libgit2 bindings in Ruby

Rugged is a library for accessing libgit2 in Ruby. It gives you the speed and portability of libgit2 with the beauty of the Ruby language.


libgit2 is a pure C implementation of the Git core methods. It’s designed to be fast and portable. For more information about libgit2, check out libgit2‘s website or browse the libgit2 organization on GitHub.


Rugged is a self-contained gem. You can install it by running:

$ gem install rugged

You need to have CMake and pkg-config installed on your system to be able to build the included version of libgit2. On OS X, after installing Homebrew, you can get CMake with:

“sh $ brew install cmake

If you want to build Rugged with HTTPS and SSH support, check out the list of optional libgit2 dependencies.

If you’re using bundler and want to bundle libgit2 with Rugged, you can use the :submodules option:

“by gem ‘rugged’, git: ‘git://’, submodules: true

To load Rugged, you’ll usually want to add something like this:

“by require ‘rugged’

Use the system provided libgit2

By default, Rugged builds and uses a bundled version of libgit2. If you want to use the system library instead, you can install rugged as follows:

“em install rugged – –use-system-libraries

Or if you are using bundler:

“undle config build.rugged –use-system-libraries bundle install

However, note that Rugged does only support specific versions of libgit2.


Rugged gives you access to the many parts of a Git repository. You can read and write objects, walk a tree, access the staging area, and lots more. Let’s look at each area individually.



The repository is naturally central to Git. Rugged has a Repository class that you can instantiate with a path to open an existing repository :

“by repo =‘path/to/my/repository’)

=> #Rugged::Repository:2228536260 "path/to/my/repository/.git/"

You can create a new repository with init_at. Add a second parameter :bare to make a bare repository:

“by Rugged::Repository.init_at(‘.’, :bare)

You can also let Rugged discover the path to the .git directory if you give it a subdirectory.


=> “/Users/me/projects/repo/.git/”

Once your Repository instantiated (in the following examples, as repo), you can access or modify it.

Accessing a Repository


Does the given SHA1 exist in this repository?


=> true

Boolean repository state values:


=> false


=> true


=> false


=> false

Path accessors


=> “path/to/my/repository/.git/”


=> “path/to/my/repository/”

The HEAD of the repository.

ref = repo.head

=> #Rugged::Reference:2228467240 "refs/heads/master", target: #}>

From the returned ref, you can also access the name, target, and target SHA:

=> “refs/heads/master”

=> #Rugged::Commit:2228467250 "helpful message", tree: #}>


=> “2bc6a70483369f33f641ca44873497f13a15cde5”

Reading an object

object =‘a0ae5566e3c8a3bddffab21022056f0b5e03ef07’)

=> #Rugged::OdbObject:0x109a64780


=> 237

=> “tree 76f23f186076fc291742816721ea8c3e95567241\nparent 8e3c5c52b8f29da0adc7e8be8a037cbeaea6de6b\nauthor Vicent Mart\303\255 1333859005 +0200\ncommitter Vicent Mart\303\255 1333859005 +0200\n\nAdd Repository#blob_at\n”


=> :commit

Writing to a Repository

There’s a few ways to write to a repository. To write directly from your instantiated repository object:

“by sha = repo.write(content, type)

You can also use the Commit object directly to craft a commit; this is a bit more high-level, so it may be preferable:

“by oid = repo.write(“This is a blob.”, :blob) index = repo.index index.read_tree( index.add(:path => “”, :oid => oid, :mode => 0100644)

options = {} options[:tree] = index.write_tree(repo)

options[:author] = { :email => “”, :name => ‘Test Author’, :time => } options[:committer] = { :email => “”, :name => ‘Test Author’, :time => } options[:message] ||= “Making a commit via Rugged!” options[:parents] = repo.empty? ? [] : [].compact options[:update_ref] = ‘HEAD’

Rugged::Commit.create(repo, options)


Object is the main object class - it shouldn’t be created directly, but all of these methods should be useful in their derived classes.

“by obj = repo.lookup(sha) obj.oid # object sha obj.type # One of :commit, :tree, :blob or :tag

robj = obj.read_raw str = int = robj.len

There are four base object types in Git: blobs, commits, tags, and trees. Each of these object types have a corresponding class within Rugged.

