gotten in the tree. This allows tree traversal with struct tree_desc desc; struct name_entry entry; desc.buf = tree->buffer; desc.size = tree->size; while (tree_entry(&desc, &entry) { ... use "entry.{path, sha1, mode, pathlen}" ... } which is not only shorter than writing it out in full, it's hopefully less error prone too. [ It's actually a tad faster too - we don't need to recalculate the entry pathlength in both extract and update, but need to do it only once. Also, some callers can avoid doing a "strlen()" on the result, since it's returned as part of the name_entry structure. However, by now we're talking just 1% speedup on "git-rev-list --objects --all", and we're definitely at the point where tree walking is no longer the issue any more. ] NOTE! Not everybody wants to use this new helper function, since some of the tree walkers very much on purpose do the descriptor update separately from the entry extraction. So the "extract + update" sequence still remains as the core sequence, this is just a simplified interface. We should probably add a silly two-line inline helper function for initializing the descriptor from the "struct tree" too, just to cut down on the noise from that common "desc" initializer. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-05-30 18:45:45 +02:00
while (tree_entry(tree, &entry)) {
strcpy(path_buf + len, entry.path);
tree_entry(): new tree-walking helper function This adds a "tree_entry()" function that combines the common operation of doing a "tree_entry_extract()" + "update_tree_entry()". It also has a simplified calling convention, designed for simple loops that traverse over a whole tree: the arguments are pointers to the tree descriptor and a name_entry structure to fill in, and it returns a boolean "true" if there was an entry left to be gotten in the tree. This allows tree traversal with struct tree_desc desc; struct name_entry entry; desc.buf = tree->buffer; desc.size = tree->size; while (tree_entry(&desc, &entry) { ... use "entry.{path, sha1, mode, pathlen}" ... } which is not only shorter than writing it out in full, it's hopefully less error prone too. [ It's actually a tad faster too - we don't need to recalculate the entry pathlength in both extract and update, but need to do it only once. Also, some callers can avoid doing a "strlen()" on the result, since it's returned as part of the name_entry structure. However, by now we're talking just 1% speedup on "git-rev-list --objects --all", and we're definitely at the point where tree walking is no longer the issue any more. ] NOTE! Not everybody wants to use this new helper function, since some of the tree walkers very much on purpose do the descriptor update separately from the entry extraction. So the "extract + update" sequence still remains as the core sequence, this is just a simplified interface. We should probably add a silly two-line inline helper function for initializing the descriptor from the "struct tree" too, just to cut down on the noise from that common "desc" initializer. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-05-30 18:45:45 +02:00
if (S_ISDIR(entry.mode))
/* Match "abc/" against pathspec to
* decide if we want to descend into "abc"
* directory.
*/
tree_entry(): new tree-walking helper function This adds a "tree_entry()" function that combines the common operation of doing a "tree_entry_extract()" + "update_tree_entry()". It also has a simplified calling convention, designed for simple loops that traverse over a whole tree: the arguments are pointers to the tree descriptor and a name_entry structure to fill in, and it returns a boolean "true" if there was an entry left to be gotten in the tree. This allows tree traversal with struct tree_desc desc; struct name_entry entry; desc.buf = tree->buffer; desc.size = tree->size; while (tree_entry(&desc, &entry) { ... use "entry.{path, sha1, mode, pathlen}" ... } which is not only shorter than writing it out in full, it's hopefully less error prone too. [ It's actually a tad faster too - we don't need to recalculate the entry pathlength in both extract and update, but need to do it only once. Also, some callers can avoid doing a "strlen()" on the result, since it's returned as part of the name_entry structure. However, by now we're talking just 1% speedup on "git-rev-list --objects --all", and we're definitely at the point where tree walking is no longer the issue any more. ] NOTE! Not everybody wants to use this new helper function, since some of the tree walkers very much on purpose do the descriptor update separately from the entry extraction. So the "extract + update" sequence still remains as the core sequence, this is just a simplified interface. We should probably add a silly two-line inline helper function for initializing the descriptor from the "struct tree" too, just to cut down on the noise from that common "desc" initializer. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-05-30 18:45:45 +02:00
