Class column NULL vs matching subtype

[jeffdefacto]

I've noticed that the different themes are inconsistent in how the class column is handled. In some layers there are no nulls in the class field and it mirrors subtype when there is not a more specific value. E.g. when subtype = 'civic' then class can also be 'civic'

However, Divisions and Transportation seem to not have this mirroring and instead allow null values for class. The Buildings theme currently has a mixture of both situations.

We should have a consistent standard for how this field is populated.

[vcschapp]

@jeffdefacto Can you give specific examples? A table with at least one example for every theme and feature type would be super helpful, especially if it includes: theme, type, feature ID, subtype, and class.

(Note that some feature types do not have a class property and consequently a NULL would be expected in the class column for that feature type in the Parquet. But I'm not sure if that's what you're getting at.)

[jeffdefacto]

Here are the subtype and class counts for the water theme. All features have a class even if it just mirrors the subtype. In the case of ocean and river subtypes there is only the mirrored class tag in use so there is not much value added by the class.

2024-08-20.0.theme=base.type=water.csv

On the other hand, in transportation segments, rail and water have no classes and are left blank.

2024-08-20.0.theme=transportation.type=segment.csv

Within buildings, there are instances of both of these approaches. For example, subtype civic has both features with a class of civic and a blank class.

2024-08-20.0.theme=buildings.type=building.csv

Here are the other relevant layers for reference as well.

2024-08-20.0.theme=base.type=infrastructure.csv
2024-08-20.0.theme=base.type=land_use.csv
2024-08-20.0.theme=base.type=land.csv
2024-08-20.0.theme=divisions.type=division_area.csv
2024-08-20.0.theme=divisions.type=division_boundary.csv
2024-08-20.0.theme=divisions.type=division.csv
2024-08-20.0.theme=places.type=place.csv

I can provide specific feature ids as well if that would be helpful but the aggregate data is probably the clearest examples.