Altair - Bar Chart Domain Cropping Error when adding Color

Question

I have a bar chart which shows percentage values for several languages.

chart = alt.Chart(df_lang_perc).mark_bar().encode(
    x=alt.X('Lang:N', sort='-y', title='Language'),
    y=alt.Y('perc_pos:Q', title='Percentage of Positive Reviews', scale=alt.Scale(domain=[0.7,1]), axis=alt.Axis(format='%')),
)

chart.display()

In the DataFrame, each language is categorized into a market. When I try to add color by language, Altair has a bug cropping the bars to the correct length.

chart = alt.Chart(df_lang_perc).mark_bar().encode(
    x=alt.X('Lang:N', sort='-y', title='Language'),
    y=alt.Y('perc_pos:Q', title='Percentage of Positive Reviews', scale=alt.Scale(domain=[0.7,1]), axis=alt.Axis(format='%')),
    color=alt.Color('market')
)

chart.display()

Is this a know issue and is there a workaround?

Minimal reproducible example:

df_lang_perc = pd.DataFrame({
    'Lang': ['brazilian', 'czech', 'english', 'latam', 'thai'],
    'perc_pos': [0.936053, 0.906603, 0.881611, 0.918832, 0.904748],
    'market': ['Latin_America', 'Europe', 'Emglish', 'Latin_America', 'Asia']
})

Answer 1

I am not sure why this happens when you group by color, but it is related to clipping chart elements outside the axes, which you can force by using mark_bar(clip=True) as a workaround for your chart.

Edit: I actually think the error is in that the blue bars are being cut automatically (not sure if this is an Altair issue, or the frontend such as JupyterLab, etc). I believe bars are always extended to zero in VegaLite unless explicitly clipped to discourage zooming in on small difference unless it is done carefully and intentionally. This is also how that charts behave in the VegaLite online editor, where both the colored and not colored bars require the explicit clip=True to avoid extending beneath the x-axis.