h3

> What an amazing tool that can completely change function names when it converts from C to Go. How can one read the code of the benchmark, then switch into virulent sarcasm mode without trying to understand the code? And seeing "+1" comments without any effort to understand is also disheartening. The blog post had a link about the Go helper functions the author used. It lands on https://github.com/akhenakh/goh3/blob/main/h3.go This shows that the `FromGeo()` function used by the Go benchmark is a helper that calls transpiled functions. The benchmark code itself was of course not transpiled, so the sarcasm was unneeded and wrong. If anyone wants to dig in deeper, the C function `latLngToCell()` calls 2 functions, see https://github.com/uber/h3/blob/master/src/h3lib/lib/h3Index... The Go function `XgeoToH3()` allocates a TLS then calls the

Looks really awesome! Very quick responses. I would be interested in seeing the implementation of the service, interesting choice going with Swift. I'm guessing your using something like Vapor for hosting the API? How are the files designed? I'm guessing you have some cheap way of mapping (lat, long) into a file location? Maybe just by truncating the coordinates to your precision and mapping them onto the file? Using some fancy-pants Uber Hexagons[0]. How is the forecast delivered? Hmm! Many questions :-). I've been thinking lately of similar (but not for weather) "geographical" API's, and how to store/query data cheaply, so this interested me :-) [0]: https://github.com/uber/h3

Hexagons. There's a library available at https://github.com/uber/h3 that lets you partition the earth into hexagons (minus the required 12 pentagons) at various resolutions. It's spiffy. I had the hardest (read impossible) time when I first took a look at it getting it to compile into a mobile game. It worked fine in desktop, but not on mobile. So I punted and converted the whole thing into C#. It's a hot mess currently, but it works. However, the folks who created H3 added functionality and some optimizations that I want, so I'm rebuilding it, partly with dealing with some of the issues that have come up with my game during development. It's coming along smoothly, though the unit tests are what's slowing me down at this moment. Once that's done, I'm going back to my game. Covering the world in brightly colored hexagons as reality st

The dual of a geodesic polyhedron, e.g. an icosahedron whose faces have been subdivided is in fact a common and excellent choice for geospatial applications, with https://github.com/uber/h3 being a good implementation. You can also subdivide the faces of a cube into smaller squares with a quadtree, e.g. https://s2geometry.io/ The Disdyakis triacontahedron's faces are probably too long and skinny for most geospatial applications though.

Why do you consider H3 to be on life support? It's basically a finished spec with actively developed implementations. https://github.com/uber/h3

I hate Uber, but I now opened this issue suggesting a way to massively increase the performance of this anyway: https://github.com/uber/h3/issues/258 (Note: by "massively" I mean pushing it to O(1) for the fundamental coordinate operations. So, perhaps read that as "most massively".)

Converting lat/longs to Uber's H3 [1][2] system might prove useful here. Because H3 indexes are hierarchical, you can easily truncate them to obtain less precision. And then convert them back into lat/longs after truncation. [1] https://github.com/uber/h3 [2] https://eng.uber.com/h3/

I used this a lot when I worked at Uber. Very similar but maps to variably sized Hexagons which were nice for things like surge/dispatch algorithms: https://github.com/uber/h3

Shameless plug: why not use H3 (https://github.com/uber/h3)?

Yep, Dr. Sahr came to the office a few times when we were first getting the original version done. He implemented the core of the algorithm and we took on many of the utility methods, most of them in https://github.com/uber/h3/blob/master/src/h3lib/lib/algos.c Then we dug in on code formatting, performance tuning (we've removed almost all of the H3IndexFat struct representation usage and switched most things to bitwise operations), and testing coverage. As for the bindings, I don't know which will be open sourced, so I can't say more, but we had asked Dr. Sahr to make sure the API itself never allocates memory, it must be passed in allocated memory, so it's not too hard to make bindings that work with both manual memory management and garbage-collected languages. Making the emscripten-compiled front-end Javascript "binding" was a lot of f

There are pentagons: > The first H3 resolution (resolution 0) consists of 122 cells (110 hexagons and 12 icosahedron vertex-centered pentagons), referred to as the base cells. From: https://github.com/uber/h3/blob/9df3940473e2f0446cf38e22443c...

Thanks for that link. Also found this which goes into some more detailed comparison with rectangular grids: https://github.com/uber/h3/blob/master/docs/doxyfiles/usecas...

I have never heard of this library before, which is surprising, because I've been looking hard for such a hexagonal privacy-fencing system for a long time! I can't find anything that explains how the partitioning in geohex works, but the Uber system makes it clear that they are using Dr. Sahr's partitioning scheme: https://github.com/uber/h3/blob/master/docs/doxyfiles/h3inde...