Collapsing a git repository’s history

You can remove the binary bloat and keep the rest of your history. Git allows you to reorder and ‘squash’ prior commits, so you can combine just the commits that add and remove your big binary files. If the adds were all done in one commit and the removals in another, this will be much easier than dealing with each file.

$ git log --stat       # list all commits and commit messages 

Search this for the commits that add and delete your binary files and note their SHA1s, say 2bcdef and 3cdef3.

Then to edit the repo’s history, use rebase -i command with its interactive option, starting with the parent of the commit where you added your binaries. It will launch your $EDITOR and you’ll see a list of commits starting with 2bcdef:

$ git rebase -i 2bcdef^    # generate a pick list of all commits starting with 2bcdef
# Rebasing zzzzzz onto yyyyyyy 
# 
# Commands: 
#  pick = use commit 
#  edit = use commit, but stop for amending 
#  squash = use commit, but meld into previous commit 
# 
# If you remove a line here THAT COMMIT WILL BE LOST.
#
pick 2bcdef   Add binary files and other edits
pick xxxxxx   Another change
  .
  .
pick 3cdef3   Remove binary files; link to them as external resources
  .
  .

Insert squash 3cdef3 as the second line and remove the line which says pick 3cdef3 from the list. You now have a list of actions for the interactive rebase which will combine the commits which add and delete your binaries into one commit whose diff is just any other changes in those commits. Then it will reapply all of the subsequent commits in order, when you tell it to complete:

$ git rebase --continue

This will take a minute or two.
You now have a repo that no longer has the binaries coming or going. But they will still take up space because, by default, Git keeps changes around for 30 days before they can be garbage-collected, so that you can change your mind.
If you want to remove them now:

$ git reflog expire --expire=1.minute refs/heads/master
      #all deletions up to 1 minute  ago available to be garbage-collected
$ git fsck --unreachable      # lists all the blobs(files) that will be garbage-collected
$ git prune
$ git gc                      

Now you’ve removed the bloat but kept the rest of your history.

Leave a Comment