Custom annotations

[janik-martin]

Hi, I was trying to add my own custom annotations to some of my classes. Something like:

...
default_prefix: moucmn
...
classes:
  sample class:
    ...
    annotations:
      moucmn:my_annotation:
        tag: moucmn:my_annotation
        value: some value

but it skips the export with message:
INFO:root:Ignoring namespace error: Unknown CURIE prefix: moucmn

I found this same approach with Biolink Model which is a huge schema (and source of some advanced features knowledge for me :)
For example:

...
default_prefix: biolink
...
subclass of:
  is_a: related to at concept level
  ...
  annotations:
      biolink:canonical_predicate:
        tag: biolink:canonical_predicate
        value: true
      biolink:opposite_of:
        tag: biolink:opposite_of
        value: biolink:superclass of

I am not sure if I am doing something wrong, or this is not possible with linkml and biolink is an exception. Note that using 'standard' well known annotations works fine..

Thanks for any suggestions.

[cmungall]

Can I just clarify, you are seeing this behavior in gen-owl?

…

On Wed, Apr 20, 2022 at 1:10 AM janik-martin ***@***.***> wrote: Hi, I was trying to add my own custom annotations to some of my classes. Something like: ... default_prefix: moucmn ... classes: sample class: ... annotations: moucmn:my_annotation: tag: moucmn:my_annotation value: some value but it skips the export with message: INFO:root:Ignoring namespace error: Unknown CURIE prefix: moucmn I found this same approach with Biolink Model <https://biolink.github.io/biolink-model/> which is a huge schema (and source of some advanced features knowledge for me :) For example: ... default_prefix: biolink ... subclass of: is_a: related to at concept level ... annotations: biolink:canonical_predicate: tag: biolink:canonical_predicate value: true biolink:opposite_of: tag: biolink:opposite_of value: biolink:superclass of I am not sure if I am doing something wrong, or this is not possible with linkml and biolink is an exception. Note that using 'standard' well known annotations works fine.. Thanks for any suggestions. — Reply to this email directly, view it on GitHub <#709>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAMMOKTNPXD5PLYHHEHYPTVF63WZANCNFSM5T3GGT3A> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[cmungall]

Yes, it would be good to include translation of annotations

Some caveats - the annotation system is intentionally quite loose - you can use unprefixed tags, and the mapping of these to URIs is unspecified. The values are modeled as simple strings, again without any typing.

Also, there is a syntactic trick that is optional but useful, you can say

 annotations:
    tag: value

this maybe looks a little odd when prefixed tags are used

[janik-martin]

Sorry, not sure if I get the answer. But I did some digging in, and now I'm pretty sure using my own prefix in annotations won't work.
I am now mainly into owl gen and jsonld context gen, so I don't know the impact to other parts of export, but do you think it would be reasonable to export annotations without the prefix using the default_prefix (they are skipped in owl-gen right now)?

[cmungall]

yes, definitely.

I think that makes sense, along with using untyped literals as values, as default.

however, we can also have a check that determines if the tag corresponds to an existing slot, and if so, use that interpretation (its slot_uri, and translate to literal or uri based on the range of that slot)

[janik-martin]

however, we can also have a check that determines if the tag corresponds to an existing slot, and if so, use that interpretation (its slot_uri, and translate to literal or uri based on the range of that slot)

Actually I tried that yesterday and failed (but my knowing of linkml source code is very limited so far). In owlgen I tried using _prop_uri method to transform non-prefixed annotations into their URIs to add them into the RDF graph.
The problem is with slots containing spaces in their names (which is a common practice in linkml), Annotation tags need to be URIorCURIEs which e.g. "sample annotation" tag is NOT (this fails with ValueError: sample annotation is not a valid URI or CURIE) while parsing schema.

On the other hand, using its normalised underscore version e.g. sample_annotation then fails in _prop_uri method, where the dictionary schema.slots has the original name (with spaces) as key, so the lookup fails.

So it seems to me, that without a major rework, is is impossible to check the existence of the slot in the schema.. Please say I'm wrong and push me into the right direction :)

Export the annotations unchecked, using literals as values seems possible though. This would be the same level of strictness as with the prefixed ones - using schema:whateverMadeOfAnnotation as a tag will export just fine..

[cmungall]

part of the problem here is that owlgen is an 'older style generator', schema introspection is easier with the newer schemaview-style generators

I know this info is not so useful to you in isolation, I will try and get back on this...