Merge git repository in subdirectory

user_375842 0
Viewed 33458

I'd like to merge a remote git repository in my working git repository as a subdirectory of it. I'd like the resulting repository to contain the merged history of the two repositories and also that each file of the merged-in repository retain its history as it was in the remote repository. I tried using the subtree strategy as mentioned in How to use the subtree merge strategy, but after following that procedure, although the resulting repository contains indeed the merged history of the two repositories, individual files coming from the remote one haven't retained their history (`git log' on any of them just shows a message "Merged branch...").

Also I don't want to use submodules because I do not want the two combined git repositories to be separate anymore.

Is it possible to merge a remote git repository in another one as a subdirectory with individual files coming from the remote repository retaining their history?

Thanks very much for any help.

EDIT: I'm currently trying out a solution that uses git filter-branch to rewrite the merged-in repository history. It does seem to work, but I need to test it some more. I'll return to report on my findings.

EDIT 2: In hope I make myself more clear I give the exact commands I used with git's subtree strategy, which result in apparent loss of history of the files of the remote repository. Let A be the git repo I'm currently working in and B the git repo I'd like to incorporate into A as a subdirectory of it. It did the following:

git remote add -f B <url-of-B>
git merge -s ours --no-commit B/master
git read-tree --prefix=subdir/Iwant/to/put/B/in/ -u B/master
git commit -m "Merge B as subdirectory in subdir/Iwant/to/put/B/in."

After these commands and going into directory subdir/Iwant/to/put/B/in, I see all files of B, but git log on any one of them shows just the commit message "Merge B as subdirectory in subdir/Iwant/to/put/B/in." Their file history as it is in B is lost.

What seems to work (since I'm a beginner on git I may be wrong) is the following:

git remote add -f B <url-of-B>
git checkout -b B_branch B/master  # make a local branch following B's master
git filter-branch --index-filter \ 
   'git ls-files -s | sed "s-\t\"*-&subdir/Iwant/to/put/B/in/-" |
        GIT_INDEX_FILE=$GIT_INDEX_FILE.new \
                git update-index --index-info &&
        mv "$GIT_INDEX_FILE.new" "$GIT_INDEX_FILE"' HEAD 
git checkout master
git merge B_branch

The command above for filter-branch is taken from git help filter-branch, in which I only changed the subdir path.

git
8 Answers

I wanted to

  1. keep a linear history without explicit merge, and
  2. make it look like the files of the merged repository had always existed in the subdirectory, and as a side effect make git log -- file work without --follow.

Step 1: Rewrite history in the source repository to make it look like all files always existed below the subdirectory.

Create a temporary branch for the rewritten history.

git checkout -b tmp_subdir

Then use git filter-branch as described in How can I rewrite history so that all files, except the ones I already moved, are in a subdirectory?:

git filter-branch --prune-empty --tree-filter '
if [ ! -e foo/bar ]; then
    mkdir -p foo/bar
    git ls-tree --name-only $GIT_COMMIT | xargs -I files mv files foo/bar
fi'

Step 2: Switch to the target repository. Add the source repository as remote in the target repository and fetch its contents.

git remote add sourcerepo .../path/to/sourcerepo
git fetch sourcerepo

Step 3: Use merge --onto to add the commits of the rewritten source repository on top of the target repository.

git rebase --preserve-merges --onto master --root sourcerepo/tmp_subdir

You can check the log to see that this really got you what you wanted.

git log --stat

Step 4: After the rebase you’re in “detached HEAD” state. You can fast-forward master to the new head.

git checkout -b tmp_merged
git checkout master
git merge tmp_merged
git branch -d tmp_merged

Step 5: Finally some cleanup: Remove the temporary remote.

git remote rm sourcerepo
user_3812810

Say you want to merge repository a into b (I'm assuming they're located alongside one another):

cd a
git filter-repo --to-subdirectory-filter a
cd ..
cd b
git remote add a ../a
git fetch a
git merge --allow-unrelated-histories a/master
git remote remove a

For this you need git-filter-repo installed (filter-branch is discouraged).

An example of merging 2 big repositories, putting one of them into a subdirectory: https://gist.github.com/x-yuri/9890ab1079cf4357d6f269d073fd9731

More on it here.

user_524990

Similar to hfs' answer I wanted to

  • keep a linear history without explicit merge and
  • make it look like the files of the merged repository had always existed in the subdirectory, and as a side effect make git log -- file work without --follow.

However, I chose the more modern filter-repo (assuming the new repo exists and is checked out):

git clone git@host/repo/old.git
cd old
git checkout -b tmp_subdir
git filter-repo --to-subdirectory-filter old

cd ../new
git remote add old ../old
git fetch old
git rebase --rebase-merges --onto main --root old/tmp_subdir --committer-date-is-author-date

you might need to fix conflicts (manually) or change the rebase command to include --merge -s recursive -X theirs if you want to try solving it with theirs version:

git rebase --rebase-merges --onto main --root old/tmp_subdir --committer-
date-is-author-date --merge -s recursive -X theirs

you end up on a detached HEAD, so create a new branch and merge it to main note that modern repositories should not use a "master" branch but a "main"

branch for a more inclusive language.
git checkout -b old_merge
git checkout main
git merge old_merge

cleanup

git branch -d old_merge
git remote rm old
user_3862010
Related