| FazBrowse GitHub Viewer | Trending | | Home |
| Tools: [Download Repo ZIP] [Original HTTPS Page] |
Sorry, something went wrong.
| # No triangulation specified, so use matplotlib._qhull to obtain | ||
| # Delaunay triangulation. | ||
| self.triangles, self._neighbors = _qhull.delaunay(x, y, sys.flags.verbose) | ||
| self.triangles, self._neighbors = _qhull.delaunay(self.x, self.y, |
There was a problem hiding this comment.
This is a good catch. It is functionally identical but will be slightly faster in some circumstances so we should do it even if this PR is never merged.
Sorry, something went wrong.
There was a problem hiding this comment.
I pulled this out into #32417
Sorry, something went wrong.
Accept arrays of any shape so we can do our own validation and provide better error messages. This restores most of the test_triangulation test suite to the pre-nanobind version.
| Back | FazBrowse Home | New Git URL |
PR summary
I'm tinkering with nanobind to re-acquaint myself with C++ after a 20+ year hiatus. This is a draft pull request so I can run CI on my experiments.
Results
Binary Size Comparison
(darwin, arm64)
(Split Mode)
(Split Mode)
Results: ~1.33M saved per wheel by switching to nanobind and ~1.88M if we use Split Mode.
Note: Split Mode adds a dependency on nanobind-backend. This is an 80K-450K download depending on architecture and is shared amongst all projects in an environment that use nanobind split mode.
Compilation Time Comparison
AI Disclosure
PR quality check