Why is Seaborn creating an extra category in my data? [duplicate]

Question

I am trying to plot some simple data with Seaborn 0.9.0 under Python 3.6.5. The data is just two points with a different classification from each other. The classification itself is simply 1 or 2. However when I plot it with Seaborn, the legend shows three types: 0, 1 and 2.

import numpy
import seaborn
import pandas
from matplotlib import pyplot

X = numpy.array([
    [-1, -1, 1],
    [1, 1, 2]
])

data = pandas.DataFrame(X, columns=('x','y','type'))

seaborn.scatterplot(data=data, x='x', y='y', hue='type')

pyplot.show()

The resulting plot shows:

I have also tried this without Pandas, just using eg x=X[:,0], y=X[:,1], hue=X[:,2], but the result is the same.

The Seaborn docs say this about the hue argument:

Can be either categorical or numeric, although color mapping will behave differently in latter case.

But they do not clarify what "categorical" means, or what the behaviour is, or how it is different. I've also read the categorical data plotting tutorial, but haven't found an answer.

Using strings like '1' and '2' in the data just results in an error:

AttributeError: 'str' object has no attribute 'view'

Why is there an extra "type" of 0 in the legend? And, for later, how can I have more meaningful category labels?

---

Reading the categorical data plotting tutorial some more, I found this:

If your data have a pandas Categorical datatype, then the default order of the categories can be set there. If the variable passed to the categorical axis looks numerical, the levels will be sorted. But the data are still treated as categorical and drawn at ordinal positions on the categorical axes (specifically, at 0, 1, …) even when numbers are used to label them:

This half-explains what's happening here (not why there's an extra 0 category), but even using Pandas categorical type doesn't help. Adding

data['type'] = data['type'].astype('category')

...converts this data to the categorical type, but Seaborn still gives an error:

TypeError: data type not understood

Answer 1

You ran indeed into "numeric" colormapping here, meaning seaborn will try to use a meaningful (to itself) number of subset of the data to create a legend from it. This will at least be 3 different colors.

This may become more obvious when replacing the number 2 in the array with something large, e.g. 900

The solution here is indeed to activate the "categorical" mapping. The legend argument of scatterplot can take three values

legend : “brief”, “full”, or False, optional
How to draw the legend. If “brief”, numeric hue and size variables will be represented with a sample of evenly spaced values. If “full”, every group will get an entry in the legend. If False, no legend data is added and no legend is drawn.

So kind of unintuitively (at least in this case) you can set

legend="full"

to get a legend entry for every unique value in the hue column (and hence one less than using "brief").

seaborn.scatterplot(data=data, x='x', y='y', hue='type', legend="full")

Note that using strings as categories will work, but those strings cannot be convertable to numbers.

import numpy
import seaborn
import pandas
from matplotlib import pyplot

X = numpy.array([
    [-1, -1, "A"],
    [ 1,  1, "B"]])

data = pandas.DataFrame(X, columns=('x','y','type'))

seaborn.scatterplot(data=data, x='x', y='y', hue='type', legend="brief")

pyplot.show()