For big projects with multiple contributors where nobody is
really cracking the whip on Github maintenance, there is a tendency for the
branch count to just keep climbing. I
have seen cases where it tops 100 branches, and it becomes quite painful to
clean up. In cases like that, I have
developed the following method for handling it:
First, you need local copies of all the remote
branches. This is not as easy as just
doing a fetch origin, as that just pulls over an underlying awareness of the
fact that the branches exist. You do not
actually create local copies of all those branches.
So we run the following command:
for remote in `git branch -r`; do git branch
--track ${remote#origin/} $remote; done
This creates local copies of all the branches including one
called HEAD, which is not a real branch and is dangerous. So let’s get rid of that by renaming it to something else and then deleting it.
git branch -m HEAD bogusBranch
git branch –D bogusBranch
We now want to delete all the branches that are already
merged. We can get a list of them by
running:
git branch --merged master | grep -v master
If we like the list, we now delete them from the origin:
git branch --merged master | grep -v master |
xargs -n 1 git push --delete origin
And we now delete the same list locally:
git branch --merged master | grep -v master |
xargs -n 1 git branch -D
And we have now cleaned up our repository by removing
everything that has already been merged to master.
If something is not on the merged list, and we want to know
what changes are not merged, we get a list of un-merged commits
from that branch:
git cherry -v master <unmergedBranchName>
Now we checkout that local branch:
git checkout unmergedBranchName
And we do a whatchanged to see what the difference is:
git whatchanged -m -n 1 -p <SHA of target
commit>
