git-archive for files changed in revision range

git-archive for files changed in revision range

[Graham Cox]

[-- Attachment #1: Type: text/plain, Size: 1141 bytes --]

Hiya,

We have a small project that is being managed in a Git repository (MSysGit to 
be exact) - mostly for backups and so on. The project is a mod for the 
computer game Civilization 4. (Actually a mod of a mod, but still...)  As 
such, to release the mod to other people to actually use the only thing that 
needs to be released is all of the files that have actually been changed. (The 
actual git repository contains ~700MB of files, the vast majority of which 
haven't changed since the initial import and so don't need to be downloaded by 
people).

I've managed to make it produce an archive that contains only the files that 
have changed by using a combination of git-archive and git-whatchanged, along 
with grep and sed, but it's kinda unwieldly. Is there a better way of doing 
this?

The command line I used was something like (This is mostly from memory):
git-archive --format=zip . `git-whatchanged <start>..HEAD --pretty=oneline 
  | grep '^:' | sed 's/^.*\t//'` > release.zip

To produce a zip containing all of the modified and added files for the 
revision range <start>..HEAD.

-- 
Graham Cox


[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: git-archive for files changed in revision range

[Johannes Schindelin]

Hi,

On Wed, 13 Feb 2008, Graham Cox wrote:

> We have a small project that is being managed in a Git repository 
> (MSysGit to be exact) - mostly for backups and so on. The project is a 
> mod for the computer game Civilization 4. (Actually a mod of a mod, but 
> still...)  As such, to release the mod to other people to actually use 
> the only thing that needs to be released is all of the files that have 
> actually been changed. (The actual git repository contains ~700MB of 
> files, the vast majority of which haven't changed since the initial 
> import and so don't need to be downloaded by people).
> 
> I've managed to make it produce an archive that contains only the files 
> that have changed by using a combination of git-archive and 
> git-whatchanged, along with grep and sed, but it's kinda unwieldly. Is 
> there a better way of doing this?
> 
> The command line I used was something like (This is mostly from memory):
> git-archive --format=zip . `git-whatchanged <start>..HEAD --pretty=oneline 
>   | grep '^:' | sed 's/^.*\t//'` > release.zip
> 
> To produce a zip containing all of the modified and added files for the 
> revision range <start>..HEAD.

Mebbe

	git archive --format=zip HEAD \
		$(git diff-tree -r --name-only --diff-filter=AM \
			<start>..HEAD)

Hmm?

Ciao,
Dscho

Re: git-archive for files changed in revision range

[Jeff King]

On Wed, Feb 13, 2008 at 01:03:04PM +0000, Graham Cox wrote:

> The command line I used was something like (This is mostly from memory):
> git-archive --format=zip . `git-whatchanged <start>..HEAD --pretty=oneline 
>   | grep '^:' | sed 's/^.*\t//'` > release.zip

That will list files multiple times if they were modified in more than
one commit. And really, there's no need to walk the history. You really
are just comparing against two points (your baseline and your current
state). Walking the history will also erroneously include files which
changed, but then reverted back to their original state (though I expect
that is the uncommon case).

How about:

  # mark our start point for this and future releases
  git tag baseline `git rev-list HEAD | tail -n 1`

  # find files which differ between then and now; but we only need
  # to care about added and modified files, since deleted ones
  # don't need to be shipped
  git diff --diff-filter=AM --raw baseline HEAD | awk '{print $6}' >files

  # and archive
  git archive --format=zip HEAD `cat files`

You can even use diff's "-z" option if you have filenames that are hard
to quote, but I will leave that as an exercise to the reader.

-Peff

Re: git-archive for files changed in revision range

[Johannes Sixt]

Graham Cox schrieb:
> The command line I used was something like (This is mostly from memory):
> git-archive --format=zip . `git-whatchanged <start>..HEAD --pretty=oneline 
>   | grep '^:' | sed 's/^.*\t//'` > release.zip
> 
> To produce a zip containing all of the modified and added files for the 
> revision range <start>..HEAD.

git diff-tree -r --name-only -z <start> HEAD | \
    xargs -0 git archive --format=zip HEAD > release.zip

-- Hannes