Plot the distribution of commit frequency for multiple submissions (WIP)

analyse_grades.py

# WIP! Have not tested with multiple submissions yet!

from pathlib import Path
import warnings

import matplotlib.pyplot as plt
import pandas as pd


def get_dates_from_file(dates_file) -> pd.Series:
    # Is this the best way? Should the index values be strings?
    timestamps = pd.read_csv(dates_file, header=None, names=["timestamp"])
    timestamps.timestamp = pd.to_datetime(timestamps.timestamp)
    timestamps["date"] = timestamps.timestamp.apply(lambda x: f"{x:%Y-%m-%d}")
    commits_per_day = timestamps.groupby("date").size()
    return commits_per_day


if __name__ == "__main__":
    # Before running this, we need to generate the dates for each repository.
    # Specifically we need `git log --format="%ci" > dates`
    # to be run in the right directory for each submission,
    # and moved up to the "top level" of the submission

    # Create a DataFrame to hold the number of commits per day for all students.
    # Indexed by date, each column corresponds to a student.
    all_numbers = pd.DataFrame()

    for submission in submission_ids:
        # Run git log if needed or retrieve file
        student_file = Path(f"Participant_{submission}_assignsubmission_file_") / "dates"
        # Assuming everyone has a repo...
        commits_per_day = get_dates_from_file(student_file)
        commits_per_day.name = submission
        try:
            all_numbers = all_numbers.join(commits_per_day, how="outer", sort=True)
        except:
            warnings.warn(f"Could not merge data for student {submission}")
    # This will leave NaNs for dates where no data existed, we'd prefer 0
    all_numbers = all_numbers.fillna(0).astype(int)

    # Plot
    # f = plt.figure()
    # ax = f.add_subplot(1, 1, 1)
    # ax.imshow([[c[1] for c in commits_per_day.items()]])
    # ax.set_aspect(20)
    # f.savefig("heatmap.png")
    plt.imshow(all_numbers.to_numpy())  # needs some massaging
    plt.savefig("heatmap.png")