git/Documentation/git-cat-file.txt
Jeff King 0750bb5b51 cat-file: support "unordered" output for --batch-all-objects
If you're going to access the contents of every object in a
packfile, it's generally much more efficient to do so in
pack order, rather than in hash order. That increases the
locality of access within the packfile, which in turn is
friendlier to the delta base cache, since the packfile puts
related deltas next to each other. By contrast, hash order
is effectively random, since the sha1 has no discernible
relationship to the content.

This patch introduces an "--unordered" option to cat-file
which iterates over packs in pack-order under the hood. You
can see the results when dumping all of the file content:

  $ time ./git cat-file --batch-all-objects --buffer --batch | wc -c
  6883195596

  real	0m44.491s
  user	0m42.902s
  sys	0m5.230s

  $ time ./git cat-file --unordered \
                        --batch-all-objects --buffer --batch | wc -c
  6883195596

  real	0m6.075s
  user	0m4.774s
  sys	0m3.548s

Same output, different order, way faster. The same speed-up
applies even if you end up accessing the object content in a
different process, like:

  git cat-file --batch-all-objects --buffer --batch-check |
  grep blob |
  git cat-file --batch='%(objectname) %(rest)' |
  wc -c

Adding "--unordered" to the first command drops the runtime
in git.git from 24s to 3.5s.

  Side note: there are actually further speedups available
  for doing it all in-process now. Since we are outputting
  the object content during the actual pack iteration, we
  know where to find the object and could skip the extra
  lookup done by oid_object_info(). This patch stops short
  of that optimization since the underlying API isn't ready
  for us to make those sorts of direct requests.

So if --unordered is so much better, why not make it the
default? Two reasons:

  1. We've promised in the documentation that --batch-all-objects
     outputs in hash order. Since cat-file is plumbing,
     people may be relying on that default, and we can't
     change it.

  2. It's actually _slower_ for some cases. We have to
     compute the pack revindex to walk in pack order. And
     our de-duplication step uses an oidset, rather than a
     sort-and-dedup, which can end up being more expensive.
     If we're just accessing the type and size of each
     object, for example, like:

       git cat-file --batch-all-objects --buffer --batch-check

     my best-of-five warm cache timings go from 900ms to
     1100ms using --unordered. Though it's possible in a
     cold-cache or under memory pressure that we could do
     better, since we'd have better locality within the
     packfile.

And one final question: why is it "--unordered" and not
"--pack-order"? The answer is again two-fold:

  1. "pack order" isn't a well-defined thing across the
     whole set of objects. We're hitting loose objects, as
     well as objects in multiple packs, and the only
     ordering we're promising is _within_ a single pack. The
     rest is apparently random.

  2. The point here is optimization. So we don't want to
     promise any particular ordering, but only to say that
     we will choose an ordering which is likely to be
     efficient for accessing the object content. That leaves
     the door open for further changes in the future without
     having to add another compatibility option.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2018-08-13 13:48:31 -07:00

