Stats in Git

From: Marius Storm-Olsen <marius@trolltech.com>
Date: Sep 2, 2007, 14:49 UTC
Message-ID: <46DACD93.9000509@trolltech.com>

I was checking out the performance situation with Git on Windows, and found out that the Posix stat functions on Windows are just obscenely slow. We really can't use them, at least on in Git. So, I made a patch for the MinGW version, which I'll post right after this mail.

However, while look at that whole stat'ing situation in git, I saw that doing 'git status' actually stats all the files _thrice_! Yup, that's not 1 time, or 2 times, but actually 3(!) times before 'git status' is content! I know that git-status is a script, so I think this clearly indicates that git-status is a prime candidate for a built-in ;-)

I haven't looked into details as to why it stats the files so many times. I guess someone more experienced in Git core could give an opinion, if by writing git-status as a builtin it would be possible to only stat the files once. It would have a huge impact on Windows where stats are inheritly much slower than on Linux.

By applying the diff below, you can see for yourself what happens when
you stat the repo created with Moe's script:
    mkdir bummer
    cd bummer
    for ((i=0;i<100;i++)); do
    mkdir $i && pushd $i;
    for ((j=0;j<1000;j++)); do
    echo "$j" >$j; done; popd;
    done

$ git status 2>&1 | wc -l 300137

Fast on Linux now, but still quite slow on Windows..

-- .marius

diff --git a/git-compat-util.h b/git-compat-util.h
index ca0a597..6b6405c 100644
--- a/git-compat-util.h
+++ b/git-compat-util.h
@@ -369,4 +369,23 @@ static inline int strtoul_ui(char const *s, int base, unsigned int *result)
 	return 0;
 }

+static inline int git_lstat(const char *file_name, struct stat *buf)
+{
+	fprintf(stderr, "lstat: %s\n", file_name);
+	return lstat(file_name, buf);
+}
+static inline int git_fstat(int fd, struct stat *buf)
+{
+	fprintf(stderr, "fstat: %d\n", fd);
+	return fstat(fd, buf);
+}
+static inline int git_stat(const char *file_name, struct stat *buf)
+{
+	fprintf(stderr, "stat: %s\n", file_name);
+	return stat(file_name, buf);
+}
+#define lstat(x,y) git_lstat(x,y)
+#define fstat(x,y) git_fstat(x,y)
+#define stat(x,y) git_stat(x,y)
+
 #endif