Andres Rios Tascon
Andres Rios Tascon
There's two more fixes that I'll add to this PR. Currently, if you have an `std::variant` where one of the types is an empty struct, something goes wrong and it...
> There's two more fixes that I'll add to this PR. Currently, if you have an std::variant where one of the types is an empty struct, something goes wrong and...
> Currently not covered: > > * records with no fields (allowed in Awkward, just needs RNTuple → Awkward translation) > * invalid variants: ??? I added a test file...
@jpivarski Here's a follow up on the above discussion. For empty structs, ROOT has no issues. Following our discussion on Slack from a while ago, all I had to do...
I'll add tests once https://github.com/scikit-hep/scikit-hep-testdata/pull/155 gets merged and released. > That's invalid, and Awkward Array may complain about it at any time. Yeah, the thing is that type tags are...
Sounds good, thank you! I'll go ahead and implement this and I'll let you know once it's ready.
@jpivarski how do you feel about the changes in ba422a5 to support invalid variants? The advantages over what we discussed over Slack are that it is very straightforward and doesn't...
Thank you @jpivarski, that makes sense. I made the changes so that now invalid variants end up as `None`. I think this PR is ready for review now.
Hmm for some reason it sometimes doesn't work on Windows. I'll look into it.
> I don't see any test failures with Windows. Is that something that you encountered outside of CI? The failures on Windows were for a98dc19. It turned out that NumPy...