https://user-images.githubusercontent.com/1664407/211030226-f2c855b0-cb55-461a-bf98-4e990ac52162.mov

Repro near the Lea bridge in Hackney: https://acteng.github.io/atip/scheme.html?authority=Hackney#18.6/51.5620442/-0.0459354 I've hit this elsewhere too.

Ideas: visualize the circle around the cursor, and the circles we snap to. Make sure we're not snapping so eagerly when zoomed in like this?

I'm passively looking into a solution for my need for an offline routing engine. My research can be read here: https://github.com/IsraelHikingMap/Site/issues/1636 Generally speaking the following are my requirements:

1. I should be able to run this routing engine in Android and iOS, or using Ionic capacitor/webview (pure javascript/WASM)
2. This routing engine should allow profiles, my use case is: Hike, Bike, 4 wheel drive profiles
3. The results from the routing engine should have elevation data
4. The setup time should be minimal - preferably using a pre-process step where the data is saved in a format the routing engine can read fast
5. [Optional] If the routing engine can receive RGB Terrain tiles and MVT tiles and create a route using this input the need for a dedicated file might not be necessary as I have this data available offline/in memory/cached.
6. [Optional] Allow running this on a server using a docker container to allow similar results when doing the routing while offline and while online

My current setup: We are maintaining a docker image of GraphHopper here: https://github.com/IsraelHikingMap/graphhopper-docker-image-push This image can do two things mainly:

1. Download an OSM pbf file and create a directory, this directory has the graph data in a format graphhopper can read fast
2. Serve a routing endpoint on the server

What we do is basically run every hour a build of the graph directory and restart a container to use the newly created graph directory (there is another container that runs in parallel to avoid a time where this service is unavailable).

I thinks this concept is a solid one - you pre-process the data without affecting anyone, once the data is processed you can read it fast and use it fast for the routing engine. If such a directory/file can be made available to download and use on the client device (using fetch, or a native file transfer) it would be great.

I'll be happy to help although I have little to no knowledge in Rust, but after using my app in a two days hike where I needed this offline routing feature I've matured to the point where this is probably the next thing for me to invest in.

As a side note, there's a format I read somewhere called openLR which might be interesting to look into for this package: https://github.com/itinero/OpenLR

This what I can see as a full implementation, having said that, it might be good to start with a "poor man's routing" where things just snap to a route while you are offline (some kind of fallback) and later on make it better, or not. I have the same concept when it comes to search where the search in the backend is using elasticsearch and when offline I use minisearch.

Main project in Github: https://github.com/IsraelHikingMap/Site Website: https://israelhiking.osm.org.il/

osm2streets may still filter out private roads, service roads, driveways, etc in ways that don't quite match how people here want to use it. Should we just use the osm_reader::Document, look through ways, and look for highway?

Already doing horrible CSS things to override stuff in ATIP. It'd be simpler to optionally have a callback that constructs the DOM elements for both states?

Most of the logic of this plugin is in the frontend half, not the routing. Why limit it to MapLibre? https://github.com/JamesLMilner/terra-draw is much more solid foundation to build on