Spatial SQL

The mapular-geo server is DuckDB with spatial and H3 extensions. Use it to query, aggregate, and export Parquet files downloaded by mapular-mpoi.

No credentials needed. Works with files in /tmp, $TMPDIR, and ~/Downloads.

Tools

query(sql)

Run a DuckDB SQL statement. Returns a file path and preview rows.

export(sql, format)

Export SQL results to parquet, csv, or geojson.

geocode(address)

Convert a free-form address to lat/lon. Returns lat, lon, display_name, confidence, and provider.

Examples

Filter by category

SELECT name, latitude, longitude
FROM read_parquet('/tmp/pois.parquet')
WHERE overture_group_secondary = 'health'
LIMIT 20

Proximity query

SELECT name, latitude, longitude,
  ST_Distance(
    ST_Point(longitude, latitude),
    ST_Point(13.377, 52.516)
  ) AS dist_deg
FROM read_parquet('/tmp/pois.parquet')
WHERE ST_DWithin(
  ST_Point(longitude, latitude),
  ST_Point(13.377, 52.516),
  0.005
)
ORDER BY dist_deg

H3 aggregation

SELECT
  h3_latlng_to_cell(latitude, longitude, 7) AS h3_index,
  COUNT(*) AS poi_count
FROM read_parquet('/tmp/pois.parquet')
GROUP BY 1
ORDER BY poi_count DESC
LIMIT 10

Export to GeoJSON

export(
  sql="SELECT name, latitude, longitude FROM read_parquet('/tmp/pois.parquet') LIMIT 100",
  format="geojson"
)

Geocode an address

geocode("Brandenburg Gate, Berlin")
→ { lat: 52.5163, lon: 13.3777, confidence: "high" }

File path restrictions

query restricts file reads to the system temp directory (/tmp or $TMPDIR) and ~/Downloads. To allow an additional path, set MAPULAR_GEO_DATA_DIR=/path/to/data in the server's environment.