overcommit is a tool to manage and configure Git hooks.

In addition to supporting a wide variety of hooks that can be used across multiple repositories, you can also define hooks specific to a repository which are stored in source control. You can also easily add your existing hook scripts without writing any Ruby code.

• Requirements
  • Dependencies
• Limitations
• Installation
  • Automatically Install Overcommit Hooks
• Usage
• Continuous Integration
• Configuration
  • Hook Options
  • Hook Categories
  • Gemfile
  • Plugin Directory
  • Signature Verification
• Built-In Hooks
  • CommitMsg
  • PostCheckout
  • PostCommit
  • PostMerge
  • PostRewrite
  • PreCommit
  • PrePush
  • PreRebase
• Repo-Specific Hooks
  • Adding Existing Git Hooks
• Security
• Contributing
• Community
• Changelog
• License

This project aims to support the following Ruby runtimes on both *nix and Windows:

• MRI 2.2+

Some of the hooks have third-party dependencies. For example, to lint your SCSS files, you're going to need our scss_lint gem.

Depending on the hooks you enable/disable for your repository, you'll need to ensure your development environment already has those dependencies installed. Most hooks will display a warning if a required executable isn't available.

If you are using Bundler to manage your Ruby gem dependencies, you'll likely want to use the gemfile option to control which gem versions are available during your hook runs.

overcommit is installed via RubyGems. It is strongly recommended that your environment support running gem install without requiring sudo privileges. Using a Ruby version manager like rbenv or rvm can help here.

Once you have an environment that allows you to install gems without sudo, run:

gem install overcommit

You can then run the overcommit command to install hooks into repositories.

mkdir important-project
cd important-project
git init
overcommit --install

After running overcommit --install, any existing hooks for your repository which Overcommit will replace will be backed up. You can restore everything to the way it was by running overcommit --uninstall.

If you want to use overcommit for all repositories you create/clone going forward, add the following to automatically run in your shell environment:

export GIT_TEMPLATE_DIR="$(overcommit --template-dir)"

The GIT_TEMPLATE_DIR provides a directory for Git to use as a template for automatically populating the .git directory. If you have your own template directory, you might just want to copy the contents of overcommit --template-dir to that directory.

Once you've installed the hooks via overcommit --install, they will automatically run when the appropriate hook is triggered.

The overcommit executable supports the following command-line flags:

Sometimes a hook will report an error that for one reason or another you'll want to ignore. To prevent these errors from blocking your commit, you can include the name of the relevant hook in the SKIP environment variable, e.g.

SKIP=RuboCop git commit

If you would prefer to specify a whitelist of hooks rather than a blacklist, use the ONLY environment variable instead.

ONLY=RuboCop git commit

Use this feature sparingly, as there is no point to having the hook in the first place if you're just going to ignore it. If you want to ensure a hook is never skipped, set the required option to true in its configuration. If you attempt to skip it, you'll see a warning telling you that the hook is required, and the hook will still run.

If you have scripts that execute git commands where you don't want Overcommit hooks to run, you can disable Overcommit entirely by setting the OVERCOMMIT_DISABLE environment variable.

OVERCOMMIT_DISABLE=1 ./my-custom-script

Overcommit automatically colorizes its output based on whether it is outputting to a TTY. However, you can manually enable/disable color by setting the OVERCOMMIT_COLOR environment variable.

OVERCOMMIT_COLOR=0 git commit

You can run the same set of hooks that would be executed in a pre-commit hook against your entire repository by running overcommit --run. This makes it easy to have the checks verified by a CI service such as Travis CI, including custom hooks you've written yourself.

The --run flag works by creating a pre-commit context that assumes all the files in your repository have changed, and follows the same rules as a normal pre-commit check. If any hook fails with an error, it will return a non-zero exit code.

Overcommit provides a flexible configuration system that allows you to tailor the built-in hooks to suit your workflow. All configuration specific to a repository is stored in .overcommit.yml in the top-level directory of the repository.

