Mark T. B. Carroll

(Basically, from the point of view of developing in Java, it's generally not fun to try to interpret the Zarr v2 spec. :confused: Compressors are another thorny bit.)

I was surprised by that too. I'm adapting it from https://github.com/ome/omero-ms-zarr/blob/6add6804/spec.md#user-content-color, don't know why it needs stringifying though.

I think the image's pixel values have to fit the `dtype` as at https://zarr.readthedocs.io/en/stable/spec/v2.html#data-type-encoding.

Aha, that'll be it, thank you. Could be we decide to opt for 1/0 for split masks, it's an easy change to make to the microservice. I wonder if that...

For `image/123.zarr/0/` we have the directory listing controlled by `omero.ms.zarr.folder.layout` so we know, alongside the `.zarray`, if to list flattened chunks or the top level of the nested chunk directories....

Note also https://github.com/ome/omero-ms-zarr/blob/v0.1.0/src/main/java/org/openmicroscopy/ms/zarr/ZarrDataVerticle.java#L77.