Thinning a repository

Hi everyone,

Is there a quick way to reproduce the effect of a shallow clone on a local repository that doesn't involve filter-branch and/or re-clone?

My motivation is to reduce the local size of repositories I'm only following, by trimming the history without prejudice to a [N] set of last commits. It feels stupid that the quickest way I'm aware of right now to achieve this is to "git clone --depth N ..." again. filter-branch is ridiculously slow, as it iterates through history.

I've tried using graft points, but the combination of:

echo [sha] > .git/info/grafts git reflog expire --expire=0 --all git repack -Ad

doesn't really save any space and/or reduce the object count as I would expect. It means there's probably still reachable?

I'd really love to have a 'git thin [depth]' subcommand to perform the above however. I don't really want to have to iterate through refs just to check if they are still reachable within [n] commits just to delete them.

Thanks for any pointer.

On Sun, Nov 30, 2014 at 01:18:34PM +0100, Yuri D'Elia wrote:

I'm curious, why is it a bad thing to do a re-clone? If you clone your local repo it would be really fast.

On 11/30/2014 01:34 PM, Fredrik Gustafsson wrote:

I see no reason to involve the workspace. I would also want to keep the current remote/origin as it is.

I expect a 'thin' command to be faster by not touching the workspace at all.