Bug report

https://matplotlib.org/api/_as_gen/matplotlib.pyplot.contourf.html#matplotlib.pyplot.contourf
As per the documentation for contour plots, when you pass "levels" as a keyword argument, it sets an instance attribute which breaks the _contour_level_args function.

Code for reproduction

import matplotlib.pyplot as plt

x = np.arange(-1, 1, 0.1)
y = np.arange(-1, 1, 0.1)

X, Y = np.meshgrid(x, y)

Z = (1-X)**2 + Y**2 # or some other function
fig, ax = plt.subplots()
CS = ax.contourf(X, Y, Z, levels=10)

The issue is in the _contour_level_args function in contour.py, where self.levels is not none because you're passing levels as a keyword argument, which seems to indicate that levels is parsed strangely?

Matplotlib version

• Operating system: Mojave
• Matplotlib version: 3.0.2
• Matplotlib backend (print(matplotlib.get_backend())):
• Python version:
• Jupyter version (if applicable):
• Other libraries:

I installed with conda