.githooker

tl;dr:

• simple setup, maintenance and handling of git-hooks across teams and projects
• common git-hook tasks as declarative configuration inside your git repository (optional)

Why?

.githooker shall avoid duplicating code across multiple repositories used for git-hook-ish tasks like evaluating the list of staged files and fire actions accordingly in a pre-commit hook. The setup of such a git-hook is basically turned into a declarative configuration with .githooker. Of course, any arbitrary executable/script is also handled by .githooker.

Moreover, .githooker provides commands to simply manage git-hooks in an interactive CLI manner. The following output of .githooker/help shows/explains all available commands:

Last but not least, git and bash are the only dependencies necessary for this "git-hook helper" in the hope to provide highly versatile use across virtually all project languages.

Note: This shall not stop anyone from using the general idea of generic_hooks.sh with python, ruby, groovy or the like instead of bash within .githooker.

How odes it work?

.githooker is added as a git submodule to your repository. Then a .githooks/ directory is manually created, which holds arbitrary git-hooks and/or mentioned git-hook declarations. Furthermore, .githooker handles symbolic linking from hooks in .git/hooks/pre-commit to .githooks/pre-commit.sh.

To use declarative configuration one needs to source .githooker/generic_hooks.sh in each hook script. Then following two commands are available:

run_command_once "$expression" "$command"

run_command_for_each_file "$expression" "$command"

• $expression - which evaluates list of staged files (line-by-line)

• $command - which will be executed when expression matches

The expression can hold an extension, a filename or even path relative to repository. Moreover, one or multiple expressions can be configured. It's basically a | grep -e, only difference is that a comma ',' becomes an or to put some syntactic sugar to the declarations. Of course, one can always pass any arbitrary valid expression.

".xml"         # matches any *.xml file

"foo.json"     # matches specific filename

"./foo/bar.py" # matches specific file

".c,.cpp"      # matches *.c or *.cpp files

The difference between run_command_{once,for_each_file} is that *_once runs only once if the expression matches. *_for_each_file on the other hand executes a command for each file and adds the file as an argument to the command, so the actual line in generic_hooks.sh looks like:

    $command $changed_file

How to setup?

1st) initial setup

git submodule add https://github.com/boddenberg-it/.githooker .githooker

Create an actual pre-commit hook script in .githooks/pre-commit.sh with the content you want to run, e.g.:

#!/bin/bash
source .githooker/generic_hooks.sh

run_command_once ".py" "./ci/python_related_smoke_tests.sh"

run_command_for_each_file ".xml" "xmllint"

After creating hook simply commit changes and push them - done!

2nd) local setup

After pulling latest changes initialize the git submodule via:

git submodule init

Alternatively, cloning submodules directly when initially cloning super project works via:

git clone --recurse-submodules "$PROJECT_URL"

Then run .githooker/interactive to configure all available hooks in interactive mode as seen in following screenshot:

Or explicitly enable pre-commit hook by executing .githooker/enable pre-commit.

Note: One can also pass the full path to hook script or filename with extension in case you want to use the OS auto-completion feature.

Test setup

First of all .githooker test suites can be invoked in two different ways/environments.

• .githooker/test when using .githooker as submodule
• ./tests/run_tests.sh when developing .githooker, i.e. having it cloned as repo.

Please note that .githooker/test shall be only invoked in a clean git state, i.e. having no staged changes. Loss of changes may occur otherwise.

Test suites can be seen in following example output:

The run_tests.sh script, which is invoked in both of above ways provides general functions for setUp/tearDown tasks and invokes all *_test.sh scripts by sourcing them. Some of these scripts may use an additional helper or expect scripts, which all start with _*. Both expect scripts are only necessary to evaluate output generated by .githooker. But they are only invoked when expect is available to ensure functional tests run regardless of expect.

.githooker/
└── tests
    ├── _counter_for_run_once_test.sh
    ├── _interactive.exp
    ├── _notification_test.exp
    ├── disable_test.sh
    ├── enable_test.sh
    ├── generic_hooks_tests.sh
    ├── interactive_test.sh
    ├── list_test.sh
    ├── pass_hook_path_with_extension_test.sh
    └── run_tests.sh

Travis CI

The .travis.yml file declares all distros as env: -distro=* configuration. Then all distro specific Dockerfile.* are built in parallel using same test script run_githooker_testsuites.sh, which runs test suites in both scenarios.

.githooker/
├── .travis.yml
└── tests
    ├── _docker
    │   ├── Dockerfile.alpine
    │   ├── Dockerfile.archlinux
    │   ├── Dockerfile.centos
    │   ├── Dockerfile.debian
    │   ├── Dockerfile.ubuntu
    │   └── run_githooker_testsuites.sh
    └── test_dockerfiles.sh # tests docker commands used in travis CI locally - sequentially though.

The pipeline runs after a new change has been introduced. Furthermore, it runs every 24 hours if no change occurred to give feedback about OS compatibility. Simply click the Travis badge to see test runs for each OS.

Feebdack

Please send questions, feedback, bugs or suggestions to githooker@boddenberg.it.