Make facet error more informative

[joelostblom]

Currently if we charts use different data and are trying to be faceted, altair raises ValueError: Facet charts require data to be specified at the top level. I don't think it is immediately clear to everyone what this means, and this PR is an attempt to make the error message more instructive.

Not that there will still be situations like the below that raises the error although the same data is technically used, just respecified as a new variable.

alt.layer(
    alt.Chart(data.movies.url).mark_rule().encode(x='IMDB_Rating:Q'),
    alt.Chart(data.movies.url).mark_rule().encode(x='IMDB_Rating:Q')
).facet(
    'Major_Genre:N'
)

I am not sure if altair should be capable of figuring out that this is the same as

source = data.movies.url
alt.layer(
    alt.Chart(source).mark_rule().encode(x='IMDB_Rating:Q'),
    alt.Chart(source).mark_rule().encode(x='IMDB_Rating:Q')
).facet(
    'Major_Genre:N'
)

and avoid the error, maybe those comparisons will be too complex for large data frames and non-portable across dataframe implementations (although many have some type of equality test built-in like pandas).

[mattijn]

Could these datasets within a facet be consolidated as is done for other data?

See this function: https://github.com/altair-viz/altair/blob/main/altair/vegalite/v5/api.py#L63

[joelostblom]

Yeah I think that is a good thing to explore, but maybe in a separate PR? I opened #3265

[mattijn]

Thanks!