What this tool does
GeoParquet files are not human-readable. Unlike GeoJSON, you cannot open one in a text editor and see coordinates, and most everyday GIS desktop tools still do not open the format directly. That leaves a gap between having a .parquet file and actually knowing what is in it: how many features, what geometry type, which columns, what area it covers.
This viewer closes that gap. Drop in a file and it reads the Parquet container, checks the file's own GeoParquet metadata to find the geometry column and its encoding, decodes every WKB (Well-Known Binary) value back into real coordinates, and draws the result on a map. Click any feature to see its full attribute table, adjust the color and style to make patterns easier to see, and export everything back out as a GeoJSON file when you need it in a format more tools already understand. No file is uploaded anywhere; the Parquet reader is a WebAssembly build that runs entirely on your machine.
How the viewer reads a GeoParquet file
The process runs in three stages. First, a WebAssembly build of the Rust Parquet library reads the raw column data and hands it over as an Apache Arrow table, the same in-memory format used by GeoPandas and DuckDB internally. Second, the viewer reads the table's schema metadata for a key named "geo", the small JSON block every GeoParquet file carries that names the primary geometry column, states its encoding, and lists which geometry types are present. If that key is missing, the file is a plain Parquet file, not GeoParquet, and the viewer says so rather than guessing.
Third, the viewer walks every row of the geometry column, which holds raw WKB byte values, and decodes each one back into coordinates: a one-byte flag for byte order, a four-byte code for the geometry type, then the coordinates themselves as pairs of 8-byte numbers. The same decoder also recognizes the SRID flag PostGIS adds to its EWKB output and the Z/M dimension markers from the ISO WKB standard, so files exported from a spatial database read correctly too. The properties in every other column are read the same way Arrow already stores them and attached to each feature, and the whole thing becomes a standard GeoJSON FeatureCollection ready to draw on the map.
Step-by-step guide
Click "Upload" and choose a .parquet file, drag and drop one onto the map, or click "Sample Data" to load a small ready-made example.
The file is decoded entirely in your browser. The File Info panel shows the feature count, geometry type, geometry column name, CRS, and file size straight from the file's own metadata.
Click any point, line, or polygon on the map to open a popup listing every property stored for that feature.
Use the Style panel to change the color, line thickness, and fill opacity applied to every feature on the map.
Click "Download as GeoJSON" to export every decoded feature as a standard GeoJSON file.
Real-world use cases
Where opening a GeoParquet file directly, without writing any code, actually helps.
- Spot-checking a data pipeline's output. If a script or ETL job writes GeoParquet files, dragging the latest output straight into this viewer is a fast way to confirm the geometry and attributes look right before trusting the file downstream.
- Exploring an open dataset before writing any code. Datasets like Overture Maps or the collections on Source Cooperative ship as GeoParquet. Downloading a small extract and viewing it here shows exactly what columns and geometry types are inside, before setting up DuckDB or GeoPandas.
- Verifying a GeoJSON-to-GeoParquet conversion. After converting a file with a tool like this site's GeoJSON to GeoParquet converter, loading the result back into this viewer confirms the round trip preserved every feature and property correctly.
- Reviewing data before publishing it as open data. Municipal and research teams increasingly publish GeoParquet alongside GeoJSON. A quick visual check here catches obvious problems, like an empty geometry column or a bounding box in the wrong hemisphere, before the file goes live.
- Teaching or demonstrating the GeoParquet format. Because the file info panel shows the primary geometry column, CRS, and encoding straight from the file's own metadata, this is a convenient way to show students or colleagues what actually sits inside a GeoParquet file.
- Recovering usable data from a file with no other viewer at hand. Not every GIS desktop tool reads GeoParquet yet. When you just need to see what is in a .parquet file right now, or pull it back out as GeoJSON for a tool that only understands that format, this saves installing anything.
Frequently asked questions
What kind of file can I open here?
Any file that follows the GeoParquet specification: a regular Apache Parquet file with a schema-level "geo" metadata key describing a geometry column. Files produced by GeoPandas, DuckDB's spatial extension, ogr2ogr, Apache Sedona, or this site's own GeoJSON to GeoParquet converter all work.
Which geometry encoding does the viewer support?
WKB (Well-Known Binary), which is the default and by far the most common encoding used by GeoParquet files today. GeoParquet 1.1 also defines separate native Arrow encodings for each geometry type; files using those are not supported yet, and the viewer will tell you clearly if it hits one.
Does it support Point, Line, and Polygon files, including multi-part geometries?
Yes. Point, LineString, Polygon, MultiPoint, MultiLineString, and MultiPolygon are all decoded correctly, including files that mix more than one geometry type in the same column.
What happens to 3D (Z) or measured (M) coordinates?
The reader recognizes both the PostGIS EWKB style and the ISO WKB style of marking Z and M dimensions, and skips those extra values correctly so the rest of the geometry still decodes. Since GeoJSON output and the map are inherently 2D, only X and Y are kept for display.
How does the viewer know which column holds the geometry?
It reads the "primary_column" field from the file's "geo" metadata, which every valid GeoParquet file must include. This is also how the viewer reports the CRS and confirms the encoding before attempting to decode anything.
What if my file has no "geo" metadata at all?
Then it is a plain Parquet file, not a GeoParquet file, and the viewer will say so rather than guessing. A geometry column with no accompanying metadata could be WKB, WKT, or something else entirely, so there is no safe way to decode it automatically.
Is there a file size limit?
Uploads are capped at 100 MB. Very large files can also be slow to render on the map because every feature becomes a Leaflet layer; for files with hundreds of thousands of rows, expect the initial render to take a few seconds.
Can I get my data back out as GeoJSON?
Yes. The "Download as GeoJSON" button exports every decoded feature, geometry and properties included, as a standard GeoJSON FeatureCollection you can open in any GIS tool.
Is my file uploaded to a server?
No. Reading the Parquet file, decoding WKB geometry, and rendering the map all happen locally in your browser using WebAssembly. The file never leaves your device.
Reading runs entirely in your browser using Apache Arrow and a WebAssembly build of parquet-wasm. No file is ever uploaded to a server. Map tiles © OpenStreetMap contributors, drawn with Protomaps (hosted by KONAKTIVE), with CARTO as the fallback.
