The Git Gem

• Summary
• v2.x Release
• Install
• Major Objects
• Errors Raised By This Gem
• Specifying And Handling Timeouts
• Examples
• Ruby version support policy
• License

Summary

The git gem provides a Ruby interface to the git command line.

Get started by obtaining a repository object by:

• opening an existing working copy with Git.open
• initializing a new repository with Git.init
• cloning a repository with Git.clone

Methods that can be called on a repository object are documented in Git::Base

v2.x Release

git 2.0.0 has recently been released. Please give it a try.

If you have problems with the 2.x release, open an issue and use the 1.x version instead. We will do our best to fix your issues in a timely fashion.

JRuby on Windows is not yet supported by the 2.x release line. Users running JRuby on Windows should continue to use the 1.x release line.

The changes in this major release include:

• Added a dependency on the activesupport gem to use the deprecation functionality
• Create a policy of supported Ruby versions to support only non-EOL Ruby versions
• Create a policy of supported Git CLI versions (released 2020-12-25)
• Update the required Ruby version to at least 3.0 (released 2020-07-27)
• Update the required Git command line version to at least 2.28
• Update how CLI commands are called to use the process_executer gem which is built on top of Kernel.spawn. See PR #684 for more details on the motivation for this implementation.

The master branch will be used for 2.x development. If needed, fixes for 1.x version will be done on the v1 branch.

Install

Install the gem and add to the application's Gemfile by executing:

bundle add git

to install version 1.x:

bundle add git --version "~> 1.19"

If bundler is not being used to manage dependencies, install the gem by executing:

gem install git

to install version 1.x:

gem install git --version "~> 1.19"

Major Objects

Git::Base - The object returned from a Git.open or Git.clone. Most major actions are called from this object.

Git::Object - The base object for your tree, blob and commit objects, returned from @git.gtree or @git.object calls. the Git::AbstractObject will have most of the calls in common for all those objects.

Git::Diff - returns from a @git.diff command. It is an Enumerable that returns Git::Diff:DiffFile objects from which you can get per file patches and insertion/deletion statistics. You can also get total statistics from the Git::Diff object directly.

Git::Status - returns from a @git.status command. It is an Enumerable that returns Git:Status::StatusFile objects for each object in git, which includes files in the working directory, in the index and in the repository. Similar to running 'git status' on the command line to determine untracked and changed files.

Git::Branches - Enumerable object that holds Git::Branch objects. You can call .local or .remote on it to filter to just your local or remote branches.

Git::Remote- A reference to a remote repository that is tracked by this repository.

Git::Log - An Enumerable object that references all the Git::Object::Commit objects that encompass your log query, which can be constructed through methods on the Git::Log object, like:

git.log.max_count(:all).object('README.md').since('10 years ago').between('v1.0.7','HEAD').map{ |commit| commit.sha}

A maximum of 30 commits are returned if max_count is not called. To get all commits that match the log query, call max_count(:all).

Note that git.log.all adds the --all option to the underlying git log command. This asks for the logs of all refs (basically all commits reachable by HEAD, branches, and tags). This does not control the maximum number of commits returned. To control how many commits are returned, you should call max_count.

Git::Worktrees - Enumerable object that holds Git::Worktree objects.

Errors Raised By This Gem

The git gem will only raise an ArgumentError or an error that is a subclass of Git::Error. It does not explicitly raise any other types of errors.

It is recommended to rescue Git::Error to catch any runtime error raised by this gem unless you need more specific error handling.

begin# some git operationrescueGit::Error=>eputs"An error occurred: #{e.message}"end

See Git::Error for more information.

Specifying And Handling Timeouts

The timeout feature was added in git gem version 2.0.0.

A timeout for git command line operations can be set either globally or for specific method calls that accept a :timeout parameter.

The timeout value must be a real, non-negative Numeric value that specifies a number of seconds a git command will be given to complete before being sent a KILL signal. This library may hang if the git command does not terminate after receiving the KILL signal.

When a command times out, it is killed by sending it the SIGKILL signal and a Git::TimeoutError is raised. This error derives from the Git::SignaledError and Git::Error.

If the timeout value is 0 or nil, no timeout will be enforced.

If a method accepts a :timeout parameter and a receives a non-nil value, the value of this parameter will override the global timeout value. In this context, a value of nil (which is usually the default) will use the global timeout value and a value of 0 will turn off timeout enforcement for that method call no matter what the global value is.

