“Maybe the Shapefile was right after all” is right after all
The more that I work with broad datasets, collected from many different means and used in many different contexts, the more I think that you really only need the “multi” instantiation of polygons, lines, and points. Usually, due to database typecasting, you want to coerce any heterogeneous datatypes into the same datatype. Since it’s much simpler to cast Polygons upwards into MultiPolygons, I’m rarely seeing any Polygons or Lines, even in datasets where the majority of features are single-component.
Overall, Multi- datatypes are more general than their single-component counterparts. So, exploiting that, I’m seeing that most of the data I see across our workflows gets cast upwards into Multi-geometries, even when there is only one component.
It’d be so much more succinct to just rename Multi-geometries as their singly-instanced geometries, and then use “length” or “component” counts to determine how many elements the geometry collection contained. I know that I’ll be doing this consistently in my research database, at least.
imported from: yetanothergeographer