Richard Fairhurst

My C++ isn't great either tbh! I suspect stxxl would be easiest as it's pretty much a drop-in replacement and it's easily configurable for your own disk/memory needs.

As a current benchmark: producing an OpenMapTiles-schema mbtiles from great-britain-latest.osm.pbf (around 1GB) requires 22GB memory and takes around 2 hours on my slowish machine. I'm sure it's possible to reduce...

Wow, interesting! Tried it on a couple of small extracts. On a Wiltshire extract it works well, though doesn't output any point geometries (e.g. placenames), just lines/polygons. On an Oxfordshire...

> And why not the linestring/multipolygon ? I think that might be an artefact of an earlier refactoring - originally everything was stored as *10000000 so they could be an...

It is - more than I expected. I did wonder whether it might be worth having a hybrid approach - keeping the existing tileIndex for most objects (which are small),...

I like the principle. Possibly a pipeline like .pbf -> tilemaker -> .mvt -> .geojson, and then we parse the geojson to make sure the features are correct. Testing the...

One possibility that springs to mind is that tilemaker's .pbf reader only reads "dense nodes". There's another way of writing nodes into a .pbf file, which I've never seen in...

The most recent version is the one in this repository, systemed/tilemaker . The Docker build is not formally supported but you're welcome to try it and report any failures you...

Thanks - will take a look when I've got a moment.

Do you have a working MBGL style that I can test this with (to save having to reverse engineer one)?