Support for zipped zarr files

[vlevasseur073]

Hello,

is it foreseen to add the support of zipped zar files (on local filesystem as well as on cloud storage)? It is very valuable in the case of cloud storage. Regarding the python library, it can open zipped files from cloud storage if the zarr has zero compression, and the archive must not containt the root folder.

Many thanks in advance for feedbacks,
Vincent

[nhz2]

Yes, you can read from a zipped Zarr file with Zarr.ZipStore.

Here is an example in the tests, where ppythonzip is a path to the file:

Zarr.jl/test/python.jl

Lines 240 to 243 in 8d65246

	g = zopen(Zarr.ZipStore(Mmap.mmap(ppythonzip)))
	@test g isa Zarr.ZGroup
	@test g.attrs["groupatt"] == "Hi"
	a1 = g["a1"]

I imagine the goal with reading from cloud storage would be to only download the needed parts of the zip file. This should be possible, but I don't know enough about cloud storage to say how easy this would be to do.

[vlevasseur073]

Many thanks for your prompt answer. For local filesystem I implemented something based on p7zip_jll, unzipping the file on a temporary folder, but this way looks way better. I will check this soon.
Regarding zip file on cloud storage, it is indeed useful with lazy loading to access only the structure of the whole product. I don't know much neither but I think it basically reads the .zmetadata to build the template.

[mkitti]

I would take a look at the Zarr v3 sharding codec which allows many chunks to be consolidated into a few (or one) file.
https://zarr-specs.readthedocs.io/en/latest/v3/codecs/sharding-indexed/index.html

I intend to implement this feature here in the near future.