Graph Shapes and Assertions

[multimeric]

A user may want to constrain a graph in ways that cannot be expressed using node and property shapes. For example:

• There must be at least one entity of type schema:Thing in the graph. Equivalently, there must be at least one triple of the form ?subject rdf:type schema:Thing
• There must be no more than N triples in the graph
• No IRIs beginning with some prefix should exist in the graph

Therefore, I wonder if there has been discussion of the idea of graph shapes. Since all of these can be expressed readily using SPARQL ASK queries, I'm envisaging something like this:

:myGraphShape
	a sh:GraphShape ;
      	sh:ask "ASK { ?s a schema:Thing }" .

How can I propose this as part of SHACL and is it in scope?

[HolgerKnublauch]

Counting instances of classes can be expressed through a constraint with sh:inversePath rdf:type at the class itself.

A pattern to express graph-wide constraints is to use something like sh:targetNode and then normal SPARQL constraints attached to that shape.

So technically, the scenarios that you want to cover can already be expressed with normal node and property shapes.

[multimeric]

Thanks for the reply!

Counting instances of classes can be expressed through a constraint with sh:inversePath rdf:type at the class itself.

If the class is neither defined nor used in the data graph, will this still work? An example would be wonderful.

A pattern to express graph-wide constraints is to use something like sh:targetNode and then normal SPARQL constraints attached to that shape.

I've been trying this, but the issue is that you need to guarantee that a given IRI will always be in the graph which in my use cases I can't do.

[HolgerKnublauch]

Example: https://www.w3.org/wiki/SHACL/Examples

If you have sh:targetNode X then no triple for X actually needs to exist anywhere. The shape will be validated no matter what. Therefore, attaching a shape to the graph node itself will cause it to be evaluated exactly once.

[multimeric]

Great, thank you! This worked nicely for my use case, and I suppose this targetNode trick will work for running a singular ASK validation query over the graph as well. It's not wildly intuitive to do it this way but that's only a minor issue.

[TallTed]

No IRIs beginning with some prefix should exist in the graph

Note that no IRIs beginning with any prefix ever exist in the graph. IRIs in the graph begin with the namespace to which the prefix is attached, e.g., there are no instances of schema:Thing in the graph; rather, there are instances of https://schema.org/thing.