zarr-python
zarr-python copied to clipboard
C/F ordering in naming
In the process of investigating an analogous format ( https://github.com/zarr-developers/zarr/issues/231 ), one of the points raised was the current chunk naming uses C-ordering. Though others have used F-ordering. Given this, was wondering if it would be reasonable to make the chunk naming order configurable.
As a possible proposal, would it be reasonable to add an optional parameter to .zarray
, which specifies the naming order, but simply have this default to the existing C-order for compatibility. As such this should hopefully be a compatible (i.e. non-breaking) change to the format.
@constantinpape, do you know if this is still relevant? This came up in relation to Zarr/N5 compatibility, but don't know if this is still an issue
cc @joshmoore (for vis)
And further cc @d-v-b
Thinking about the pros and cons of expanding the configuration surface area to cover this. As a "pro", it would erode some of the differences between n5 and zarr, as n5 uses F order for the chunk IDs. Although other differences would still remain, like the concrete names of the metadata files, I think progress towards unification is nice.
I can't really think of any cons right now besides the added burden on implementations and the generic cost of expanding what a config covers. So I'm +1 on this idea.
I also still think this is relevant and I like your initial proposal from https://github.com/zarr-developers/zarr-python/issues/232#issue-291746455.