git_commit_graph.py

from scipy.interpolate import interp1d
import numpy as np

import shlex
import subprocess as sp

import prettyplotlib as ppl
from matplotlib import pyplot as plt
from matplotlib.ticker import FuncFormatter
import io
from pandas import read_csv

cmd = "git log --no-merges --date=short --pretty='format:%ad\t%H\t%aN'"
p = sp.Popen(shlex.split(cmd), stdout=sp.PIPE)
# p.wait() will likely just hang if the log is long enough because the
# stdout buffer will fill up
stdout, _ = p.communicate()

table = read_csv(io.StringIO(stdout.decode('utf-8')), sep='\t',
                 names=['date', 'hash', 'author'], index_col=0,
                 parse_dates=True).sort()
table = table.to_period(freq='W')
commits_per_period = table.hash.groupby(level=0).aggregate(len)

dates = [p.start_time.date() for p in commits_per_period.index]
ncommits = commits_per_period.values

fn = interp1d(range(len(dates)), ncommits, 'cubic', bounds_error=False)

fig, ax = plt.subplots(1)
fig.set_size_inches((8, 2))
x = np.linspace(0, len(dates), 1000)
ppl.fill_between(x, 0, fn(x))
ax.set_title('Number of commits to Astropy')
ax.set_xlim(0, len(dates))
ax.set_ylim(0, max(ncommits) + 0.1 * max(ncommits))
ax.xaxis.set_ticks(np.linspace(0, len(dates) - 1, 8)[1:-1])

def formatter(x, p):
    if x >= len(dates):
        return ''
    return dates[int(x)].strftime('%b %Y')

formatter = FuncFormatter(formatter)
ax.xaxis.set_major_formatter(formatter)
ax.xaxis.tick_bottom()
ax.yaxis.tick_left()
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)

# There must be a btter way to do this...
plt.setp(plt.xticks()[1], rotation=30)
fig.savefig('graph.png', bbox_inches='tight')