Skip to content

Datasets

Everything measured on this site, downloadable under Creative Commons Attribution 4.0.

These are the files the measured pages are built from. Each one is written by the script that takes the measurement, so a figure quoted on a page and the figure in the file behind it cannot drift apart. Everything here is published under Creative Commons Attribution 4.0, so you can copy it, change it and use it commercially with credit. This list last changed on .

FileWhat it holdsFormat
corpus-manifest.jsonSixty-one small GeoJSON files, each isolating one edge case, with the clause each comes from and whether the specification calls it valid.JSON
corpus.csvThe same manifest as one row per fixture, for a spreadsheet or a dataframe.CSV
geojson.schema.jsonRFC 7946 as a draft 2020-12 schema, with the rules it cannot express listed in the document.JSON Schema
geojson.strict.schema.jsonThe same schema with closed objects and required properties, shaped for constrained decoding, which is the trick that makes a model emit only text matching a schema.JSON Schema
schema-variants.jsonWhat the strict variant refuses that the canonical one accepts, and why.JSON
simplification.jsonEight simplification backends over one dataset at four tolerances, with how many shared borders each one pulls apart.JSON
formats.jsonOne dataset written to eleven formats, and twenty-three probes converted out and back to see what each format drops.JSON
spatial-operations.jsonTwenty-four code samples for six operations, with the output each one printed when it was run.JSON
file-size.jsonWhat each optimisation saved on one dataset, uncompressed and gzipped.JSON
token-cost.jsonTokens per position by coordinate precision and vertex count, measured with tiktoken.JSON

Where these come from

Each file is the output of a page that explains what was measured and how. The simplification numbers come from simplification compared, and the format numbers from format conversion compared. The corpus and both schemas are described on the JSON Schema page, which measures one against the other.