When writing your own configuration, it will automatically extend the default configuration, so you only need to specify your configuration with respect to the default. In order to enable/disable hooks, you can add the following to your repo-specific configuration file:

PreCommit:
  RuboCop:
    enabled: true
    command: ['bundle', 'exec', 'rubocop'] # Invoke within Bundler context

Individual hooks expose both built-in configuration options as well as their own custom options unique to each hook. The following table lists all built-in configuration options:

In addition to the built-in configuration options, each hook can expose its own unique configuration options. The AuthorEmail hook, for example, allows you to customize the regex used to check commit author emails via the pattern option—useful if you want to enforce that developers use a company email address for their commits. This provides incredible flexibility for hook authors as you can make your hooks sufficiently generic and then customize them on a per-project basis.

Hook configurations are organized into categories based on the type of hook. So pre-commit hooks are located under the PreCommit option, and post-commit hooks are located under PostCommit. See the default configuration for a thorough example.

Within a hook category, there is a special type of hook configuration that applies to all hooks in the category. This configuration looks like a normal hook configuration, except it has the name ALL:

PreCommit:
  ALL:
    problem_on_unmodified_line: warn
    requires_files: true
    required: false
    quiet: false

  SomeHook:
    enabled: true

  ...

The ALL configuration is useful for when you want to DRY up your configuration, or when you want to apply changes across an entire category of hooks.

Note that array configuration options (like include/exclude) in the special ALL hook section are not merged with individual hook configurations if custom ones are defined for the hook. Any custom configuration option for include/exclude will replace the ALL hook's configuration. If you want to have a global list of default exclusions and extend them with a custom list, you can use YAML references, e.g.

PreCommit:
  ALL:
    exclude: &default_excludes
      - 'node_modules/**/*'
      - 'vendor/**/*'
  MyHook:
    exclude:
      - *default_excludes
      - 'another/directory/in/addition/to/default/excludes/**/*'

Again, you can consult the default configuration for detailed examples of how the ALL hook can be used.

You may want to enforce the version of Overcommit or other gems that you use in your git hooks. This can be done by specifying the gemfile option in your .overcommit.yml.

The gemfile option tells Overcommit to load the specified file with Bundler, the standard gem dependency manager for Ruby. This is useful if you would like to:

• Enforce a specific version of Overcommit to use for all hook runs (or to use a version from the master branch that has not been released yet)
• Enforce a specific version or unreleased branch is used for a gem you want to use in your git hooks

Loading a Bundler context necessarily adds a startup delay to your hook runs as Bundler parses the specified Gemfile and checks that the dependencies are satisfied. Thus for projects with many gems this can introduce a noticeable delay.

The recommended workaround is to create a separate Gemfile in the root of your repository (call it .overcommit_gems.rb), and include only the gems that your Overcommit hooks need in order to run. Generate the associated lock file by running:

bundle install --gemfile=.overcommit_gems.rb

...and commit .overcommit_gems.rb and the resulting .overcommit_gems.rb.lock file to your repository. Set your gemfile option to .overcommit_gems.rb, and you're all set.

Using a smaller Gemfile containing only the gems used by your Overcommit hooks significantly reduces the startup delay in your hook runs. It is thus the recommended approach unless your project has a relatively small number of gems in your Gemfile.

You can change the directory that project-specific hooks are loaded from via the plugin_directory option. The default directory is .git-hooks.

If you prefer to have your hooks be completely silent unless there is a problem, you can set the top-level quiet option to true. Note that if you have many hooks or slow hooks this may not be desirable, as you don't get visual feedback indicating the general progress of the hook run.

Overcommit runs hooks in parallel by default, with a number of concurrent workers equal to the number of logical cores on your machine. If you know your particular set of hooks would benefit from higher/lower number of workers, you can adjust the global concurrency option. You can define single-operator mathematical expressions, e.g. %{processors} * 2, or %{processors} / 2.

concurrency: '%{processors} / 4'

Note that individual hooks can specify the number of processors they require with the processors hook option. See the hook options section for more details.

You can disable manual verification of signatures by setting verify_signatures to false. See the Security section for more information on this option and what exactly it controls.

Currently, Overcommit supports the following hooks out of the box—simply enable them in your .overcommit.yml.

Note: Hooks with a * are enabled by default.

