Overture Buildings & Places (Cloud-Native Geo experiments)

This data is currently just a set of experiments using the new overture maps. See their github repo for more info. My hope is that this can try out some better ways to distribute the data, and to build an ecosystem of tools and data distributions on top of the core.

There are two main folders that I recommend starting with:

• geoparquet-country-quad/ is the Overture Building dataset, repartitioned by country, split by quadkey when larger than ~2 gigs, distributed as GeoParquet with row size group of 20,000.
• places-geoparquet-country/ is the Overture Places (POI's) dataset, repartitioned by country. There was no need to split by quadkey as the whole dataset is only ~3 gigabytes.

There's also a PMTiles of the buildings dataset in the root directory. You can check it out at https://observablehq.com/d/2dd90b64498c3864. I do not have a PMTiles up of the places dataset, but you can get one from https://r2-public.protomaps.com/protomaps-sample-datasets/overture-pois.pmtiles, and view that PMTiles at https://bdon.github.io/overture-tiles/places.html. I

If you just want to use Overture data those are the two to focus on. The other folders are a number of experiments and interim processing steps.

• geoparquet-refined/ is the same files as the original but turned into Geoparquet, with columns added for iso_country and quadkey, and data ordered by quadkey.
• duckdb/ contains a single DuckDB database. It's pretty easy to make this yourself, it is just all the data in geoparquet refined, loaded up into a DuckDB. I'm not sure if distributing DuckDB databases directly is useful. But the file is definitely useful to partition data in different ways.
• geoparquet-country-quad-2 I recommend ignoring. It's some alternate partitioning, and the dataset is incomplete. But until I get a set of different repartitioned datasets I'll keep it.