Commit Objects

“by commit = repo.lookup(‘a0ae5566e3c8a3bddffab21022056f0b5e03ef07’)

=> #Rugged::Commit:2245304380


=> “Add Repository#blob_at\n”


=> Sat Apr 07 21:23:25 -0700 2012



=> #Rugged::Tree:2245269740


=> [#]

You can also write new objects to the database this way:

“by author =

Rugged::Commit.create(r, :author => author, :message => “Hello world\n\n”, :committer => author, :parents => [2cb831a8aea28b2c1b9c63385585b864e4d3bad1], :tree => some_tree, :update_ref => “HEAD”) #=> “f148106ca58764adc93ad4e2d6b1d168422b9796”

Tag Objects

“by tag = repo.lookup(tag_sha)

object = sha = str = tag.target_type # :commit, :tag, :blob str = # “v1.0” str = tag.message person = tag.tagger

Tree Objects

“by tree = repo.lookup(‘779fbb1e17e666832773a9825875300ea736c2da’)

=> #Rugged::Tree:2245194360

number of tree entries


tree[0] # or… tree.first # or… tree.get_entry(0)


The tree object is an Enumerable, so you can also do stuff like this:

“by tree.each { |e| puts e[:oid] } tree.sort { |a, b| a[:oid] <=> b[:oid] }.map { |e| e[:name] }.join(‘:’)

And there are some Rugged-specific methods, too:

“by tree.each_tree { |entry| puts entry[:name] } # list subdirs tree.each_blob { |entry| puts entry[:name] } # list only files

You can also write trees with the TreeBuilder:

“by oid = repo.write(“This is a blob.”, :blob) builder = builder « { :type => :blob, :name => “”, :oid => oid, :filemode => 0100644 }

options = {} options[:tree] = builder.write

options[:author] = { :email => “”, :name => ‘Test Author’, :time => } options[:committer] = { :email => “”, :name => ‘Test Author’, :time => } options[:message] ||= “Making a commit via Rugged!” options[:parents] = repo.empty? ? [] : [].compact options[:update_ref] = ‘HEAD’

Rugged::Commit.create(repo, options)

Blob Objects

Blob objects represent the data in the files of a Tree Object.

“by blob = repo.lookup(‘e1253910439ea902cf49be8a9f02f3c08d89ac73’) blob.content # => Gives you the content of the blob.

Streaming Blob Objects

There is currently no way to stream data from a blob, because libgit2 itself does not (yet) support streaming blobs out of the git object database. While there are hooks and interfaces for supporting it, the default file system backend always loads the entire blob contents into memory.

If you need to access a Blob object through an IO-like API, you can wrap it with the StringIO class. Note that the only advantage here is a stream-compatible interface, the complete blob object will still be loaded into memory. Below is an example for streaming a Blob using the Sinatra framework:


Sinatra endpoint

get “/blobs/:sha” do repo = blob = repo.lookup params[:sha]

headers({ “Vary” => “Accept”, “Connection” => “keep-alive”, “Transfer-Encoding” => “chunked”, “Content-Type” => “application/octet-stream”, })

stream do |out| do |chunk| out « chunk end end end

Commit Walker

Rugged::Walker is a class designed to help you traverse a set of commits over a repository.

You first push head SHAs onto the walker, and then call next to get a list of the reachable commit objects one at a time. You can also hide() commits if you are not interested in anything beneath them (useful in situations like when you’re running something like git log master ^origin/master).

“by walker = walker.sorting(Rugged::SORT_TOPO | Rugged::SORT_REVERSE) # optional walker.push(hex_sha_interesting) walker.hide(hex_sha_uninteresting) walker.each { |c| puts c.inspect } walker.reset

Index (“staging”) area

We can inspect and manipulate the Git Index as well. To work with the index inside an existing repository, instantiate it by using the Repository.index method instead of manually opening the Index by its path.

“by index =

Re-read the index file from disk.


Count up index entries.

count = index.count

The collection of index entries.


Iterating over index entries.

index.each { |i| puts i.inspect }

Get a particular entry in the index.




Stage. Also updates existing entry if there is one.


Stage. Create ientry from file in path, updates the index.



You can access references through the Rugged::ReferenceCollection object returned by Repository#references.

“by ref = repo.references[refs/heads/master]

sha = ref.target_id str = ref.type # :direct str = # “refs/heads/master”

You can also easily iterate over all references:

“by repo.references.each do |ref| puts end

Or only over references that match the given pattern (glob):

“by repo.references.each(“refs/tags/“) do |ref| puts end

It is also easy to create, update, rename or delete a reference:

“by ref = repo.references.create(“refs/heads/unit_test”, some_commit_sha)

repo.references.update(ref, new_sha) # or… repo.references.update(“refs/heads/unit_test”, new_sha)

repo.references.rename(ref, “refs/heads/blead”) # or… repo.references.rename(“refs/heads/unit_test”, “refs/heads/blead”)

repo.references.delete(ref) # or… repo.references.delete(“refs/heads/unit_test”) # or…

Finally, you can access the reflog for any branch:

“by ref = repo.references[refs/heads/master] entry = ref.log.first sha = entry[:id_old] sha = entry[:id_new] str = entry[:message] prsn = entry[:committer]


The Rugged::BranchCollection object returned by Repository#branches will help you with all of your branch-related needs.

