Brandon Liu

hmm, yeah I was worried about this. Do you depend on reading the generated `tilestats` created by tippecanoe? If not, can you use `--no-tile-stats` to see if that gets under...

This is resolved as of spec v3, which supports an arbitrary size of JSON metadata and separates out essential fields (minzoom, maxzoom, tile type, compression) into a separate header.

This is addressed in the upcoming v3 design: https://github.com/protomaps/PMTiles/blob/spec-v3/spec/v3_design_doc.md 1. This goes one step further than combining x and y by combining z,x,y into a single value. 2. I think...

see notes in https://github.com/protomaps/PMTiles/issues/41 as well. experimentally - running https://github.com/protomaps/PMTiles/tree/spec-v3/spec/converter on the z14 planetiler output can get to an effective

* I'm leaning towards making the index and tile compression be configurable, with the compression type stored in header. Brotli does not have a breadth of implementations across different languages...

Yes, that decoding routine is roughly right, though you can use a for-loop and an accumulator: https://github.com/protomaps/PMTiles/blob/spec-v3/spec/converter/tile_id.go#L71 I think the dense format would complicate the encoding routine, which is currently...

Thinking a bit out loud, the way I would probably do this is to re-use RunLength for dense sequences: a positive RunLength N means that N (offset,length) pairs follow corresponding...

@msbarry has benchmarked the decoding/encoding routine when the Hilbert code was still in here https://github.com/onthegomap/planetiler/pull/266/files#diff-9e784e32e5ed72b6cd67c89789e4b4f0fe3082810b9f839b357b0ccd11de58d4 and the decoding step was trivial relative to other things happening.

There is no concept of zoom in leaves as that is implied by the TileId. A directory might cross zoom levels; this can be as many as eight zoom levels...

There isn't much overlap between the OSM PBF format and PMTiles, as PMTiles is restricted purely to Z/X/Y addressing and not geometry or geographic coordinates - for that you might...