How to create a legend from your data

Matplotlib can easily generate a legend for our axes. It does so by using the label= argument in the .bar() call (or any other axes-generating plot call):

import matplotlib.pyplot as plt
from data import polls

poll_titles = [poll[0] for poll in polls]
poll_men = [poll[1] for poll in polls]
poll_women = [poll[2] for poll in polls]

poll_x_coordinates = range(len(polls))

figure = plt.figure(figsize=(6, 6))
figure.subplots_adjust(bottom=0.35)
axes = figure.add_subplot()

axes.bar(
    poll_x_coordinates,
    poll_men,
    label="Men"
)
axes.bar(
    poll_x_coordinates,
    poll_women,
    bottom=poll_men,
    label="Women"
)

plt.xticks(poll_x_coordinates, poll_titles, rotation=30, ha="right")
plt.show()

Next up, we tell the axes to draw a legend using the data available:

import matplotlib.pyplot as plt
from data import polls

poll_titles = [poll[0] for poll in polls]
poll_men = [poll[1] for poll in polls]
poll_women = [poll[2] for poll in polls]

poll_x_coordinates = range(len(polls))

figure = plt.figure(figsize=(6, 6))
figure.subplots_adjust(bottom=0.35)
axes = figure.add_subplot()

axes.bar(
    poll_x_coordinates,
    poll_men,
    label="Men"
)
axes.bar(
    poll_x_coordinates,
    poll_women,
    bottom=poll_men,
    label="Women"
)

axes.legend()

plt.xticks(poll_x_coordinates, poll_titles, rotation=30, ha="right")
plt.show()

That was the first and simplest way of defining the legend. However, sometimes it can be useful to manually assign a legend item to individual plots. We can do this by first creating variables for each plot that we want to include in the legend. We don't need the label= argument for this:

import matplotlib.pyplot as plt
from data import polls

poll_titles = [poll[0] for poll in polls]
poll_men = [poll[1] for poll in polls]
poll_women = [poll[2] for poll in polls]

poll_x_coordinates = range(len(polls))

figure = plt.figure(figsize=(6, 6))
figure.subplots_adjust(bottom=0.35)
axes = figure.add_subplot()

men_plot = axes.bar(
    poll_x_coordinates,
    poll_men
)
women_plot = axes.bar(
    poll_x_coordinates,
    poll_women,
    bottom=poll_men
)

plt.xticks(poll_x_coordinates, poll_titles, rotation=30, ha="right")
plt.show()

Then we can call axes.legend() but with some arguments: the plots and a label for each plot:

import matplotlib.pyplot as plt
from data import polls

poll_titles = [poll[0] for poll in polls]
poll_men = [poll[1] for poll in polls]
poll_women = [poll[2] for poll in polls]

poll_x_coordinates = range(len(polls))

figure = plt.figure(figsize=(6, 6))
figure.subplots_adjust(bottom=0.35)
axes = figure.add_subplot()

men_plot = axes.bar(
    poll_x_coordinates,
    poll_men
)
women_plot = axes.bar(
    poll_x_coordinates,
    poll_women,
    bottom=poll_men
)

axes.legend((men_plot, women_plot), ("Men", "Women"))

plt.xticks(poll_x_coordinates, poll_titles, rotation=30, ha="right")
plt.show()

Doing this means that the legend is defined in one place, which might be preferrable in terms of readability.