New EntityType value to identify composed entities

[gouttegd]

There has been quite a few discussions about whether it should be possible to represent mappings involving “complex” or “composed” or “post-coordinated” entities – where at least one side of the mapping is not a single entity but a composition of several entities — and if so, how to represent such mappings.

For example:

• Proposal for accommodating complex mappings #36
• Mapping involving post-coordinated subjects or objects #108
• [New metadata element]: sparql_expansion #180

In #108 in particular, the discussion seems to have stabilized around the general idea of encoding the post-coordinated entity within the subject (or object) ID, using a construct such as, for example, obo.pattern:anatomical_trait?anatomical_entity=UBERON:123&disease=PATO:123 (or various variations of it). This has led to the proposal for URI Expressions, which is the exact same idea but using a form of JSON encoding rather than URI parameters.

Regardless of how the post-coordinated entities are ultimately encoded within a ID, the main benefit of this general approach is that it requires almost no change to SSSOM at all. Subject and object IDs in SSSOM are always opaque strings, so the fact that one ID happens to encode a post-coordinated entity using some elegant/ugly hack is completely irrelevant. SSSOM implementations can just pass the ID as it is, and leave it to applications built on top of them to eventually “unpack” the ID (stop treating it as opaque) and get access to the separate entities that made up the post-coordinated one.

But if this approach is used (in fact that approach can already be used right now, precisely because it does not require any explicit support in SSSOM), SSSOM should at least allow to explicitly indicate that the subject or object of a mapping is a post-coordinated entity.

SSSOM allows to indicate the type of “entity being mapped” with the subject_type and object_type slots, which accept a value from the EntityType enumeration. Currently, there is no value in that enumeration that can be suitable to mark the type of a post-coordinated entity; it would be wrong to say that a post-coordinated entity is a “owl class”, or a “owl object property”, or a “skos concept”, or a “rdfs resource”, or any of the other available type. A post-coordinated entity is a sui generis type of entity, which should have its own representation in the EntityType enumeration.

I therefore propose that a new value, tentatively named composed entity, be added to the EntityType enumeration.

[matentzn]

Full agree and thank you for suggesting! composed seems to be a more precise way to put it than "complex", so I agree with the suggestion