Have you checked your vector tiles lately?
summary
Vector tiles often carry lots of unused geometry and properties that make mid-zoom tiles balloon; a practical workflow shrinks those peaks by linting tiles against the style, shaving fields and layers until the rendered output is identical. The team builds an API around a Mapbox-style shaver to compare tile linting before/after sizes, finds examples where a single layer (place names) adds ~1.2 MB to a tile, and demonstrates clear flattening of the tile-size curve for NYC; results feed into CI (GitHub Actions) to automate fixes and savings.
AI-written summary · about
chapters (8)
- 0:00Introduction to vector tiles
- 1:28Why tile size matters
- 3:12Tile size patterns explained
- 6:44Cartography and tile optimization
- 7:26Introducing the linting tool
- 10:55Insights from tile linting
- 13:21Automating tile optimization
- 14:28Q&A and conclusion
related talks
16mRaster tiles to vector tiles: how to cross the gulf
22mUnder the hood: Offline tilepacks v2
37mStreamline data pipelines with incremental tileset updates in the Mapbox Tiling Service (MTS)
15mMaking maps work on mobile
21mPoint-in-polygon lookups with vector tiles: quick, effective, and no server needed!
49mClustering in the Mapbox Tiling Service
46mCreating vector tilesets with the Mapbox Tiling Service
50mBehind the scenes of Mapbox Satellite imagery