Get the total number of results for a string match over time from a git repository. The example here uses 'def test_' to fetch time series data for the addition of unit tests to a code base.

git-sum-string-over-time.py

#!/usr/bin/env python2.7
import itertools
import subprocess
import time

def soi(s):
    #if isinstance(s, int):
    #    return s
    return int(s.split(':')[-1].rstrip())

def count_pattern(pattern):
    old_dates = set()

    p = subprocess.Popen(['git', 'log', '-S', pattern],
                         stdout=subprocess.PIPE)

    for line in p.stdout:
        if not line.startswith('commit'):
            continue

        commit = line.split(' ')[-1].rstrip()

        # Get the date -- usually only 2-3 lines following
        for tries in range(5):
            line = p.stdout.next()
            if line.startswith('Date:'):
                date = line.split(':', 1)[-1].lstrip().rstrip()
                tries = 0
                break
        if tries > 0:
            raise Exception("Date not found for commit.")

        time.sleep(.05)
        # Git-grep cannot take a ref-spec, so checkout the commit.
        subprocess.Popen(['git', 'checkout', '-f', commit],
                         stdout=subprocess.PIPE,
                         stderr=subprocess.PIPE)

        # Get pattern matches
        m = subprocess.Popen(['git', 'grep', '-c', pattern],
                             stdout=subprocess.PIPE)
        matches = sum(itertools.imap(soi, m.stdout))

        # This should really apply the timezone, but we don't
        # really need that level of precision here...
        ts = time.strptime(date.rsplit(' ', 1)[0])

        # year / mon / day formatting (easier to sort)
        local_date = "%s/%s/%s" % (ts.tm_year, ts.tm_mon, ts.tm_mday)

        if local_date in old_dates:
            continue
        old_dates.add(local_date)

        print "%s,%s" % (local_date, matches)

count_pattern('def test_')