Iterate over all branches:

“by repo.branches.each_name().sort

=> [master, origin/HEAD, origin/master, origin/packed]


=> [master]


=> [origin/HEAD, origin/master, origin/packed]

Look up branches and get attributes:

“by branch = repo.branches[master] # => ‘master’ branch.canonical_name # => ‘refs/heads/master’

Look up the id for the target of a branch:

“by repo.branches[master].target_id

=> “36060c58702ed4c2a40832c51758d5344201d89a”

Creation and deletion:

“by branch = repo.branches.create(“test_branch”, “HEAD”)

repo.branches.rename(“test_branch”, “new_branch”) # or… repo.branches.rename(“refs/heads/test_branch”, “new_branch”) # or… repo.branches.rename(ref, “new_branch”) # or…

repo.branches.delete(“test_branch”) # or… repo.branches.delete(“refs/heads/test_branch”) # or… repo.branches.delete(ref) # or…


There are various ways to get hands on diffs:


Diff between two subsequent commits

diff_commits = commit_object.parents[0].diff(commit_object)

Diff between two tree objects

diff_trees = tree_object_a.diff(tree_object_b)

Diff between index/staging and current working directory

diff_index = repository.index.diff

Diff between index/staging and another diffable (commit/tree/index)

diff_index_diffable = repository.index.diff(some_diffable)

When you already have a diff object, you can examine it:


Get patch


Get delta (faster, if you only need information on what files changed)

diff.each_delta{ |d| puts d.inspect }

Rugged::Diff::Delta:70144372137380 {old_file: {:oid=“0000000000000000000000000000000000000000”, :path=>“foo1”, :size=>0, :flags=>6, :mode=>0, new_file: , similarity: 0, status: :added>

Rugged::Diff::Delta:70144372136540 {old_file: {:oid=“81b68f040b120c9627518213f7fc317d1ed18e1c”, :path=>“txt1”, :size=>14, :flags=>6, :mode=>33188, new_file: , similarity: 0, status: :deleted>

Rugged::Diff::Delta:70144372135780 {old_file: {:oid=“a7bb42f71183c162efea5e4c80597437d716c62b”, :path=>“txt2”, :size=>17, :flags=>6, :mode=>33188, new_file: , similarity: 0, status: :modified>

Detect renamed files

Note that the status field changed from :added/:deleted to :renamed

diff.find_similar! diff.each_delta{ |d| puts d.inspect }

Rugged::Diff::Delta:70144372230920 {old_file: {:oid=“81b68f040b120c9627518213f7fc317d1ed18e1c”, :path=>“txt1”, :size=>14, :flags=>6, :mode=>33188, new_file: , similarity: 100, status: :renamed>

Rugged::Diff::Delta:70144372230140 {old_file: {:oid=“a7bb42f71183c162efea5e4c80597437d716c62b”, :path=>“txt2”, :size=>17, :flags=>6, :mode=>33188, new_file: , similarity: 0, status: :modified>

Merge one diff into another (mutating the first one)


Write a patch into a file (or any other object responding to write)

Note that the patch as in diff.patch will be written, it won’t be applied

file =‘/some/file’, ‘w’) diff.write_patch(file) file.close

Config files

It’s also easy to read and manipulate the Git config file data with Rugged.


Read values


Set values

repo.config[] = true

Delete values


General methods

Rugged also includes a general library for handling basic Git operations. One of these is converting a raw sha (20 bytes) into a readable hex sha (40 characters).

“by Rugged.hex_to_raw(‘bfde59cdd0dfac1d892814f66a95641abd8a1faf’)

=> “\277\336Y\315\320\337\254\035\211(\024\366j\225d\032\275\212\037\257”


 Alternative backends

You can store bare repositories in alternative backends instead of storing on disk. (see redbadger/rugged-redis for an example of how a rugged backend works).

“by a_backend = ‘setting’, opt2: ‘setting’)

repo = Rugged::Repository.init_at(‘repo_name’, :bare, backend: a_backend)


repo = Rugged::Repository.bare(‘repo_name’, backend: a_backend)


Fork libgit2/rugged on GitHub, make it awesomer (preferably in a branch named for the topic), send a pull request.


Simply clone and install:

$ git clone
$ cd rugged
$ bundle install
$ rake compile
$ rake test


We encourage you to use StackOverflow for any questions or concerns regarding Rugged. Please tag your questions with the rugged keyword.

For bug reports, please open a ticket on the GitHub issue tracker.



MIT. See LICENSE file.