如何避免饼图标签在 MatPlotLib ver.2.0.2 中重叠?

问题描述

There were a lot of questions posted regarding labels overlap for pie chart plotting. However, I couldn't find automated solution except for converting them to the legend. This solution doesn't work for me as I have a lot of values (around 60), and conversion to a legend will make the plot look very messy and unclear. So my question, if I want to label pie wedges around the pie, is the any automated solution for MatPlotLib version 2.0.2 that enables labels to have a good spacing (no overlapping) ? The only solution I found is to do it manually with annotation(). Please see below the script with dummy values. Also is it possible to connect the wedge of the pie to the related label with an arrow?

I use Python 2.7 and MatPlotLib 2.0.2

Thanks,

Example 1 (overlapping labels)

Example 2 (manually corrected)

import pylab 
import matplotlib.pyplot as plt
fig, ax = plt.subplots()
l = ax.pie([1,0.2, 0.3,0.1,0.4,7,50,35,5,2],#[0, 0.1, 0, 0.1,0,0.1,0,0.1,0,0.1],
            #labels=("one","two","three made up sentences","four is also ther","five becomes a sentence","six it is","seven long", "eight long sent", "nine, as bla bel mo","ten is also short"),
            labels=("","","","","","six it is","seven long", "eight long sent", "nine, as bla bel mo","ten is also short"),
            colors=("b","g","r","y", "b","g","r","y","g","black"),
            startangle =20,
            radius=1,
            frame=True,   # Plot axes frame with the chart if true.
            labeldistance = 1.1 ) #returns a list of matplotlib.patches.Wedge objects

l2 = ax.pie([1,0.2, 0.3,0.1,0.4,7,50,35,5,2],#[0, 0.1, 0, 0.1,0,0.1,0,0.1,0,0.1],
            colors=("r","g","b","w", "g","b","y","r","w","black"),
            startangle =20,
            radius=1-0.7,
            frame=True)   # Plot axes frame with the chart if true.

coor = [t.get_position() for t in l[1]]

ax.axis('equal')
plt.annotate( 'one was very short now ext', xy= (coor[0][0], coor[0][1]) )      # https://kite.com/docs/python/matplotlib.pyplot.annotate
plt.annotate( 'two long sentense',          xy= (coor[1][0], coor[1][1]) ) 
plt.annotate('three things to say' ,        xy= (coor[2][0], coor[2][1]+0.02) ) 
plt.annotate( 'four main tasks to do',      xy= (coor[3][0], coor[3][1]+0.04) ) 
plt.annotate( 'five reasons to avoid',      xy= (coor[4][0], coor[4][1]+0.06 ))   

plt.savefig('test_draft.pdf')
plt.show()

解决方案

You could experiment with rotating the text. The following determines an angle for each wedge and then rotates the annotation text accordingly. Some extra tweaks are needed depending on the angle to determine the text alignment and also to ensure the text does not appear upside down. You might need to further refine this.

import pylab 
import matplotlib.pyplot as plt
import math

fig, ax = plt.subplots()

labels= [
    "one",
    "two", 
    "three made up sentences", 
    "four is also there", 
    "five becomes a sentence",
    "six it is",
    "seven long", 
    "eight long sent", 
    "nine, as bla bel mo",
    "ten is also short"]

l = ax.pie([1,0.2, 0.3,0.1,0.4,7,50,35,5,2],
            labels=[''] * len(labels),
            colors=("b","g","r","y", "b","g","r","y","g","black"),
            startangle=20,
            radius=1,
            frame=True,   # Plot axes frame with the chart if true.
            labeldistance=1.1) #returns a list of matplotlib.patches.Wedge objects

l2 = ax.pie([1,0.2, 0.3,0.1,0.4,7,50,35,5,2],#[0, 0.1, 0, 0.1,0,0.1,0,0.1,0,0.1],
            colors=("r","g","b","w", "g","b","y","r","w","black"),
            startangle=20,
            radius=1-0.7,
            frame=True)   # Plot axes frame with the chart if true.

for label, t in zip(labels, l[1]):
    x, y = t.get_position()
    angle = int(math.degrees(math.atan2(y, x)))
    ha = "left"
    va = "bottom"

    if angle > 90:
        angle -= 180

    if angle < 0:
        va = "top"

    if -45 <= angle <= 0:
        ha = "right"
        va = "bottom"

    plt.annotate(label, xy=(x,y), rotation=angle, ha=ha, va=va, size=8)

ax.axis('equal')
plt.show() 

Which would display as: