Dataset.plot

[headtr1ck]

We often use the workflow of using Datasets that contain the same "datatype" of different sensors (i.e. two cameras with different resolution that capture the same object).

A functionality to simply call ds.plot() would simplify the workflow alot.

This plot should create a grid and plot each variable in a subfigure.

Optional args/kwargs:

• variables: Sequence[Hashable] which variables to plot, or the order
• nrows
• ncols

This simple idea would only work for 1D and 2D data.
One could go even further and add the col argument to plot a series of plots for each variable and thus support 3D data.

[max-sixty]

I agree this would be useful! There's some challenge in finding an approach that works universally. To the extent that each dataset is dimensioned the same, a grid seems good. But there will also be different cases.

Pandas has an approach with df.plot.hist(...), i.e. a namespace within plot.

What do others think about adding a ds.plot() with subplots?

[OriolAbril]

I am aiming to get to something similar (or at least the infrastructure to easily get something like this going). It is still very early stage, but feedback is welcome:

• Initial idea with DataArrays but more flexible than current FacetGrid when it comes to aesthetics mapping: https://xrtist.readthedocs.io/en/latest/data_organizers.html
• Adaptation to Dataset: https://xrtist.readthedocs.io/en/latest/plot_museum.html

[TomNicholas]

This crosses over a lot with the same issue in Datatree. Fundamentally users commonly fill the data structure with a simpler and easier to reason about set of data than what we developers worry it might be containing.

I also agree with @max-sixty that you don't want to be too hasty to commandeer the top-level .plot() name.