314 lines
9.9 KiB
Plaintext

git-cat-file(1)
===============
NAME
----
git-cat-file - Provide content or type and size information for repository objects
SYNOPSIS
--------
[verse]
'git cat-file' (-t [--allow-unknown-type]| -s [--allow-unknown-type]| -e | -p | <type> | --textconv | --filters ) [--path=<path>] <object>
'git cat-file' (--batch | --batch-check) [ --textconv | --filters ] [--follow-symlinks]
DESCRIPTION
-----------
In its first form, the command provides the content or the type of an object in
the repository. The type is required unless `-t` or `-p` is used to find the
object type, or `-s` is used to find the object size, or `--textconv` or
`--filters` is used (which imply type "blob").
In the second form, a list of objects (separated by linefeeds) is provided on
stdin, and the SHA-1, type, and size of each object is printed on stdout. The
output format can be overridden using the optional `<format>` argument. If
either `--textconv` or `--filters` was specified, the input is expected to
list the object names followed by the path name, separated by a single white
space, so that the appropriate drivers can be determined.
OPTIONS
-------
<object>::
The name of the object to show.
For a more complete list of ways to spell object names, see
the "SPECIFYING REVISIONS" section in linkgit:gitrevisions[7].
-t::
Instead of the content, show the object type identified by
<object>.
-s::
Instead of the content, show the object size identified by
<object>.
-e::
Exit with zero status if <object> exists and is a valid
object. If <object> is of an invalid format exit with non-zero and
emits an error on stderr.
-p::
Pretty-print the contents of <object> based on its type.
<type>::
Typically this matches the real type of <object> but asking
for a type that can trivially be dereferenced from the given
<object> is also permitted. An example is to ask for a
"tree" with <object> being a commit object that contains it,
or to ask for a "blob" with <object> being a tag object that
points at it.
--textconv::
Show the content as transformed by a textconv filter. In this case,
<object> has to be of the form <tree-ish>:<path>, or :<path> in
order to apply the filter to the content recorded in the index at
<path>.
--filters::
Show the content as converted by the filters configured in
the current working tree for the given <path> (i.e. smudge filters,
end-of-line conversion, etc). In this case, <object> has to be of
the form <tree-ish>:<path>, or :<path>.
--path=<path>::
For use with --textconv or --filters, to allow specifying an object
name and a path separately, e.g. when it is difficult to figure out
the revision from which the blob came.
--batch::
--batch=<format>::
Print object information and contents for each object provided
on stdin. May not be combined with any other options or arguments
except `--textconv` or `--filters`, in which case the input lines
also need to specify the path, separated by white space. See the
section `BATCH OUTPUT` below for details.
--batch-check::
--batch-check=<format>::
Print object information for each object provided on stdin. May
not be combined with any other options or arguments except
`--textconv` or `--filters`, in which case the input lines also
need to specify the path, separated by white space. See the
section `BATCH OUTPUT` below for details.
--batch-all-objects::
Instead of reading a list of objects on stdin, perform the
requested batch operation on all objects in the repository and
any alternate object stores (not just reachable objects).
Requires `--batch` or `--batch-check` be specified. Note that
the objects are visited in order sorted by their hashes.
--buffer::
Normally batch output is flushed after each object is output, so
that a process can interactively read and write from
`cat-file`. With this option, the output uses normal stdio
buffering; this is much more efficient when invoking
`--batch-check` on a large number of objects.
--unordered::
When `--batch-all-objects` is in use, visit objects in an
order which may be more efficient for accessing the object
contents than hash order. The exact details of the order are
unspecified, but if you do not require a specific order, this
should generally result in faster output, especially with
`--batch`. Note that `cat-file` will still show each object
only once, even if it is stored multiple times in the
repository.
--allow-unknown-type::
Allow -s or -t to query broken/corrupt objects of unknown type.
--follow-symlinks::
With --batch or --batch-check, follow symlinks inside the
repository when requesting objects with extended SHA-1
expressions of the form tree-ish:path-in-tree. Instead of
providing output about the link itself, provide output about
the linked-to object. If a symlink points outside the
tree-ish (e.g. a link to /foo or a root-level link to ../foo),
the portion of the link which is outside the tree will be
printed.
+
This option does not (currently) work correctly when an object in the
