OSM Leaderboard

OpenStreetMap contributor leaderboard — real-time edit stats with MapLibre GL 3D map, period filters, and PWA support.

DEMO

https://mapconcierge.github.io/osm-leaderboard/

Stack

• React + Vite + TypeScript + Tailwind CSS + shadcn/ui
• MapLibre GL JS + OpenFreeMap vector tiles
• TanStack Query for data fetching
• PWA (Service Worker + Web App Manifest)

Getting Started

pnpm install
pnpm --filter @workspace/osm-leaderboard run dev

Edit artifacts/osm-leaderboard/public/users.yaml to configure the contributor roster and hashtags.

Development

This project is maintained locally with Claude Code, not Replit. Clone the repo and work from a local checkout; run pnpm run typecheck before committing.

How It Works

Score formula

score = totalChanges + buildingsAdded×5 + wheelchairMapped×3 + hashtagChangesets×2

All four inputs — Changes, Buildings, Wheelchair, and Hashtags — are computed from the same set of a user's changesets, not separate queries, so they're always internally consistent.

• Changes: sum of changes_count across the user's changesets (OSM Changesets API)
• Buildings / Wheelchair: for each changeset, the app downloads its diff (/api/0.6/changeset/{id}/download) and counts building-tagged way/relation edits and wheelchair-tagged element edits in <create>/<modify> blocks. Diff results are cached in localStorage per changeset id (closed changesets are immutable, so the cache never goes stale).
• Hashtags: changesets matching any hashtag configured in users.yaml, checked against both the changeset's own hashtags tag and a lowercase substring search of the comment (not whitespace-tokenized — many Japanese-language comments have no space around a hashtag, e.g. #PLATEAUで測量).

Period filters & the 1,000-changeset cap

Daily/Weekly/Monthly/Yearly filters page back through a user's changeset history only as far as needed and are always exact. All Time pages back through full history but stops at a 1,000-changeset safety cap — some contributors have 10,000+ lifetime changesets, and downloading a diff for each one from the browser (no backend, no batching) isn't practical. This makes uncorrected "All Time" figures for very active mappers an undercount, not exhaustive.

The "All Time" HDYC correction

To compensate for that cap, public/hdyc-corrections.json stores per-user lifetime totals distilled from saved hdyc.neis-one.org ("How Did You Contribute") snapshots, which compute over a mapper's entire OSM history with no cap. HDYC has no public API, so a snapshot has to be captured manually per user (log in, search the user, "Save Page As" the rendered HTML — see pnpm --filter @workspace/scripts run extract-hdyc) and re-captured periodically to stay current. The raw saved snapshots themselves are published in HDCY2OSMlogs/ for transparency — every number in hdyc-corrections.json is traceable to a specific saved HDYC page.

For the All Time period only, the app takes Math.max(liveValue, hdycValue) — a floor, never an override, so a user who's kept mapping since their snapshot was taken still shows their fresher (higher) live numbers:

• totalChanges and buildingsAdded use this HDYC floor. No equivalent free, uncapped, per-user API exists for these — we looked at the ohsome API as a way to replace the manual snapshot step entirely, but its filter grammar has no per-user/contributor filter at all (its /users/count endpoints count distinct users editing a given area, the reverse of what's needed) and every endpoint requires a bounding box — there's no planet-wide, all-time, single-user query mode. So the manual HDYC snapshot remains the only source for these two fields.
• totalChangesets no longer depends on the HDYC snapshot: it's fetched directly and exactly from GET /api/0.6/user/{uid}.json (changesets.count in the response), which has no cap and needs no pagination — one cheap, CORS-enabled call per user (the numeric uid comes free off any changeset already fetched for that user). The HDYC floor is kept only as a fallback if that lookup fails (network error, or a profile opted out of public visibility).
• Hashtags and Wheelchair have no HDYC correction at all — HDYC's hashtag counts aren't directly comparable (see the extraction script), and HDYC doesn't track the wheelchair tag.

Mapper Level (HOT Tasking Manager)

Each contributor's card shows a Beginner/Intermediate/Advanced badge — linking out to their HOTOSM Tasking Manager profile — plus a progress meter toward the next level. It's based on lifetime changeset count (the same exact totalChangesets lookup described above), independent of the period filter currently selected. Thresholds match HOT Tasking Manager's own defaults (TM_MAPPER_LEVEL_INTERMEDIATE/TM_MAPPER_LEVEL_ADVANCED):

Level	Changesets
Beginner	0–249
Intermediate	250–499
Advanced	500+

Data Sources & License

• Contributor edit stats (changesets, changeset diffs) come from the OpenStreetMap API — © OpenStreetMap contributors, ODbL
• Code and content in this repository: CC0 1.0 Universal

Author

mapconcierge (Taichi FURUHASHI) — Furuhashi Lab, Aoyama Gakuin University

Links

• Demo: https://mapconcierge.github.io/osm-leaderboard/
• Repository: https://github.com/mapconcierge/osm-leaderboard