From: Junio C Hamano <gitster@pobox.com>
To: "Nguyễn Thái Ngọc Duy" <pclouds@gmail.com>
Cc: git@vger.kernel.org, Nicolas Pitre <nico@fluxnic.net>
Subject: Re: [PATCH v2] pack-objects: use streaming interface for reading large loose blobs
Date: Mon, 14 May 2012 08:56:47 -0700 [thread overview]
Message-ID: <7vvcjyiwn4.fsf@alter.siamese.dyndns.org> (raw)
In-Reply-To: <1336883862-9013-1-git-send-email-pclouds@gmail.com> ("Nguyễn Thái Ngọc Duy"'s message of "Sun, 13 May 2012 11:37:42 +0700")
Nguyễn Thái Ngọc Duy <pclouds@gmail.com> writes:
> + st = open_istream(sha1, &type, &sz, NULL);
> + if (!st)
> + die(_("unable to read %s"), sha1_to_hex(sha1));
> +
> + memset(&stream, 0, sizeof(stream));
> + git_deflate_init(&stream, pack_compression_level);
> +
> + if (type != OBJ_BLOB)
> + die("BUG: %s is not a blob", sha1_to_hex(sha1));
Just a comment, not an expression of preference, but if we are treating
istream interface as an add-on, I wonder it might make sense not to die in
these places, and instead fall back to the usual "read and then write"
codepath.
To put it another way, I am wondering if it makes the logic more clear if
you restructure the calling side a bit more, perhaps like this:
if (!to_reuse) {
if (!usable_delta) {
/* we know we will write as base object */
... do *NOT* do read_sha1_file() here !!!
} else if ... delta cases {
decide delta_data and z_delta_size as before
}
if (entry->z_delta_size) {
datalen = entry->z_delta_size;
hdrlen = encode_in_pack_object_header(type, size, header);
... write either OFS_DELTA or REF_DELTA here
} else {
/* base object case */
if (write_object_in_base_representation(f, sha1)) {
/* stream interface punted */
read_sha1_file();
do_compress();
write it;
}
}
} else { /* reuse case ... */
> @@ -259,9 +309,13 @@ static unsigned long write_object(struct sha1file *f,
> if (!to_reuse) {
> no_reuse:
> if (!usable_delta) {
> - buf = read_sha1_file(entry->idx.sha1, &type, &size);
> - if (!buf)
> - die("unable to read %s", sha1_to_hex(entry->idx.sha1));
> + if (entry->type == OBJ_BLOB && entry->size > big_file_threshold)
> + buf = NULL;
Two comments:
- In get_object_details(), we do this:
if (big_file_threshold <= entry->size)
entry->no_try_delta = 1;
The new code is inconsistent with respect to the boundary conditions,
when the size is exactly at the threshold.
- Magic condition "buf has NULL means we will read via streaming
interface" feels somewhat hacky and tasteless. I am borderline OK with
such a hack whose scope is clearly limited to a short function like
this one, but at least it needs to be documented. I'd rather see the
conditionals that look at !buf/buf are rewritten to use a new bool
variable similar to to_reuse and usable_delta whose name tells the
reader more explicitly what is going on, though.
next prev parent reply other threads:[~2012-05-14 15:56 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-05-12 10:26 [PATCH] pack-objects: use streaming interface for reading large loose blobs Nguyễn Thái Ngọc Duy
2012-05-12 16:51 ` Nicolas Pitre
2012-05-13 4:37 ` [PATCH v2] " Nguyễn Thái Ngọc Duy
2012-05-14 15:56 ` Junio C Hamano [this message]
2012-05-14 19:43 ` Junio C Hamano
2012-05-15 11:18 ` Nguyen Thai Ngoc Duy
2012-05-15 15:27 ` Junio C Hamano
2012-05-16 7:09 ` Nguyen Thai Ngoc Duy
2012-05-16 12:02 ` [PATCH v2 1/4] streaming: allow to call close_istream(NULL); Nguyễn Thái Ngọc Duy
2012-05-16 12:02 ` [PATCH v2 2/4] pack-objects, streaming: turn "xx >= big_file_threshold" to ".. > .." Nguyễn Thái Ngọc Duy
2012-05-18 21:05 ` Junio C Hamano
2012-05-16 12:02 ` [PATCH v2 3/4] pack-objects: refactor write_object() Nguyễn Thái Ngọc Duy
2012-05-18 21:16 ` Junio C Hamano
2012-05-19 2:43 ` Nicolas Pitre
2012-05-16 12:02 ` [PATCH v2 4/4] pack-objects: use streaming interface for reading large loose blobs Nguyễn Thái Ngọc Duy
2012-05-18 21:02 ` [PATCH v2 1/4] streaming: allow to call close_istream(NULL); Junio C Hamano
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: http://vger.kernel.org/majordomo-info.html
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=7vvcjyiwn4.fsf@alter.siamese.dyndns.org \
--to=gitster@pobox.com \
--cc=git@vger.kernel.org \
--cc=nico@fluxnic.net \
--cc=pclouds@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
Code repositories for project(s) associated with this public inbox
https://80x24.org/mirrors/git.git
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).