Bug report

Bug summary

I am trying to set the absolute quiverkey arrow-length matching a specific arrow-length of my quiver data. Unfortunately this seems to be impossible.

Code for reproduction

I've slightly modified the example from: matplotlib.org

import matplotlib.pyplot as plt
import numpy as np

X = np.arange(-10, 11, 2)
Y = np.arange(-10, 11, 2)
U, V = np.meshgrid(X, Y)

# calculate length of each arrow
data_arrow_length = np.abs(np.sqrt(np.square(U[:]) + np.square(U[:])))

# set displayed arrow length for longest arrow
displayed_arrow_length = 2

# calculate scale factor for quiver
scale_factor = np.max(data_arrow_length)/displayed_arrow_length

# mysterious arrow length for quiverkey
quiverkey_length = 200

fig, ax = plt.subplots()
q = ax.quiver(X, Y, U, V, pivot='middle',
              units='xy', width=0.1,
              scale=scale_factor, scale_units='xy')
ax.quiverkey(q, X=0.3, Y=1.1, U=quiverkey_length,
             label='Quiver key, length = 10', labelpos='E')

plt.show()

Actual outcome

The resulting figure looks like this: small scale figure
However, if I increase the figure window size the quiverkey does not scale with my data: maximized figure

Expected outcome

I would like to have the same scaling options for quiverkey as exist for quiver. Or even better: Inherit the properties from quiver so that the quiverkey parameter "U" gives an arrow with identical length regarding the scaling of the original quiver plot.

Matplotlib version

• Operating system: Ubuntu 18.04 LTS
• Matplotlib version: 3.0.0
• Matplotlib backend (print(matplotlib.get_backend())): TkAgg
• Python version: 3.6.5
• Numpy version: 1.14.3

Python installation via apt-get packages and matplotlib installation using pip.