dulwich/README

This is the dulwich project.

It aims to give an interface to git repos that doesn't call out to git
directly. It is probably going to be implemented in pure python.

Currently can read blobs, trees and commits from the files. It reads both
legacy and new headers. However it is untested for anything but the simple
case.

Can also understand a little about the repository format.

The testsuite uses the nosetests program from Turbogears, as I got annoyed
trying to set up unittest.

Open up a repo by passing it the path to the .git dir. You can then ask for
HEAD with repo.head() or a ref with repo.ref(name). Both return the SHA id
they currently point to. You can then grab this object with
repo.get_object(sha).

For the actual objects the ShaFile.from_file(filename) will return the object
stored in the file whatever it is. To ensure you get the correct type then
call {Blob,Tree,Commit}.from_file(filename). I will add repo methods to do
this for you with file lookup soon.

There is also support for creating blobs. Blob.from_string(string) will create
a blob object from the string. You can then call blob.sha() to get the sha
object for this blob, and hexdigest() on that will get its ID. There is
currently no method that allows you to write it out though.

Everything is currently done with assertions, where much of it should probably
be exceptions. This was merely done for expediency. If you hit an assertion,
it either means you have done something wrong, there is corruption, or
you are trying an unsupported operation.

If you have any comments or questions you can contact me at 
jw+debian@jameswestby.net

James Westby