ChatGPT解决这个技术问题 Extra ChatGPT

How do I discard unstaged changes in Git?

How do I discard changes in my working copy, that are not in the index?

git-clean only removes untracked files from the working tree git-scm.com/docs/git-clean
To clarify Asenar's comment above, git-clean -df can be dangerous. It will delete local untracked files (e.g. covered by a .gitignore) Read all below carefully and consider git checkout . instead
'git clean -df ' Be warned! I tried that and lost key folders that are unable to be restored... Ouch!
hitting git status gives a suggestion on how to do that! git checkout -- .
@Paulo: starting in July 2019, git status gives the suggestion: git restore. git restore is a new command exactly for this purpose. See my 2019 update.

N
Nagev

For all unstaged files in current working directory use:

git restore .

For a specific file use:

git restore path/to/file/to/revert

That together with git switch replaces the overloaded git checkout (see here), and thus removes the argument disambiguation.

If a file has both staged and unstaged changes, only the unstaged changes shown in git diff are reverted. Changes shown in git diff --staged stay intact.

Before Git 2.23

For all unstaged files in current working directory:

git checkout -- .

For a specific file:

git checkout -- path/to/file/to/revert

-- here to remove ambiguity (this is known as argument disambiguation).


This seems to be the git canonical way. i.e. exactly what git tells you to do if you type git status
Doesn't work if there are untracked files. Git says error: The following untracked working tree files would be overwritten by checkout: ....
newbie question, what does "git checkout -- ." mean semantically?
@Ninjack git checkout -- . means the same thing as git checkout ., except that you're explicit about the fact that you're not specifying the branch name. They both say checkout the HEAD version on the branch I am currently on for '.' or './'. If you do git checkout branch-name directory-or-file-name in general, you get the HEAD version of directory-or-file-name on branch branch-name.
IMO this variant is imperfect, as it doesn't handle situation when your changed repository is not on the HEAD revision at the moment of changes cleaning and you DO NOT want to update it to HEAD, and want to just clean the changes.
B
Brandon Minnick

Another quicker way is:

git stash save --keep-index --include-untracked

You don't need to include --include-untracked if you don't want to be thorough about it.

After that, you can drop that stash with a git stash drop command if you like.


@KarimSamir: The question specifically asks about changes that are not in the index. The git reset command will discard changes in the index too.
git checkout -- . is much faster
Neither the git stash, nor any variety of git checkout will discard unstaged deletes. According to the output of git status, the actual correct answer here is some flavor git reset HEAD
This pollutes the stash stack. git checkout -- . does the job with one command only.
@FelipeTonello I wish I could give you 100 rep for that comment. In my test, that seemed to do exactly what I wanted. Instead, I spent the last hour using "git checkout -- *" and working around all the untracked file errors.
L
Laurel

It seems like the complete solution is:

git clean -df
git checkout -- .

WARNING: while it won't delete ignored files mentioned directly in .gitignore, git clean -df may delete ignored files residing in folders.

git clean removes all untracked files and git checkout clears all unstaged changes.


The other two answers don't actually work, this one did.
@dval this is becues the first command removed the unindexed files and the second one removed the unstaged changes (of indexed files). So if you did not have any staged changes this it is the same as reverting to the last commit with git reset --hard
use -dff if the untracked directory is a git clone.
Be careful running git clean -df. If you don't understand what it does, you might be deleting files you mean to keep, like robots.txt, uploaded files, etc.
As @ctlockey said, the first command also delete directories if they are composed of ignored files only... Lost a whole bunch of configuration files on my project :( Be careful.
C
CB Bailey

This checks out the current index for the current directory, throwing away all changes in files from the current directory downwards.

git checkout .

or this which checks out all files from the index, overwriting working tree files.

git checkout-index -a -f

+1 This is the RIGHT ANSWER, as it correctly handles the case where some files have both staged and un-staged changes. Note that this solution DISCARDS the unstaged changes; if you wish to retain them, then you should use @greg-hewgill 's answer of git stash save --keep-index.
git checkout -- does not work if you have only one branch. git checkout . always works.
Thank you so much! Finally an answer that ALWAYS works! This may be combined with git clean to also remove untracked files.
Not being funny but this appears to be the only working answer. The higher rated answers appear to do nothing. I'm sure they do something, but they certainly don't "reset" the unstaged changes.
f
falsarella
git clean -df

Cleans the working tree by recursively removing files that are not under version control, starting from the current directory.

