git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: Creating objects manually and repack

From
Shawn Pearce <spearce@spearce.org>
Date
Aug 5, 2006, 05:46 UTC
Message-ID
<20060805054655.GB18679@spearce.org>
In-Reply-To
<20060805052135.GA18679@spearce.org>
Shawn Pearce <spearce@spearce.org> wrote:
Show 25 quoted lines
> Jon Smirl <jonsmirl@gmail.com> wrote:
> > On 8/5/06, Martin Langhoff <martin.langhoff@gmail.com> wrote:
> > >On 8/5/06, Jon Smirl <jonsmirl@gmail.com> wrote:
> > >> On 8/4/06, Linus Torvalds <torvalds@osdl.org> wrote:
> > >> > and you're basically all done. The above would turn each *,v file into 
> > >a
> > >> > *-<sha>.pack/*-<sha>.idx file pair, so you'd have exactly as many
> > >> > pack-files as you have *,v files.
> > >>
> > >> I'll end up with 110,000 pack files.
> > >
> > >Then just do it every 100 files, and you'll only have 1,100 pack
> > >files, and it'll be fine.
> > 
> > This is something that has to be tuned. If you wait too long
> > everything spills out of RAM and you go totally IO bound for days. If
> > you do it too often you end up with too many packs and it takes a day
> > to repack them.
> > 
> > If I had a way to pipe the all of the objects into repack one at a
> > time without repack doing multiple passes none of this tuning would be
> > necessary. In this model the standalone objects never get created in
> > the first place. The fastest IO is IO that has been eliminated.
> 
> I'm almost done with what I'm calling `git-fast-import`.

OK, now I'm done. I'm attaching the code. Toss it into the Makefile as git-fast-import and recompile.

I tested it with the following Perl script, feeding the Perl script a list of files that I wanted blobs for on STDIN:

	while (<>) {
		chop;
		print pack('L', -s $_);
		open(F, $_);
		my $buf;
		print $buf while read(F,$buf,128*1024) > 0;
		close F;
	}
This gave me an execution order of:
	find . -name '*.c' | perl test.pl | git-fast-import in.pack
	git-index-pack in.pack

at which point in.pack claims to be a completely valid pack with an index of in.idx. Move these into .git/objects/pack, generate trees and commits, and run git-repack -a -d. If the order you feed the objects to git-fast-import in is reasonable (do one RCS file at a time, feed most recent to least recent revisions) you may not get any major benefit from using -f during your final repack.

The code for git-fast-import could probably be tweaked to accept trees and commits too, which would permit you to stream the entire CVS repository into a single pack file. :-)

I can't help you decompress the RCS files faster, but hopefully this will help you generate the GIT pack faster. Hopefully you can make use of it!

-- 
Shawn.


#include "builtin.h"
#include "cache.h"
#include "object.h"
#include "blob.h"
#include "delta.h"
#include "pack.h"
#include "csum-file.h"

static int max_depth = 10;
static unsigned long object_count;
static int packfd;
static int current_depth;
static void *lastdat;
static unsigned long lastdatlen;
static unsigned char lastsha1[20];

static ssize_t yread(int fd, void *buffer, size_t length)
{
	ssize_t ret = 0;
	while (ret < length) {
		ssize_t size = xread(fd, (char *) buffer + ret, length - ret);
		if (size < 0) {
			return size;
		}
		if (size == 0) {
			return ret;
		}
		ret += size;
	}
	return ret;
}

static ssize_t ywrite(int fd, void *buffer, size_t length)
{
	ssize_t ret = 0;
	while (ret < length) {
		ssize_t size = xwrite(fd, (char *) buffer + ret, length - ret);
		if (size < 0) {
			return size;
		}
		if (size == 0) {
			return ret;
		}
		ret += size;
	}
	return ret;
}

static unsigned long encode_header(enum object_type type, unsigned long size, unsigned char *hdr)
{
	int n = 1;
	unsigned char c;

	if (type < OBJ_COMMIT || type > OBJ_DELTA)
		die("bad type %d", type);

	c = (type << 4) | (size & 15);
	size >>= 4;
	while (size) {
		*hdr++ = c | 0x80;
		c = size & 0x7f;
		size >>= 7;
		n++;
	}
	*hdr = c;
	return n;
}

static void write_blob (void *dat, unsigned long datlen)
{
	z_stream s;
	void *out, *delta;
	unsigned char hdr[64];
	unsigned long hdrlen, deltalen;

	if (lastdat && current_depth < max_depth) {
		delta = diff_delta(lastdat, lastdatlen,
			dat, datlen,
			&deltalen, 0);
	} else
		delta = 0;

	memset(&s, 0, sizeof(s));
	deflateInit(&s, zlib_compression_level);

	if (delta) {
		current_depth++;
		s.next_in = delta;
		s.avail_in = deltalen;
		hdrlen = encode_header(OBJ_DELTA, deltalen, hdr);
		if (ywrite(packfd, hdr, hdrlen) != hdrlen)
			die("Can't write object header: %s", strerror(errno));
		if (ywrite(packfd, lastsha1, sizeof(lastsha1)) != sizeof(lastsha1))
			die("Can't write object base: %s", strerror(errno));
	} else {
		current_depth = 0;
		s.next_in = dat;
		s.avail_in = datlen;
		hdrlen = encode_header(OBJ_BLOB, datlen, hdr);
		if (ywrite(packfd, hdr, hdrlen) != hdrlen)
			die("Can't write object header: %s", strerror(errno));
	}

