When certain commands are run, Git searches the
.git/hooks directory for suitable hook scripts which are then executed if found. You’ll find a small set of example scripts there (you can activate them by renaming them to remove the
.sample prefix and setting their executable bit), and a complete list of hooks can be found in the githooks(5) man page.
This article suggests a handful of hooks which can streamline development and help improve your efficiency.
There’s really no excuse, yet I’ll admit even I’ve done this once or twice before: commit syntactically broken code. But what if a lint check could be performed automatically as part of the commit process to safeguard from this? And if you run a lint check manually before each commit, automating it can protect against the occasional moment of forgetfulness.
The following shell code can be saved as
.git/hooks/pre-commit (or appended if pre-commit hook code already exists) to trigger an automatic check whenever a commit is made:
#!/bin/bash git diff --cached --name-status --diff-filter=ACMR | while read STATUS FILE; do if [[ "$FILE" =~ ^.+(php|inc)$ ]]; then php -l "$FILE" 1> /dev/null if [ $? -ne 0 ]; then echo "Aborting commit due to files with syntax errors" >&2 exit 1 fi fi done
git diff reports on what’s changed between commits, and the options above return only the files that have been added (A), copied (C), modified (M), or renamed (R) in the staged commit. The files with an extension of
.inc are targeted for a lint check, and a failed check will exit the script with a non-zero return code and thus abort the commit itself. Your code may be buggy, but at least you know it’s got all of its semicolons.
Spell-Check Commit Messages
Developers are supposed to be meticulous and have an eye for detail (at least that’s what we all put on our CV when looking for a job), so a commit message with bad spelling can be embarrassing.
Commit messages give the reader a broad idea of what changes were made in a commit and thus have the potential to be read by a large audience, from other developers to compliance auditors. Even if you code only for yourself, writing professional messages is a good habit to develop, and automating a “second set of eyes” can help reduce speling erors in your comit messeges.
The following code can be saved as
.git/hooks/post-commit (or appended); it calls Aspell and outputs a list of suspect words. If there are errors, you can immediately fix the commit message by running
git commit --amend.
#!/bin/bash ASPELL=$(which aspell) if [ $? -ne 0 ]; then echo "Aspell not installed - unable to check spelling" >&2 exit fi AWK=$(which awk) if [ $? -ne 0 ]; then echo "Awk not installed - unable to filter spelling errors" >&2 WORDS=$($ASPELL list < "$1") else WORDS=$($ASPELL list < "$1" | $AWK '!_[$0]++') fi if [ -n "$WORDS" ]; then echo "Possible spelling errors found in commit message: " $WORDS >&2 fi
You can also compile a supplemental dictionary using identifiers extracted from your project’s source code (perhaps triggered with a post-checkout hook) and pass it to Aspell with
--extra-dicts to reduce the number of false positives. Parsing code and managing dictionary files is beyond the scope of this article though, so I’ll leave it as an exercise for you to explore later.
I don’t know of any scientific research that definitively proves formatting standards reduce cognitive friction when reading code. Most of the standards irritatingly target only low-hanging fruit anyway: capitalize something this way, place your braces here and not there, etc. Standards that enforce good architectural design and address specific interoperability concerns have more merit in my opinion.
But if you’re working with people who think 81 characters on a line is the eighth deadly sin, it might help smooth political friction by making sure your code conforms to an adopted formatting standard. The following code can be used as a post-commit hook (
.git/hooks/post-commit) to automatically check for formatting violations.
#!/bin/bash DIR=$(git rev-parse --show-toplevel) EMAIL=$(git config user.email) TMPFILE=$(mktemp) git diff --cached --name-status --diff-filter=ACMR | while read STATUS FILE; do if [[ "$FILE" =~ ^.+(php|inc)$ ]]; then phpcs --standard=zend "$FILE" >> $TMPFILE fi done if [ -s $TMPFILE ]; then mailx -s "PHPCS: Coding Standards Violations" $EMAIL < $TMPFILE fi rm $TMPFILE
The script uses PHP CodeSniffer to scan each PHP file in the commit for violations and then sends an email listing the violations that were found. If you’re using one of those buzzword productivity approaches like Getting Things Done to manage your email, you can set up an email filter that routes the received emails directly to your TODO folder to make sure you fix the violations in a timely manner.
Automatically Run Composer
So far we’ve seen hooks that focus on commits and the development process, but there are also some useful hooks that can be executed at other points in a Git workflow, even deployment. Suppose you have a remote repository set up on a production server, and pushing to it is your approach to deployment.
Further suppose that you use Composer to manage your application’s dependencies and you only keep your own code in your repository. This means you still have additional work to do after a push, either copying the dependencies over manually or ssh-ing into the server to run
To make your life easier, putting this in the remote repository’s
.git/hooks/post-receive file for a post-receive hook will run Composer automatically.
#!/bin/bash DIR=$(git rev-parse --show-toplevel) if [ -e "$DIR/composer.json" ]; then if [ -d "$DIR/vendor" ]; then composer.phar install else composer.phar update fi fi
In this article I shared a handful of hooks that can hopefully streamline how you develop your applications and make you more efficient. Feel free to share whether you are taking advantage of this powerful feature or not, and maybe even some of your own hooks in the comments below.
Here are a few additional resources for hints, tips, and inspiration:
- Git – Customizing Git: Git Hooks
- Mark’s Blog – Missing Git Hooks Documentation
- Meta Magic – Managing Project, User, and Global Git Hooks
- Ben Kulbertis – Synchronizing a MySQL Database with Git and Git Hooks
Timothy Boronczyk is a native of Syracuse, New York, where he lives with no wife and no cats. He has a degree in Software Application Programming, is a Zend Certified Engineer, and a Certified Scrum Master. By day, Timothy works as a developer at ShoreGroup, Inc. By night, he freelances as a writer and editor. Timothy enjoys spending what little spare time he has left visiting friends, dabbling with Esperanto, and sleeping with his feet off the end of his bed.