-d: Remove untracked directories in addition to untracked files

-f: Force (might be not necessary depending on clean.requireForce setting)

Run git help clean to see the manual


why this answer doesn't have all the votes? answered back in the 2011 and still correct.
This should be the accepted answer.
p
prosoitos

2019 update

You can now discard unstaged changes in one tracked file with:

git restore <file>

and in all tracked files in the current directory (recursively) with:

git restore .

If you run the latter from the root of the repository, it will discard unstaged changes in all tracked files in the project.

Notes

git restore was introduced in July 2019 and released in version 2.23 as part of a split of the git checkout command into git restore for files and git switch for branches.

git checkout still behaves as it used to and the older answers remain perfectly valid.

When running git status with unstaged changes in the working tree, this is now what Git suggests to use to discard them (instead of git checkout -- as it used to prior to v2.23).

As with git checkout -- ., this only discards changes in tracked files. So Mariusz Nowak's answer still applies and if you want to discard all unstaged changes, including untracked files, you could run, as he suggests, an additional git clean -df.


I wanted to revert my unstaged changes only without affecting newly added files so a git restore . worked perfectly. Thanks.
I did git restore <filename> and it worked perfectly.
Worked fine for me.
According to the man page git restore . restores all files in current directory, not in the whole repository.
You are right. Thanks! I just tested it and indeed, that is the case. It is, however, recursive. So when run from the root of the project, it applies to the whole repository. I will edit my answer.
B
Ben

My favorite is

git checkout -p

That lets you selectively revert chunks.

See also:

git add -p

I love the ability to see the actual change before it's discarded.
This is what I use. git checkout -p and then "a" to accept all.
I've never thought about. That -p adds a nice extra layer of safety. Combine it with git clean -d to actually answer OP.
E
ErikMD

Since no answer suggests the exact option combination that I use, here it is:

git clean -dxn .  # dry-run to inspect the list of files-to-be-removed
git clean -dxf .  # REMOVE ignored/untracked files (in the current directory)
git checkout -- . # ERASE changes in tracked files (in the current directory)

This is the online help text for the used git clean options:

-d

Remove untracked directories in addition to untracked files. If an untracked directory is managed by a different Git repository, it is not removed by default. Use -f option twice if you really want to remove such a directory.

-x

Don’t use the standard ignore rules read from .gitignore (per directory) and $GIT_DIR/info/exclude, but do still use the ignore rules given with -e options. This allows removing all untracked files, including build products. This can be used (possibly in conjunction with git reset) to create a pristine working directory to test a clean build.

-n

Don’t actually remove anything, just show what would be done.

-f

If the Git configuration variable clean.requireForce is not set to false, Git clean will refuse to delete files or directories unless given -f, -n, or -i. Git will refuse to delete directories within the .git subdirectory or file, unless a second -f is given.


+1 for this solution. Regarding your remark that "git checkout . needs to be done in the root of the repo", maybe you might mention we can just do git reset --hard instead? (which is actually equivalent to git reset --hard HEAD and should work whichever is the current directory...)
Also regarding the first command git clean -dfx, here is a tip I use to be on the safe side before running it: just run git clean -d -x -n before, to display the list of files-to-be-removed, then confirm the operation by running git clean -d -x -f (I put the argument -n, resp. -f in the end to be able to quickly change it in a terminal)
Quick note that this is unreversable, and if you have files in .gitignore you will lose them. So consider backing up your project before this.
@MartinG I just took the opportunity to incorporate my two suggestions, including the one that adds a "dry-run" step (as better safe than sorry!). Anyway, feel free to amend my edit if need be!
How do I delete untracked files that show up in git status without deleting untracked files that are in .gitignore?
2
2540625

If you merely wish to remove changes to existing files, use checkout (documented here).

git checkout -- .

