In the example scatter_piecharts.py, segments of the unit disk are approximated through polygons and used as markers. By chance, each segment has a coordinate with modulus 1, so that it is not rescaled. As soon as one starts to play with the example a bit, e.g. changing the proportions of the first two segments into
the second, green segment is rescaled, which it should not be for our purposes.
To counter that effect, one needs to apply a segment-specific scale, as is done here:
The resulting code looks like this (there's probably a nicer waying of writing this):
"""
This example makes custom 'pie charts' as the markers for a scatter plotqu
Thanks to Manuel Metz for the example
"""
import math
import numpy as np
import matplotlib.pyplot as plt
# first define the ratios
r1 = 0.2 # 20%
r2 = r1 + 0.4 # 40%
# define some sizes of the scatter marker
sizes = [60,80,120]
# calculate the points of the first pie marker
#
# these are just the origin (0,0) +
# some points on a circle cos,sin
x = [0] + np.cos(np.linspace(0, 2*math.pi*r1, 10)).tolist()
y = [0] + np.sin(np.linspace(0, 2*math.pi*r1, 10)).tolist()
xy1 = list(zip(x,y))
s1 = max(max(x),max(y))
# ...
x = [0] + np.cos(np.linspace(2*math.pi*r1, 2*math.pi*r2, 10)).tolist()
y = [0] + np.sin(np.linspace(2*math.pi*r1, 2*math.pi*r2, 10)).tolist()
xy2 = list(zip(x,y))
s2 = max(max(x),max(y))
x = [0] + np.cos(np.linspace(2*math.pi*r2, 2*math.pi, 10)).tolist()
y = [0] + np.sin(np.linspace(2*math.pi*r2, 2*math.pi, 10)).tolist()
xy3 = list(zip(x,y))
s3 = max(max(x),max(y))
fig, ax = plt.subplots()
ax.scatter( np.arange(3), np.arange(3), marker=(xy1,0),
s=[ s1*s1*x for x in sizes ], facecolor='blue' )
ax.scatter( np.arange(3), np.arange(3), marker=(xy2,0),
s=[ s2*s2*x for x in sizes ], facecolor='green' )
ax.scatter( np.arange(3), np.arange(3), marker=(xy3,0),
s=[ s3*s3*x for x in sizes ], facecolor='red' )
plt.show()
In the example scatter_piecharts.py, segments of the unit disk are approximated through polygons and used as markers. By chance, each segment has a coordinate with modulus 1, so that it is not rescaled. As soon as one starts to play with the example a bit, e.g. changing the proportions of the first two segments into
the second, green segment is rescaled, which it should not be for our purposes.
To counter that effect, one needs to apply a segment-specific scale, as is done here:
The resulting code looks like this (there's probably a nicer waying of writing this):