How to combine two different plots using seaborn.objects?

[paw-lu]

When using the newish seaborns.objects inferface—which is great by the way—is there a cannonical way to combine two plots? Best way I can think of now is to create an Axes and plot both of them on it.

If this is the best way, what's the easiest way to assign a label to each line here so that there's a legend that distinguishes each line?

import numpy as np
import seaborn.objects as so

_, ax = plt.subplots(figsize=(20, 10))
rng = np.random.default_rng()
data1 = rng.weibull(1, size=5_000)
data2 = rng.weibull(1.3, size=5_000)
(
    so.Plot(data1)
    .add(so.Line(color="C0",), so.KDE(bw_adjust=2))
    .on(ax)
    .plot()
)
(
    so.Plot(data2)
    .add(so.Line(color="C1"), so.KDE(bw_adjust=2))
    .on(ax)
    .plot()
)

Thanks!

[paw-lu]

Quick aside: I really like altair's API on combining plots, where if you have chart1 and chart2, you can combine them via chart1 + chart2, have them side by side via chart1 | chart2, stacked using chart1 & chart2 etc.

[mwaskom]

Is what you're looking for here a layered plot?

(
    so.Plot()
    .add(so.Line(color="C0"), so.KDE(bw_adjust=2), x=data1)
    .add(so.Line(color="C1"), so.KDE(bw_adjust=2), x=data2)
)

[paw-lu]

Is what you're looking for here a layered plot?

Yep, this works perfectly. Embarassingly I didn't realize that you could initialize so.Plot() with no args, then pass x to add instead with the data.

[paw-lu]

Thanks a lot!

[mwaskom]

Cool, also note that the "layer legend" aspect is still WIP: #3046

[tburnett]

Please add a comment about this use case to the Plot documentation. And perhaps to the tutorial as well. I had an issue with overplotting a line on a histogram. With only the x source specified in the constructor, it was not possible to then set x and y in the add.
But overlayed plots without legends are almost useless for presentations or publications.