Provide example to read GeoParquet files (ogr2ogr) under Linux and Windows

[Stefan (sfkeller)]

It would be nice to have also an example, of how to use ogr2ogr to read GeoParquet files on Windows and Linux, since GDAL/OGR is a gold standard of open source converters.

My first attempts have been failing under Windows when reading "PARQUET:/vsis3/overturemaps..." already. Using UBUNTU Linux it works - even if it's snake-slow.

Below is what I have tried.

--Stefan

Trying to extract GeoParquet file from S3 storage using GDAL/OGR on Windows

0. Installation and Launch (Windows)

• Installed latest QGIS-Version 3.42.3-Münster (which includes GDAL/OGR)

• Launched OSGeo4W.bat

SET AWS_NO_SIGN_REQUEST=YES

ogrinfo --version
=> GDAL 3.10.3, released 2025/04/01

ogrinfo --formats | findstr /C:"Parquet"
=> Parquet -vector- (rw+v): (Geo)Parquet (*.parquet)

1. Get 100 places from within bbox of Switzerland

ogr2ogr -f GPKG ovm_places_ch_100.gpkg -spat 5.9 45.8 10.5 47.8 "PARQUET:/vsis3/overturemaps-us-west-2/release/2024-08-20.0/theme=places/type=place" -limit 100

=> ERROR 1: Unable to open datasource `PARQUET:/vsis3/overturemaps-us-west
=> (UBUNTU: Works, but duration ist about 15 min.)

2. Get all restaurants within bbox of Switzerland:

ogr2ogr -f GPKG ovm_restaurants_ch.gpkg -spat 5.9 45.8 10.5 47.8 "PARQUET:/vsis3/overturemaps-us-west-2/release/2024-08-20.0/theme=places/type=place" -where ""categories.primary" = 'restaurant'"

=> ERROR 1: Unable to open datasource `PARQUET:/vsis3/overturemaps-us-west
=> (UBUNTU: Works, but duration ist about 15 min.)

[Dana Bauer (danabauer)]

Looks like the problem might be the way you're constructing the path to the data. Try this:

ogr2ogr --config AWS_NO_SIGN_REQUEST=yes -spat 5.9 45.8 10.5 47.8  -f GPKG ovm_restaurants_ch.gpkg /vsis3/overturemaps-us-west-2/release/2024-08-20.0/theme=places/type=place

[Stefan (sfkeller)]

Thx Dana Bauer (@danabauer) - You made my day.