Inconsistent indexing of subplots returned by `pandas.DataFrame.plot` when changing plot kind

Question

I know that, this issue is known and was already discussed. But I am encountering a strange behaviour, may be someone has idea why: When I run this:

plot = df.plot(kind="box", subplots=True, layout= (7,4), fontsize=12, figsize=(20,40))
fig = plot[0].get_figure()
fig.savefig(str(path) + "/Graphics/" + "Boxplot_all.png")
plot

It works just fine.

When I change the kind plot to ="line", it gives that known error... but why? I don't get it...

plot = df.plot(kind="line", subplots=True, layout= (7,4), fontsize=12, figsize=(20,40))
fig = plot[0].get_figure()
fig.savefig(str(path) + "/Graphics/" + "Line_all.png")
plot

Thanks for your ideas and hints. Cheers Dave

Answer 1

This seems to be an inconsistency in Pandas. According to their docs, indeed, the DataFrame's method .plot() should return

matplotlib.axes.Axes or numpy.ndarray of them

This is true if you choose the kind="line" option:

>>> df = pd.DataFrame(np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]]),
                      columns=['a', 'b', 'c'])

>>> plot=df.plot(subplots=True, layout=(2,2),kind="line")

>>> type(plot)
numpy.ndarray

>>> plot.shape
(2, 2)

But not with kind="box", where you get a pandas Series:

>>> plot=df.plot(subplots=True, layout=(2,2),kind="box")

>>> type(plot)
pandas.core.series.Series

>>> plot.shape
(3,)

So, if using kind="line", you have to access a 2D array, so you should use:

fig = plot[0,0].get_figure()