benchmark Command¶

Benchmark GeoParquet conversion performance across different methods.

Quick Reference¶

gpio benchmark INPUT_FILE [OPTIONS]

Description¶

Tests available conversion methods (DuckDB, GeoPandas with Fiona/PyOGRIO, GDAL ogr2ogr) on an input geospatial file and reports time and memory usage.

Available Converters¶

Converter	Description	Install
`duckdb`	DuckDB spatial extension	Always available
`geopandas_fiona`	GeoPandas with Fiona engine	`geopandas`, `fiona`
`geopandas_pyogrio`	GeoPandas with PyOGRIO engine	`geopandas`, `pyogrio`
`gdal_ogr2ogr`	GDAL ogr2ogr CLI	System GDAL installation

Install all optional converters:

uv pip install geoparquet-io[benchmark]
# or: pip install geoparquet-io[benchmark]

Options¶

Option	Default	Description
`--iterations N`	3	Number of iterations per converter
`--converters LIST`	all available	Comma-separated list of converters to run
`--output-json PATH`	-	Save results to JSON file
`--keep-output DIR`	temp (cleaned up)	Directory to save converted files
`--warmup/--no-warmup`	enabled	Run warmup iteration before timing
`--format table\\|json`	table	Output format
`--quiet`	-	Suppress progress output

Examples¶

Basic Benchmark¶

gpio benchmark input.geojson

Runs all available converters with 3 iterations each.

Specific Converters¶

gpio benchmark input.shp --converters duckdb,geopandas_pyogrio --iterations 5

Save Results¶

gpio benchmark input.gpkg --output-json results.json --keep-output ./output

Saves JSON results and keeps the converted Parquet files.

JSON Output¶

gpio benchmark input.geojson --format json

Output¶

Table Format (default)¶

======================================================================
BENCHMARK RESULTS
======================================================================

File: ARG.geojson
  Format: .geojson
  Features: 3,486,802
  Size: 1120.15 MB
  Geometry: LINESTRING

Converter                 Time (s)           Memory (MB)
-------------------------------------------------------------
DuckDB                    29.751 +/- 0.443   0.0 +/- 0.0
GeoPandas (PyOGRIO)       59.957 +/- 1.078   1196.7 +/- 0.0

Fastest: DuckDB (29.751s)

JSON Format¶

Includes environment info, file metadata, raw results per iteration, and aggregated statistics.

Interpreting Results¶

Time: Mean elapsed seconds +/- standard deviation
Memory: Peak memory usage in MB (Python tracemalloc for in-process, psutil for external)
DuckDB shows 0 MB because it manages memory outside Python's allocator