Prepend Git commit message with partial branch name

Question

My current branch naming convention is as such:

ticket-45-my-new-feature-branch-description

I am currently using this code in my .git/hooks/prepare-commit-msg file to prepend every commit message with the branch name as such:

BRANCH_NAME=$(git branch 2>/dev/null | grep -e ^* | tr -d ' *')
if [ -n "$BRANCH_NAME" ] && [ "$BRANCH_NAME" != "master" ]; then
    echo "[$BRANCH_NAME] $(cat $1)" > $1
fi

End result:

[ticket-45-my-new-feature-branch-description] test commit

What I'm trying to accomplish is output like this:

[ticket-45] test commit

Brownie points if we can capitalize it:

[TICKET-45] test commit

I would love to keep my descriptive branch names, but truncate the prepended text in commit messages. I'm sure I have to use some regex, but I really don't know how to accomplish that. I should mention that we have several projects going on at once, so branch names are different, like so:

ticket-123-branch-name
abc-22-my-branch
ua-11-my-feature

The only thing in common is the fact that I need everything before the second '-'.

Any help is greatly appreciated!!!

Answer 1

Ok, first, this:

BRANCH_NAME=$(git branch 2>/dev/null | grep -e ^* | tr -d ' *')

is major overkill :-) Use:

branch=$(git symbolic-ref --short HEAD) || ...

to get the current branch name. The part after || is "what to do if you're not on a branch" (i.e., if you're in "detached head" mode)—you'll have to decide that for yourself. (Your current code sets BRANCH_NAME to the empty string; to do that, you don't even need the || part, but you may want to add -q, or a 2>/dev/null, to avoid the "fatal:" message from symbolic-ref.)

The rest is just basic scripting. In bash you can use regex's directly, in old sh you can invoke expr or sed. Both sed and tr can upper-case-ify, but sed can do a regex too, so it looks like a good candidate:

$ trimmed=$(echo $branch | sed -e 's:^\([^-]*-[^-]*\)-.*:\1:' -e \
    'y/abcdefghijklmnopqrstuvwxyz/ABCDEFGHIJKLMNOPQRSTUVWXYZ/')
$ echo $trimmed
TICKET-45

Last, it's slightly dangerous to do:

echo "some stuff $(cat $1)" > $1

as you're depending on the shell to expand the $(cat $1) before it truncates the output file for the > $1 part. (Obviously it works, but you're subject to shell vagaries.) Better to use a temporary file, or maybe another sed but in-place:

sed -i .bak -e "1s:^:[$trimmed] :" $1
# or use -i '', but note minor warning in sed man pages

The above is only tested piecemeal, but should work.