Skip to content

Download & CDN

There are three ways to get a file. They are not interchangeable.

jsDelivr CDN

Fast, globally cached, correct CORS headers. Best for browsers.

https://cdn.jsdelivr.net/gh/andresgmg/World-GeoJSON@main/<path>

20 MB hard limit

jsDelivr refuses to serve files larger than 20 MB. Chile's communes file is 70 MB, so the CDN route simply does not exist for it — you get an error, not a slow download. Use raw GitHub for large files, or use the simplified preview files, which are built specifically to stay well under this limit.

Pin a tag in production. @main follows the default branch, so a boundary correction merged upstream changes what your application receives, silently.

https://cdn.jsdelivr.net/gh/andresgmg/World-GeoJSON@v1.0.0/<path>

https://cdn.jsdelivr.net/gh/andresgmg/World-GeoJSON@main/<path>

Raw GitHub

https://raw.githubusercontent.com/andresgmg/World-GeoJSON/main/<path>

Works for files of any size and sends Access-Control-Allow-Origin: *, so browser fetch works. But it is not a CDN: no edge caching, and it is rate-limited. Fine for development, scripts and server-side downloads; poor for production browser traffic.

Git

For working with the data locally or in a pipeline.

The repository's packed size is only about 8.5 MB — the pretty-printed JSON compresses roughly 9:1 — so a full clone is faster than the file sizes suggest. What is slow is checking out 148 MB of working tree you may not need.

git clone https://github.com/andresgmg/World-GeoJSON.git

git clone --filter=blob:none --sparse https://github.com/andresgmg/World-GeoJSON.git
cd World-GeoJSON
git sparse-checkout set data/earth/CHL

--filter=blob:none makes the clone fetch file contents lazily, so you download only the blobs you actually check out.

git clone --filter=blob:none --no-checkout https://github.com/andresgmg/World-GeoJSON.git
cd World-GeoJSON
git sparse-checkout set --no-cone '/*' '!/data/**' '/data/**/manifest.json'
git checkout main

Gets every manifest without a single GeoJSON file. This is exactly what this site's CI does — the documentation build reads manifests only.

Checksums

Each dataset page publishes a SHA-256 for its file. Verify a download with:

Get-FileHash .\CHL_ADM1.geojson -Algorithm SHA256

sha256sum CHL_ADM1.geojson

Line endings will break your checksum

Git normalises line endings on checkout. On Windows a .geojson checked out as CRLF is one byte per line larger than the same file on Linux — for Chile's communes file that is a 1.8 MB difference, and a completely different hash.

The repository's .gitattributes pins *.geojson to LF precisely so that published checksums match everywhere. If your hash disagrees, check git config core.autocrlf before assuming the file is corrupt.