mikkeli
8ee3d4cde5
Verified on the Orin. A Shikoku extract yields only admin_level=4 prefecture rows with no country record, and drive_on_right=1 on every one, so tiles built against it route as right-hand traffic. The full Japan extract yields the JP country row with drive_on_right=0. The failure is silent, so every build now verifies the country row before tiles are considered usable. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013kj58Jugkic783b8ZrtTVZ
1.4 KiB
1.4 KiB
type
| type |
|---|
| default |
Map minting guidelines
1. Artifacts are never committed
- Routing tiles and basemaps are large binaries rebuilt from upstream. This repo holds the scripts that produce them and a report of the procedure, nothing else.
- Every build records the source PBF's date. An artifact whose provenance is unknown is an artifact to rebuild.
2. Build the admin database first, and from the full extract
valhalla_build_adminsis what determines driving side, and it is a separate step before the tile build.- Always build it from
japan-latest.osm.pbf, never a sub-region. Driving side is a country-level property. A Shikoku extract yields onlyadmin_level=4prefecture rows, no country record, anddrive_on_right=1on every one — measured, not theorised. - The admin build and the tile build are separate invocations against the same config, so routing tiles may still come from a small regional extract while the admin database comes from the whole country.
- Verify, do not assume. Every build checks that a country-level row exists and
reports
drive_on_right=0before the tiles are considered usable. This failure is silent otherwise: routing simply comes out mirrored.
3. Tiles ship as a directory, not a tar
- Once for RAM on an 8 GB board, once because
rsyncdeltas a tree of changed files where a multi-gigabyte tar forces a full checksum at both ends.