Git is a DAG of snapshots, with each node on the graph representing a commit. Each commit can have 'n' parent commits.
Given any two commits, is there a single, concise way to discern the "order" of these two in the DAG. git rev-list
seems to be the most promising, but I can't seem to find the right incantation.
Ideally, I'd have something like the following
$ git related hash1 hash2
hash1 is ancestor of hash2
OR
hash2 is ancestor of hash1
OR
hash1 unrelated to hash2
OR
hash1 is equal to hash2
Use git merge-base --is-ancestor <commit1> <commit2>
There is more than one way to find the answer to this. The simplest is to use
git merge-base --is-ancestor <commit> <commit>
From the documentation for git merge-base
:
--is-ancestor
Check if the first <commit>
is an ancestor of the second <commit>
, and exit with status 0 if true, or with status 1 if not. Errors are signaled by a non-zero status that is not 1.
Other options
git log with triple dot ...
notation
Another option is to use git log
and use triple dot notation ...
to tell Git to output the set union of the child commits, minus the set intersection. Basically, it tells you how a set of commits have diverged from each other:
$ git log --oneline --graph --left-right \
--first-parent --decorate <commit1>...<commit2>
The above command will show you commits that are reachable from commit1
or commit2
, but not both, i.e. C1 UNION C2 - C1 INTERSECTION C2, in terms of set operations.
If neither commit is a parent of the other, you'll see the child commits of both, but if one is an ancestor of the other, you'll only see the output for the descendant commit, since the ancestor is contained in the path of the descendant, and is thus excluded from the output.
You can read more about git log
and triple dot notation from the following resources:
- git-log(1).
- gitrevisions(1): Specifying Ranges.
- Revision Selection.
git branch --contains option
git-rev-list(1) seems like it could be used to answer this. Another way is to simply attach temporary branch labels to the commits you want to test, and then use the --contains
option for git branch
:
git branch --contains <commit-to-test>
The output will be all branches that contain the commit somewhere in their commit tree, so by using a temporary branch on the other commit, you can see if the commit you're testing is an ancestor.
From the documentation:
--contains [<commit>]
Only list branches which contain the specified commit (HEAD if not specified).
The following shell script might do the trick:
if git rev-list $SHA1 | grep -q $SHA2 ; then echo "$SHA2 is ancestor of $SHA1"
elif git rev-list $SHA2 | grep -q $SHA1 ; then echo "$SHA1 is ancestor of $SHA2"
else echo "$SHA1 unrelated to $SHA2" ; fi
Or, to neatly wrap it up into a git alias:
git config --global alias.related '!function git_related() { if git rev-list $1 | grep -q $2 ; then echo "$2 is ancestor of $1" ; elif git rev-list $2 | grep -q $1 ; then echo "$1 is ancestor of $2" ; else echo "$1 unrelated to $2" ; fi } ; git_related $1 $2'
if (( $(git rev-list $1..$2|wc -l) == 0 )); then echo "$2 is ancestor of $1"
elif (( $(git rev-list $2..$1|wc -l) == 0 )); then echo "$1 is ancestor of $2"
else echo "$1 and $2 are unrelated"
fi
git log --oneline -1 OLD_SHA..NEW_SHA
Iff this gives you some log, then OLD_SHA is parent of NEW_SHA.
To build on @helmbert's excellent git related
alias, here's a version which also accepts branch names (or HEAD etc) as arguments, rather than just commit ids:
git config --global alias.related '!function git_related() { commit1=`git log -n 1 --format="%h" $1` ; commit2=`git log -n 1 --format="%h" $2` ; if git rev-list $commit1 | grep -q $commit2 ; then echo "$2 is ancestor of $1" ; elif git rev-list $commit2 | grep -q $commit1 ; then echo "$1 is ancestor of $2" ; else echo "$1 unrelated to $2" ; fi } ; git_related $1 $2'