Encoding point Z-order in Altair charts

import altair as alt
from IPython.display import HTML
from sklearn.datasets import make_blobs
import pandas as pd

Just making some data here for demonstration.

centers = [[1, 1], [-1, -1], [1, -1]]
X, y = make_blobs(n_samples=2000, centers=centers, cluster_std=1)

df_points = (
    pd.DataFrame(X, columns=['a', 'b'])
    .assign(label=pd.Categorical(y, ordered=True, categories=[0, 1, 2]))
)

df_points.head()

	a	b	label
0	-1.214458	-1.672904	2
1	0.271734	-1.012541	2
2	2.086623	2.339706	0
3	1.918859	0.593477	2
4	-0.617844	-1.212739	1

The order encoding in combination with mark_point allows you to control the z-order of points in the chart. You can set the sorting order of categorical data via categories in pd.Categorical (see above).

axis = alt.Axis(titleFontSize=18, titlePadding=20, titleFontWeight='normal', labelFontSize=12)
scale = {'range': ['lightgray','lightgray','steelblue']}

c = alt.Chart(df_points, width=400, height=400).mark_point(fillOpacity=0.25, strokeOpacity=0.75, size=50).encode(
    x=alt.X('a', axis=axis),
    y=alt.Y('b', axis=axis),
    color=(
        alt.Color('label:N')
        .scale(**scale)
        .legend(orient='top-left', titleFontSize=16, labelFontSize=12)
    ),
    fill=alt.Fill('label:N').scale(**scale),
1    order=alt.Order('label', sort='ascending'),
)

with alt.renderers.enable("default"):
    html = c._repr_mimebundle_()["text/html"]

s = html.replace("display: flex;", "display: flex;\njustify-content: center;")

display(HTML(s))

1

This is the order encoding. It forces the blue points to the top.