Adding `size=X` to `sns.scatterplot` adds an entry to the legend (as well as changing the pointsize)

[beyarkay]

Minimum working example:

import numpy as np
import seaborn as sns
import pandas as pd

data = pd.DataFrame()
data['x'] = np.random.random((100,))
data['y'] = np.random.random((100,))
data['hue'] = np.random.choice(['hue1', 'hue2', 'hue3'], (100,))
sns.scatterplot(
    data=data,
    x='x',
    y='y',
    hue='hue',
    size=1,
)

Result:

I'm not entirely sure what's going on here, and it's possible that it's an error on my part. There seems to be many different types of behaviour depending on the usage of hue, size, and s:

sns.scatterplot(..., size=1, ...) (item added to legend)

sns.scatterplot(..., hue='hue', size=1, ...) (item added to legend)

sns.scatterplot(..., hue='hue', s=1, ...) (no extra legend item)

sns.scatterplot(..., hue='hue', ...)(no extra legend item)

I'd expect that adding size=1 would not add an item 1 to the legend. Not sure if this is a bug or if I'm using the interface wrong though.

[mwaskom]

size is a mapping variable, like hue. When you pass a scalar, it gets broadcast, so it's equivalent to mapping with a vector of all 1s.

[beyarkay]

Okay thanks. I'm not sure why it gets added as an item in the legend though?

[mwaskom]

Why wouldn't it? The legend annotates the mapping variables, and you have a size mapping variable.