Closed sbailey closed 10 months ago
@rosteen thanks for proactively fixing the codestyle issues and merging this. I've been busy enough with my day job that I hadn't even looked at what was failing, fearing it would be worse and take more time than I had available.
This PR fixes a bug in the Spectrum1D.read / write default tabular-fits format, which was dropping the .mask attribute. It adds tests demonstrating the problem and failing before the fix, and with the fixes these tests now pass.
Detail: due to astropy/astropy#11963, by default int8 masks are written as bool thus losing information, so here I upcast int8 masks to int16 with the result that the dtype changes in the round trip but at least information isn't lost.