Contract tests suite verifying json-ld output of renku log command

[jachro]

As a Renku owner, I'd like to be sure that json-ld output generated by the renku log command does meet the contract for the Renku KG data.

Acceptance criteria:

• a suite of contract test verifying json-ld serialization of all types of entities meets the established contract.

[Panaetius]

As discussed on Slack, SHACL will be used to validate RDF/JSON-LD output

[Panaetius]

Similar to #547 and can be done at the same time

[Panaetius]

Tracking JSON-LD Issues:

• rdflib-jsonld is not maintained anymore (last commit 1 year ago) and should be replaced, possibly with pyld
• Old project metadata still uses http://xmlns.com/foaf/0.1/Project' instead of http://schema.org/Project, same with schema:name and foaf:name. This is handled/hidden by jsonld.s with the translate parameter, but there's no migration for it.
• Dataset Files can have invalid id's when filename contain colons :, e.g. blob/9609b88114ac588ac815276d5ba6061590950bc2/data/Pyndl: Naive discriminative learning in python/pyndl-v0.6.0.zip. Since the colon is unescaped, it is interpreted as an IRI instead of a node id. This leads to validation errors, though renku itself seems to work with it . Files with : in the name can't be added through dataset add (which itself seems like a bug, filenames should always be escaped in metadata...), but can come from imports. Plus previously imported datasets used name instead of display_name, so it can be a leftover of dataset names containing : which there is no migration for.
• Everything has a reference to project and as such project is in every json-ld graph. This should be replaced with an IRI instead of the full project data, for deduplication

...more to come ...

[rokroskar]

This is handled/hidden by jsonld.s with the translate parameter, but there's no migration for it.

Why would you need a migration? The two are semantically equivalent.

[rokroskar]

Everything has a reference to project and as such project is in every json-ld graph. This should be replaced with an IRI instead of the full project data, for deduplication

👍 - we should do this in other places too, e.g. creator - right now we are duplicating this metadata in a lot of places. One issue is that for properties like creator that are not at the top-level, it's not clear to me when or by whom the actual node (representing the creator) would be generated.