strcpy(path_buf + len + entry.pathlen, "/");
if (!pathspec_matches(paths, down))
;
tree_entry(): new tree-walking helper function This adds a "tree_entry()" function that combines the common operation of doing a "tree_entry_extract()" + "update_tree_entry()". It also has a simplified calling convention, designed for simple loops that traverse over a whole tree: the arguments are pointers to the tree descriptor and a name_entry structure to fill in, and it returns a boolean "true" if there was an entry left to be gotten in the tree. This allows tree traversal with struct tree_desc desc; struct name_entry entry; desc.buf = tree->buffer; desc.size = tree->size; while (tree_entry(&desc, &entry) { ... use "entry.{path, sha1, mode, pathlen}" ... } which is not only shorter than writing it out in full, it's hopefully less error prone too. [ It's actually a tad faster too - we don't need to recalculate the entry pathlength in both extract and update, but need to do it only once. Also, some callers can avoid doing a "strlen()" on the result, since it's returned as part of the name_entry structure. However, by now we're talking just 1% speedup on "git-rev-list --objects --all", and we're definitely at the point where tree walking is no longer the issue any more. ] NOTE! Not everybody wants to use this new helper function, since some of the tree walkers very much on purpose do the descriptor update separately from the entry extraction. So the "extract + update" sequence still remains as the core sequence, this is just a simplified interface. We should probably add a silly two-line inline helper function for initializing the descriptor from the "struct tree" too, just to cut down on the noise from that common "desc" initializer. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-05-30 18:45:45 +02:00
else if (S_ISREG(entry.mode))
hit |= grep_sha1(opt, entry.sha1, path_buf);
else if (S_ISDIR(entry.mode)) {
char type[20];
struct tree_desc sub;
void *data;
tree_entry(): new tree-walking helper function This adds a "tree_entry()" function that combines the common operation of doing a "tree_entry_extract()" + "update_tree_entry()". It also has a simplified calling convention, designed for simple loops that traverse over a whole tree: the arguments are pointers to the tree descriptor and a name_entry structure to fill in, and it returns a boolean "true" if there was an entry left to be gotten in the tree. This allows tree traversal with struct tree_desc desc; struct name_entry entry; desc.buf = tree->buffer; desc.size = tree->size; while (tree_entry(&desc, &entry) { ... use "entry.{path, sha1, mode, pathlen}" ... } which is not only shorter than writing it out in full, it's hopefully less error prone too. [ It's actually a tad faster too - we don't need to recalculate the entry pathlength in both extract and update, but need to do it only once. Also, some callers can avoid doing a "strlen()" on the result, since it's returned as part of the name_entry structure. However, by now we're talking just 1% speedup on "git-rev-list --objects --all", and we're definitely at the point where tree walking is no longer the issue any more. ] NOTE! Not everybody wants to use this new helper function, since some of the tree walkers very much on purpose do the descriptor update separately from the entry extraction. So the "extract + update" sequence still remains as the core sequence, this is just a simplified interface. We should probably add a silly two-line inline helper function for initializing the descriptor from the "struct tree" too, just to cut down on the noise from that common "desc" initializer. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-05-30 18:45:45 +02:00
data = read_sha1_file(entry.sha1, type, &sub.size);
if (!data)
die("unable to read tree (%s)",
tree_entry(): new tree-walking helper function This adds a "tree_entry()" function that combines the common operation of doing a "tree_entry_extract()" + "update_tree_entry()". It also has a simplified calling convention, designed for simple loops that traverse over a whole tree: the arguments are pointers to the tree descriptor and a name_entry structure to fill in, and it returns a boolean "true" if there was an entry left to be gotten in the tree. This allows tree traversal with struct tree_desc desc; struct name_entry entry; desc.buf = tree->buffer; desc.size = tree->size; while (tree_entry(&desc, &entry) { ... use "entry.{path, sha1, mode, pathlen}" ... } which is not only shorter than writing it out in full, it's hopefully less error prone too. [ It's actually a tad faster too - we don't need to recalculate the entry pathlength in both extract and update, but need to do it only once. Also, some callers can avoid doing a "strlen()" on the result, since it's returned as part of the name_entry structure. However, by now we're talking just 1% speedup on "git-rev-list --objects --all", and we're definitely at the point where tree walking is no longer the issue any more. ] NOTE! Not everybody wants to use this new helper function, since some of the tree walkers very much on purpose do the descriptor update separately from the entry extraction. So the "extract + update" sequence still remains as the core sequence, this is just a simplified interface. We should probably add a silly two-line inline helper function for initializing the descriptor from the "struct tree" too, just to cut down on the noise from that common "desc" initializer. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-05-30 18:45:45 +02:00
