Thomas Robitaille
Thomas Robitaille
Yes I think that those interpolation techniques basically fail if there is a single NaN in the input image.
Is there a reference stating that this is actually public API behaviour we can rely on or is in a CPython implementation detail?
I don't really mind but does this really solve the issue in general? We still don't have control over hypothesis's dependency pinning?
@Carifio24 - it looks like the test failures are actually related: https://github.com/glue-viz/glue-vispy-viewers/actions/runs/18658021138/job/54702008413?pr=401 Most of the Qt jobs are failing. The ones that pass are all Windows weirdly. Would you be...
I agree with @jakevdp - we should avoid even importing GPL code.
> Indeed, if this means that affiliated packages also cannot be GPL'd I don't think anyone said that - affiliated packages can be released under any license? (since astropy doesn't...
I have found where the issue is coming from - working on a fix
@neutrinoceros - now that https://github.com/astropy/astropy/pull/18236 is merged, could you rebase and remove the exceptions for astropy/table? (and similarly for the WCS PR)
For the record, here are the benchmarks we are running: https://github.com/astropy/astropy-benchmarks/blob/main/benchmarks/table.py which benchmark(s) is/are the most relevant to look at here? (as @neutrinoceros you can see the results in the...