[PATCH/RFC v3 resend 0/4] fast-import: Let importers retrieve blobs

From: Jonathan Nieder <jrnieder@gmail.com>
Date: Nov 28, 2010, 19:41 UTC
Message-ID: <20101128194131.GA19998@burratino>
In-Reply-To: <1287147256-9457-2-git-send-email-david.barr@cordelta.com>

[resending since git@vger doesn't seem to have accepted the previous copy. Sorry for the noise.]

David Barr wrote:

Show 5 quoted lines

> So introduce another way: a "cat-blob" command introduced in the
> command stream requests for fast-import to print a blob to stdout
> or a file descriptor specified by the argument --cat-blob-fd.
> 
> The output uses the same format as "git cat-file --batch".

I am very fond of this patch. Still, the fact remains that until this command is implemented by some other fast-import backend, it is hard to know what git-specific concepts are encoded in its current implementation. (I am in particular worried about what should be gauranteed about the blob_identifier in the

	blob blob_identifier 1823

line and whether

	cat-blob blob_name
	cat-blob :11

are going to overlap and cause trouble for some backends.)

On the other hand, development of svn-fe continues to benefit from cat-blob and its cousins ls-tree and ls[1]. and there has been brief discussion of using cat-blob to make cvs2git more friendly. So here's a reroll, since it seems clear that this feature should be in future versions of fast-import in some form.

Thoughts welcome, as always (even as simple as "I have a bad feeling about this" or "everything in this patch looks ready to go").

Patches based against v1.7.0.7 for no particular reason.

David Barr (1):
  fast-import: let importers retrieve blobs

Jonathan Nieder (3):
  fast-import: stricter parsing of integer options
  fast-import: clarify documentation of "feature" command
  fast-import: Allow cat-blob requests at arbitrary points in stream

 Documentation/git-fast-import.txt |   76 ++++++++---
 fast-import.c                     |  128 ++++++++++++++++--
 t/t9300-fast-import.sh            |  267 ++++++++++++++++++++++++++++++++++++-
 3 files changed, 442 insertions(+), 29 deletions(-)

[1] ls-tree commit_name "path/to/file" would have output format 100644 blob blob_identifier path/to/file and within a commit command, ls "path/to/file" would produce output in the same format.