BUG: IntervalIndex.get_indexer raising for read only array by phofl · Pull Request #53703 · pandas-dev/pandas (original) (raw)

So I guess the pattern is for fused types prefer ndarray (unless that gets fixed/implemented) elsewhere memview?

And that's also what we already do in most places. For example in groupby.pyx, you will see that we typically use ndarray for the input values (which use a fused type and can be const), while mostly use memoryviews for output values (which are writeable) or labels/mask (which don't use fused types).
(although it's certainly not exact, I see some cases using ndarray that could use memoryview)