sha1_to_hex(entry.sha1));
sub.buf = data;
hit |= grep_tree(opt, paths, &sub, tree_name, down);
free(data);
}
}
return hit;
}
static int grep_object(struct grep_opt *opt, const char **paths,
struct object *obj, const char *name)
{
if (obj->type == TYPE_BLOB)
return grep_sha1(opt, obj->sha1, name);
if (obj->type == TYPE_COMMIT || obj->type == TYPE_TREE) {
struct tree_desc tree;
void *data;
int hit;
data = read_object_with_reference(obj->sha1, tree_type,
&tree.size, NULL);
if (!data)
die("unable to read tree (%s)", sha1_to_hex(obj->sha1));
tree.buf = data;
hit = grep_tree(opt, paths, &tree, name, "");
free(data);
return hit;
}
die("unable to grep from object of type %s", typename(obj->type));
}
static const char builtin_grep_usage[] =
"git-grep <option>* <rev>* [-e] <pattern> [<path>...]";
int cmd_grep(int argc, const char **argv, char **envp)
{
int hit = 0;
int cached = 0;
int seen_dashdash = 0;
struct grep_opt opt;
Add "named object array" concept We've had this notion of a "object_list" for a long time, which eventually grew a "name" member because some users (notably git-rev-list) wanted to name each object as it is generated. That object_list is great for some things, but it isn't all that wonderful for others, and the "name" member is generally not used by everybody. This patch splits the users of the object_list array up into two: the traditional list users, who want the list-like format, and who don't actually use or want the name. And another class of users that really used the list as an extensible array, and generally wanted to name the objects. The patch is fairly straightforward, but it's also biggish. Most of it really just cleans things up: switching the revision parsing and listing over to the array makes things like the builtin-diff usage much simpler (we now see exactly how many members the array has, and we don't get the objects reversed from the order they were on the command line). One of the main reasons for doing this at all is that the malloc overhead of the simple object list was actually pretty high, and the array is just a lot denser. So this patch brings down memory usage by git-rev-list by just under 3% (on top of all the other memory use optimizations) on the mozilla archive. It does add more lines than it removes, and more importantly, it adds a whole new infrastructure for maintaining lists of objects, but on the other hand, the new dynamic array code is pretty obvious. The change to builtin-diff-tree.c shows a fairly good example of why an array interface is sometimes more natural, and just much simpler for everybody. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-06-20 02:42:35 +02:00
struct object_array list = { 0, 0, NULL };
const char *prefix = setup_git_directory();
const char **paths = NULL;
int i;
memset(&opt, 0, sizeof(opt));
opt.pattern_tail = &opt.pattern_list;
opt.regflags = REG_NEWLINE;
/*
* If there is no -- then the paths must exist in the working
* tree. If there is no explicit pattern specified with -e or
* -f, we take the first unrecognized non option to be the
* pattern, but then what follows it must be zero or more
* valid refs up to the -- (if exists), and then existing
* paths. If there is an explicit pattern, then the first
* unrecocnized non option is the beginning of the refs list
* that continues up to the -- (if exists), and then paths.
*/
while (1 < argc) {
const char *arg = argv[1];
argc--; argv++;
if (!strcmp("--cached", arg)) {
cached = 1;
continue;
}
if (!strcmp("-a", arg) ||
!strcmp("--text", arg)) {
opt.binary = GREP_BINARY_TEXT;
continue;
}
if (!strcmp("-i", arg) ||
!strcmp("--ignore-case", arg)) {
opt.regflags |= REG_ICASE;
continue;
}
if (!strcmp("-I", arg)) {
opt.binary = GREP_BINARY_NOMATCH;
continue;
}
if (!strcmp("-v", arg) ||
!strcmp("--invert-match", arg)) {
opt.invert = 1;
continue;
}
if (!strcmp("-E", arg) ||
!strcmp("--extended-regexp", arg)) {
opt.regflags |= REG_EXTENDED;
continue;
}
if (!strcmp("-F", arg) ||
!strcmp("--fixed-strings", arg)) {
opt.fixed = 1;
continue;
}
if (!strcmp("-G", arg) ||
!strcmp("--basic-regexp", arg)) {
opt.regflags &= ~REG_EXTENDED;
continue;
}
if (!strcmp("-n", arg)) {
opt.linenum = 1;
continue;
}
if (!strcmp("-H", arg)) {
/* We always show the pathname, so this
* is a noop.
