fix: status's range is string instead of uriorcurie

[Silvanoc]

The slot status declares it's range to be uriorcurie, but in fact bare strings like testing or unstable are being used in the metamodel itself (more details in the below linked issue). This patch aligns the specification with its current use.

Fixing it to remain uriorcurie would require multiple changes with bigger impact than this change. They might get implemented in the future.

Fixes: linkml/linkml#3760

[matentzn]

@Silvanoc can you redo the PR with just the model changes? I will do the generation of files on a slightly different cadance (it also makes it clearer which files have actially changed other than linkml_model/model/schema/meta.yaml)

[Silvanoc]

@Silvanoc can you redo the PR with just the model changes? I will do the generation of files on a slightly different cadance (it also makes it clearer which files have actially changed other than linkml_model/model/schema/meta.yaml)

@matentzn sure! Guess why I have two commits with exactly that split 😉 I'll drop the 2nd one and done.

[noelmcloughlin]

Could we test this change against OWL generator?

While investigating linkml/linkml#3803 I tried this patch but found OwlSchemaGenerator.add_metadata() is silently dropping status annotation from OWL output due to this owlgen.py conditional

if metaslot_curie.startswith("linkml:"):
    # only mapped properties
    continue

This skips metaslots whose URI falls back to the native linkml: namespace - and after deleting slot_uri: bibo:status, the slot's URI becomes linkml:status, and every element's status value vanishes from the generated OWL.

https://github.com/linkml/linkml/blob/8e443f15faa4673acc3345923ea84199f73e7f50/packages/linkml/src/linkml/generators/owlgen.py#L379

[Silvanoc]

@noelmcloughlin it might be related with issue linkml/linkml#3752. Please give the attached PR a try.

[noelmcloughlin]

Seems different bug... let me clarify

this PR alone is problematic. After removing slot_uri: bibo:status, slot's URI becomes linkml:status, and every element (ArrayExpression, apply_to, range_expression, etc.) status value vanishes from generated OWL (on my branch, I confirmed zero "testing" occurrences in my output, versus ~32 bibo:testing triples using alternative uriocurie approach). So a sibling PR forowlgen.py is needed to address fallout

[Silvanoc]

Hmm, this PR is not changing status itself, only its range. Since you have tried it out, I assume that the explanation is not precise, but the problematic effect exists.

I'll switch this PR to draft until I've clarified the situation and, it needed, provided the required sibling PR for owlgen.py.

[Silvanoc]

@noelmcloughlin I've realized this issue trying to make SchemaView.expand_curie strict (trying to expand something that is neither a URI -no expansion needed- nor a CURIE -can be expanded) on PR linkml/linkml#3805 (specifically commit linkml/linkml@85fb0c9). Apparently this and that PRs are somehow coupled together.

[noelmcloughlin]

yes, three lines in owlgen.py are coupled to this PR

if metaslot_curie.startswith("linkml:"):
    # only mapped properties
    continue