Add "type_designator_mapping" slot.

[gouttegd]

Add a new type_designator_mapping slot.

That slot is intended to be used in a slot_definition for a slot that is set as the type designator slot for the class it belongs to.

It allows to define custom mappings to match a type designator value with the type to be designated. Such mappings can be used to provide more ways to reference a given class in a type designator value (beyond the default way of using the class name, URI, or CURIE).

This implements the idea evoked in linkml/linkml#3701.

[gouttegd]

@cmungall This proposal has the following differences compared to what we discussed in linkml/linkml#3701:

(A) The keys in the type_designator_mapping map are expected to be simple, “shell-like” patterns (where the only special characters are *, standing for anything, and ?, standing for any single character), rather than full-fledged regular expressions.

So for example, the map entry that would act as the ultimate fallback (recognising anything) would be just *, instead of .*.

This is because I am not convinced we need the expressivity of full regular expressions here – this seems “overkill” to me.

However I don’t have too much of a strong opinion here. If you (or anyone else for that matter) believe regular expressions would really be beneficial here, that’s fine with me, happy to change.

(B) I introduced the notion of “longest match”, in that the designated type to be effectively used has to be found by searching for the longest matching entry in the type_designator_mapping map.

This is because the type_designator_map is serialised as a dictionary. Because of that, we cannot rely on the order of keys to avoid a premature match.

Consider the example coming from the linkml/linkml#3701 ticket:

type_designator_mapping:
  TC: TargetingComponent
  LC: LoadingComponent
  "*": FrobnicatorComponent

There is no intrinsic guarantee that the * key will be tried last! If the value of the type designator is, say, LC and the * key is tried before the LC key, then LC will be found to match the * entry (because it does, since * matches anything), and the object will be assumed to be a FrobnicatorComponent instead of the intended LoadingComponent.

The “longest match” rule avoids this situation. LC would match the key LC, and because it’s a longer (more precise) pattern, it would take precedence over the shorter * pattern.