Docs
Docs › Formats › GeoParquet

GeoParquet

GeoParquet is a compact columnar format that stores vector geometries alongside tabular attributes. It is now a common cloud-native interchange format for vector data: the Overture Maps Foundation publishes its global releases in GeoParquet, and tools such as DuckDB (with spatial), GeoPandas, BigQuery, Athena and Synapse can query GeoParquet datasets directly.

File extensions: .parquet, .geoparquet · Read: ✓ · Write: ✓ · Multi-layer: no (one layer per file)

CLI examples

mapshaper roads.parquet -info
mapshaper roads.parquet -o roads.geojson
mapshaper roads.geojson -o roads.parquet

Format-specific input options

There are no GeoParquet-specific -i options.

Format-specific output options

  • compression=snappy|zstd|none — select Parquet column compression. The default is snappy. Use none to write uncompressed Parquet data. Snappy is designed for fast compression and decompression at moderate compression ratios. ZSTD is slower to compress but typically produces smaller files and is often preferred when file size matters more than write speed (e.g. files published for download or stored in a cloud bucket).
  • level= — set the ZSTD compression level when using compression=zstd. Valid values are integers from 1 to 22. If omitted, the ZSTD library default is used.

Practical notes

  • Mapshaper imports one geometry column per file (the GeoParquet primary_column when present). Additional geometry columns are currently ignored.
  • Mapshaper writes a single WKB geometry column named geometry.
  • There is no fixed maximum GeoParquet file size. Limits are memory-driven:
    • In the regular CLI, the limit is your Node heap. For very large files, use mapshaper-xl, which starts with an 8 GB heap by default (and can be increased, e.g. mapshaper-xl 16gb ...).
    • In the web app, very large imports (often in the hundreds of MB, depending on browser/device) can run out of memory and crash the tab. For large GeoParquet files, prefer the CLI / mapshaper-xl.

External resources

Edit this page on GitHub