index is specified (e.g. `:link` instead of `HEAD:link`) rather than
one in the tree.
+
This option cannot (currently) be used unless `--batch` or
`--batch-check` is used.
+
For example, consider a git repository containing:
+
--
f: a file containing "hello\n"
link: a symlink to f
dir/link: a symlink to ../f
plink: a symlink to ../f
alink: a symlink to /etc/passwd
--
+
For a regular file `f`, `echo HEAD:f | git cat-file --batch` would print
+
--
ce013625030ba8dba906f756967f9e9ca394464a blob 6
--
+
And `echo HEAD:link | git cat-file --batch --follow-symlinks` would
print the same thing, as would `HEAD:dir/link`, as they both point at
`HEAD:f`.
+
Without `--follow-symlinks`, these would print data about the symlink
itself. In the case of `HEAD:link`, you would see
+
--
4d1ae35ba2c8ec712fa2a379db44ad639ca277bd blob 1
--
+
Both `plink` and `alink` point outside the tree, so they would
respectively print:
+
--
symlink 4
../f
symlink 11
/etc/passwd
--
OUTPUT
------
If `-t` is specified, one of the <type>.
If `-s` is specified, the size of the <object> in bytes.
If `-e` is specified, no output, unless the <object> is malformed.
If `-p` is specified, the contents of <object> are pretty-printed.
If <type> is specified, the raw (though uncompressed) contents of the <object>
will be returned.
BATCH OUTPUT
------------
If `--batch` or `--batch-check` is given, `cat-file` will read objects
from stdin, one per line, and print information about them. By default,
the whole line is considered as an object, as if it were fed to
linkgit:git-rev-parse[1].
You can specify the information shown for each object by using a custom
`<format>`. The `<format>` is copied literally to stdout for each
object, with placeholders of the form `%(atom)` expanded, followed by a
newline. The available atoms are:
`objectname`::
The 40-hex object name of the object.
`objecttype`::
The type of the object (the same as `cat-file -t` reports).
`objectsize`::
The size, in bytes, of the object (the same as `cat-file -s`
reports).
`objectsize:disk`::
The size, in bytes, that the object takes up on disk. See the
note about on-disk sizes in the `CAVEATS` section below.
`deltabase`::
If the object is stored as a delta on-disk, this expands to the
40-hex sha1 of the delta base object. Otherwise, expands to the
null sha1 (40 zeroes). See `CAVEATS` below.
`rest`::
If this atom is used in the output string, input lines are split
at the first whitespace boundary. All characters before that
whitespace are considered to be the object name; characters
after that first run of whitespace (i.e., the "rest" of the
line) are output in place of the `%(rest)` atom.
If no format is specified, the default format is `%(objectname)
%(objecttype) %(objectsize)`.
If `--batch` is specified, the object information is followed by the
object contents (consisting of `%(objectsize)` bytes), followed by a
newline.
For example, `--batch` without a custom format would produce:
------------
<sha1> SP <type> SP <size> LF
<contents> LF
------------
Whereas `--batch-check='%(objectname) %(objecttype)'` would produce:
------------
<sha1> SP <type> LF
------------
If a name is specified on stdin that cannot be resolved to an object in
the repository, then `cat-file` will ignore any custom format and print:
------------
<object> SP missing LF
------------
If --follow-symlinks is used, and a symlink in the repository points
outside the repository, then `cat-file` will ignore any custom format
and print:
------------
symlink SP <size> LF
<symlink> LF
------------
The symlink will either be absolute (beginning with a /), or relative
to the tree root. For instance, if dir/link points to ../../foo, then
<symlink> will be ../foo. <size> is the size of the symlink in bytes.
If --follow-symlinks is used, the following error messages will be
displayed:
------------
<object> SP missing LF
------------
is printed when the initial symlink requested does not exist.
------------
dangling SP <size> LF
<object> LF
------------
is printed when the initial symlink exists, but something that
it (transitive-of) points to does not.
------------
loop SP <size> LF
<object> LF
------------
is printed for symlink loops (or any symlinks that
require more than 40 link resolutions to resolve).
------------
notdir SP <size> LF
<object> LF
------------
is printed when, during symlink resolution, a file is used as a
directory name.
CAVEATS
-------
Note that the sizes of objects on disk are reported accurately, but care
should be taken in drawing conclusions about which refs or objects are
responsible for disk usage. The size of a packed non-delta object may be
much larger than the size of objects which delta against it, but the
choice of which object is the base and which is the delta is arbitrary
and is subject to change during a repack.
Note also that multiple copies of an object may be present in the object
database; in this case, it is undefined which copy's size or delta base
will be reported.
GIT
---
Part of the linkgit:git[1] suite