If I understood correctly, this is automatically done if a non-None `geometry` is provided after #58, right @sphamba ?
Basically if df_attrs is a GeoDataFrame, we are thus using GeoDataFrame.to_parquet, that already takes care of writing the geometry columns. Should we use "geoarrow" instead of the default "WKB" (see the geometry_encoding argument at https://geopandas.org/en/stable/docs/reference/api/geopandas.GeoDataFrame.to_parquet.html)?
In any case, I doubt this is any performance bottleneck for most use cases and I would like to avoid any premature optimization, but I can (i) test the effects with a proper benchmark and (ii) add a to_parquet_kwargs argument to the to_tstore methods which is forwarded to df_attrs.to_parquet.
Originally posted by @martibosch in #19 (comment)
Basically if
df_attrsis aGeoDataFrame, we are thus usingGeoDataFrame.to_parquet, that already takes care of writing the geometry columns. Should we use "geoarrow" instead of the default "WKB" (see thegeometry_encodingargument at https://geopandas.org/en/stable/docs/reference/api/geopandas.GeoDataFrame.to_parquet.html)?In any case, I doubt this is any performance bottleneck for most use cases and I would like to avoid any premature optimization, but I can (i) test the effects with a proper benchmark and (ii) add a
to_parquet_kwargsargument to theto_tstoremethods which is forwarded todf_attrs.to_parquet.Originally posted by @martibosch in #19 (comment)