seaborn.objects share x or y = False keeps all axes categorical

[subsurfaceiodev]

Hi! As you may see, the x axis on the left subplot of the figure is displayed as categorical even though its data is numerical. Expected behavior would be for the x axis to be numerical and show the correct distance between x=2 and x=100.

import pandas as pd
import seaborn.objects as so

dataset = pd.DataFrame(dict(
    x=[1, 2, 10, 'A'],
    y=[1, 2, 2, 1],
    group=['g1', 'g1', 'g1', 'g2'],
    col=['c1', 'c1', 'c1', 'c2'],
))

p = (
    so.Plot(dataset,
            x='x',
            y='y',
            color='group',
            )
    .facet('col')
    .add(so.Dot())
    .share(x=False, y=True)
)
p.show()

[frfeng]

Without looking at the source code, my guess is seaborn uses the data types from Pandas dtype (in this case x is an object rather than numerical because of the value A) and does not make further assumptions on the data type for each faceted data. Maybe there are reasons to not do that?

[mwaskom]

As documented, Plot.share allows each facet to have different axis limits and ticks. However, the facets still share a scale, and so either need to treat the data as categorical or numerical, but not both. The facet operation show you slices of a variable, not distinct variables.

Depending on the nature of your data you may be able to use Plot.pair, which creates subplots for distinct variables.

[subsurfaceiodev]

We know seaborn is not meant to be an exact alternative to plotly express or viceversa, but the following code produces with px what one would expect:

pex = px.scatter(
    data_frame=dataset,
    x='x',
    y='y',
    color='group',
    facet_col='col'
)