在分面 Altair 图表中重复轴

Question

Given a simple, faceted chart like:

import altair as alt

data = alt.Data(values = [
    { "category" : "a", "x" : 1, "y" : 2 },
    { "category" : "a", "x" : 2, "y" : 4 },
    { "category" : "b", "x" : 1, "y" : 3 },
    { "category" : "b", "x" : 2, "y" : 5 }
])

alt.Chart(data).mark_point().encode(x = "x:Q", y = "y:Q").facet(
    row = "category:O"
)

How do you have the x axis appear for each subchart, rather than just once at the bottom? This is to improve readability when there are a lot of subcharts...

Answer 1

Unfortunately, there's no way to make the x-axis appear in multiple charts when using the row encoding. As a workaround, you can manually vconcat charts based on filtered data:

chart = alt.Chart(data).mark_point().encode(x="x:Q", y="y:Q")

alt.vconcat(
    chart.transform_filter(alt.datum.category == 'a'),
    chart.transform_filter(alt.datum.category == 'b')
)

To avoid writing out the column values manually, you can generate the different subcharts using Python tools; for example, this is equivalent to the above:

df = pd.DataFrame.from_records([
    { "category" : "a", "x" : 1, "y" : 2 },
    { "category" : "a", "x" : 2, "y" : 4 },
    { "category" : "b", "x" : 1, "y" : 3 },
    { "category" : "b", "x" : 2, "y" : 5 }
])

chart = alt.Chart(df).mark_point().encode(x="x:Q", y="y:Q")

alt.vconcat(
    *(chart.transform_filter(alt.datum.category == val)
      for val in df['category'].unique())
)

Answer 2

Here's a simple way to repeat the axis, with a drawback:

import altair as alt

data = alt.Data(values = [
    { "category" : "a", "x" : 1, "y" : 2 },
    { "category" : "a", "x" : 2, "y" : 4 },
    { "category" : "b", "x" : 1, "y" : 3 },
    { "category" : "b", "x" : 2, "y" : 5 }
])

alt.Chart(data).mark_point().encode(x = "x:Q", y = "y:Q").facet(
    row = "category:O"
).resolve_scale(x='independent')

But: If the facets do not have identical x-axis ranges, this will also de-couple them! There is, to my knowledge, no simple way of just repeating the axis without making the facets independent. But you can always fix the range to achieve that.