You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
A recent thread on Python idea pointed out that Json.dumps does not by default respect the json spec (at least on some number like NaN/Inf if should refuse to encode). we might want to be careful in v3 implmentation.
Issue #141 is directly related. Besides changes to the spec specifically making it possible to encode such values somehow, this should probably be moved to zarr-python, do you agree @joshmoore?
I'm not sure. I think this is something that all the v2 implementations will or do struggle with, no? Ok to mark this as v2 for the moment, and we perhaps reach out to the other implementations to see where they stand?
Activity
jstriebel commentedon Nov 16, 2022
Issue #141 is directly related. Besides changes to the spec specifically making it possible to encode such values somehow, this should probably be moved to zarr-python, do you agree @joshmoore?
joshmoore commentedon Nov 17, 2022
I'm not sure. I think this is something that all the v2 implementations will or do struggle with, no? Ok to mark this as v2 for the moment, and we perhaps reach out to the other implementations to see where they stand?
e.g. JuliaIO/Zarr.jl#97
jakirkham commentedon Nov 17, 2022
We already check arrays are loadable by different implementations. Maybe we extend this to metadata?