Closed dabreegster closed 10 months ago
Hey I am going to try and take a swing at doing bbox filtering using rstar. I have used it on some projects before with good results.
Yes please! Maybe I don't understand coordinate systems well enough, but can this be done with WGS84, or do we have to project to Web Mercator or something else first?
(I won't pick this up until Monday again, no fear of merge conflicts)
I think given that we need to convert to WebMercator anyway to generate the tiles, it makes sense to
One thing that might be interesting to experiment with is generating the tiles in parallel (with rayon) and storing them somewhere intermediary before constructing the full pm tiles at the end? That's an enhancement though
@stuartlynn , there's some chatter in the georust discord about possible distance metrics that could be useful: https://discord.com/channels/598002550221963289/963346313708195880/1167857249977372733
@stuartlynn , there's some chatter in the georust discord about possible distance metrics that could be useful: https://discord.com/channels/598002550221963289/963346313708195880/1167857249977372733
Not sure that link is working for me. What were you thinking if using the different distance metrics for?
Hmm, try joining https://discord.gg/Fp2aape and checking #random
around 10/28. Someone pointed out https://github.com/vipera/cheap-ruler-rs. I think web merctor + rstar makes sense to me though.
This is an experiment to generate pmtiles directly in Rust, not using tippecanoe. It's specialized for the use case in od2net -- a bunch of tiny LineStrings representing segments of the road network. The "importance" of the roads is given by a count property in the GeoJSON data, so this should be used to figure out which features to drop at low zooms.
CC @stuartlynn
"Correctness" to use in the od2net viewer:
Performance of the tool itself:
Logistic: