Hi everyone,

We use zarr as a backend to store data generated by a pipeline. These data take the form of counts, and metadata.

We typically usexarray to write and load zarr stores intoxarray.Dataset objects.

As we begin to formalise, we realise we need a way to validate that a zarr store is what we are expecting. Specifically:

• it contains the arrays we expect
• the dtypes are as expected
• the shapes are consistent, ie array B shares dimension 0 with dimension 1 of array A.
• attributes are provided and meet expectations.
• all blocks are present (no missing data)

I've looked at pydantic and AVRO, but they don't seem to do exactly what I want. Is anyone aware of any other solutions that would enable this? Or able to provide an example?