*/
continue;
}
if (!strcmp("-l", arg) ||
!strcmp("--files-with-matches", arg)) {
opt.name_only = 1;
continue;
}
if (!strcmp("-L", arg) ||
!strcmp("--files-without-match", arg)) {
opt.unmatch_name_only = 1;
continue;
}
if (!strcmp("-c", arg) ||
!strcmp("--count", arg)) {
opt.count = 1;
continue;
}
if (!strcmp("-w", arg) ||
!strcmp("--word-regexp", arg)) {
opt.word_regexp = 1;
continue;
}
if (!strncmp("-A", arg, 2) ||
!strncmp("-B", arg, 2) ||
!strncmp("-C", arg, 2) ||
(arg[0] == '-' && '1' <= arg[1] && arg[1] <= '9')) {
unsigned num;
const char *scan;
switch (arg[1]) {
case 'A': case 'B': case 'C':
if (!arg[2]) {
if (argc <= 1)
usage(builtin_grep_usage);
scan = *++argv;
argc--;
}
else
scan = arg + 2;
break;
default:
scan = arg + 1;
break;
}
if (sscanf(scan, "%u", &num) != 1)
usage(builtin_grep_usage);
switch (arg[1]) {
case 'A':
opt.post_context = num;
break;
default:
case 'C':
opt.post_context = num;
case 'B':
opt.pre_context = num;
break;
}
continue;
}
if (!strcmp("-f", arg)) {
FILE *patterns;
int lno = 0;
char buf[1024];
if (argc <= 1)
usage(builtin_grep_usage);
patterns = fopen(argv[1], "r");
if (!patterns)
die("'%s': %s", argv[1], strerror(errno));
while (fgets(buf, sizeof(buf), patterns)) {
int len = strlen(buf);
if (buf[len-1] == '\n')
buf[len-1] = 0;
/* ignore empty line like grep does */
if (!buf[0])
continue;
add_pattern(&opt, strdup(buf), argv[1], ++lno);
}
fclose(patterns);
argv++;
argc--;
continue;
}
if (!strcmp("-e", arg)) {
if (1 < argc) {
add_pattern(&opt, argv[1], "-e option", 0);
argv++;
argc--;
continue;
}
usage(builtin_grep_usage);
}
if (!strcmp("--", arg))
break;
if (*arg == '-')
usage(builtin_grep_usage);
/* First unrecognized non-option token */
if (!opt.pattern_list) {
add_pattern(&opt, arg, "command line", 0);
break;
}
else {
/* We are looking at the first path or rev;
* it is found at argv[1] after leaving the
* loop.
*/
argc++; argv--;
break;
}
}
if (!opt.pattern_list)
die("no pattern given.");
if ((opt.regflags != REG_NEWLINE) && opt.fixed)
die("cannot mix --fixed-strings and regexp");
if (!opt.fixed)
compile_patterns(&opt);
/* Check revs and then paths */
for (i = 1; i < argc; i++) {
const char *arg = argv[i];
unsigned char sha1[20];
/* Is it a rev? */
if (!get_sha1(arg, sha1)) {
struct object *object = parse_object(sha1);
if (!object)
die("bad object %s", arg);
Add "named object array" concept We've had this notion of a "object_list" for a long time, which eventually grew a "name" member because some users (notably git-rev-list) wanted to name each object as it is generated. That object_list is great for some things, but it isn't all that wonderful for others, and the "name" member is generally not used by everybody. This patch splits the users of the object_list array up into two: the traditional list users, who want the list-like format, and who don't actually use or want the name. And another class of users that really used the list as an extensible array, and generally wanted to name the objects. The patch is fairly straightforward, but it's also biggish. Most of it really just cleans things up: switching the revision parsing and listing over to the array makes things like the builtin-diff usage much simpler (we now see exactly how many members the array has, and we don't get the objects reversed from the order they were on the command line). One of the main reasons for doing this at all is that the malloc overhead of the simple object list was actually pretty high, and the array is just a lot denser. So this patch brings down memory usage by git-rev-list by just under 3% (on top of all the other memory use optimizations) on the mozilla archive. It does add more lines than it removes, and more importantly, it adds a whole new infrastructure for maintaining lists of objects, but on the other hand, the new dynamic array code is pretty obvious. The change to builtin-diff-tree.c shows a fairly good example of why an array interface is sometimes more natural, and just much simpler for everybody. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-06-20 02:42:35 +02:00
add_object_array(object, arg, &list);
continue;
}
if (!strcmp(arg, "--")) {
i++;
seen_dashdash = 1;
}
break;
}
/* The rest are paths */
if (!seen_dashdash) {
int j;
for (j = i; j < argc; j++)
verify_filename(prefix, argv[j]);
}
if (i < argc)
paths = get_pathspec(prefix, argv + i);
else if (prefix) {
paths = xcalloc(2, sizeof(const char *));
paths[0] = prefix;
paths[1] = NULL;
}
