extract multiple directories using git-filter-branch

Question

I have a big repository which currently contains multiple projects in top level subfolders, say /a, /b, /c, and /d.

Answer 1

Use

git filter-branch -f --prune-empty --tree-filter 'bash preserve-only.sh a b' -- --all

where preserve-only.sh is:

IFS=':'
GLOBIGNORE="$*"
rm -rf *

This should remove everything but a and b from all commits of all branches, which should be the same as extracting exactly the given directories.

To complete the actual split you can use a filter like rm -rf a b to get all the changes not extracted in the first run.

---

Update: While trying to speed things up using --index-filter I came to an even easier solution:

git filter-branch -f --prune-empty --index-filter \
  'git rm --cached -r -q -- . ; git reset -q $GIT_COMMIT -- a b' -- --all

This just removes everything and afterwards restores the given directories afterwards.

Answer 2

After searching around and trying the proposed solutions, it seems like the recommended way of doing is now with git-filter-repo (see here)

git filter-repo --path a --path b

Answer 3

I'm not aware of any better way than tree-filter for this. So you already have all the information you need. Now just do it!

Start by creating your two branches:

git branch br1
git branch br2

Now for each branch, check it out, then filter it using the tree-filter.

You could then split them out to separate directories either by pushing them out, or by cloning or pulling them in.

Answer 4

I prefer this

git filter-branch -f --prune-empty --tree-filter "ls -I a -I b | xargs rm -rf"  -- --all