Overture July 2024 Canadian Addresses

issue.md

DuckDB v1.0.0 1f98600c2c

$ cd '/mnt/j/gis/Global/overture/2024_07/theme=addresses/type=address'
$ COUNTRY=CA
$ echo "COPY (
              SELECT h3_cell_to_boundary_wkt(
                          h3_latlng_to_cell(bbox.ymax, bbox.xmax, 5))::geometry geom,
                      COUNT(*)
              FROM read_parquet('*.parquet')
              WHERE country = '$COUNTRY'
              group by 1
          ) TO '../../addresses.$COUNTRY.gpkg'
              WITH (FORMAT GDAL,
                    DRIVER 'GPKG',
                    LAYER_CREATION_OPTIONS 'WRITE_BBOX=YES')" | ~/duckdb

Floating point exception

Using S3 as a source works fine, it's just with a local file that the issue comes up. I'll raise a ticket with DuckDB. duckdb/duckdb-spatial#368

For the JSON issue, scan the whole dataset, it'll be one rogue JSON record. It's unlikely it's in the first 10 records.

Looks like direct from S3 also resolves the JSON issue also. That's weird.

D SELECT   country,
           COUNT(*) AS num_addrs,
           sources->0->'dataset' AS source
  FROM     READ_PARQUET('s3://overturemaps-us-west-2/release/2024-07-22.0/theme=addresses/type=address/*.parquet')
  WHERE country = 'US' GROUP BY 1, 3
  ORDER BY 1, 2 DESC;
100% ▕████████████████████████████████████████████████████████████▏
┌─────────┬───────────┬────────┐
│ country │ num_addrs │ source │
│ varchar │   int64   │  json  │
├─────────┼───────────┼────────┤
│ US      │  78078341 │ "NAD"  │
└─────────┴───────────┴────────┘