Dynamic label with number of points in scatterplot based on transform selection in Altair?

Question

tl;dr: is there a way to get a dynamic count of points in filtered plots in Altair?

---

In the 'filtering by selection' example for Altair, it is possible to update a scatterplot based on a selection in a linked bar chart.

import altair as alt
from vega_datasets import data

cars = data.cars()

click = alt.selection_point(encodings=['color'])

scatter = alt.Chart(cars).mark_point().encode(
    x='Horsepower:Q',
    y='Miles_per_Gallon:Q',
    color='Origin:N'
).transform_filter(
    click
)

hist = alt.Chart(cars).mark_bar().encode(
    x='count()',
    y='Origin',
    color=alt.condition(click, 'Origin', alt.value('lightgray'))
).add_params(
    click
)

scatter & hist

However, it would be useful to show how many data points remain in the scatterplot after filtering. How can this be achieved?

It is possible to get a static count of data in a dataframe and use that for a label, which I tried with another approach eg:

import altair as alt
import pandas as pd

# Sample data
data = pd.DataFrame({'values': [1, 2, 3, 4, 5, 1, 2, 3]})
total_count = data['values'].sum()

# Create data frame for total count
count_data = pd.DataFrame({'label': ['Total Count'], 'count': [total_count]})

# Combine data
combined_data = pd.concat([data, count_data])

# Create histogram with text for total count
hist = alt.Chart(combined_data).mark_bar().encode(
    x='values:Q',
    y='count()',
)

text = hist.mark_text(
    color='black',
    baseline='middle'
).encode(
    x=alt.value(200),  # Adjust position as needed
    y=alt.value(40),
    text='count:Q',
)

# Display the chart
hist + text

This adds a rather ugly label:

Is there a way to append a dynamic label with the number of points present in the filtered scatterplot? An external element would be fine, but I am quite new to Altair and haven't figured that bit out yet despite some searching.

Answer 1

To show a dynamic count of filtered data in Altair, use transform_aggregate

Showing how many data points are present on a plot after filtering can be done with a combination of transform_aggregate to retrive a count, and text marks.

The aggregate transformation can be added on to the the plot which is getting filtered, eg: filtered_plot.transform_aggregate(count='count(*)'). The Text mark can then be applied as you would any label.

Example using iris sepal data

Note: adding the text with count adds an 'undefined' field in the legend, which I have yet to find a way to remove

Code:

import altair as alt
from vega_datasets import data  # Example dataset library

# Sample data - iris sepal length
source = data('iris')

# Create the scatterplot - eg length by width, coloured by species
scatterplot = alt.Chart(source).mark_point().encode(
    x='sepalLength:Q',
    y='sepalWidth:Q',
    color='species:N'
)

# Create a selection to filter points
selection = alt.selection_point(encodings=['color'], resolve='global')

# count 
countplot = alt.Chart(source).mark_bar().encode(
    y='species:N',
    x='count()',
    color=alt.condition(selection, 'species:N', alt.ColorValue('gray'))
).add_params(selection)


# Filter scatterplot based on selection
filtered_scatterplot = scatterplot.transform_filter(selection)

# Calculate count of filtered points
filtered_count = filtered_scatterplot.transform_aggregate(
    count='count(*)')

# Display count as text
text = filtered_count.mark_text(
    color='black',
    fontSize=14,
    baseline='middle'
).encode(
    x=alt.value(100),  # Adjust position as needed
    y=alt.value(10),
    text='count:Q',
)

# Display the chart with linked selection and count
filtered_scatterplot & text & countplot