	s.avail_out = deflateBound(&s, s.avail_in);
	s.next_out = out = xmalloc(s.avail_out);
	while (deflate(&s, Z_FINISH) == Z_OK)
		/* nothing */;
	deflateEnd(&s);

	if (ywrite(packfd, out, s.total_out) != s.total_out)
		die("Failed writing compressed data %s", strerror(errno));

	free(out);
	if (delta)
		free(delta);
}

static void init_pack_header ()
{
	const char* magic = "PACK";
	unsigned long version = 2;
	unsigned long zero = 0;

	version = htonl(version);

	if (ywrite(packfd, (char*)magic, 4) != 4)
		die("Can't write pack magic: %s", strerror(errno));
	if (ywrite(packfd, &version, 4) != 4)
		die("Can't write pack version: %s", strerror(errno));
	if (ywrite(packfd, &zero, 4) != 4)
		die("Can't write 0 object count: %s", strerror(errno));
}

static void fixup_header_footer ()
{
	SHA_CTX c;
	char hdr[8];
	unsigned char sha1[20];
	unsigned long cnt;
	char *buf;
	size_t n;

	if (lseek(packfd, 0, SEEK_SET) != 0)
		die("Failed seeking to start: %s", strerror(errno));

	SHA1_Init(&c);
	if (yread(packfd, hdr, 8) != 8)
		die("Failed reading header: %s", strerror(errno));
	SHA1_Update(&c, hdr, 8);

fprintf(stderr, "%lu objects\n", object_count);
	cnt = htonl(object_count);
	SHA1_Update(&c, &cnt, 4);
	if (ywrite(packfd, &cnt, 4) != 4)
		die("Failed writing object count: %s", strerror(errno));

	buf = xmalloc(128 * 1024);
	for (;;) {
		n = xread(packfd, buf, 128 * 1024);
		if (n <= 0)
			break;
		SHA1_Update(&c, buf, n);
	}
	free(buf);

	SHA1_Final(sha1, &c);
	if (ywrite(packfd, sha1, sizeof(sha1)) != sizeof(sha1))
		die("Failed writing pack checksum: %s", strerror(errno));
}

int main (int argc, const char **argv)
{
	packfd = open(argv[1], O_RDWR|O_CREAT|O_TRUNC, 0666);
	if (packfd < 0)
		die("Can't create pack file %s: %s", argv[1], strerror(errno));

	init_pack_header();
	for (;;) {
		unsigned long datlen;
		int hdrlen;
		void *dat;
		char hdr[128];
		unsigned char sha1[20];
		SHA_CTX c;

		if (yread(0, &datlen, 4) != 4)
			break;

		dat = xmalloc(datlen);
		if (yread(0, dat, datlen) != datlen)
			break;

		hdrlen = sprintf(hdr, "blob %lu", datlen) + 1;
		SHA1_Init(&c);
		SHA1_Update(&c, hdr, hdrlen);
		SHA1_Update(&c, dat, datlen);
		SHA1_Final(sha1, &c);

		write_blob(dat, datlen);
		object_count++;
		printf("%s\n", sha1_to_hex(sha1));
		fflush(stdout);

		if (lastdat)
			free(lastdat);
		lastdat = dat;
		lastdatlen = datlen;
		memcpy(lastsha1, sha1, sizeof(sha1));
	}
	fixup_header_footer();
	close(packfd);

	return 0;
}
Previous: Shawn Pearce
Message 36 of 36 in “Creating objects manually and repack”
  1. Jon SmirlAug 4, 2006
  2. Jeff KingAug 4, 2006
  3. Linus TorvaldsAug 4, 2006
  4. Jon SmirlAug 4, 2006
  5. Linus TorvaldsAug 4, 2006
  6. Linus TorvaldsAug 4, 2006
  7. Jon SmirlAug 4, 2006
  8. Jon SmirlAug 4, 2006
  9. Jon SmirlAug 4, 2006
  10. Linus TorvaldsAug 4, 2006
  11. Jon SmirlAug 4, 2006
  12. A Large Angry SCMAug 4, 2006
  13. Jon SmirlAug 4, 2006
  14. Linus TorvaldsAug 4, 2006
  15. Linus TorvaldsAug 4, 2006
  16. Rogan DawesAug 4, 2006
  17. Jon SmirlAug 4, 2006
  18. Linus TorvaldsAug 4, 2006
  19. Jon SmirlAug 4, 2006
  20. Linus TorvaldsAug 4, 2006
  21. Linus TorvaldsAug 4, 2006
  22. Junio C HamanoAug 4, 2006
  23. Linus TorvaldsAug 4, 2006
  24. Carl WorthAug 4, 2006
  25. Junio C HamanoAug 4, 2006
  26. Carl WorthAug 4, 2006
  27. Carl WorthAug 4, 2006
  28. Jakub NarebskiAug 4, 2006
  29. Junio C HamanoAug 4, 2006
  30. Jakub NarebskiAug 4, 2006
  31. Martin LanghoffAug 5, 2006
  32. Jon SmirlAug 5, 2006
  33. Shawn PearceAug 5, 2006
  34. Jon SmirlAug 5, 2006
  35. Shawn PearceAug 5, 2006
  36. Shawn PearceAug 5, 2006

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.