Would it be possible to structure https://wheels.pantsbuild.org/simple/ such that the pip dependency lookup is a bit more efficient? Right now, pip compares the dependency name to each link in the index:

2024-11-27T10:21:51,324 Fetched page <https://wheels.pantsbuild.org/simple> as text/html; charset=utf-8
  2024-11-27T10:21:51,325   Skipping link: wrong project name (not attrs): <https://github.com/pantsbuild/pants/releases/download/release_2.23.0/pantsbuild.pants-2.23.0-cp39-cp39-macosx_10_15_x86_64.whl> (from <https://wheels.pantsbuild.org/simple/>)
  2024-11-27T10:21:51,325   Skipping link: wrong project name (not attrs): <https://github.com/pantsbuild/pants/releases/download/release_2.23.0/pantsbuild.pants-2.23.0-cp39-cp39-macosx_11_0_arm64.whl> (from <https://wheels.pantsbuild.org/simple/>)

(+ 100s of lines) Instead, the links could be grouped by project (pantsbuild.pants)? This comparison alone takes ~100ms per package, which adds up quickly.

It’s certainly possible to change the structure to something compatible-but-faster. It is published from: https://github.com/pantsbuild/wheels.pantsbuild.org

My attempt at the multi-level index: https://github.com/pantsbuild/wheels.pantsbuild.org/pull/1