Do we need to validate the method-based attribute setting separately?

[joelostblom]

In #1629, Jake introduced a comment about validating the method-based attribute setting with the schema:

https://github.com/altair-viz/altair/blob/88b706f4afbef00df7c2f724d42f0e001846a3b7/tools/schemapi/schemapi.py#L637-L641

I am not sure if we need to do this as it seems sufficient that the spec is already validated via the usual validation approach. Does anyone see a use for a separate validation here or can we safely remove this comment?

[ChristopherDavisUCI]

My understanding (and I could be way off) when I saw this was that it was just to have more useful error messages. For example, I tried the following (where it should instead be scheme="turbo")

import altair as alt
from vega_datasets import data

cars = data.cars.url

alt.Chart(cars).mark_circle().encode(
    color=alt.Color('Acceleration:Q').scale("turbo")
)

and I believe in the ~100 line error message, the word scale never appears.

So to my understanding, the idea makes sense but definitely is not a prerequisite to a Release Candidate.

[joelostblom]

The traceback is definitely longer when using .scale vs scale=alt.Scale, but it does end in the same error message for me, which does point the user to what is wrong:

SchemaValidationError: Invalid specification

        altair.vegalite.v5.schema.channels.Color->scale, validating 'anyOf'

        'turbo' is not valid under any of the given schemas

It would be nice with a shorter error message though and it makes sense that validating earlier could shorten it. I wonder if validating an additional time would be affected by #2744