rapidsai / cuxfilter

GPU accelerated cross filtering with cuDF.
https://docs.rapids.ai/api/cuxfilter/stable/
Apache License 2.0
273 stars 67 forks source link

[IMPROVEMENT] Add wheel builds to cuxfilter #491

Closed AjayThorve closed 1 year ago

AjayThorve commented 1 year ago

Follow this guide to add wheels builds to cuxfilter.

cc @exactlyallan @bdice

edit: updated checklist as per suggestion

bdice commented 1 year ago

A couple notes:

  1. You only need to use scikit-build if you have C++ built components. You can think of scikit-build as a bridge between Python build tools and CMake (for CUDA/C++ components). So you can skip that here.
  2. I shared these links in Slack but figured they could be useful here too:

Your best model to follow might be pure Python packages like dask-cudf because those are migrated but don't use scikit-build.

Also maybe there is a good opportunity for some cleanup of the Python code as a whole, I saw some bits while reviewing the build system like Makefile and record.txt that seem safe to remove.