A Dumb Git Object Parser.

git-object-parser.py

#!/usr/bin/env python
#
#  git-obj-parser.py
#  ~~~~~~~~~~~~~~~~~
#  Run from a git work dir, creates a list of all objects with their
#  contents in a format like:
#
#  > BLOB  1b9f426a8407ffee551ad2993c5d7d3780296353
#  -----------------------------------------------
#  > TREE  098e6de29daf4e55f83406b49f5768df9bc7d624
#  BLOB    1b9f426a8407ffee551ad2993c5d7d3780296353        README
#  -----------------------------------------------
#

import os
import glob
import re
import zlib

BASE_PATH = os.path.join('.git', 'objects')
TYPES = ['blob', 'tree', 'commit']


def tree(obj):
    """Handle tree objects"""

    tree_pattern = re.compile(r'''
            (?P<octal>[0-9]+)      # Octal
            \s                     # Space
            (?P<file>[^\0]+)       # Filename
            \0                     # Null
            (?P<sha>.{20})         # SHA
        ''', re.VERBOSE)

    obj = re.split('tree \d+\0', obj)[1]
    matches = tree_pattern.finditer(obj)
    for match in matches:
        obj_type = 'tree' if match.group('octal') == '40000' else 'blob'
        print '{type}\t{sha}\t{name}'.format(
            type=obj_type.upper(),
            name=match.group('file'),
            sha=match.group('sha').encode('hex'))


def commit(obj):
    """Parse commit objects"""
    commit_pattern = re.compile(r'''
        (?P<type>(tree|parent))  # Either a tree or a parent object
        \s                       # Space
        (?P<sha>[a-f0-9]{40})    # SHA1
    ''', re.VERBOSE)

    obj = re.split(b'commit \d+\0', obj)[1]
    matches = commit_pattern.finditer(obj)
    for match in matches:
        print '{type}\t{sha}'.format(
            type=match.group('type').upper(), sha=match.group('sha'))


def blob(obj):
    """Handle blob objects"""
    # the content of blobs doesn't really matter
    pass


def main():
    for obj in glob.glob(os.path.join(BASE_PATH, '**', '*')):
        info = {}
        info['name'] = ''.join(obj[len(BASE_PATH):].split('/'))
        with open(obj, 'r') as f:
            message = zlib.decompress(f.read())

        for obj_type in TYPES:
            if message.startswith(obj_type):
                info['type'] = obj_type
                break

        if info['type'] == 'commit':
            print '> {}\t{}'.format('COMMIT', info['name'])
            commit(message)
        if info['type'] == 'tree':
            print '> {}\t{}'.format('TREE', info['name'])
            tree(message)
        if info['type'] == 'blob':
            print '> {}\t{}'.format('BLOB', info['name'])
        print '-----------------------------------------------'

if __name__ == '__main__':
    main()