matplotlib two legends out of bar plot

Question

My question is very similar to matplotlib two legends out of plot. There's an answer which works fine for common line plots.
I faced a problem with copying the solution for bar plots...
The problem is that in the given solution l1, l2, ... are <matplotlib.lines.Line2D and if i do the same trick for bar-plot, it cannot infer the colors...

Code:

import matplotlib.pyplot as plt
import numpy as np

bar_data_cost = np.random.rand(4,11)
bar_data_yield =  np.random.rand(4,11)
cmap_yield = plt.cm.Greens(np.linspace(0.2, 1, len(bar_data_cost)))
cmap_costs = plt.cm.Oranges(np.linspace(0.2, 1, len(bar_data_cost)))

fig, axes = plt.subplots(nrows=1, ncols=1, figsize=(20,8))

ax1 = axes
y_offset_yield = np.zeros(len(bar_data_yield[0]))
y_offset_cost = np.zeros(len(bar_data_cost[0]))
index1 = np.arange(len(bar_data_yield[1])) - 0.2
index2 = np.arange(len(bar_data_yield[1])) + 0.2

for row in range(len(bar_data_yield)):
    b1 = ax1.bar(left=index1, width=0.4, height=bar_data_yield[row], bottom=y_offset_yield, color=cmap_yield[row])
    y_offset_yield = bar_data_yield[row]

for row in range(len(bar_data_yield)):
    b2 = ax1.bar(left=index2, width=0.4, height=bar_data_cost[row], bottom=y_offset_cost, color=cmap_costs[row])
    y_offset_cost = bar_data_cost[row]


fig.legend(b1, grouped_dataset.index.levels[0], fontsize=16, loc="upper right")
fig.legend(b2, grouped_dataset.index.levels[0], fontsize=16, loc="center right")

But no `grouped_dataset` is defined. But I found your issue! Answering now... — Parfait, Aug 20 '18 at 18:11

score 4 · Accepted Answer · answered Aug 20 '18 at 18:20

Currently, your legend outputs only the last b1 and b2 from for loops since they are re-assigned with each iteration. In posted link, a tuple of lines are passed in first argument of legend. Hence, pass a list of b1 and list of b2 into legend calls after appending bars iteratively.

Below demonstrates with seeded data for reproducibility and substitutes your grouped_dataset.index.levels[0] as this is unknown from your post.

import matplotlib.pyplot as plt
import numpy as np

np.random.seed(82018)

...

b1_list = []
for row in range(len(bar_data_yield)):
    index1 = np.arange(len(bar_data_yield[row])) - 0.2
    b1_list.append(ax1.bar(left=index1, width=0.4, height=bar_data_yield[row], 
                           bottom=y_offset_yield, color=cmap_yield[row]))
    y_offset_yield = bar_data_yield[row]

b2_list = []
for row in range(len(bar_data_yield)):
    index2 = np.arange(len(bar_data_yield[row])) + 0.2
    b2_list.append(ax1.bar(left=index2, width=0.4, height=bar_data_cost[row], 
                           bottom=y_offset_cost, color=cmap_costs[row]))
    y_offset_cost = bar_data_cost[row]

fig.legend(b1_list, list('ABCD'), fontsize=16, loc="upper right")
fig.legend(b2_list, list('WXYZ'), fontsize=16, loc="center right")

plt.show()
plt.clf()
plt.close()

matplotlib two legends out of bar plot

1 Answers1