ChatGPT解决这个技术问题 Extra ChatGPT

Hard reset of a single file

git

I currently have three modified files in my working directory. However I want one of them to be reset to the HEAD status.

In SVN, I'd use svn revert <filename> (followed by svn update <filename> if needed) but in Git I should use git reset --hard. However this command cannot operate on a single file.

Is there any way in Git to discard the changes to a single file and overwrite it with a fresh HEAD copy?

git checkout below is the answer. In git, "revert" is something you do to a commit. "Revert" replays the inverse of a historical commit into your working directory, so you can make a new commit that "undoes" the reverted commit. I find this is a frequent point of confusion for people coming to git from svn.
If you are interested why you cannot do hard reset with paths, check out my answer there.
This question assumes, that one knows what a Hard reset is.
See also stashing one file: stackoverflow.com/questions/3040833/…

C
Community

You can use the following command:

git checkout HEAD -- my-file.txt

... which will update both the working copy of my-file.txt and its state in the index with that from HEAD.

-- basically means: treat every argument after this point as a file name. More details in this answer. Thanks to VonC for pointing this out.


More complete answer. +1 ;) For the '--', see also stackoverflow.com/questions/6561142/… (and, more generally, stackoverflow.com/questions/1192180/…)
Also, don't forget you can reference a previous commit with HEAD~1 to indicate the penultimate commit.
You can leave out HEAD if you are at the head of the current branch - see norbauer.com/rails-consulting/notes/…
Any insight why the reset command (as it says) "cannot do hard reset with paths", and then why the checkout command is not (cannot be?) used for hard-resetting the whole set? (I mean why it has been designed so.)
@cxw Unfortunately, this is not quite true. From the man page of git checkout: "Overwrite paths in the working tree by replacing with the contents in the index or in the <tree-ish>". I.e. if <tree-ish> is omitted, whatever content in the index will be used to update the working tree. This may or may not differ from HEAD.
A
AxeEffect

Since Git 2.23 (August 2019) you can use restore (more info):

git restore pathTo/MyFile

The above will restore MyFile on HEAD (the last commit) on the current branch.

If you want to get the changes from other commit you can go backwards on the commit history. The below command will get MyFile two commits previous to the last one. You need now the -s (--source) option since now you use master~2 and not master (the default) as you restore source:

git restore -s master~2 pathTo/MyFile

You can also get the file from other branch!

git restore -s my-feature-branch pathTo/MyFile

The easiest way so far. Unfortunately this answer is not getting enough attention.
It should be noted that the git documentation still marks this command as experimental.
git: 'restore' is not a git command. git version 2.8.2.windows.1
@ButtleButkus 2.8 was released 2016-03-28. This answer already mentions it requires 2.23+.
@refaelio Overwriting local changes is the entire point. Is there some other operation that this command might be confused with?
A
Asclepius

Reset to head:

To hard reset a single file to HEAD:

git checkout @ -- myfile.ext

Note that @ is short for HEAD. An older version of git may not support the short form.

Reset to index:

To hard reset a single file to the index, assuming the index is non-empty, otherwise to HEAD:

git checkout -- myfile.ext

The point is that to be safe, you don't want to leave out @ or HEAD from the command unless you specifically mean to reset to the index only.


What's up with the "--" before myfile.ext?
@LanceKind As I understand, that's used to demarcate a list of filenames which follows it. Without it, there are cases when git interprets the arguments incorrectly.
Not just filenames. The widely used convention separates options from positional arguments in many utilities. See man bash page. Also mentioned in this answer: unix.stackexchange.com/a/187548/142855
Conventionally, the -- is used to tell the program I've finished specifying "options", and from here on, everything will be a positional argument.. Conventionally, "options" are the tokens like --recursive which can appear in any order, or even be combined together in their short form, like with rm -rf. On the contrary, "positional arguments" are much more like arguments passed to a function in a programming language: their position in the list of tokens defines what exactly the program is going to do with them (these are often filenames). -- removes ambiguity as to which is which.
This is not "hard" resetting anything it just applies changes from some commit and requires additional commit. On branch level hard reset, however, your changes are simply reverted to the commit you wanted to reset to without any additional commit
S
Soviut

To revert to upstream/master do:

git checkout upstream/master -- myfile.txt

W
Worthwelle

Reference to HEAD is not necessary.

git checkout -- file.js is sufficient


M
Milad Rahimi

You can use the following command:

git checkout filename

If you have a branch with the same file name you have to use this command:

git checkout -- filename

This will not "hard reset" the file - it only copies the index state to the working tree. A "hard reset" would first reset the index.
H
HariKishore K

you can use the below command for reset of single file

git checkout HEAD -- path_to_file/file_name

List all changed files to get path_to_file/filename with below command

git status

A
ADDYQU

You can use the following command:

git reset -- my-file.txt

which will update both the working copy of my-file.txt when added.


Does not change the content of the modified file, as requested.
That's not specifically the point @ADDQU. The question is how to "hard reset" a file, not to remove it from the staged list.
@Rafael You are correct, but I want to let you know that there is a way too.
k
kmiklas

A simple, easy, hands-on, way to get you out of hot water, especially if you're not so comfortable with git:

View the log of your file git log myFile.js commit 1023057173029091u23f01w276931f7f42595f84f Author: kmiklas Date: Tue Aug 7 09:29:34 2018 -0400 JIRA-12345 - Refactor with new architecture. Note hash of file: 1023057173029091u23f01w276931f7f42595f84f Show the file using the hash. Make sure it's what you want: git show 1023057173029091u23f01w276931f7f42595f84f:./myFile.js Redirect file to a local copy git show 1023057173029091u23f01w276931f7f42595f84f:./myFile.js > myFile.07aug2018.js Back up your current file. cp myFile.js myFile.bak.js Open both files in your favorite text editor. vim myFile.js vim myFile.07aug2018.js Copy n' paste code from myFile.07aug2018.js to myFile.js, and save. Commit and push myFile.js Again view the log, and confirm that your file is properly in place. Tell your clients to pull the latest, happily watch it work with the old version in place.

Not the sexiest, or most git-centric solution, and definitely a "manual" reset/reversion, but it works. It requires minimal knowledge of git, and doesn't disturb the commit history.


This answer is far more complex and error prone than any of the solutions which predate it by years.
Why should anyone use this solution!? the true answer is only a simple command.
This is the optimal solution for certain circumstances. It has merit for those in a bind, where git is malfunctioning, and for those who need a work-around.
@kmiklas out of curiosity, what certain circumstances is this ideal?
"It requires minimal knowledge of git" - this is a (huge) downside, not an upside