How to apply color scale by field range?

[aborruso]

In vega-lite I'm able to apply color scale by field range, I'm able to use the field values of color field to color my chart (this is the sample code).

I'm not able to rewrite in Python/altair, the same of this JSON.

    "color": {
      "type": "nominal",
      "field": "label",
      "scale": {"range": {"field": "color"}},
      "title": ""
    }

It's not this

chart=alt.Chart(df).mark_rect().encode(
    x='x:O',
    y='y:O',
    color=alt.Color('label:N',scale=range('color:N'))
)

[ChristopherDavisUCI]

If you ever want to compare your Altair code to the Vega-Lite code, a good trick is to evaluate print(chart.to_json()), which can sometimes help with spotting the problem (maybe not as much in this particular example).

I believe this will work:

color=alt.Color('label:N',scale=alt.Scale(range={"field": "color"}))

@mattijn Do you see a nicer way than explicitly specifying this dictionary, {"field": "color"}?

In a soon-to-be-released version of Altair, you can skip the alt.Scale portion:

color=alt.Color('label:N').scale(range={"field": "color"})

There is also an option to use the raw color values directly, but I couldn't figure out a nice way to also get your labels to show up with this approach: https://altair-viz.github.io/user_guide/customization.html#raw-color-values

alt.Chart(df).mark_rect().encode(
    x='x:O',
    y='y:O',
    color=alt.Color('color:N', scale=None)
)

[aborruso]

If you ever want to compare your Altair code to the Vega-Lite code, a good trick is to evaluate print(chart.to_json()), which can sometimes help with spotting the problem (maybe not as much in this particular example).

I know, but here I had the right vega code, and so I needed a function that did the opposite (print(chart.to_python())).

color=alt.Color('label:N',scale=alt.Scale(range={"field": "color"}))

It works, thank you very much