Little script to get a time series of code churn (insertions, deletions) on a git repo

git_code_churn.py

from dateutil import parser
import subprocess
import os
import re
import sys

import numpy as np

from pandas import *

repo_path = '/home/wesm/code/pandas'
githist = ('git log --pretty=format:\"%h %ad | %s%d [%an]\" --date=short ' +
           repo_path + ' > githist.txt')

def rungithist():
    os.system(githist)

def get_commit_history():
    # return TimeSeries

    rungithist()

    githist = open('githist.txt').read()
    os.remove('githist.txt')

    sha_date = []
    for line in githist.split('\n'):
        sha_date.append(line.split()[:2])

    shas, dates = zip(*sha_date)

    dates = [parser.parse(d) for d in dates]

    return Series(dates, shas)

def get_commit_churn(sha, prev_sha):
    stdout = subprocess.Popen(['git', 'diff', sha, prev_sha, '--stat'],
                              stdout=subprocess.PIPE).stdout

    statline = stdout.read().split('\n')[-2]

    match = re.match('.*\s(.*)\sinsertions.*\s(.*)\sdeletions', statline)

    insertions = int(match.group(1))
    deletions = int(match.group(2))

    return insertions, deletions


def get_code_churn(commits):
    shas = commits.index

    prev = shas[0]

    insertions = [np.nan]
    deletions = [np.nan]

    for cur in shas[1:]:
        i, d = get_commit_churn(cur, prev)
        insertions.append(i)
        deletions.append(d)

        prev = cur

    return DataFrame({'insertions' : insertions,
                      'deletions' : deletions}, index=shas)

if __name__ == '__main__':
    commits = get_commit_history()
    churn = get_code_churn(commits)

    by_date = churn.groupby(commits).sum()

    # clean out days where I touched Cython

    by_date = by_date[by_date.sum(1) < 5000]