Warning: This list represents the list of hooks available on the master branch. Please consult the change log to view which hooks have not been released yet.

commit-msg hooks are run against every commit message you write before a commit is created. A failed hook prevents a commit from being created. These hooks are useful for enforcing policies on your commit messages, e.g. ensuring a task ID is included for tracking purposes, or ensuring your commit messages follow proper formatting guidelines.

• *CapitalizedSubject
• *EmptyMessage
• GerritChangeId
• HardTabs
• MessageFormat
• RussianNovel
• *SingleLineSubject
• SpellCheck
• *TextWidth
• *TrailingPeriod

post-checkout hooks run after a successful git checkout, or in other words any time your HEAD changes or a file is explicitly checked out.

• BowerInstall
• BundleInstall
• ComposerInstall
• IndexTags
• NpmInstall
• SubmoduleStatus
• YarnInstall

post-commit hooks run after a commit is successfully created. A hook failing in this case does not prevent the commit since it has already occurred; however, it can be used to alert the user to some issue.

• BowerInstall
• BundleInstall
• Commitplease
• ComposerInstall
• GitGuilt
• IndexTags
• NpmInstall
• SubmoduleStatus
• YarnInstall

post-merge hooks run after a git merge executes successfully with no merge conflicts. A hook failing in this case does not prevent the merge since it has already occurred; however, it can be used to alert the user to some issue.

• BowerInstall
• BundleInstall
• ComposerInstall
• IndexTags
• NpmInstall
• SubmoduleStatus
• YarnInstall

post-rewrite hooks run after a commit is modified by a git commit --amend or git rebase. A hook failing in this case does not prevent the rewrite since it has already occurred; however, it can be used to alert the user to some issue.

• BowerInstall
• BundleInstall
• ComposerInstall
• IndexTags
• NpmInstall
• SubmoduleStatus
• YarnInstall

pre-commit hooks are run after git commit is executed, but before the commit message editor is displayed. If a hook fails, the commit will not be created. These hooks are ideal for syntax checkers, linters, and other checks that you want to run before you allow a commit to even be created.

pre-commit hooks currently do not support hooks with side effects (such as modifying files and adding them to the index with git add). This is a consequence of Overcommit's pre-commit hook stashing behavior to ensure hooks are run against only the changes you are about to commit.

Without Overcommit, the proper way to write a pre-commit hook would be to extract the staged changes into temporary files and lint those files instead of whatever contents are in your working tree (as you don't want unstaged changes to taint your results). Overcommit takes care of this for you, but to do it in a generalized way introduces this limitation. See the thread tracking this issue for more details.

• *AuthorEmail
• *AuthorName
• BerksfileCheck
• *BrokenSymlinks
• BundleAudit
• BundleCheck
• BundleOutdated
• *CaseConflicts
• ChamberSecurity
• CoffeeLint
• Credo
• CssLint
• Dogma
• EsLint
• ExecutePermissions
• Fasterer
• FixMe
• Flay
• Foodcritic
• ForbiddenBranches
• GoLint
• GoVet
• Hadolint
• LicenseFinder
• HamlLint
• HardTabs
• Hlint
• HtmlHint
• HtmlTidy
• ImageOptim
• JavaCheckstyle
• Jscs
• JsHint
• JsLint
• Jsl
• JsonSyntax
• KtLint
• LicenseHeader
• LineEndings
• LocalPathsInGemfile
• Mdl
• *MergeConflicts
• NginxTest
• PhpCs
• PhpCsFixer
• PhpLint
• PhpStan
• Pronto
• PuppetLint
• PuppetMetadataJsonLint
• Pycodestyle
• Pydocstyle
• Pyflakes
• Pylint
• PythonFlake8
• RakeTarget
• RailsBestPractices
• RailsSchemaUpToDate
• Reek
• RuboCop
• RubyLint
• SwiftLint
• Scalariform
• Scalastyle
• ScssLint
• SemiStandard
• ShellCheck
• SlimLint
• Sqlint
• Standard
• Stylelint
• TrailingWhitespace
• TravisLint
• TsLint
• Vint
• W3cCss
• W3cHtml
• XmlLint
• XmlSyntax
• YamlLint
• YamlSyntax
• YardCoverage
• YarnCheck

pre-push hooks are run during git push, after remote refs have been updated but before any objects have been transferred. If a hook fails, the push is aborted.

