H. Thomson Comer

Having chatted with @trxcllnt about `arrow` yesterday for a while, I'm inclined to go with your Option #3 of potentially making each of the `xy` geometry columns into a Union...

I'm working on a pyarrow implementation of GeoArrow with geoPandas and cuSpatial. Today I'm dropping all my custom logic for building arrow buffers, https://github.com/rapidsai/cuspatial/issues/582. When that's finished I'll be dropping...

Heya! http://www.github.com/rapidsai/cuspatial/pull/585 has most of the final implementation of our GeoArrow. I wanted to share with you https://github.com/rapidsai/cuspatial/blob/8718c149ec56da2c408f7b7214f927f89e51ba5d/python/cuspatial/cuspatial/geometry/pygeoarrow.py, the `PyGeoArrow` spec for each nested list. These lists ended up not...

I'm excited to see this spec developing!

> These tests must accept a parameter `lib` indicating whether to run with `cudf` or `pandas`. The test will then be called twice, once with cudf and once with pandas....

@isVoid Please move your attention to #712 instead of this PR. Sorry for giving you the runaround on these PRs, let's chat about them this morning?

I'm really referring to the python interface in this issue, which was my oversight. There isn't a parallel for `GeoSeries` yet at the C++ level. In the general case the...

@KaiWang1337 it would be a fantastic contribution if you could submit your fix as a PR for us to review! Thanks!

There's tons of great material already in this thread. I can contribute one of the GPU developers' motivations for originally designing the specification as `xyxyxy` instead of `xxx`, `yyy`, which...

Oh, @harrism also explains that we wanted vectorized memory access using xy interleaving https://developer.nvidia.com/blog/cuda-pro-tip-increase-performance-with-vectorized-memory-access/