elivagarShortbread vector tiles at planet scale

Reads OSM PBF, writes PMTiles v3. The full planet in 9m32s on a 16-core desktop, peaking at 12.7 GB of RAM.

Get Started

CLI Reference

GitHub

Planet on a Desktop

A 90.5 GB planet PBF to a 58.7 GiB archive in 571.7s on 16 cores, peaking at 12.7 GB of RAM. 269.8M tiles addressed, 52.2M unique, 2.52 billion features.

Full Shortbread Schema

All 26 layers, z0 through z14, with ocean coastlines served from a durable precomputed world artifact that most runs never recompute.

Gated on What Renderers See

The standing correctness gate decodes every tile the way MapLibre does, groups rings with its own classifyRings, and tessellates with earcut. Zero deviant polygons is the bar.

What it does

elivagar turns an OpenStreetMap PBF into a PMTiles v3 archive carrying the Shortbread schema: 26 layers, zoom 0 to 14, one file you can serve directly.

elivagar run denmark-latest.osm.pbf -o denmark.pmtiles \
  --ocean z0-z7:data/simplified-water-polygons-split-3857/simplified_water_polygons.shp \
  --ocean z8-z14:data/water-polygons-split-3857/water_polygons.shp \
  --ocean data/ocean-tiles.pmtiles

Why the memory number matters

Tile generation at planet scale is usually a question of how much RAM you can rent. elivagar is built the other way round: every stock in the pipeline is bounded by construction, so the peak does not track the input size. The planet build above used 12.7 GB on a host that had 30.5 GiB available.

That constraint is why the pipeline looks the way it does - an external merge sort partitioned by Hilbert tile-id range, a byte-budgeted claim window in assemble, a relation-block buffer that spills at a cap, and a memory-mapped way index rather than an in-RAM one.

Honest comparisons

planetiler's published planet table lists 2h38m on a 16 cpu / 32 GB machine, and 19 minutes on 192 cores / 720 GB. Three things belong next to any such number, and this project would rather state them than be corrected later:

Our figure is for enriched input. The pbfhogg preprocessing pass is not in it; end to end the honest number is roughly 20 minutes.
The profiles differ. Shortbread here, OpenMapTiles there. Not the same work.
The planet snapshots differ. Ours is seq 4912, 2026-02-23.

Treat 571.7s as a measurement of this pipeline, not as a record claim.