zarr v3 and _strip_protocol() - which doesn't strip the protocol

[fjansson]

I'm trying to use xarray ds.to_zarr() to directly upload to a swift store (DKRZ), and I'm using zarr v3 (fresh pip install in virtual environment yesterday). I want to do something like this:

import xarray as xr
import zarr
import swiftspec
...
url='swift://swift.dkrz.de/some-path/'                                                                                                           
store = zarr.storage.FsspecStore.from_url(url, read_only=False)
ds.to_zarr(store, encoding=encoding, mode='w')

but from_url fails with ValueError: path argument to FsspecStore must not include scheme (swift://)

I posted this as an zarr-python issue here: zarr-developers/zarr-python#3201 with the answer that this happens because switfspec's _strip_protocol() doesn't actually remove swift:// from the URL.

Also when directly passing the url to ds.to_zarr(), I get the same error. Is this something that could be addressed in swiftspec, to make zarr v3 work with SWIFT?

[d70-t]

I didn't know that _strip_protocol must strip the protocol. ... and I've looked into the http protocol implementation to learn how to implement swiftspec, apparently http got an exception in zarr-python ¯\_(ツ)_/¯.

It might be possible to actually strip the protocol, but that won't really make things easier on the swiftspec side, as we'd end up having three ways to specify a path (without protocol, with swift protocol and with https protocol...), and it would probably make even more problems if we'd eventually want to tackle #6, so I'm not super inclined to fix this behaviour, if not needed (and shouldn't an underscore method be considered private anyways?).

As there seems to be some movement on the zarr-python repo, I'd wait a little and hope the issue will solve itself before trying to think of ways to do it on this side.

[observingClouds]

@fjansson is this now fixed with the upstream changes?

[fjansson]

@observingClouds yes, it was fixed with zarr-developers/zarr-python#3343, I'm closing this issue.