100% stacked area/histogram in matplotlib with dates on X axis

Question

I would like to take the best of this and this questions. Namely, I have a DataFrame that contains the test name, date of execution and outcome. And I want to showcase how the percentage of failed cases decreased over time.

My data looks like this:

TestName;Date;IsPassed
test1;12/8/2016 9:44:30 PM;0
test1;12/8/2016 9:39:00 PM;0
test1;12/8/2016 9:38:29 PM;1
test1;12/8/2016 9:38:27 PM;1
test2;12/8/2016 5:05:02 AM;1
test3;12/7/2016 8:58:36 PM;0
test3;12/7/2016 8:57:19 PM;1
test3;12/7/2016 8:56:15 PM;1
test4;12/5/2016 6:50:49 PM;0
test4;12/5/2016 6:49:50 PM;0
test4;12/5/2016 3:23:09 AM;1
test4;12/4/2016 11:51:29 PM;1

And I was using this code to plot the cases separately:

fig, ax = plt.subplots()
passed = tests[tests.IsPassed == 1]
failed = tests[tests.IsPassed == 0]
passed_dates = mdates.date2num(passed.Date.astype(datetime))
failed_dates = mdates.date2num(failed.Date.astype(datetime))
ax.hist(passed_dates, bins=10, color='g')
ax.hist(failed_dates, bins=10, color='r')
ax.xaxis.set_major_locator(mdates.AutoDateLocator())
ax.xaxis.set_major_formatter(mdates.DateFormatter('%d.%m.%y'))
plt.show()

But now I would like to

1. Divide the time span into a configurable amount of buckets
2. Count the amount of test runs per bucket (without for loops, as there is a lot of entries in the dataframe)
3. Plot either a 100% area chart or the stacked histogram per each bucket, so that the amount from step 2 is 100%

The problem for me right now is that the perfectly working solution with the hist() takes care of summing up automatically, and I don't see a way to pass the Y axis to it.

Update

Here is what I'd like to accomplish (taken from another source):

Answer 1

Using the argument stacked = True allows you to provide several arrays as input to plt.hist.

ax.hist([passed_dates, failed_dates], bins=10, stacked=True, label=["passed", "failed"])

Using relative counts requires to divide by the number of absolute counts per bin. This functionality is not built into the hist function. You would need to calculate the histograms manually and then plot the result as stacked bar plots.

from __future__ import division
import matplotlib.pyplot as plt
import matplotlib.dates
import datetime
import numpy as np
import pandas as pd

dates = pd.date_range("2016/01/01","2016/06/01" )
dates2 = pd.date_range("2016/02/01","2016/03/17", freq="18H")
dates = dates.append(dates2)

passed = np.round(np.random.rand(len(dates))+0.231).astype(np.int8)
tests = pd.DataFrame({"Date" : dates, "IsPassed": passed})

fig, ax = plt.subplots()
passed = tests[tests.IsPassed == 1]
failed = tests[tests.IsPassed == 0]
all_dates = matplotlib.dates.date2num(tests.Date.astype(datetime.datetime))
passed_dates = matplotlib.dates.date2num(passed.Date.astype(datetime.datetime))
failed_dates = matplotlib.dates.date2num(failed.Date.astype(datetime.datetime))

hist, bins = np.histogram(all_dates, bins=10)
histpassed, bins_ = np.histogram(passed_dates, bins=bins)
histfailed, bins__ = np.histogram(failed_dates, bins=bins)

binwidth=bins[1]-bins[0]
ax.bar(bins[:-1]+binwidth/2., histpassed/hist, width=binwidth*0.8, label="passed")
ax.bar(bins[:-1]+binwidth/2., histfailed/hist, width=binwidth*0.8, bottom=histpassed/hist, label="failed")

ax.xaxis.set_major_locator(matplotlib.dates.AutoDateLocator())
ax.xaxis.set_major_formatter(matplotlib.dates.DateFormatter('%d.%m.%y'))
ax.legend()
fig.autofmt_xdate()
plt.savefig(__file__+".png")
plt.show()