Altair - 在分面图中绘制圆、椭圆、注释

Question

Is it possible to plot lines and geometric shapes inside an Altair chart that follow the units of measure of the chart x and y variables? The chart may be faceted, and the shapes and lines depend on the data in each specific facet.

A reproducible example:

import pandas as pd
import numpy as np
import altair as alt
alt.renderers.enable('notebook')

# make some data to test
N = 1000
df = pd.DataFrame({
    'x1': np.random.normal(0, 1, N),
    'x2': np.random.normal(0, 1, N),
    'facet': np.random.choice(list('ABCDEFGHI'), N),
})

# derived variables 
df['y1'] = np.where(np.sqrt(df['x1']**2 + df['x2']**2) > 2, 'F', 'P')
df['y2'] = 0.5*df['x2'] + 2.0 + np.random.normal(0, .5, N)
df['color'] = np.where(df['y1'].eq('F'), 'red', 'green')

# custom color map
domain = ['F', 'P']
range_ = ['red', 'green']

# create and render the chart
p1 = alt.Chart(df).mark_circle(opacity=1, size=15).encode(
    alt.X('x1', scale=alt.Scale(domain=(-4, 4))),
    alt.Y('x2', scale=alt.Scale(domain=(-4, 4))),
    color=alt.Color('y1', scale=alt.Scale(domain=domain, range=range_)),
    facet='facet'
)

# set some additional properties
p1.properties(width=150, height=150, columns=3).resolve_scale()

Which produces the following output:

Q1: Is is possible to draw a circle centered at 0, 0 with radius=2 in each of the charts as illustrated in the 2nd facet?

The circle units of measure would be the same as the x and y units of measure. In this case, x and y may represent a physical linear measurement, where lenght per pixel is uniform in x and yie it could be a dart board.

The analogy may be to R lattice xyplot aspect = 'iso'. See aspect description here: https://rdrr.io/cran/lattice/man/xyplot.html

Q2: Is it possible to add a text annotation in one corner of the plot that counts the number of "F" (red) in each chart?

Q3: For cases where the axis are not "iso" and have different units of measure, can an line and ellipse (say, a 95% density ellipse) be plotted as illustrated in the 2nd facet below? Perhaps with the slope and intercept of the line of fit annotated on the chart?

Example:

# create and render the chart
p1 = alt.Chart(df).mark_point().encode(
    x='x2',
    y='y2',
    facet='facet'
)

# set some additional properties
p1.properties(width=150, height=150, columns=3).resolve_scale()

In R Lattice, these types of visualizations were accomplished through use of 'aspect', some specific lattice/grid functions and in some cases writing a custom 'panel' (facet) function, that had access to the indexes of the data in each facet and could run linear models in each facet and display the the results.

Answer 1

There is no mechanism in Altair for this kind of annotation, short of creating data which underlies the desired annotations and drawing them as normal chart layers.

Support for general chart annotations is an open issue in Vega and Vega-Lite, the rendering libraries that Altair uses.