Weird behavior in brush zoom in nominal variable and double vertical concatenated grouped bar chart.

[marcocarotta]

Hi everyone, I am kinda new to the library but I really appreciate the work, it's an amazing tool!

I am trying to do a similar graph as in this example here but with a grouped bar with xoffset. The weird is appearing as soon you try to brush the lower chart to get a bigger view up.

Screen.Recording.2025-05-08.144150.mp4

I would be glad also to contribute to the library if you want to show me how.

I am having a weird behavior when brushing in the lower chart. Here is the code:

import altair as alt
import pandas as pd
import random

# Disable max rows limit
alt.data_transformers.disable_max_rows()

number_of_points = 20

# simulate random data
data = {
    'distance': list(range(number_of_points)),
    'delay_a': [random.randint(-100, 100) for _ in range(number_of_points)],
    'delay_b': [random.randint(-100, 100) for _ in range(number_of_points)],
}
# melt into long format
data = pd.melt(pd.DataFrame(data), id_vars=['distance'], value_vars=['delay_a', 'delay_b'], var_name='solver', value_name='delay')

# Correctly set the outlier using .loc
data.loc[1, 'delay'] = 1000  # Add an outlier for testing

source = pd.DataFrame(data)

# Variable to set distance type
distance_type = 'O'  # Set to 'N' for nominal or 'Q' for quantitative

## Brush for selection (x-encoding only)
brush = alt.selection_interval(encodings=['x'])

## Base scatter plot
base = alt.Chart(source).mark_bar().encode(
    x=alt.X(f'distance:{distance_type}', title='Distance'),
    y='delay:Q',
    xOffset=alt.XOffset('solver:N'),
    color=alt.Color('solver:N')
).properties(
    width=900,
    height=450
)

## Upper chart zoomed in by the brush selection on x-axis
upper = base.encode(
    alt.X(f'distance:{distance_type}', scale=alt.Scale(domain=brush)),
)

## Lower chart which controls the brush selection (overview)
lower = base.properties(
    height=60
).add_params(brush)

## Combine the charts
final_chart = alt.vconcat(upper, lower).resolve_legend(
    color="independent"
).configure_view(
    stroke=None
)

final_chart.display()

help!