Closed mattwthompson closed 2 months ago
After looking at it a bit, I think the point of this (admittedly strange) test IS to run with different random inputs each time. So I don't think this change is needed, the (weird) test is (marginally) better without it, and the code is more readable in the old form.
But I am still bothered by how weird this test is. So an alternative to closing this PR would be making the test not-weird by hardcoding a random-ish conformer.
Yeah it's a bit confusing, just trusting the "rules" here. It may never get removed but following their (admittedly weak) suggestion prevents this from being a confusing deprecation warning in the future
Context https://docs.astral.sh/ruff/rules/numpy-legacy-random/#numpy-legacy-random-npy002