Add "named object array" concept We've had this notion of a "object_list" for a long time, which eventually grew a "name" member because some users (notably git-rev-list) wanted to name each object as it is generated. That object_list is great for some things, but it isn't all that wonderful for others, and the "name" member is generally not used by everybody. This patch splits the users of the object_list array up into two: the traditional list users, who want the list-like format, and who don't actually use or want the name. And another class of users that really used the list as an extensible array, and generally wanted to name the objects. The patch is fairly straightforward, but it's also biggish. Most of it really just cleans things up: switching the revision parsing and listing over to the array makes things like the builtin-diff usage much simpler (we now see exactly how many members the array has, and we don't get the objects reversed from the order they were on the command line). One of the main reasons for doing this at all is that the malloc overhead of the simple object list was actually pretty high, and the array is just a lot denser. So this patch brings down memory usage by git-rev-list by just under 3% (on top of all the other memory use optimizations) on the mozilla archive. It does add more lines than it removes, and more importantly, it adds a whole new infrastructure for maintaining lists of objects, but on the other hand, the new dynamic array code is pretty obvious. The change to builtin-diff-tree.c shows a fairly good example of why an array interface is sometimes more natural, and just much simpler for everybody. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-06-20 02:42:35 +02:00
if (!list.nr)
return !grep_cache(&opt, paths, cached);
if (cached)
die("both --cached and trees are given.");
Add "named object array" concept We've had this notion of a "object_list" for a long time, which eventually grew a "name" member because some users (notably git-rev-list) wanted to name each object as it is generated. That object_list is great for some things, but it isn't all that wonderful for others, and the "name" member is generally not used by everybody. This patch splits the users of the object_list array up into two: the traditional list users, who want the list-like format, and who don't actually use or want the name. And another class of users that really used the list as an extensible array, and generally wanted to name the objects. The patch is fairly straightforward, but it's also biggish. Most of it really just cleans things up: switching the revision parsing and listing over to the array makes things like the builtin-diff usage much simpler (we now see exactly how many members the array has, and we don't get the objects reversed from the order they were on the command line). One of the main reasons for doing this at all is that the malloc overhead of the simple object list was actually pretty high, and the array is just a lot denser. So this patch brings down memory usage by git-rev-list by just under 3% (on top of all the other memory use optimizations) on the mozilla archive. It does add more lines than it removes, and more importantly, it adds a whole new infrastructure for maintaining lists of objects, but on the other hand, the new dynamic array code is pretty obvious. The change to builtin-diff-tree.c shows a fairly good example of why an array interface is sometimes more natural, and just much simpler for everybody. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-06-20 02:42:35 +02:00
for (i = 0; i < list.nr; i++) {
struct object *real_obj;
Add "named object array" concept We've had this notion of a "object_list" for a long time, which eventually grew a "name" member because some users (notably git-rev-list) wanted to name each object as it is generated. That object_list is great for some things, but it isn't all that wonderful for others, and the "name" member is generally not used by everybody. This patch splits the users of the object_list array up into two: the traditional list users, who want the list-like format, and who don't actually use or want the name. And another class of users that really used the list as an extensible array, and generally wanted to name the objects. The patch is fairly straightforward, but it's also biggish. Most of it really just cleans things up: switching the revision parsing and listing over to the array makes things like the builtin-diff usage much simpler (we now see exactly how many members the array has, and we don't get the objects reversed from the order they were on the command line). One of the main reasons for doing this at all is that the malloc overhead of the simple object list was actually pretty high, and the array is just a lot denser. So this patch brings down memory usage by git-rev-list by just under 3% (on top of all the other memory use optimizations) on the mozilla archive. It does add more lines than it removes, and more importantly, it adds a whole new infrastructure for maintaining lists of objects, but on the other hand, the new dynamic array code is pretty obvious. The change to builtin-diff-tree.c shows a fairly good example of why an array interface is sometimes more natural, and just much simpler for everybody. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-06-20 02:42:35 +02:00
real_obj = deref_tag(list.objects[i].item, NULL, 0);
if (grep_object(&opt, paths, real_obj, list.objects[i].name))
hit = 1;
}
return !hit;
}