Ability to associate a sh:severity to a sh:SPARQLConstraint

[tfrancart]

Right now it seems impossible to associate a severity level to a SPARQL based constraint.
It would be nice to be able to associate a sh:severity to a sh:SPARQLConstraint

[ajnelson-nist]

I just hit this the other day too.

I suppose the issue is there is a mental similarity in crafting a SPARQL constraint as there is a whole property shape. So, when I put the effort in to craft a SPARQL constraint, I want to attach a severity to its triggering---even if I'm attaching the constraint to a sh:NodeShape.

I think there's some nuance I'm missing between "constraints" and "constraint components" that might help explain this better.

[mgberg]

I also agree, I've run into that issue several times in the past. As I generally create SPARQL constraints for Node Shapes, it seems like you should be able to attach a severity to it in the same way you would a property shape.

[HolgerKnublauch]

In 1.2 we now have the ability to set the severity etc using reification. Is that not enough or do we need a 3rd way?

[tfrancart]

In 1.2 we now have the ability to set the severity etc using reification. Is that not enough or do we need a 3rd way?

Well, reification is simply not needed here. The SPARQL constraint is an entity by itself, just like a property shape. It could be associated with a severity directly.

[HolgerKnublauch]

Yes, not needed but already exists. So in theory adding a new mechanism isn't needed here, and that's what I am questioning.

[tfrancart]

Yes, not needed but already exists. So in theory adding a new mechanism isn't needed here, and that's what I am questioning.

Reification should be used when the subject is an assertion. Here the intended subject is not an assertion, it is an entity (of type sh:SPARQLConstraint). In a sense, this entity is already a reification of the SPARQL query.
Or does it mean that in SHACL 1.2, all severities are expected to be expressed using reification, including on NodeShapes and PropertyShapes (or rather, on sh:property constraints) ? (sorry, I didn't follow all the discussions)

[HolgerKnublauch]

It will be legal to add a reification such as

ex:MyShape 
    sh:sparql [
        sh:select "..."
    ] {| sh:severity sh:Warning |}

regardless of whether we also support

ex:MyShape
    sh:sparql [
        sh:select "..." ;
        sh:severity sh:Warning ;
    ] 

The other option that remains is to attach it to the shape

ex:MyShape
    sh:severity sh:Warning ;
    sh:sparql [
        sh:select "..." ;
    ] 

Where to stop? Maybe if we do what you suggest, people would also expect

ex:MyShape
    sh:node [
        sh:class ex:SomeClass ;
        sh:severity sh:Warning ;
    ] 

because it also just looks like a constraint object.

I would caution against adding too many ways of making the same statements.

[TallTed]

I would caution against adding too many ways of making the same statements.

Are they really the same statement, made multiple ways; rephrased, if you will?

If so, it seems like this should be clearly demonstrated if not proven, if not as part of our spec for implementers, in some documentation/guide for end users, such that when they want to use one of the phrasings that we decided not to support, they can find the equivalent phrasing that we did find worthy.

But it doesn't feel to me like these are the same statement, and even if they are, I wonder why they shouldn't all be supported.