Allow for resolve_scale to work with latitude and longitude

[thomascamminady]

I came across this when trying to plot the running tracks I ran this year. For each run, I have latitude and longitude data and I want to plot the track. Each column should be a new route. And since tracks happen at various places across many countries, I need the scale to be resolved independently.

Here's an MWE in altair that works:

import altair as alt
import polars as pl

base = (
    alt.Chart(
        pl.concat(
            [
                pl.DataFrame(
                    {
                        "x": [0.1, 0.2, 0.3, 0.4, 0.5],
                        "y": [0.1, 0.2, 0.3, 0.4, 0.5],
                        "id": ["1", "1", "1", "1", "1"],
                    }
                ),
                pl.DataFrame(
                    {
                        "x": [1.1, 1.2, 1.3, 1.4, 1.5],
                        "y": [1.1, 1.2, 1.3, 1.4, 1.5],
                        "id": ["2", "2", "2", "2", "2"],
                    }
                ),
            ]
        )
    )
    .mark_line()
    .encode(column="id")
)

(
    base.encode(
        x=alt.X("x").scale(zero=False),
        y=alt.Y("y").scale(zero=False),
    )
    .resolve_scale(x="independent", y="independent")
    .show()
)

However, this obviously distorts the tracks. I should have used latitude and longitude instead of x and y. So I tried this:

(
    base.encode(
        latitude=alt.Latitude("x"),
        longitude=alt.Longitude("y"),
    )
    .resolve_scale(latitude="independent", longitude="independent")
    .show()
)

This errors because resolve_scale does not accept latitude or longitude as inputs.

Is there a fundamental issue as for why resolve_scale can't work with latitude / longitude? Could this be added?

I first wanted to open this under altair but I guess this would have moved here eventually? Sorry for only providing the python code.

Thanks a lot for the great work you're doing!

[kjgoodrick]

I believe this is related to the issue #3729 which prevents faceting on geoshape plots.

As a workaround you can manually concatenate each plot. When doing this you will also encounter #9321 which causes concatenated geoshape plots to have shared projections, so it is also necessary to slightly change the projection on each loop.

import altair as alt
import polars as pl

base_df = pl.concat(
    [
        pl.DataFrame(
            {
                "x": [0.1, 0.2, 0.3, 0.4, 0.5],
                "y": [0.1, 0.2, 0.3, 0.4, 0.5],
                "id": ["1", "1", "1", "1", "1"],
            }
        ),
        pl.DataFrame(
            {
                "x": [1.1, 1.2, 1.3, 1.4, 1.5],
                "y": [1.1, 1.2, 1.3, 1.4, 1.5],
                "id": ["2", "2", "2", "2", "2"],
            }
        ),
    ]
)

base = (
    alt.Chart(base_df)
    .mark_line()
    .encode(
        latitude=alt.Latitude("x"),
        longitude=alt.Longitude("y"),
    )
)

alt.concat(
    *(
        base.transform_filter(alt.datum.id == run_id).project(
            precision=0.707 + (count / 1e6)
        )
        for count, run_id in enumerate(base_df["id"].unique())
    )
)

[thomascamminady]

Thanks for the workaround!
Happy to see this issue closed if it is ultimately going to be resolved by #3729 .

[domoritz]

If there is an easier solution that doesn't involve trying to fully fix facte for geoshape, I am happy to review a pull request. Happy to reopen this if this is something you can help with.