• Brakeman
• Minitest
• PhpUnit
• ProtectedBranches
• Pytest
• PythonNose
• RakeTarget
• RSpec
• TestUnit

pre-rebase hooks are run during git rebase, before any commits are rebased. If a hook fails, the rebase is aborted.

• MergedCommits

Out of the box, overcommit comes with a set of hooks that enforce a variety of styles and lints. However, some hooks only make sense in the context of a specific repository.

At Brigade, for example, we have a number of simple checks that we run against our code to catch common errors. For example, since we use RSpec, we want to make sure all spec files contain the line require 'spec_helper'.

Inside our repository, we can add the file .git-hooks/pre_commit/ensure_spec_helper.rb in order to automatically check our spec files:

module Overcommit::Hook::PreCommit
  class EnsureSpecHelper < Base
    def run
      errors = []

      applicable_files.each do |file|
        if File.read(file) !~ /^require 'spec_helper'/
          errors << "#{file}: missing `require 'spec_helper'`"
        end
      end

      return :fail, errors.join("\n") if errors.any?

      :pass
    end
  end
end

The corresponding configuration for this hook would look like:

PreCommit:
  EnsureSpecHelper:
    enabled: true
    description: 'Checking for missing inclusion of spec_helper'
    include: '**/*_spec.rb'

You might already have hook scripts written which you'd like to integrate with Overcommit right away. To make this easy, Overcommit allows you to include your hook script in your configuration without writing any Ruby code. For example:

PostCheckout:
  CustomScript:
    enabled: true
    required_executable: './bin/custom-script'

So long as a command is given (either by specifying the command option directly or specifying required_executable) a special hook is created that executes the command and appends any arguments and standard input stream that would have been passed to the regular hook. The hook passes or fails based on the exit status of the command.

While Overcommit can make managing Git hooks easier and more convenient, this convenience can come at a cost of being less secure.

Since installing Overcommit hooks will allow arbitrary plugin code in your repository to be executed, you expose yourself to an attack where checking out code from a third party can result in malicious code being executed on your system.

As an example, consider the situation where you have an open source project. An attacker could submit a pull request which adds a post-checkout hook that executes some malicious code. When you fetch and checkout this pull request, the post-checkout hook will be run on your machine, along with the malicious code that you just checked out.

Overcommit attempts to address this problem by storing a signature of your configuration and all hook plugin code since the last time it ran. When the signature changes, a warning is displayed alerting you to which plugins have changed. It is then up to you to manually verify that the changes are not malicious, and then continue running the hooks.

The signature is derived from the contents of the plugin's source code itself and any configuration for the plugin. Thus a change to the plugin's source code or your local repo's .overcommit.yml file could result in a signature change.

In typical usage, your plugins usually don't change too often, so this warning shouldn't become a nuisance. However, users who work within proprietary repositories where all developers who can push changes to the repository already have a minimum security clearance may wish to disable this check.

While not recommended, you can disable signature verification by setting verify_signatures to false in your .overcommit.yml file.

Regardless of whether you have verify_signatures disabled for your project, if you are running Overcommit for the first time you will need to sign your configuration with overcommit --sign. This needs to happen once so Overcommit can record in your local git repo's configuration (outside of source control) that you intend to enable/disable verification. This way if someone else changes verify_signatures you'll be asked to confirm the change.

We love contributions to Overcommit, be they bug reports, feature ideas, or pull requests. See our guidelines for contributing to best ensure your thoughts, ideas, or code get merged.

All major discussion surrounding Overcommit happens on the GitHub issues list.

You can also follow @git_overcommit on Twitter.

If you're interested in seeing the changes and bug fixes between each version of overcommit, read the Overcommit Changelog.

This project is released under the MIT license.

The Overcommit logo is adapted from the Git Logo by Jason Long, and is licensed under the Creative Commons Attribution 3.0 Unported License.