Hybrid PairGrid/FacetGrid?

I am a really big fan of FacetGrid and the new PairPlot.

One sort of plot that I often like to create involves columns for different variables (e.g., a PairGrid) and rows for different values of single variable (e.g., a FacetGrid).

It would be nice for Seaborn to support these sort of “hybrid” Pair/Facet grids natively. You could even identify which is the case for each aesthetic automatically (likely in a new class) simply by whether the input arguments are strings (e.g., hue='species') or list-like (e.g., hue=['temperature', 'precipitation']).

My current work around is to use turn the different variables into a single variable via DataFrame.stack() and then use FacetGrid, but this is awkward (the axis labels require some manual fixes).

[nmusolino]

Here is a StackOverflow question asking how to do exactly this: https://stackoverflow.com/questions/30623721/plot-multiple-dataframe-columns-in-seaborn-facetgrid

[mwaskom]

This has been implemented in the objects interface for (experimental) release in v0.12:

(
    so.Plot(penguins, x="body_mass_g", color="species")
    .pair(y=["bill_length_mm", "bill_depth_mm"])
    .facet(col="sex")
    .add(so.Scatter())
)

I think it’s unlikely that this will appear as a standalone class ala PairGrid/FacetGrid … one never knows, but for now I’m going to call this completed.