Collectives™ on Stack Overflow

Find centralized, trusted content and collaborate around the technologies you use most.

Teams

Q&A for work

Connect and share knowledge within a single location that is structured and easy to search.

To get the shortened commit hash, use the %h format specifier:

git log --pretty=format:'%h' -n 1
%H represents the long commit hash. Also, -1 can be used directly in place of -n 1.
                This is a bad/ incorrect way of doing it because this method will give you the wrong hash if you have a detached head. For example if the current commit is 12ab34... and the previous commit was 33aa44... then if i do 'git checkout 33aa44'  and then I run your command I will still be getting back 12ab34... despite my head actually pointing to 33aa44...
– theQuestionMan
                Jul 17, 2017 at 0:03
                @theQuestionMan I don't experience the behavior you describe; git checkout 33aa44; git log -n 1 gives me 33aa44. What version of git are you using?
– outofculture
                Jul 19, 2017 at 17:32
                @AmedeeVanGasse, ah! I HAD NO IDEA this is a toilet analogy! I've been seeing porcelain in the git man pages for years, but had NO idea it was referring to a toilet! The porcelain is the toilet, and it's "closer to the user" (who figuratively sits on this toilet) than the plumbing, which is lower-level and farther from the user--ie: below the "porcelain"! Mind blown.
– Gabriel Staples
                Feb 21, 2021 at 7:26
                This is the correct answer, since it works even if you checkout a specific commit instead of HEAD.
– Parsa
                Feb 22, 2019 at 19:16
                @Parsa: when checking out a specific commit HEAD points to this commit rather than a named branche know as detached head.
– ChristofSenn
                Jan 28, 2020 at 20:46
                @Parsa Your (mistakenly upvoted?) comment is misleading as the accepted answer of jakub-narębski 'git rev-parse HEAD' works after checking out somewhere else.
– spawn
                May 12, 2022 at 14:54
                If two git commit hashes are needed, such as one from the branch you are currently working with and a master branch, you could also use git rev-parse FETCH_HEAD if you need the hash for the master commit that you merged into your current branch.  e.g. if you have branches master and feature/new-feature for a given repo., while on feature/new-feature you could use git fetch origin master && git merge FETCH_HEAD and then git rev-parse --short FETCH_HEAD if you needed the commit hash from the master you just merged in for any scripts you may have.
– Mike
                Sep 3, 2018 at 21:43
git show -s --format=%h
The -s flag is same as --no-patch and stands for "Suppress diff output".
Click here for more git show examples.
For completeness, since no one has suggested it yet.  .git/refs/heads/master is a file that contains only one line: the hash of the latest commit on master.  So you could just read it from there.
Or, as a command:
cat .git/refs/heads/master
Update:
Note that git now supports storing some head refs in the pack-ref file instead of as a file in the /refs/heads/ folder.  https://www.kernel.org/pub/software/scm/git/docs/git-pack-refs.html
                .git/HEAD typically points to a ref, if you have a SHA1 in there, you are in detached head mode.
– eckes
                Apr 9, 2013 at 1:48
                This isn't very robust compared to other approaches, in particular because it assumes that there is a .git subdirectory, which is not necessarily the case. See the --separate-git-dir flag in the git init man page.
– jub0bs
                Dec 29, 2014 at 17:44
