pygit2 - libgit2 bindings in Python

pygit2 is a set of Python bindings to the libgit2 linkable C Git library. The supported versions of Python are 2.6, 2.7, 3.1 and 3.2

You can find an up-to-date api documentation at http://libgit2.github.com/pygit2.

Through this text Python 3 is used for the inline examples. Also, the Python 3 terminology is used (for instance we say text strings instead of unicode strings).

Installing and running

First you need to install the latest version of libgit2. You can find platform-specific instructions to build the library in the libgit2 website:

http://libgit2.github.com

Also, make sure you have Python 2.6+ installed together with the Python development headers.

When those are installed, you can install pygit2:

$ git clone git://github.com/libgit2/pygit2.git
$ cd pygit2
$ python setup.py install
$ python setup.py test

Building on *nix (including OS X)

If you installed libgit2 and pygit2 in one of the usual places, you should be able to skip this section and just use the generic pygit2 installation steps described above. This is the recommended procedure.

Shared libraries packaged by your distribution are usually in /usr/lib. To keep manually installed libraries separate, they are usually installed in /usr/local/lib. If you installed libgit2 using the default installation procedure (e.g. without specifying CMAKE_INSTALL_PREFIX), you probably installed it under /usr/local/lib. On some distributions (e.g. Ubuntu), /usr/local/lib is not in the linker's default search path (see the ld man page for details), and you will get errors like:

$ python -c 'import pygit2'
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "pygit2/__init__.py", line 29, in <module>
    from _pygit2 import *
ImportError: libgit2.so.0: cannot open shared object file: No such file or directory

The following recipe shows how to install libgit2 and pygit2 on these systems. First, download and install libgit2 (following the instructions in the libgit2 README.md):

$ git clone git://github.com/libgit2/libgit2.git
$ mkdir libgit2/build
$ cd libgit2/build
$ cmake ..
$ cmake --build .
  $ sudo cmake --build . --target install
$ cd ../..

Now, download and install pygit2. You will probably have to set the LIBGIT2 environment variable so the compiler can find the libgit2 headers and libraries:

$ git clone git://github.com/libgit2/pygit2.git
$ cd pygit2
$ export LIBGIT2="/usr/local"
$ export LDFLAGS="-Wl,-rpath='$LIBGIT2/lib',--enable-new-dtags $LDFLAGS"
$ python setup.py build
$ sudo python setup.py install

This compiles the pygit2 libraries with a RUNPATH, which bakes extra library search paths directly into the binaries (see the ld man page for details). With RUNPATH compiled in, you won't have to use LD_LIBRARY_PATH. You can check to ensure RUNPATH was set with readelf:

$ readelf --dynamic build/lib.linux-x86_64-3.2/_pygit2.cpython-32.so | grep PATH
 0x000000000000000f (RPATH)              Library rpath: [/usr/local/lib]
 0x000000000000001d (RUNPATH)            Library runpath: [/usr/local/lib]

Building on Windows

pygit2 expects to find the libgit2 installed files in the directory specified in the LIBGIT2 environment variable.

In addition, make sure that libgit2 is build in "__cdecl" mode. The following recipe shows you how to do it, assuming you're working from a bash shell:

$ export LIBGIT2=C:/Dev/libgit2
$ git clone git://github.com/libgit2/libgit2.git
$ cd libgit2
$ mkdir build
$ cd build
$ cmake .. -DSTDCALL=OFF -DCMAKE_INSTALL_PREFIX=$LIBGIT2 -G "Visual Studio 9 2008"
$ cmake --build . --config release --target install
$ ctest -v

At this point, you're ready to execute the generic pygit2 installation steps described above.

