Reputation: 8055
How can I do the following in Git?
My current branch is branch1 and I have made some local changes. However, I now realize that I actually meant to be applying these changes to branch2. Is there a way to apply/merge these changes so that they become local changes on branch2 without committing them on branch1?
Upvotes: 714
Views: 162664
Reputation: 1430
WARNING: Not for git newbies.
Similar to chakrit's answer, I run into this situation quite often: Working on a feature branch, I find a bug and want to fix it. But the fix belongs on the main
branch, not my-feature
. The whole sequence to get the changes into main
is 7 or more git
commands, which is really quite annoying and easy to get wrong
Since I could not find a script to do it, I wrote it myself. Simply place it somewhere in $PATH
(e.g. /usr/local/bin
or /$HOME/.local/bin
or something), then you can do the following:
# currently working on branch `my-feature`
$ git add some-file # use git add -p if you want only some changes
$ git commit-branch main --rebase -m 'Fixed some nasty bug in some-file'
It will then print some progress messages:
Committing your staged changes to branch 'main'.
+ git checkout --quiet HEAD~0
+ git commit --quiet -m 'Fixed some nasty bug in some-file'
++ git rev-parse HEAD
+ commit_hash=82513091473646a09d541893b8bd60a0f98b765d
+ git stash push --quiet
+ git checkout --quiet main
+ git cherry-pick --quiet 82513091473646a09d541893b8bd60a0f98b765d
[main 1c5d96e] Fixed some nasty bug in some-file
Date: Mon Feb 6 15:04:03 2023 +0100
1 file changed, 2 insertions(+)
+ git checkout --quiet my-feature
+ git rebase --quiet main
+ git stash pop --quiet
+ set +x
Success.
Here's the source code for the file git-commit-branch
. Don't forget to do chmod +x
after placing it in $PATH
. The script is also on github: https://github.com/fritzw/git-utils. Feel free to suggest improvements.
The way it works is as follows:
If any command fails it will simply stop there and print some info to help you recover the situation. If you want more details, look at the comments and the sequence of git commands at the end of the script.
#!/usr/bin/env bash
set -o errexit
set -o nounset
usage() {
echo "Usage: git commit-branch <target-branch> [--rebase|-r] [ <git-commit-options>... ]"
echo ""
echo "Commits your staged changes to <target-branch>, discarding them from your current branch."
echo "Use --rebase or -r to rebase your current branch on the new commit in <target-branch>,"
echo "and thus include the changes in your current branch as well."
echo ""
echo "Example usage working on branch my-feature:"
echo " git add some-file"
echo " git commit-branch main --rebase -m 'Fixed a bug in some-file'"
}
if [[ $# -lt 1 ]]; then
usage
exit 1
fi
target_branch="$1"; shift # Remove first argument
if ! git rev-parse --verify "$target_branch" >/dev/null; then
echo "fatal: '$target_branch' is not a branch in this git repository."
usage
exit 1
fi
rebase_command=''
if [[ $# -gt 0 ]] && [[ "$1" == "-r" || "$1" == "--rebase" ]]; then
rebase_command="git rebase --quiet $target_branch"
shift # Remove -r/--rebase argument
fi
current_branch="$(git branch --show-current)"
if ! [[ "$current_branch" ]]; then
echo "fatal: Unable to determine current branch. You must be on a branch to use git commit-branch."
exit 1
fi
commit_hash='not-committed-yet'
print_error_message() {
set +x
echo
echo "Something went wrong in the last command. :-("
echo "Your unstaged changes and untracked files should be in the last stash."
echo "Your previously staged changes should be in the following commit: $commit_hash"
echo "Please check which commands were executed and try to undo them manually."
echo
}
echo "Committing your staged changes to branch '$target_branch'."
trap 'print_error_message' ERR # Print some hopefully helpful info if something fails
set -x # Print all executed commands
git checkout --quiet 'HEAD~0' # Go into 'detached HEAD' state to avoid changing current branch
git commit --quiet "$@" # Create temporary commit
commit_hash="$(git rev-parse HEAD)" # Note temporary commit ID
git stash push --include-untracked --quiet # Save all other changes from working tree
git checkout --quiet "$target_branch" # Move to target branch
git cherry-pick --quiet "$commit_hash" # Apply changes from temporary commit to target branch
git checkout --quiet "$current_branch" # Switch back to current branch
$rebase_command # Execute git rebase if --rebase flag is present
git stash pop --quiet # Re-apply untracked changes to working tree
set +x # Stop printing executed commands
echo "Success."
if ! [[ "$rebase_command" ]]; then
echo ""
echo "If you want to include those changes in your current branch, you can run:"
echo " git stash; git rebase $target_branch; git stash pop"
echo "or"
echo " git stash; git merge $target_branch; git stash pop"
echo ""
fi
Upvotes: 0
Reputation: 2315
I have found this answer useful.
However, as that thread is closed and is unable to make comments I have an issue with that answer.
When I applied git checkout other_branch
I got the following error
error: pathspec 'other_branch' did not match any file(s) known to git
So instead of applying the command, I rather use which fixed my problem
git branch other_branch
git checkout other_branch
Upvotes: 0
Reputation: 2044
When you created e.g a new file. VSCode e.g displays a U
next to the file in the file explorer.
When you make changes to files that were committed to the repo earlier (in previous commits).
So imagine you are on branch A
, but you want to commit only changes to existing files to branch A
, while the newly created file (untracked) should be committed to a new branch B
. It's possible to use stashing with a few tricks, explained step by step.
