ENH: support writing + filtered reading from bbox columns in GeoParquet

[jorisvandenbossche]

GeoParquet 1.1 is adding a new feature to the spec: bbox columns ("bbox covering encoding", https://github.com/opengeospatial/geoparquet/blob/main/format-specs/geoparquet.md#covering)

We should add support for that to geopandas:

• Writing (to_parquet): optionally compute and add this column to written file (behind some option, eg write_bbox_covering=True/False?)
• Reading (read_parquet):
  • Allow the user to specify a bbox to filter (similarly as read_file), and automatically translate this to a pyarrow filter predicate.
  • By default drop the bbox struct column when reading and converting to pandas? I would say yes (converting a struct column to pandas is expensive), but would there be a use case for providing an option to keep it?

[martinfleis]

This would be nice for 1.0.

I would also drop it by default but you may want to be able to read it in case of some debugging.

[nicholas-ys-tan]

anyone currently working on this one? This one looks interesting and would love to try tackling it

[jorisvandenbossche]

Definitely feel free to give it a try! And I can certainly give you some pointers or answer questions if needed

[nicholas-ys-tan]

Thanks Joris, apologies I haven't actioned yet, have been a bit under the weather. Aiming to have a draft PR up by end of this week.