Tippecanoe replaced

[joeblew999]

I managed to replace the Tippecanoe c++ dependency with a golang one . There is an interface so both can be toggled and tests to make we the golang matches the c++

Also got it almost running inside Cloudflare as a worker also , but the SQLite dependency is blocking me . You can compile golang to WASM to run inside Cloudflare. The code can also still just run as normal go binary too . I might also add wazero support but I doubt it’s wanted ?

If this is useful I can try to do a PR …

Welcome any help or tips around this

[bdon]

Can you explain what project and dependency you are referring to?

[joeblew999]

https://github.com/felt/tippecanoe

https://github.com/joeblew999/ubuntu-website/blob/main/internal/airspace/gotiler/gotiler.go replaces the need for the c++ code from tippecanoe

https://github.com/joeblew999/ubuntu-website/blob/main/internal/airspace/tiler/tippecanoe.go uses tippecanoe if installed.

I am pretty new to PMTiles, so maybe I dont know what I dont know in terms of the problem I am solving, so feel free to question me.

[bdon]

Hi, thanks for open sourcing this, it looks useful.

The code you have now applies Douglas-Peucker simplification and drops any features that becomes zero length/area. This is a reasonable solution for small datasets but Tippecanoe does much more sophisticated merging and dropping selection which leads to better generalization of dense features at low zooms.

Also, I do not believe the Clip function in https://github.com/paulmach/orb/tree/master/clip is sufficient for guaranteeing output polygons can always be displayed correctly in MapLibre GL, for that they need to both Simple and Valid (non-intersecting). You always need a snap rounding step which is implemented either in Java JTS or https://github.com/mapbox/wagyu - I don't know of a pure Golang library that can handle all edge cases, if you can demonstrate an existing or new Go library that does this it would be very useful.

Your solution looks great for small datasets of points/lines but Tippecanoe is still the recommended tool for very large datasets. Is that helpful in figuring out in how the project fits in?

[joeblew999]

That’s really useful . Thanks for taking the time to look .

Happy to break this out to its own repo too and implement any suggestions like you have brought up.

I was not sure if anyone wanted this