No branch is specified, so it checks out the current branch.

The double-hyphen (--) tells Git that what follows should be taken as its second argument (path), that you skipped specification of a branch.

The period (.) indicates all paths.

If you want to remove files added since your last commit, use clean (documented here):

git clean -i 

The -i option initiates an interactive clean, to prevent mistaken deletions.

A handful of other options are available for a quicker execution; see the documentation.

If you wish to move changes to a holding space for later access, use stash (documented here):

git stash

All changes will be moved to Git's Stash, for possible later access.

A handful of options are available for more nuanced stashing; see the documentation.


This will exactly convert your changes and discard newly added files from previous commit.
will it remove the newly added files? or just undo the changes in the old unstaged files?
F
Forhadul Islam

The easiest way to do this is by using this command:

This command is used to discard changes in working directory -

git checkout -- .

https://git-scm.com/docs/git-checkout

In git command, stashing of untracked files is achieved by using:

git stash -u

http://git-scm.com/docs/git-stash


Twice I've come here, read this answer, and forgotten the . at the end. To future me: the period is essential!
I needed to get rid of all local changes in a sub directory, without blowing away every other change. This answer helped a lot, thanks
Please describe what the two commands do. It's really unhelpful to have no explanation.
excellent. the checkout does in one command what the most popular one does in two. can also be followed up with git clean -fd to clean files not in the index.
f
fontno

I really found this article helpful for explaining when to use what command: http://www.szakmeister.net/blog/2011/oct/12/reverting-changes-git/

There are a couple different cases:

If you haven't staged the file, then you use git checkout. Checkout "updates files in the working tree to match the version in the index". If the files have not been staged (aka added to the index)... this command will essentially revert the files to what your last commit was. git checkout -- foo.txt If you have staged the file, then use git reset. Reset changes the index to match a commit. git reset -- foo.txt

I suspect that using git stash is a popular choice since it's a little less dangerous. You can always go back to it if you accidently blow too much away when using git reset. Reset is recursive by default.

Take a look at the article above for further advice.


J
Joshua Kunzmann

If you aren't interested in keeping the unstaged changes (especially if the staged changes are new files), I found this handy:

git diff | git apply --reverse

D
Dorian

As you type git status, (use "git checkout -- ..." to discard changes in working directory) is shown.

e.g. git checkout -- .


Downvoted because it doesn't help to quickly discard all files. The three dots indicate that you are required to list all the files. This is especially bad if you need to discard tons of files at once, eg. during a large merge after you have staged all the modifications you like to keep
Of course, the correct command is "git checkout -- ." a single dot. In the comment, the three dots were a grammatical thing, to indicate there are many other options that could have been used..
A
Asped

You can use git stash - if something goes wrong, you can still revert from the stash. Similar to some other answer here, but this one also removes all unstaged files and also all unstaged deletes:

git add .
git stash

if you check that everything is OK, throw the stash away:

git stash drop

The answer from Bilal Maqsood with git clean also worked for me, but with the stash I have more control - if I do sth accidentally, I can still get my changes back

UPDATE

I think there is 1 more change (don't know why this worked for me before):

git add . -A instead of git add .

without the -A the removed files will not be staged


B
Bijan

git checkout -f

man git-checkout:

-f, --force

When switching branches, proceed even if the index or the working tree differs from HEAD. This is used to throw away local changes.

When checking out paths from the index, do not fail upon unmerged entries; instead, unmerged entries are ignored.


This would discard changes in the index!! (And the OP requires to leave them as is.)
N
Nick

Instead of discarding changes, I reset my remote to the origin. Note - this method is to completely restore your folder to that of the repo.

So I do this to make sure they don't sit there when I git reset (later - excludes gitignores on the Origin/branchname)

NOTE: If you want to keep files not yet tracked, but not in GITIGNORE you may wish to skip this step, as it will Wipe these untracked files not found on your remote repository (thanks @XtrmJosh).

git add --all

Then I

git fetch --all

Then I reset to origin

git reset --hard origin/branchname

That will put it back to square one. Just like RE-Cloning the branch, WHILE keeping all my gitignored files locally and in place.

Updated per user comment below: Variation to reset the to whatever current branch the user is on.

git reset --hard @{u}

This is my preferred option, but why do you add all changes first? So far as I'm aware this just modifies the directory listing in Git files, while using git reset --hard, this will be lost anyway while the directories will still be removed.
I dont on mac or linux, github windows powershell sometimes leaves the files there after reset. I think its because git reset sets all files in the repo to its original state. If theyre not added, theyre not touched. The desktop client then will pickup the "hey this file is in here and needs to be committed"
Sense made. I don't use Windows so haven't seen that issue (haven't used Windows for the last few months at least, don't remember much before that - it's one huge regrettable blur). Might be worth noting the rationale in your main answer :)
I ran across this issue on a Mac too now. If the file is not tracked in the Repo sometimes git reset doesnt touch it. I cant really isolate the "WHY" but when that happens, if I reset, and i still have 1 uncommitted file or two, i add --all and reset --hard again
A nice little variation of this I like is git reset --hard @{u} which resets the branch to wherever the current remote-tracking branch is
a
artur

Tried all the solutions above but still couldn't get rid of new, unstaged files.

Use git clean -f to remove those new files - with caution though! Note the force option.


s
siddhantsomani

To do a permanent discard: git reset --hard

To save changes for later: git stash


B
Ben Wilde

Just use:

git stash -u

Done. Easy.

If you really care about your stash stack then you can follow with git stash drop. But at that point you're better off using (from Mariusz Nowak):

git checkout -- .
git clean -df

Nonetheless, I like git stash -u the best because it "discards" all tracked and untracked changes in just one command. Yet git checkout -- . only discards tracked changes, and git clean -df only discards untracked changes... and typing both commands is far too much work :)


Note: git stash -u will soon (Git 2.14.x/2.15, Q3 2017) evolve a bit: stackoverflow.com/a/46027357/6309
If i get the question of the OP correct the indexed files should be kept. Only unstage changes should be removed. So is should be git stash -k in my opinion.
D
DrB

simply say

git stash

It will remove all your local changes. You also can use later by saying

git stash apply 

or git stash pop


K
Khem Raj Regmi

you have a very simple git command git checkout .


G
GlassGhost

This works even in directories that are; outside of normal git permissions.

sudo chmod -R 664 ./* && git checkout -- . && git clean -dfx

Happened to me recently


Beware though, that the git ignored content will not retain it's original permissions! Hence it can cause a security risk.
@twicejr You're wrong, please read git help clean "-d Remove untracked directories in addition to untracked files."
Why did you set all your files to be world read/write? Not good practice.
Setting all permissions to 664 makes a lot of assumptions about what kind of permissions the project needs. I think using that part of the command will cause issues for some people.
j
jdgregson

No matter what state your repo is in you can always reset to any previous commit:

git reset --hard <commit hash>

This will discard all changes which were made after that commit.


This will also discard everything in the index (not just things not in the index), which is beyond what the OP is asking for.
v
vivekporwal04
cd path_to_project_folder  # take you to your project folder/working directory 
git checkout .             # removes all unstaged changes in working directory

L
Lahiru Jayaratne

In my opinion,

git clean -df

should do the trick. As per Git documentation on git clean

git-clean - Remove untracked files from the working tree

Description

Cleans the working tree by recursively removing files that are not under version control, starting from the current directory. Normally, only files unknown to Git are removed, but if the -x option is specified, ignored files are also removed. This can, for example, be useful to remove all build products. If any optional ... arguments are given, only those paths are affected.

Options

-d Remove untracked directories in addition to untracked files. If an untracked directory is managed by a different Git repository, it is not removed by default. Use -f option twice if you really want to remove such a directory. -f --force If the Git configuration variable clean.requireForce is not set to false, git clean will refuse to run unless given -f, -n or -i.


t
tjb

Another way to get rid of new files that is more specific than git clean -df (it will allow you to get rid of some files not necessarily all), is to add the new files to the index first, then stash, then drop the stash.

This technique is useful when, for some reason, you can't easily delete all of the untracked files by some ordinary mechanism (like rm).


b
bbarker

What follows is really only a solution if you are working with a fork of a repository where you regularly synchronize (e.g. pull request) with another repo. Short answer: delete fork and refork, but read the warnings on github.

I had a similar problem, perhaps not identical, and I'm sad to say my solution is not ideal, but it is ultimately effective.

I would often have git status messages like this (involving at least 2/4 files):

$ git status
# Not currently on any branch.
# Changes to be committed:
#   (use "git reset HEAD <file>..." to unstage)
#
#       modified:   doc/PROJECT/MEDIUM/ATS-constraint/constraint_s2var.dats
#       modified:   doc/PROJECT/MEDIUM/ATS-constraint/parsing/parsing_s2var.dats
#
# Changes not staged for commit:
#   (use "git add <file>..." to update what will be committed)
#   (use "git checkout -- <file>..." to discard changes in working directory)
#
#       modified:   doc/PROJECT/MEDIUM/ATS-constraint/constraint_s2Var.dats
#       modified:   doc/PROJECT/MEDIUM/ATS-constraint/parsing/parsing_s2Var.dats

A keen eye will note that these files have dopplegangers that are a single letter in case off. Somehow, and I have no idea what led me down this path to start with (as I was not working with these files myself from the upstream repo), I had switched these files. Try the many solutions listed on this page (and other pages) did not seem to help.

I was able to fix the problem by deleting my forked repository and all local repositories, and reforking. This alone was not enough; upstream had to rename the files in question to new filenames. As long as you don't have any uncommited work, no wikis, and no issues that diverge from the upstream repository, you should be just fine. Upstream may not be very happy with you, to say the least. As for my problem, it is undoubtedly a user error as I'm not that proficient with git, but the fact that it is far from easy to fix points to an issue with git as well.


V
Vadim Kotov

I had a weird situation where a file is always unstaged, this helps me to resolve.

git rm .gitattributes git add -A git reset --hard


t
twicejr

When you want to transfer a stash to someone else:

# add files
git add .  
# diff all the changes to a file
git diff --staged > ~/mijn-fix.diff
# remove local changes 
git reset && git checkout .
# (later you can re-apply the diff:)
git apply ~/mijn-fix.diff

[edit] as commented, it ís possible to name stashes. Well, use this if you want to share your stash ;)


Actually Git stash can have a title. For instance git stash save "Feature X work in progress".
P
Pau

You could create your own alias which describes how to do it in a descriptive way.

I use the next alias to discard changes.

Discard changes in a (list of) file(s) in working tree

discard = checkout --

Then you can use it as next to discard all changes:

discard .

Or just a file:

discard filename

Otherwise, if you want to discard all changes and also the untracked files, I use a mix of checkout and clean:

Clean and discard changes and untracked files in working tree

cleanout = !git clean -df && git checkout -- .

So the use is simple as next:

cleanout

Now is available in the next Github repo which contains a lot of aliases:

https://github.com/GitAlias/gitalias