To set a global timeout, use the Git.config object:

Git.config.timeout=nil# a value of nil or 0 means no timeout is enforcedGit.config.timeout=1.5# can be any real, non-negative Numeric interpreted as number of seconds

The global timeout can be overridden for a specific method if the method accepts a :timeout parameter:

repo_url='https://github.com/ruby-git/ruby-git.git'Git.clone(repo_url)# Use the global timeout valueGit.clone(repo_url,timeout: nil)# Also uses the global timeout valueGit.clone(repo_url,timeout: 0)# Do not enforce a timeoutGit.clone(repo_url,timeout: 10.5)# Timeout after 10.5 seconds raising Git::SignaledError

If the command takes too long, a Git::TimeoutError will be raised:

beginGit.clone(repo_url,timeout: 10)rescueGit::TimeoutError=>ee.result.tapdo |r|
r.class#=> Git::CommandLineResultr.status#=> #<Process::Status: pid 62173 SIGKILL (signal 9)>r.status.timeout?#=> truer.git_cmd# The git command ran as an array of stringsr.stdout# The command's output to stdout until it was terminatedr.stderr# The command's output to stderr until it was terminatedendend

Examples

Here are a bunch of examples of how to use the Ruby/Git package.

Require the 'git' gem.

require'git'

Git env config

Git.configuredo |config|
# If you want to use a custom git binaryconfig.binary_path='/git/bin/path'# If you need to use a custom SSH scriptconfig.git_ssh='/path/to/ssh/script'end

NOTE: Another way to specify where is the git binary is through the environment variable GIT_PATH

Here are the operations that need read permission only.

g=Git.open(working_dir,:log=>Logger.new(STDOUT))g.indexg.index.readable?g.index.writable?g.repog.dir# ls-tree with recursion into subtrees (list files)g.ls_tree("HEAD",recursive: true)# log - returns a Git::Log object, which is an Enumerator of Git::Commit objects# default configuration returns a max of 30 commitsg.logg.log(200)# 200 most recent commitsg.log.since('2 weeks ago')# default count of commits since 2 weeks ago.g.log(200).since('2 weeks ago')# commits since 2 weeks ago, limited to 200.g.log.between('v2.5','v2.6')g.log.each{|l| putsl.sha}g.gblob('v2.5:Makefile').log.since('2 weeks ago')g.object('HEAD^').to_s# git show / git rev-parseg.object('HEAD^').contentsg.object('v2.5:Makefile').sizeg.object('v2.5:Makefile').shag.gtree(treeish)g.gblob(treeish)g.gcommit(treeish)commit=g.gcommit('1cc8667014381')commit.gtreecommit.parent.shacommit.parents.sizecommit.author.namecommit.author.emailcommit.author.date.strftime("%m-%d-%y")commit.committer.namecommit.date.strftime("%m-%d-%y")commit.messagetree=g.gtree("HEAD^{tree}")tree.blobstree.subtreestree.children# blobs and subtreesg.rev_parse('v2.0.0:README.md')g.branches# returns Git::Branch objectsg.branches.localg.current_branchg.branches.remoteg.branches[:master].gcommitg.branches['origin/master'].gcommitg.grep('hello')# implies HEADg.blob('v2.5:Makefile').grep('hello')g.tag('v2.5').grep('hello','docs/')g.describe()g.describe('0djf2aa')g.describe('HEAD',{:all=>true,:tags=>true})g.diff(commit1,commit2).sizeg.diff(commit1,commit2).statsg.diff(commit1,commit2).name_statusg.gtree('v2.5').diff('v2.6').insertionsg.diff('gitsearch1','v2.5').path('lib/')g.diff('gitsearch1',@git.gtree('v2.5'))g.diff('gitsearch1','v2.5').path('docs/').patchg.gtree('v2.5').diff('v2.6').patchg.gtree('v2.5').diff('v2.6').eachdo |file_diff|
putsfile_diff.pathputsfile_diff.patchputsfile_diff.blob(:src).contentsendg.worktrees# returns Git::Worktree objectsg.worktrees.countg.worktrees.eachdo |worktree|
worktree.dirworktree.gcommitworktree.to_sendg.config('user.name')# returns 'Scott Chacon'g.config# returns whole config hash# Configuration can be set when cloning using the :config option.# This option can be an single configuration String or an Array# if multiple config items need to be set.#g=Git.clone(git_uri,destination_path,:config=>['core.sshCommand=ssh -i /home/user/.ssh/id_rsa','submodule.recurse=true'])g.tags# returns array of Git::Tag objectsg.show()g.show('HEAD')g.show('v2.8','README.md')Git.ls_remote('https://github.com/ruby-git/ruby-git.git')# returns a hash containing the available references of the repo.Git.ls_remote('/path/to/local/repo')Git.ls_remote()# same as Git.ls_remote('.')Git.default_branch('https://github.com/ruby-git/ruby-git')#=> 'master'

And here are the operations that will need to write to your git repository.

