Python script for parsing lines changed per commit for a git repository and saving as time-series data in a CSV file

git_stats_csv.py

import argparse
import subprocess
import re
import csv

from datetime import datetime
from zoneinfo import ZoneInfo
from enum import Enum
from operator import add

# Pattern for paths and files to include, use git pathspec
pathspec = '*.tex'

# Timezone to parse dates and Unix timestamps in
timezone_name = "Europe/Helsinki"

# For pretty logging
class bcolors(Enum):
    WARNING = '\033[93m'
    ENDC = '\033[0m'

# CLI arguments
parser = argparse.ArgumentParser(description="Parses lines of code changed for all commits in current branch to a CSV file")
parser.add_argument('-s', '--start', dest='start_date', type=datetime.fromisoformat,
                    help='Start date (ISO 8601 -like format, e.g. "2022-06-01 14:00")')
parser.add_argument('-d', '--sum-day', dest='sum_day', action=argparse.BooleanOptionalAction,
                    help='Sum the commits of each day and remove time from dates (also removes time zones)')
parser.add_argument('--strip-timezone', dest='strip_timezone', action=argparse.BooleanOptionalAction,
                    help='Strip time zones from the output CSV')
parser.add_argument('-o', '--output', dest='output_path', type=str,
                    help='Path to output file, e.g. "data.csv"')

args = parser.parse_args()

timez = ZoneInfo(timezone_name)
start_date: datetime = args.start_date.astimezone(timez)
sum_day: bool = args.sum_day
strip_timezone: bool = args.strip_timezone
output_path: str = args.output_path

print(f'Start date: {args.start_date}')

# Get a reference count from rev-list so that we can compare commit count parsed from git log
rev_list_cmd = f"git rev-list --count HEAD '{pathspec}'"
rev_list = subprocess.run(rev_list_cmd, shell=True, check=True, capture_output=True, text=True)

log_cmd = f"git log --stat --format=\"%at\" '{pathspec}'"
logs = subprocess.run(log_cmd,
                   shell=True, check=True, capture_output=True, text=True)

pattern = re.compile(r'\n(?=\d{10})', re.MULTILINE)

commits = re.split(pattern, logs.stdout)
print(f'Commits: {len(commits)}')

if int(rev_list.stdout) != len(commits):
    print(f'{bcolors.WARNING}WARNING: Commit count from `{rev_list_cmd}` and parsed commits from `{log_cmd}` do not match!{bcolors.ENDC}')

timestamp_pattern = r'\d{10,}'
change_pattern = r'\d+ files? changed,( (?P<insertions>\d+) insertions?\(\+\))?,?( (?P<deletions>\d+) deletions?\(\-\))?'

for i, val in enumerate(commits):
    timestamp = datetime.fromtimestamp(int(re.match(timestamp_pattern, val)[0]), timez)
    match = re.search(change_pattern, val)
    insertions = int(match.group('insertions')) if match.group('insertions') else 0
    deletions = int(match.group('deletions')) if match.group('deletions') else 0
    # print(f'{timestamp.strftime("%Y-%m-%d %H:%M %z")}, insertions: {insertions}, deletions: {deletions}')
    commits[i] = [timestamp, insertions, deletions]

if args.start_date:
    commits = list(filter(lambda row: row[0] > start_date, commits))

print(f'Filtered commits: {len(commits)}')

if sum_day:
    new_commits = {}
    for commit in commits:
        date = commit[0].date()
        new_commits[date] = list(map(add, new_commits.get(date, [0, 0]), commit[1:3]))
    commits = [[date, insertions, deletions] for date, (insertions, deletions) in new_commits.items()]
elif strip_timezone:
    for commit in commits:
        commit[0] = commit[0].replace(tzinfo=None)

header = ['Datetime', 'Insertions', 'Deletions']

with open(output_path, 'w', encoding='UTF8', newline='') as f:
    writer = csv.writer(f)
    writer.writerow(header)
    writer.writerows(commits)