[tlambert03]

invalid_multiscale_transform_output and invalid_multiscale_transform_input_output in parse_attributes/v0.6/core/invalid/image have become identical, and both seem totally valid to me (they pass image.schema from ngff-spec, and I can't find a rule in the prose that they break).

I think this might have been accidental in #22: the path-to-name and string-to-object updates there also rewrote the values these two cases were testing. This restores what they have upstream, so _output has "output": {"path": "intrinsic"} and _input_output has "input": "s0", "output": "intrinsic". Both fail the schema again.

[kevinyamauchi]

Thanks, @tlambert03 ! I agree that the current cases are valid and thus incorrect as an "invalid" case. I also agree that the change makes the cases invalid for the

This would be fine to merge as-is for me as it is for sure more correct. However, I am not 100% confident that I know which rule the cases were originally intended to test, so I think it would be good to get input from @clbarnes .

As a side-note, this is a situation where having descriptions for the cases (#46 ) or having codes for the rules in the spec would be helpful.

[clbarnes]

Thanks folks, certainly possible these got swept up with bulk updates. It looks like there aren't equivalent validate_zarr cases for these tests - perhaps there should be? That might be a good unit test, actually: checking whether each parse_attributes has a validate_zarr and that the parse_attributes metadata appears somewhere in that hierarchy.