Problem publishing multi-level datasets in viewer runnning in Jupyter notebook

[forman]

Describe the bug

Multi-level datasets from S3 storage added to a xcube Viewer instance in Jupyter cause server 404 errors for many endpoints. The reason is that the default dataset identifiers contain a slash (/) causing invalid URLs not found by xcube Server. This happens although xcube Viewer correctly URL-encodes the dataset identifiers when constructing server request URLs.

A possible quick-fix for the problem is to encode slashes by a different character in the Viewer.add_dataset() method. The solution can be verified by adding a multi-level dataset with an explicit identifier. This works:

viewer.add_dataset(mlds, ds_id="valid_id_without_slash")

To Reproduce

See screenshot below.

Expected behavior

Datasets that have dataset identifers with slashes should work with xcube Server when added to the xcube Viewer instance.

Screenshot

Additional context

xcube on DeepESDL platform.

[forman]

This is likely a regression that occurs only in Jupyter environments, see

• PR now also serve nested datasets #588 and issue xcube server cannot serve nested datasets #579
• Viewer: xcube viewer to support nested datasets xcube-viewer#191, xcube viewer cannot serve nested datasets xcube-viewer#190

That time, the issue was solved by correctly URL-encoding the dataset identifier URL part.
For some reason, the server URLs are decode twice so that the forward slash again appears and causes the 404.