Ben Dichter

Ben Dichter

Research Software Engineer and Founder of CatalystNeuro.

Git Timesheet

2 minute read

Published:

I recently faced a situation where I needed to assess the amount of work done by each member of a team on a project that has spanned over a year. That project has a git repo, and I could see when each person made a commit. I decided to break it down by weeks. Whenever a person submitted any commit to the repo on any branch, I counted them as working on the project for that week. Of course this is imperfect- someone could work a lot and make no commits for that week and someone could have submitted a commit but might have worked very little. Still, this seems like the most fair way to assess work I could think of.

The code will work on any locally cloned git repo. skip allows you to remove contributors, and is ideal for handling bots. author_map allows you to tranform handles. This is ideal if members of your team make some contributions through PRs from a clone of the repo and some of their PRs through GitHub directly, or if they have multiple usernames.

import os
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import tqdm
import datetime
import matplotlib

def git_timesheet(git_dir, skip=None, author_map=None):
    
    if skip is None:
        skip = [
            "dependabot[bot]",
            "!git for-each-ref --format='%(refname:short)' `git symbolic-ref HEAD`",
        ]
    
    if author_map is None:
        author_map = dict()

    os.system(f"git --git-dir {git_dir}/.git log --all --numstat --pretty=format:'--%h--%ad--%aN' --no-renames > git.log")

    commits = pd.read_csv("git.log", sep="\u0012", header=None, names=['raw'])

    commit_marker = commits[commits['raw'].str.startswith("--",na=False)]
    commit_info = commit_marker['raw'].str.extract(r"^--(?P<sha>.*?)--(?P<date>.*?)--(?P<author>.*?)$", expand=True)
    commit_info['date'] = pd.to_datetime(commit_info['date'])

    file_stats_marker = commits[~commits.index.isin(commit_info.index)]
    file_stats = file_stats_marker['raw'].str.split("\t", expand=True)
    file_stats = file_stats.rename(columns={0: "insertions", 1: "deletions", 2: "filename"})
    file_stats['insertions'] = pd.to_numeric(file_stats['insertions'], errors='coerce')
    file_stats['deletions'] = pd.to_numeric(file_stats['deletions'], errors='coerce')

    commit_data = commit_info.reindex(commits.index).fillna(method="ffill")
    commit_data = commit_data[~commit_data.index.isin(commit_info.index)]
    commit_data = commit_data.join(file_stats)

    # get total authors and weeks
    all_authors = commit_data["author"].unique()
    all_authors = list(np.unique([author_map.get(x, x) for x in all_authors if x not in skip]))
    
    dates = commit_data["date"]
    start = dates.min()
    stop = dates.max()

    n_weeks = (stop-start).days // 7

    timesheet = np.zeros((len(all_authors), n_weeks))

    # iterate over commits and timesheet per week
    for week_n in tqdm.trange(n_weeks):
        week_start = start + datetime.timedelta(7 * (week_n-1))
        week_stop = start + datetime.timedelta(7 * week_n)
        commit_data_for_week = commit_data[(week_start < commit_data["date"]) & (commit_data["date"] < week_stop)]
        authors_for_week = commit_data_for_week["author"].unique()
        # handle different usernames
        authors_for_week = list(np.unique([author_map.get(x, x) for x in authors_for_week]))
        for i, author in enumerate(all_authors):
            if author in authors_for_week:
                timesheet[i, week_n] = 1

    fig, ax = plt.subplots(figsize=(30, 10))
    ax.imshow(timesheet, cmap="Greys")
    ax.set_yticks(range(len(all_authors)))
    _ = ax.set_yticklabels(all_authors)
    ax.set_xlabel("weeks")

    plt.minorticks_on()
    plt.gca().xaxis.set_minor_locator(matplotlib.ticker.MultipleLocator(1))
    plt.gca().yaxis.set_minor_locator(matplotlib.ticker.MultipleLocator(1))
    plt.grid(which="both", linewidth=0.25, color="k")

I developed a function for parsing the git log and creating a visualization of weeks worked by each member. The repo I used this for is private, so I will demonstrate it on a separate repo from CatalystNeuro that is public.

git_timesheet(
    "path/to/nwb-conversion-tools",
    author_map={
        "bendichter": "Ben Dichter",
        "luiz": "Luiz Tauffer",
        "luiztauffer": "Luiz Tauffer",
        "CodyCBakerPhD": "Cody Baker",
        "h-mayorquin": "Heberto Mayorquin",
        "sbuergers": "Steffen Bürgers",
        "weiglszonja": "Szonja Weigl",
    }
)

git-timesheet

You May Also Enjoy

The new Code Crafter package for Python AST transformations

2 minute read

Published:

I built a new tool! Code Crafter is a Python library designed for manipulating Python source code through Abstract Syntax Tree (AST) transformations. This tool simplifies the process of programmatically editing Python code, allowing developers to find and modify specific data structures such as lists, dictionaries, and sets within their code.

Stacked Step Plot

1 minute read

Published:

Matplotlib provides a stackplot, which stacks area, and a step plot, which provides steps, but there is no stacked step plot. This is useful when you have some accumulating resource that have different types that begin and end all at once. I personally had a need for this when I wanted to visualize my man-month commitment to my company’s funded projects over the next 7 years.

Grouped Bar Plot

1 minute read

Published:

I often have a need to plot a grouped bar plot. Matplotlib provides this example, which is helpful, but not quite generalizable enough for my needs, as it only shows how to group 2 categories together. Here is a generalization of that tutorial that was very helpful for me and I hope is helpful for others as well.

Spiral Plot

2 minute read

Published:

Let’s use for example Google Trends results for the search term “gifts.” Google offers this plot: