Using multiple columns to define "color"

Let’s say I have a dataframe with an x, y, and two categories:

df = pd.DataFrame(dict(
    x=[1,2,3],
    cat1=['red', 'red', 'blue'],
    cat2=['apple', 'apple', 'banana'],
    y=[10,11,12]
))

I’d like a plot with 2 lines: “red,apple” and “blue,banana”, with “red,apple” having two points, and “blue,banana” having 1.

I can do this by combining the two columns into one in the dataframe, but this seems like a very common use case that should be natively supported.

[lizhoula]

How can I encode color with a column that specifies the color? I have this column and what to every bar to use this color instead of the auto one.

[jakevdp]

You can do this with a Calculate Transform; for example:

alt.Chart(df).transform_calculate(
    cat="datum.cat1 + '-' + datum.cat2"
).mark_point().encode(
    x='x:Q',
    y='y:Q',
    color='cat:N'
)