It has classes for the index and data parts. It supports lookup of an object name in the index, and then access to the object in the data part using the offset returned from the index lookup.
There are many problems with it so far.
* The mmap in python has no offset, so the whole files are mapped. * There is no support for delta objects. * There is no consistency checking. * The code is not hooked up to provide a simple API. * The code is not hooked in to the repo, so that objects are still not retrieved from packs.