Usage

The repository

Everything starts by opening an existing repository:

>>> from pygit2 import Repository
>>> repo = Repository('pygit2/.git')

Or by creating a new one:

>>> from pygit2 import init_repository
>>> bare = False
>>> repo = init_repository('test', bare)

These are the basic attributes of a repository:

Repository.path    -- path to the Git repository
Repository.workdir -- path to the working directory, None in the case of
                      a bare repo

Git objects

In the first place Git is a key-value storage system. The values stored are called objects, there are four types (commits, trees, blobs and tags), for each type pygit2 has a Python class:

# Get the last commit
>>> head = repo.head

# Show commits and trees
>>> commit
<pygit2.Commit object at 0x7f9d2f3000b0>
>>> commit.tree
<pygit2.Tree object at 0x7f9d2f3000f0>

These four classes (Commit, Tree, Blob and Tag) inherit from the Object base class, which provides shared behaviour. A Git object is identified by a unique object id, which is a binary byte string; this is often represented as an hexadecimal text string:

>>> commit.oid
b'x\xde\xb5W\x8d\x01<\xdb\xdf\x08o\xa1\xd1\xa3\xe7\xd9\x82\xe8\x88\x8f'
>>> commit.hex
'78deb5578d013cdbdf086fa1d1a3e7d982e8888f'

The API of pygit2 accepts both the raw object id and its hexadecimal representation, the difference is done based on its type (a byte or a text string).

This is the common interface for all Git objects:

Object.type       -- one of the GIT_OBJ_COMMIT, GIT_OBJ_TREE,
                     GIT_OBJ_BLOB or GIT_OBJ_TAG constants
Object.oid        -- the object id, a byte string 20 bytes long
Object.hex        -- hexadecimal representation of the object id, a text
                     string 40 chars long
Object.read_raw() -- returns the byte string with the raw contents of the
                     of the object

Objects can not be modified once they have been created.

Commits

A commit is a snapshot of the working dir with meta informations like author, committer and others.:

Commit.author    -- the author of the commit
Commit.committer -- the committer of the commit
Commit.message   -- the message, a text string
Commit.tree      -- the tree object attached to the commit
Commit.parents   -- the list of parent commits

Signatures

The author and committer attributes of commit objects are Signature objects:

>>> commit.author
<pygit2.Signature object at 0x7f75e9b1f5f8>

This is their interface:

Signature.name   -- person's name
Signature.email  -- person's email address
Signature.time   -- unix time
Signature.offset -- offset from utc in minutes

Creating commits

Commits can be created by calling the create_commit method of the repository with the following parameters:

>>> author = Signature('Alice Author', 'alice@authors.tld')
>>> committer = Signature('Cecil Committer', 'cecil@committers.tld')
>>> tree = repo.TreeBuilder().write()
>>> repo.create_commit(
... 'refs/heads/master', # the name of the reference to update
... author, committer, 'one line commit message\n\ndetailed commit message',
... tree, # binary string representing the tree object ID
... [] # list of binary strings representing parents of the new commit
... )
'#\xe4<u\xfe\xd6\x17\xa0\xe6\xa2\x8b\xb6\xdc35$\xcf-\x8b~'

Trees

A tree is a sorted collection of tree entries. It is similar to a folder or directory in a file system. Each entry points to another tree or a blob. A tree can be iterated, and partially implements the sequence and mapping interfaces:

# Number of entries
>>> tree = commit.tree
>>> len(tree)
6

# Iteration
>>> for entry in tree:
...     print(entry.hex, entry.name)
...
7151ca7cd3e59f3eab19c485cfbf3cb30928d7fa .gitignore
c36f4cf1e38ec1bb9d9ad146ed572b89ecfc9f18 COPYING
32b30b90b062f66957d6790c3c155c289c34424e README.md
c87dae4094b3a6d10e08bc6c5ef1f55a7e448659 pygit2.c
85a67270a49ef16cdd3d328f06a3e4b459f09b27 setup.py
3d8985bbec338eb4d47c5b01b863ee89d044bd53 test

# Get an entry by name
>>> entry = tree['pygit2.c']
>>> entry
<pygit2.TreeEntry object at 0xcc10f0>

# Get the object the entry points to
>>> blob = repo[entry.oid]
>>> blob
<pygit2.Blob object at 0xcc12d0>

This is the interface of a tree entry:

