Dynamically add legend for arrows in matplotlib

Question

import matplotlib.pyplot as plt
def plot_arrow(arrow_type):
    if arrow_type == 'good_arrow':
        arrow_color = 'g'
    ar = plt.arrow(x, y, dx , dy, label=arrow_type, fc=arrow_color)
    plt.legend([ar,], [arrow_type,])

The above callback function is used to draw arrows in a plot. I need to add legend to them. These arrows are drawn dynamically. The above code will have only one legend. I tried solutions like this but those do not work for matplotlib.pyplot.arrow

Edit:

Problem: I have two colored arrows but only one legend

Plot example:

don't draw the legend in this function. instead, return the artists and create the legend in the code block that is calling this function. — Paul H, Aug 03 '18 at 21:53
Is the problem to create a legend for an arrow (actually showing an arrow instead of a rectangle)? Or is it to update a legend? — ImportanceOfBeingErnest, Aug 03 '18 at 22:03
the problem is to create a new legend. I have two types of arrows but only one legend — user009122, Aug 03 '18 at 22:04
You only have one legend entry because you only provide one legend handle to `legend`. To create a legend with two handles, use e.g. `plt.legend([ar, ar2], [arrow_type1, arrow_type2])`. This would require you to store the arrows `ar` and `ar2` somewhere, e.g. in a list outside the callback function. — ImportanceOfBeingErnest, Aug 03 '18 at 22:10
@ImportanceOfBeingErnest Hmm. I tried avoiding that. This would be called over 1000 times. However I would have only a few colors (5 to 10). Also, I would have multiple arrows of the same color. Would maintaining list of arrows draw a legend for each arrow and not each color? I have updated the image — user009122, Aug 03 '18 at 22:20
You would need some logic to produce the list of handles you want to appear in the legend. `numpy.unique()` can come handy for such purposes. — ImportanceOfBeingErnest, Aug 03 '18 at 22:28
Okay. So storing the list of handles is probably the only way? — user009122, Aug 03 '18 at 22:30
No, you can in principle recover them from the axes each time the callback is called. That is for sure much more complicated and expensive though, so I did not propose it. — ImportanceOfBeingErnest, Aug 03 '18 at 22:55

Gilfoyle · Accepted Answer · 2018-08-04T07:53:23.467

2

Why not just use quiver() in your case? See here.

import matplotlib.pyplot as plt
import numpy as np

plt.figure()

def rand(n):
    return 2.0*np.random.rand(n)-1.0

def plt_arrows(n, c, l, xpos, ypos):
    X, Y = rand(n), rand(n)
    U, V = 4.*rand(n), 4.*rand(n)
    Q = plt.quiver(X, Y, U, V, angles="xy", scale=30, label=l, color=c)
    qk = plt.quiverkey(Q, xpos, ypos, 2, label=l, labelpos='N', labelcolor=c)

plt_arrows(20, "g", "good", 1.05, 0.9)
plt_arrows(20, "r", "bad", 1.05, 0.8)

plt.show()

edited Aug 04 '18 at 07:53

answered Aug 03 '18 at 22:13

Gilfoyle

3,282
3
47
83

Thanks Samuel. I would have arrows in the order of 100. So I would want a legend for a color as opposed to each arrow. And also I was trying to avoid storing handles. My function is a callback – user009122 Aug 03 '18 at 22:22

Dynamically add legend for arrows in matplotlib

1 Answers1