I am using Git and working on master branch. This branch has a file called app.js.
I have an experiment branch in which I made a bunch of changes and tons of commits. Now I want to bring all the changes done only to app.js from experiment to master branch.
How do I do that?
Once again I do not want a merge. I just want to bring all the changes in app.js from experiment branch to master branch.
git checkout master # first get back to master
git checkout experiment -- app.js # then copy the version of app.js
# from branch "experiment"
See also git how to undo changes of one file?
With the new git switch and git restore commands, that would be:
git switch master
git restore --source experiment -- app.js
By default, only the working tree is restored.
If you want to update the index as well (meaning restore the file content, and add it to the index in one command):
git restore --source experiment --staged --worktree -- app.js
# shorter:
git restore -s experiment -SW -- app.js
As Jakub Narębski mentions in the comments:
git show experiment:path/to/app.js > path/to/app.js
works too, except that, as detailed in the SO question "How to retrieve a single file from specific revision in Git?", you need to use the full path from the root directory of the repo.
Hence the path/to/app.js used by Jakub in his example.
As Frosty mentions in the comment:
you will only get the most recent state of app.js
But, for git checkout or git show, you can actually reference any revision you want, as illustrated in the SO question "git checkout revision of a file in git gui":
$ git show $REVISION:$FILENAME
$ git checkout $REVISION -- $FILENAME
would be the same is $FILENAME is a full path of a versioned file.
$REVISION can be as shown in git rev-parse:
experiment@{yesterday}:app.js # app.js as it was yesterday
experiment^:app.js # app.js on the first commit parent
experiment@{2}:app.js # app.js two commits ago
and so on.
schmijos adds in the comments:
you also can do this from a stash:
git checkout stash -- app.jsThis is very useful if you're working on two branches and don't want to commit.
Everything is much simpler, use git checkout for that.
Suppose you're on master branch, to get app.js from new-feature branch do:
git checkout new-feature path/to/app.js
// note that there is no leading slash in the path!
This will bring you the contents of the desired file. You can, as always, use part of sha1 instead of new-feature branch name to get the file as it was in that particular commit.
Note:new-feature needs to be a local branch, not a remote one.
git# check out all files in <paths> from branch <branch_name>
git checkout <branch_name> -- <paths>
Source: http://nicolasgallagher.com/git-checkout-specific-files-from-another-branch/.
See also man git checkout.
Examples:
# Check out "somefile.c" from branch `my_branch`
git checkout my_branch -- somefile.c
# Check out these 4 files from `my_branch`
git checkout my_branch -- file1.h file1.cpp mydir/file2.h mydir/file2.cpp
# Check out ALL files from my_branch which are in
# directory "path/to/dir"
git checkout my_branch -- path/to/dir
If you don't specify, the branch_name it is automatically assumed to be HEAD, which is your most-recent commit of the currently-checked-out branch. So, you can also just do this to check out "somefile.c" and have it overwrite any local, uncommitted changes:
# Check out "somefile.c" from `HEAD`, to overwrite any local, uncommitted
# changes
git checkout -- somefile.c
# Or check out a whole folder from `HEAD`:
git checkout -- some_directory
# General form
git show my_branch_or_commit_hash:my_file.cpp > any/path/my_file.cpp
# Example: check out `main.cpp` from 3 commits ago in your currently-checked-out
# branch (3 commits prior to `HEAD`, or `HEAD~3`) into a temporary directory
mkdir ../temp
git show HEAD~3:main.cpp > ../temp/main_old.cpp
Source where I learned this: @Jakub Narębski's answer to: git-checkout older revision of a file under a new name
git merge, git cherry-pick, git rebase, or git revert changes?Well, in that case, you better do this:
# Keep `--theirs` for all conflicts within this file
git checkout --theirs -- path/to/some/file
# OR: keep `--ours` for all conflicts within this file
git checkout --ours -- path/to/some/file
OR:
# Keep `--theirs` for all conflicts within files inside this dir
git checkout --theirs -- path/to/some/dir
# OR: keep `--ours` for all conflicts within files inside this dir
git checkout --ours -- path/to/some/dir
Do NOT do the regular checkout form in the previous section before this, unless that's what you really want to do. See the "WARNING WARNING WARNING" section in my answer here: Who is "us" and who is "them" according to Git?.
DEALING WITH path does not have our version or path does not have their version ERRORS:
If you ever see errors like this:
error: path 'path/to/some/dir/file1.cpp' does not have our version # OR error: path 'path/to/some/dir/file1.cpp' does not have their version
...when running the commands above, then you simply need to git rm those files first and then try the git checkout --ours or git checkout --theirs command again. See my answer here for a detailed explanation of these commands, including a form to automatically find and delete those errored files for you: git checkout --ours when file spec includes deleted file.
In this case, git checkout my_branch -- some_file_or_dir is NOT enough, because if you have files in the specified directory which exist in your currently-checked-out branch or commit but do NOT exist in my_branch, then you'd like them to be deleted locally, but git checkout does NOT delete any files which exist locally but not on the specified commit, rather, it only overwrites files locally with their versions from the specified commit. So, to also delete files locally which should not be there, so that what you end up with locally is an exact copy of what you have on my_branch, you must do the following:
git reset my_branch -- path/to/some/file_or_dir
git checkout-index -fa
git clean -fd
git commit -m "hard reset path/to/some/file_or_dir to its state \
as it was at my_branch"
See my own answer here for more details on this: Why git can't do hard/soft resets by path?.
git checkout in my answer here: Who is "us" and who is "them" according to Git?.This is a pragmatic approach which doesn't directly answer the OP, but some have found useful:
If the branch in question is on GitHub, then you can navigate to the desired branch and file using any of the many tools that GitHub offers, then click 'Raw' to view the plain text, and (optionally) copy and paste the text as desired.
I like this approach because it lets you look at the remote file in its entirety before pulling it to your local machine.
Another way is to create a patch with the differences and apply it in the master branch For instance. Let's say the last commit before you started working on app.js is 00000aaaaa and the commit containg the version you want is 00000bbbbb
The you run this on the experiment branch:
git diff 00000aaaaa 00000bbbbb app.js > ~/app_changes.git
This will create a file with all the differences between those two commits for app.js that you can apply wherever you want. You can keep that file anywhere outside the project
Then, in master you just run:
git apply ~/app_changes.git
now you are gonna see the changes in the projects as if you had made them manually.
mastermay diverge fromexperiment, e.g. can contain changes merged from other branches which would be lost in file is just copied. P.S. on the other hand he does not want to merge, but as far as I knowgit mergeterm is for branch only, not for specific file, so no contradiction here. – Alexei Martianov