TreeEntry.name        -- name of the tree entry
TreeEntry.oid         -- the id of the git object
TreeEntry.hex         -- hexadecimal representation of the oid
TreeEntry.filemode    -- the Unix file attributes
TreeEntry.to_object() -- returns the git object (equivalent to repo[entry.oid])

Diff

A diff shows the changes between trees, an index or the working dir:

# Diff two trees
>>> t0 = repo.head.tree
>>> t1 = repo.head.parents[0].tree
>>> diff = t1.diff(t0)
>>> diff

# Diff a tree with the index
>>> tree = repo.head.tree
>>> diff = tree.diff(repo.index)

# Diff a tree with the current working dir
>>> tree = repo.head.tree
>>> diff = tree.diff()

The interface for a diff:

Diff.changes          -- Dict of 'files' and 'hunks' for every change
Diff.patch            -- a patch for every changeset
Diff.merge            -- Merge two Diffs

Blobs

A blob is equivalent to a file in a file system.:

# create a blob out of memory
>>> oid  = repo.create_blob('foo bar')
>>> blob = repo[oid]

Blob.data -- the contents of the blob, a byte string

Tags

A tag is a static label for a commit. See references for more information.

References

Reference lookup:

>>> all_refs = repo.listall_references()
>>> master_ref = repo.lookup_reference("refs/heads/master")
>>> commit = repo[master_ref.oid]

Reference log:

>>> head = repo.lookup_reference('refs/heads/master')
>>> for entry in head.log():
...     print(entry.message)

The interface for RefLogEntry:

RefLogEntry.committer -- Signature of Committer
RefLogEntry.message   -- the message of the RefLogEntry
RefLogEntry.oid_old   -- oid of old reference
RefLogEntry.oid_new   -- oid of new reference

Revision parsing

You can use any of the fancy <rev> forms supported by libgit2:

>>> commit = repo.revparse_single('HEAD^')

Revision walking

You can iterate through the revision history with repo.walk:

>>> from pygit2 import GIT_SORT_TIME
>>> for commit in repo.walk(oid, GIT_SORT_TIME):
...     print(commit.hex)

The index file

Index read:

>>> index = repo.index
>>> index.read()
>>> oid = index['path/to/file'].oid    # from path to object id
>>> blob = repo[oid]                   # from object id to object

Iterate over all entries of the index:

>>> for entry in index:
...     print entry.path, entry.hex

Index write:

>>> index.add('path/to/file')          # git add
>>> del index['path/to/file']          # git rm
>>> index.write()                      # don't forget to save the changes

Status

Inspect the status of the repository:

>>> from pygit2 import GIT_STATUS_CURRENT
>>> status = repo.status()
>>> for filepath, flags in status.items():
...     if flags != GIT_STATUS_CURRENT:
...         print "Filepath %s isn't clean" % filepath

Contributing

Fork libgit2/pygit2 on GitHub, make it awesomer (preferably in a branch named for the topic), send a pull request.

TODO

See issues

Authors

The following people have contributed at least one patch to the pygit2 project (sorted alphabetically):

• Amit Bakshi
• András Veres-Szentkirályi
• Benjamin Kircher
• Bryan O'Sullivan
• Carlos Martín Nieto
• Christian Boos
• David Borowitz (Original author)
• David Versmisse
• Erik van Zijst
• Han-Wen Nienhuys
• Hugh Cole-Baker
• J David Ibáñez (Current maintainer)
• Jared Flatow
• John Szakmeister
• Josh Bleecher Snyder
• Julien Miotte
• Martin Lenders
• Nico von Geyso
• Petr Hosek
• Petr Viktorin
• Richo Healey
• Rui Abreu Ferreira
• Sarath Lakshman
• Sebastian Thiel
• Vicent Marti
• Yonggang Luo
• Zoran Zaric
• pistacchio

License

GPLv2 with linking exception. See COPYING for more details.