git switch branch without discarding local changes

Question

Alright, lets say one day we make happen to make a bunch of modifications and when we go to commit them we notice we were working on the wrong branch.

How can we force git to switch branches without discarding local changes.

I'm probably going to go about this in a naive way while I wait a reply, but I would like to know if theres a correct procedure as I'd be lying if I said this hasn't happened to me before...

• Backup changed repo
• git reset --hard
• git checkout right-branch
• Restore changes
• git commit -m "changes"

Answer 1

There are a bunch of different ways depending on how far along you are and which branch(es) you want them on.

Let's take a classic mistake:

$ git checkout master
... pause for coffee, etc ...
... return, edit a bunch of stuff, then: oops, wanted to be on develop

So now you want these changes, which you have not yet committed to master, to be on develop.

1. If you don't have a develop yet, the method is trivial:

   $ git checkout -b develop
   

   This creates a new develop branch starting from wherever you are now. Now you can commit and the new stuff is all on develop.

2. You do have a develop. See if Git will let you switch without doing anything:

   $ git checkout develop
   

   This will either succeed, or complain. If it succeeds, great! Just commit. If not (error: Your local changes to the following files would be overwritten ...), you still have lots of options.

   The easiest is probably git stash (as all the other answer-ers that beat me to clicking post said). Run git stash save or git stash push,¹ or just plain git stash which is short for save / push:

   $ git stash
   

   This commits your code (yes, it really does make some commits) using a weird non-branch-y method. The commits it makes are not "on" any branch but are now safely stored in the repository, so you can now switch branches, then "apply" the stash:

   $ git checkout develop
   Switched to branch 'develop'
   $ git stash apply
   

   If all goes well, and you like the results, you should then git stash drop the stash. This deletes the reference to the weird non-branch-y commits. (They're still in the repository, and can sometimes be retrieved in an emergency, but for most purposes, you should consider them gone at that point.)

The apply step does a merge of the stashed changes, using Git's powerful underlying merge machinery, the same kind of thing it uses when you do branch merges. This means you can get "merge conflicts" if the branch you were working on by mistake, is sufficiently different from the branch you meant to be working on. So it's a good idea to inspect the results carefully before you assume that the stash applied cleanly, even if Git itself did not detect any merge conflicts.

Many people use git stash pop, which is short-hand for git stash apply && git stash drop. That's fine as far as it goes, but it means that if the application results in a mess, and you decide you don't want to proceed down this path, you can't get the stash back easily. That's why I recommend separate apply, inspect results, drop only if/when satisfied. (This does of course introduce another point where you can take another coffee break and forget what you were doing, come back, and do the wrong thing, so it's not a perfect cure.)

---

¹The save in git stash save is the old verb for creating a new stash. Git version 2.13 introduced the new verb to make things more consistent with pop and to add more options to the creation command. Git version 2.16 formally deprecated the old verb (though it still works in Git 2.23, which is the latest release at the time I am editing this).