.git/config
Inside the .git
folder there is a config
file. Open it up and you will see something like this:
[core]
repositoryformatversion = 0
filemode = true
bare = false
logallrefupdates = true
ignorecase = true
precomposeunicode = true
[remote "origin"]
url = https://github.com/...
fetch = +refs/heads/*:refs/remotes/origin/*
[branch "main"]
remote = origin
merge = refs/heads/main
change the config file to:
[core]
repositoryformatversion = 0
filemode = true
bare = false
logallrefupdates = true
ignorecase = true
precomposeunicode = true
[alias]
stash-untracked = "!f() { \
git stash; \
git stash -u; \
git stash pop stash@{1}; \
}; f"
[remote "origin"]
url = https://github.com/...
fetch = +refs/heads/*:refs/remotes/origin/*
[branch "main"]
remote = origin
merge = refs/heads/main
Now you will be able to use the following command while you are on branch A.
git stash-untracked
You will see that the new file disappeared, if you are using a editor like VSCode (it's now stashed)
While still on branch A stage and commit the changes to the existing files:
git add .
git commit -m "committing tracked changes to current branch"
Next step is creating a new branch B (with checkout -b
you visit it immediately)
git checkout -b newBranchName
When using stash pop
the stashed changes get added to your current branch.
git stash pop
The only thing left is to stage and commit the changes on the new branch B
git add .
git commit -m "created new file"
Upvotes: 0
Reputation: 1328552
Since your files are not yet committed in branch1
:
git stash
git checkout branch2
git stash pop
or
git stash
git checkout branch2
git stash list # to check the various stash made in different branch
git stash apply x # to select the right one
Above is the longer more explicit version of rbento's answer:
git stash
git stash branch branch2
It uses:
git stash branch <branchname> [<stash>]
- Creates and checks out a new branch named
<branchname>
starting from the commit at which the<stash>
was originally created,- applies the changes recorded in
<stash>
to the new working tree and index.If that succeeds, and
<stash>
is a reference of the formstash@{<revision>}
, it then drops the<stash>
.This is useful if the branch on which you ran
git stash push
has changed enough thatgit stash apply
fails due to conflicts.
Since the stash entry is applied on top of the commit that was HEAD at the timegit stash
was run, it restores the originally stashed state with no conflicts.
As commented by benjohn (see git stash
man page):
To also stash currently untracked (newly added) files, add the argument
-u
, so:
git stash -u
Upvotes: 1000
Reputation: 11668
A shorter alternative to the accepted answer would be:
Temporarily move the changes to a stash.
git stash
Create and switch to a new branch and then pop the stash to it in just one step.
git stash branch new_branch_name
Then just add
and commit
the changes to this new branch.
Upvotes: 17
Reputation: 605
These are the steps I follow:
You can check the status and which branch you are on using:
Note: Here if you make changes in your local repo before moving to the new branch, the following steps should still work.
If "git branch" shows master, and you want to create+move to another branch:
Check branch again using "git branch" It should now show that you are in the new branch.
Now add, commit and push:
The above steps work for me in both the situation when I have made changes before moving to the new local branch or making changes after moving to the new branch. I hope it helps people running into similar situations.
Upvotes: 1
Reputation: 61558
WARNING: Not for git newbies.
This comes up enough in my workflow that I've almost tried to write a new git command for it. The usual git stash
flow is the way to go but is a little awkward. I usually make a new commit first since if I have been looking at the changes, all the information is fresh in my mind and it's better to just start git commit
-ing what I found (usually a bugfix belonging on master that I discover while working on a feature branch) right away.
It is also helpful—if you run into situations like this a lot—to have another working directory alongside your current one that always have the
master
branch checked out.
So how I achieve this goes like this:
git commit
the changes right away with a good commit message.git reset HEAD~1
to undo the commit from current branch.Sometimes later (asynchronously), or immediately in another terminal window:
cd my-project-master
which is another WD sharing the same .git
git reflog
to find the bugfix I've just made.git cherry-pick SHA1
of the commit.Optionally (still asynchronous) you can then rebase (or merge) your feature branch to get the bugfix, usually when you are about to submit a PR and have cleaned your feature branch and WD already:
cd my-project
which is the main WD I'm working on.git rebase master
to get the bugfixes.This way I can keep working on the feature uninterrupted and not have to worry about git stash
-ing anything or having to clean my WD before a git checkout
(and then having the check the feature branch backout again.) and still have all my bugfixes goes to master
instead of hidden in my feature branch.
IMO git stash
and git checkout
is a real PIA when you are in the middle of working on some big feature.
Upvotes: 13
Reputation: 793007
Stashing, temporary commits and rebasing may all be overkill. If you haven't added the changed files to the index, yet, then you may be able to just checkout the other branch.
git checkout branch2
This will work so long as no files that you are editing are different between branch1 and branch2. It will leave you on branch2 with you working changes preserved. If they are different then you can specify that you want to merge your local changes with the changes introduced by switching branches with the -m
option to checkout.
git checkout -m branch2
If you've added changes to the index then you'll want to undo these changes with a reset first. (This will preserve your working copy, it will just remove the staged changes.)
git reset
Upvotes: 89
Reputation: 9263
If it were about committed changes, you should have a look at git-rebase, but as pointed out in comment by VonC, as you're talking about local changes, git-stash would certainly be the good way to do this.
Upvotes: 2