Change size of arrows using matplotlib quiver

Question

I am using quiver from matplotlib to plot a vectorial field. I would like to change the size of the thickness of each arrow depending on the number of data which produced a specific arrow of the vector field. Therefore what I am looking for is not a general scale transformation of the arrow size, but the way to customize the thickness of the arrow in quiver one-by-one. Is it possible? Can you help me?

Answer 1

@unutbu's solution is not useful after matplotlib 2.0.0 (see this issue and this pull request). As of matplotlib 2.1.2, there seems to be no parameter of plt.quiver which officially supports one-by-one configuration of arrow widths. But some workarounds are remained.

Method 1

Just use Python's loop and the width parameter. This will be slow for large data.

import matplotlib.pyplot as plt
import numpy as np

# original code by user423805
# https://stackoverflow.com/a/6372413/5989200
xmax = 4.0
xmin = -xmax
D = 20
ymax = 4.0
ymin = -ymax

for y in np.linspace(ymin, ymax, D):
    for x in np.linspace(xmin, xmax, D):
        deg = np.arctan(y ** 3 - 3 * y - x)
        w = 0.005 * (y - ymin) / (ymax - ymin)  # just example...
        plt.quiver(x, y, np.cos(deg), np.sin(deg), width=w)

plt.show()

Method 2

This is only a workaround, but linewidths can be used if we set edgecolors.

import matplotlib.pyplot as plt
import numpy as np

# original code by user423805
# https://stackoverflow.com/a/6372413/5989200
xmax = 4.0
xmin = -xmax
D = 20
ymax = 4.0
ymin = -ymax
x = np.linspace(xmin, xmax, D)
y = np.linspace(ymin, ymax, D)
X, Y = np.meshgrid(x, y)
deg = np.arctan(Y ** 3 - 3 * Y - X)
widths = np.linspace(0, 2, X.size)
plt.quiver(X, Y, np.cos(deg), np.sin(deg), linewidths=widths, edgecolors='k')
plt.show()

Note that efiring, one of maintainers of matplotlib, said:

So please use the width kwarg together with units; linewidths is only for controlling the outline thickness, when an outline of a different color is explicitly requested.