g=Git.initGit.init('project')Git.init('/home/schacon/proj',{:repository=>'/opt/git/proj.git',:index=>'/tmp/index'})# Clone from a git urlgit_url='https://github.com/ruby-git/ruby-git.git'# Clone into the ruby-git directoryg=Git.clone(git_url)# Clone into /tmp/clone/ruby-git-cleanname='ruby-git-clean'path='/tmp/clone'g=Git.clone(git_url,name,:path=>path)g.dir#=> /tmp/clone/ruby-git-cleang.config('user.name','Scott Chacon')g.config('user.email','email@email.com')# Clone can take a filter to tell the serve to send a partial cloneg=Git.clone(git_url,name,:path=>path,:filter=>'tree:0')# Clone can take an optional loggerlogger=Logger.newg=Git.clone(git_url,NAME,:log=>logger)g.add# git add -- "."g.add(:all=>true)# git add --all -- "."g.add('file_path')# git add -- "file_path"g.add(['file_path_1','file_path_2'])# git add -- "file_path_1" "file_path_2"g.remove()# git rm -f -- "."g.remove('file.txt')# git rm -f -- "file.txt"g.remove(['file.txt','file2.txt'])# git rm -f -- "file.txt" "file2.txt"g.remove('file.txt',:recursive=>true)# git rm -f -r -- "file.txt"g.remove('file.txt',:cached=>true)# git rm -f --cached -- "file.txt"g.commit('message')g.commit_all('message')# Sign a commit using the gpg key configured in the user.signingkey config settingg.config('user.signingkey','0A46826A')g.commit('message',gpg_sign: true)# Sign a commit using a specified gpg keykey_id='0A46826A'g.commit('message',gpg_sign: key_id)# Skip signing a commit (overriding any global gpgsign setting)g.commit('message',no_gpg_sign: true)g=Git.clone(repo,'myrepo')g.chdirdonew_file('test-file','blahblahblah')g.status.changed.eachdo |file|
putsfile.blob(:index).contentsendendg.reset# defaults to HEADg.reset_hard(Git::Commit)g.branch('new_branch')# creates new or fetches existingg.branch('new_branch').checkoutg.branch('new_branch').deleteg.branch('existing_branch').checkoutg.branch('master').contains?('existing_branch')# delete remote branchg.push('origin','remote_branch_name',force: true,delete: true)g.checkout('new_branch')g.checkout('new_branch',new_branch: true,start_point: 'master')g.checkout(g.branch('new_branch'))g.branch(name).merge(branch2)g.branch(branch2).merge# merges HEAD with branch2g.branch(name).in_branch(message){# add files } # auto-commitsg.merge('new_branch')g.merge('new_branch','merge commit message',no_ff: true)g.merge('origin/remote_branch')g.merge(g.branch('master'))g.merge([branch1,branch2])g.merge_base('branch1','branch2')r=g.add_remote(name,uri)# Git::Remoter=g.add_remote(name,Git::Base)# Git::Remoteg.remotes# array of Git::Remotesg.remote(name).fetchg.remote(name).removeg.remote(name).mergeg.remote(name).merge(branch)g.fetchg.fetch(g.remotes.first)g.fetch('origin',{:ref=>'some/ref/head'})g.fetch(all: true,force: true,depth: 2)g.fetch('origin',{:'update-head-ok'=>true})g.pullg.pull(Git::Repo,Git::Branch)# fetch and a mergeg.add_tag('tag_name')# returns Git::Tagg.add_tag('tag_name','object_reference')g.add_tag('tag_name','object_reference',{:options=>'here'})g.add_tag('tag_name',{:options=>'here'})Options:
:a | :annotate:d:f:m | :message:sg.delete_tag('tag_name')g.repackg.pushg.push(g.remote('name'))# delete remote branchg.push('origin','remote_branch_name',force: true,delete: true)# push all branches to remote at one timeg.push('origin',all: true)g.worktree('/tmp/new_worktree').addg.worktree('/tmp/new_worktree','branch1').addg.worktree('/tmp/new_worktree').removeg.worktrees.prune

Some examples of more low-level index and tree operations

g.with_temp_indexdog.read_tree(tree3)# calls self.index.read_treeg.read_tree(tree1,:prefix=>'hi/')c=g.commit_tree('message')# or #t=g.write_treec=g.commit_tree(t,:message=>'message',:parents=>[sha1,sha2])g.branch('branch_name').update_ref(c)g.update_ref(branch,c)g.with_temp_workingdo# new blank working directoryg.checkoutg.checkout(another_index)g.commit# commits to temp_indexendendg.set_index('/path/to/index')g.with_index(path)do# calls set_index, then switches back afterendg.with_working(dir)do# calls set_working, then switches back afterendg.with_temp_working(dir)dog.checkout_index(:prefix=>dir,:path_limiter=>path)# do file workg.commit# commits to indexend

Ruby version support policy

This gem will be expected to function correctly on:

• All non-EOL versions of the MRI Ruby on Mac, Linux, and Windows
• The latest version of JRuby on Linux
• The latest version of Truffle Ruby on Linus

It is this project's intent to support the latest version of JRuby on Windows once the following JRuby bug is fixed:

jruby/jruby#7515

License

Licensed under MIT License Copyright (c) 2008 Scott Chacon. See LICENSE for further details.