There's always git describe as well. By default it gives you --
john@eleanor:/dev/shm/mpd/ncmpc/pkg (master)$ git describe --always
release-0.19-11-g7a68a75
                Git describe returns the first TAG reachable from a commit. How does this help me get the SHA?
– Sardaukar
                Sep 9, 2011 at 13:45
                I like git describe --long --dirty --abbrev=10 --tags it will give me something like 7.2.0.Final-447-g65bf4ef2d4 which is 447 commits after the 7.2.0.Final tag and the first 10 digest of the global SHA-1 at the current HEAD are "65bf4ef2d4". This is very good for version strings. With --long it will always add the count (-0-) and the hash, even if the tag happens to match exactly.
– eckes
                Apr 9, 2013 at 1:46
                If no tags exist then git describe --always will "show uniquely abbreviated commit object as fallback"
– Ronny Andersson
                Sep 18, 2014 at 16:57
                I use git describe --tags --first-parent --abbrev=11 --long --dirty --always. The --always option means it provides a result (hash) even if there are no tags. The --first-parent means it doesn't get confused by merge commits and only follows items on the current branch. Note also that --dirty will append -dirty to the result if the current branch has uncommitted changes.
– ingyhere
                Jan 31, 2020 at 7:05
                git-rev-list is about generating list of commit objects; it is git-rev-parse to translate object name (e.g. HEAD) into SHA-1
– Jakub Narębski
                Jun 4, 2009 at 14:13
If you need to store the hash in a variable during a script, you can use
last_commit=$(git rev-parse HEAD);
Or, if you only want the first 10 characters (like github.com does)
last_commit=$(git rev-parse --short=10 HEAD);
                Thanks. I was able to use this in a build script: now=$(date -u "+%Y-%m-%d %H:%M:%S") && last_commit=$(git rev-parse HEAD) && echo "{\"commit\": \"$last_commit\", \"build_time\": \"$now\"}" > version.json
– Ryan
                Nov 29, 2022 at 18:42
Basically, git stores the location of HEAD in .git/HEAD, in the form ref: {path from .git}.  This command reads that out, slices off the "ref: ", and reads out whatever file it pointed to.
This, of course, will fail in detached-head mode, as HEAD won't be "ref:...", but the hash itself - but you know, I don't think you expect that much smarts in your bash one-liners.  If you don't think semicolons are cheating, though...
HASH="ref: HEAD"; while [[ $HASH == ref\:* ]]; do HASH="$(cat ".git/$(echo $HASH | cut -d \  -f 2)")"; done; echo $HASH
                I formalized this to a script for my local machine.  Then, I thought, hey: the implementation I made are simple enough that it illustrates how to solve an unrelated problem (parsing arguments in raw POSIX shell scripts without external programs), but complex enough to provide a little variation and to exploit most of the features of sh.  Half an hour of documentation comments later, and here's a Gist of it: gist.github.com/Fordi/29b8d6d1ef1662b306bfc2bd99151b07
– Fordi
                Jun 29, 2016 at 14:41
                Looking at it, I made a more extensive version for detecting Git and SVN, and grabbing the git hash/svn revision.  Not a clean string this time, but easily command-line parsed, and usable as a version tag:  gist.github.com/Fordi/8f1828efd820181f24302b292670b14e
– Fordi
                Jun 29, 2016 at 15:21
I needed something a little more different: display the full sha1 of the commit, but append an asterisk to the end if the working directory is not clean. Unless I wanted to use multiple commands, none of the options in the previous answers work.
Here is the one liner that does:

