Daily Add/Del Git Log Analysis

Question

I just finished writing this Python script to calculate daily additions and subtractions from my git log to use in making pretty graphs. This is a rewrite of something I wrote previously in Perl. I have made an attempt to clean it up, but I feel that some of my list comprehensions are messy to say the least, and I'm abusing certain Python features.

#!/usr/bin/env python
# get-pmstats.py
# Henry J Schmale
# November 25, 2017
# 
# Calculates the additions and deletions per day within a git repository
# by parsing out the git log. It opens the log itself.
# Produces output as a CSV
#
# This segments out certain file wildcards

import subprocess
from datetime import datetime
from fnmatch import fnmatch

def chomp_int(val):
    try:
        return int(val)
    except ValueError:
        return 0

def make_fn_matcher(args):
    return lambda x: fnmatch(''.join(map(str, args[2:])), x)

def print_results(changes_by_date):
    print('date,ins,del')
    for key,vals in changes_by_date.items():
        print(','.join(map(str, [key, vals[0], vals[1]])))

EXCLUDED_WILDCARDS = ['*.eps', '*.CSV', '*jquery*']

changes_by_date = {}
git_log = subprocess.Popen(
    'git log --numstat --pretty="%at"',
    stdout=subprocess.PIPE,
    shell=True)

date = None
day_changes = [0, 0]
for line in git_log.stdout:
    args = line.decode('utf8').rstrip().split()
    if len(args) == 1:
        old_date = date
        date = datetime.fromtimestamp(int(args[0]))
        if day_changes != [0, 0] and date.date() != old_date.date():
            changes_by_date[str(date.date())] = day_changes
            day_changes = [0, 0]
    elif len(args) >= 3:
        # Don't count changesets for excluded file types
        if True in map(make_fn_matcher(args), EXCLUDED_WILDCARDS):
            continue
        day_changes = [sum(x) for x in zip(day_changes, map(chomp_int, args[0:2]))]
print_results(changes_by_date)

Answer 1

I would apply the following improvements:

• use collections.defaultdict to count down the number of added and deleted lines separately. This would help to improve the counting logic and avoid having to check for the old date at all
• use any() to check for the excluded wildcards
• unpack args into added, deleted lines and a filename
• switch to gitpython from subprocess

The new version of the code:

from collections import defaultdict
from datetime import datetime
from fnmatch import fnmatch

import git


EXCLUDED_WILDCARDS = ['*.eps', '*.CSV', '*jquery*']


def chomp_int(val):
    try:
        return int(val)
    except ValueError:
        return 0


repo = git.Repo(".")
git_log = repo.git.log(numstat=True, pretty="%at").encode("utf-8")

added, deleted = defaultdict(int), defaultdict(int)

for line in git_log.splitlines():
    args = line.decode('utf8').rstrip().split()

    if len(args) == 1:
        date = datetime.fromtimestamp(int(args[0])).date()
    elif len(args) >= 3:
        added_lines, deleted_lines, filename = args

        # Don't count changesets for excluded file types
        if any(fnmatch(filename, wildcard) for wildcard in EXCLUDED_WILDCARDS):
            continue

        added[date] += chomp_int(added_lines)
        deleted[date] += chomp_int(deleted_lines)

for date, added_lines in added.items():
    print(date, added_lines, deleted[date])