so.PolyFit in conjunction with so.Plot.pair: Fit values connected in wrong order?

[MrMho]

When I use so.PolyFit (with order > 1) together with so.Plot.pair like so:

import seaborn.objects as so
import numpy as np
import pandas as pd

np.random.seed(1)
a = np.random.randn(20)
b = np.random.randn(20)
y = np.random.randn(20)
df = pd.DataFrame({"a": a, "b": b, "y": y})

(
so.Plot(df, y="y")
    .pair(x=["a", "b"])
    .add(so.Dots(pointsize=8))
    .add(so.Line(color='C1', linestyle='--'), so.PolyFit(order=3))
)

I get the following output:

It looks like the orange lines connect the values of the calculated fit in the wrong order.

numpy==1.24.2
pandas==2.0.0
matplotlib==3.7.1
seaborn==0.13.2

[mwaskom]

Looks like the issue here is the pairing confusing the automated detection of plot orientation, causing the Line mark to sort its data incorrectly. I see two workarounds that produce the expected result with your example:

• Use the so.Path mark instead of so.Line (this avoids sorting — which is unnecessary after the so.PolyFit transform anyway — so it's also marginally faster)
• Pass orient="x" explicitly in the layer where you add the line fit