git describe --always --abbrev=0 --match "NOT A TAG" --dirty="*"

Result: f5366ccb21588c0d7a5f7d9fa1d3f85e9f9d1ffe*
Explanation: describes (using annotated tags) the current commit, but only with tags containing "NOT A TAG". Since tags cannot have spaces, this never matches a tag and since we want to show a result --always, the command falls back displaying the full (--abbrev=0) sha1 of the commit and it appends an asterisk if the working directory is --dirty.
If you don't want to append the asterisk, this works like all the other commands in the previous answers:

git describe --always --abbrev=0 --match "NOT A TAG"

Result: f5366ccb21588c0d7a5f7d9fa1d3f85e9f9d1ffe
                It works for me without the --match "NOT A TAG". Tested in git 2.18.0 as well as 2.7.4. Is there any situation where this argument is needed?
– Thomas
                Aug 7, 2018 at 7:24
                @Thomas it won't work if you have an annotated tag anywhere in the history of the current commit. The fake tag makes sure that the describe command does not use a tag to describe the commit,
– Rado
                Aug 8, 2018 at 5:44
git rev-parse HEAD does the trick.
If you need to store it to checkout back later than saving actual branch if any may be preferable:
cat .git/HEAD
Example output:
ref: refs/heads/master
Parse it:
cat .git/HEAD | sed "s/^.\+ \(.\+\)$/\1/g"
If you have Windows then you may consider using wsl.exe:
wsl cat .git/HEAD | wsl sed "s/^.\+ \(.\+\)$/\1/g"
Output:
refs/heads/master
This value may be used to git checkout later but it becomes pointing to its SHA. To make it to point to the actual current branch by its name do:
wsl cat .git/HEAD | wsl sed "s/^.\+ \(.\+\)$/\1/g" | wsl sed "s/^refs\///g"  | wsl sed "s/^heads\///g"
Output:
master
                While this technically works, git show is what's known as a porcelain command (i.e. user-facing), and so should not be used in scripts because its output is subject to change. The answer above (git rev-parse --short HEAD) should be used instead.
– jm3
                Mar 15, 2014 at 23:45
                @jm3 that's backwards. "Porcelain" commands have stable outputs that are intended for scripts. Search git help show for porcelain.
– John Tyree
                Jul 6, 2015 at 21:32
                @JohnTyree This is a confusing subject, but jm3 was right: porcelain commands are not meant to be parsed, but rather to be human-readable. In case you need to use a porcelain command in a script and you want to have a stable format, there's sometimes (for example with git status, push and blame) an option that does just that. Unfortunately, that option is called --porcelain, which is why this is confusing. You can find the details in this great answer by VonC
– Fabio says Reinstate Monica
                Jan 14, 2019 at 17:20
Perhaps you want an alias so you don't have to remember all the nifty details. After doing one of the below steps, you will be able to simply type:
$ git lastcommit
49c03fc679ab11534e1b4b35687b1225c365c630
Following up on the accepted answer, here are two ways to set this up:
1) Teach git the explicit way by editing the global config (my original answer):
 # open the git config editor
 $ git config --global --edit
 # in the alias section, add
 [alias]
   lastcommit = rev-parse HEAD
2) Or if you like a shortcut to teach git a shortcut, as recently commented by Adrien:
$ git config --global alias.lastcommit "rev-parse HEAD"
From here on, use git lastcommit to show the last commit's hash.
Here is one-liner in Bash shell using direct read from git files:
(head=($(<.git/HEAD)); cat .git/${head[1]})
You need to run above command in your git root folder.
This method can be useful when you've repository files, but git command has been not installed.
If won't work, check in .git/refs/heads folder what kind of heads do you have present.
If you're going for speed though, the approach mentioned by Deestan
cat .git/refs/heads/<branch-name>
is significantly faster than any other method listed here so far.
                show-ref seems to me to be the best option for scripting, since it's a plumbing command and thus guaranteed (or at least very likely) to remain stable in future releases: other answers use rev-parse, show, describe, or log, which are all porcelain commands. And in cases where speed is not of the essence, the note from the show-ref manpage applies: ‘Use of this utility is encouraged in favor of directly accessing files under the .git directory.’
– Pont
                Oct 8, 2018 at 18:21
you will see, these lines following your command.
commit d25c95d88a5e8b7e15ba6c925a1631a5357095db .. (info about your head)
d25c95d88a5e8b7e15ba6c925a1631a5357095db, is your SHA for last commit.
Pretty print of main git repo, and sub-modules:
echo "Main GIT repo:"
echo $(git show -s --format=%H) '(main)'
echo "Sub-modules:"
git submodule status | awk '{print $1,$2}'
Example output:
3a032b0992d7786b00a8822bbcbf192326160cf9 (main)
7de695d58f427c0887b094271ba1ae77a439084f sub-module-1
58f427c0887b01ba1ae77a439084947de695d27f sub-module-2
d58f427c0887de6957b09439084f4271ba1ae77a sub-module-3
How I would do it in python (based on @kenorb's bash answer)
def get_git_sha():
    # Which branch are we on?
    branch = open(".git/HEAD", "r").read()
    # Parse output "ref: refs/heads/my_branch" -> my_branch
    branch = branch.strip().split("/")[-1]
    # What's the latest commit in this branch?
    return open(f".git/refs/heads/{branch}").read().strip()
head="$(cat ".git/HEAD")"
while [ "$head" != "${head#ref: }" ]; do
  head="$(cat ".git/${head#ref: }")"
This also works over http which is useful for local package archives (I know: for public web sites it's not recommended to make the .git directory accessable):
head="$(curl -s "$baseurl/.git/HEAD")"
while [ "$head" != "${head#ref: }" ]; do
  head="$(curl -s "$baseurl/.git/${head#ref: }")"