fig.gca() vs. fig.add_subplot()

Question

Most examples of object-oriented matplotlib get an Axis object with something like

import matplotlib.pyplot as plt  fig1 = plt.figure() ax1 = fig1.add_subplot(111)  ax1.plot(...... etc. 

Which I've always found to be non-obvious, especially from a matlab-perspective. I recently found that equivalent results can be obtained via

ax1 = fig1.gca()   # "GetCurrentAxis" 

Which makes way more sense to me (possibly only due to prior Matlab use). Why is add_subplot() with a confusing 111 argument chosen as the preferred way to get an axis object? Is there any functional difference?

Thanks!

Answer 1

plt.gca gets the current axes, creating one if needed. It is only equivalent in the simplest 1 axes case.

The preferred way is to use plt.subplots (and the docs/examples are indeed lagging a bit, if you want to start contributing, updating the docs is a great place to start):

fig, ax = plt.subplots(1, 1) 

or

fig, (ax1, ax2) = plt.subplots(2, 1) 

and so on.

Answer 2

To creat 3D instance, there are three ways:

plt.gca(projection='3d')  plt.subplot(projection='3d')  fig = plt.figure() fig.add_subplot(111, projection='3d')  

Maybe the third way is more complex.