Allow passing figure handle to FacetGrid

[daliagachc]

Is your feature request related to a problem?

Sometimes i need to combine xarray Facet grids with other ax plots. It would be amazing if I could pass a created figure to the plot function. Event better a subfigure so that the possibilities are infinite!

Describe the solution you'd like

for example:

  da = xr.tutorial.open_dataset("air_temperature")['air']
  f = plt.figure()
  (
      da
      [{'time':[1,2,4]}]
      .plot.contourf(col='time',fig = f)
  )

Describe alternatives you've considered

an alternative is to manually to all plots in a created figure, but this becomes cumbersome.
I quickly checked the source code, and it does not seem very difficult to implement. mostly a modification to the
get_axis function so that it accepts an already created figure. I managed to quickly make it work in seaborn (see image below)

Additional context

No response

[mathause]

I think the usual workflow is to create the figure with gridspec and then update the created subplots and figure. Is there something that is not possible if done this way around?

[daliagachc]

thanks for the reply. Somehow i don't find an elegant way that does not involve many loops and hoops. Lets consider the following short example:
I want to produce a figure that takes the xr.tutorial.open_dataset("air_temperature") and produces a mean grouped by season and hour in the same figure:

Creating the top and bottom rows independently is super efficient with xarray:

import xarray as xr
import matplotlib.pyplot as plt

da = xr.tutorial.open_dataset("air_temperature")['air']

(
    da.groupby(da['time'].dt.season)
    .mean()
    .plot(col='season')
)

(
    da.groupby(da['time'].dt.hour)
    .mean()
    .plot(col='hour')
)

but combining both seems quite complicated to me. I wish I could pass a subfigure to the plot function in the following way

fig:plt.Figure = plt.figure()
(fig_top,fig_bot) = fig.subfigures(2,1)

(
    da.groupby(da['time'].dt.season)
    .mean()
    .plot(col='season',fig=fig_top)
)

(
    da.groupby(da['time'].dt.hour)
    .mean()
    .plot(col='hour', fig=fig_bot)
)

maybe I am missing a smart way of doing this?
cheers

[daliagachc]

I have actually manage to make it work with minimal modification of the source xarray code:

import xarray as xr
import matplotlib.pyplot as plt
import matplotlib as mpl


def _pass(*_,**__):
    pass

mpl.figure.SubFigure.tight_layout = _pass

da = xr.tutorial.open_dataset("air_temperature")['air']
fig:plt.Figure = plt.figure(
    constrained_layout=True,
    figsize = (2*4,2*2)
)

(fig_top,fig_bot) = fig.subfigures(2,1)

(
    da.groupby(da['time'].dt.season)
    .mean()
    .plot(col='season',fig=fig_top)
)

(
    da.groupby(da['time'].dt.hour)
    .mean()
    .plot(col='hour', fig=fig_bot)
)

[mathause]

Good point. Combining two FacetGrids is a good use case. My intuition would be to allow passing exactly the correct number of axes and not the figure. But then I have not heard from subfigure before now and this may also be a valid option.

f, axs = plt.subplots(2, 4)
da1.plot(col='season', ax=axs[0, :])
da2.plot(col='season', ax=axs[1, :])

[dcherian]

Seems like a good addition to me.

I'm not sure about passing Axes handles. IIRC FacetGrid does call a bunch of Figure methods, so there may be unintended consequences

[tomchor]

Is this still on the table? I think being able to pass a figure handle would be good for all the reasons mentioned above, but also it would allow us to use xmovie to plot FacetGrids out of the box since (if I'm not missing anything) xmovie asks only that fig gets passed along to the plotting function (that's probably not to waste time creating several figures per animation). At the moment this is impossible with FacetGrid.

CC @jbusecke