Description
The docstring for plt.subplot states that
Creating a new subplot with a position which is entirely inside a
pre-existing axes will trigger the larger axes to be deleted::
import matplotlib.pyplot as plt
# plot a line, implicitly creating a subplot(111)
plt.plot([1,2,3])
# now create a subplot which represents the top plot of a grid
# with 2 rows and 1 column. Since this subplot will overlap the
# first, the plot (and its axes) previously created, will be removed
plt.subplot(211)
plt.plot(range(12))
plt.subplot(212, facecolor='y') # creates 2nd subplot with yellow background
If you do not want this behavior, use the
:meth:`~matplotlib.figure.Figure.add_subplot` method or the
:func:`~matplotlib.pyplot.axes` function instead.
(which I think is a rather silly behavior but whatever -- seems inherited from MATLAB, in fact).
The actual behavior (which is also MATLAB's behavior) is to delete any axes that overlaps the newly added axes, which can be easily checked from the code
for other in fig.axes:
if other==a: continue
if bbox.fully_overlaps(other.bbox): # <-- here
byebye.append(other)
for ax in byebye: delaxes(ax)
and also by running e.g. plt.subplot(221); plt.subplot(335) (the first subplot does get deleted).
Making this a doc issue rather than requesting for the behavior to be changed because I don't think either behavior really makes sense anyways :-)
Noticed while investigating #6096 -- this is the mechanism that leads to one of the axes being deleted (note, though, that regardless of this, there still(?) seems to be an issue with incremental tight-layouting in #6096).