https://raw.githack.com/w3c/data-shapes/master/shacl12-core/index.html#example-52 :

ex:PersonShape
	a sh:NodeShape ;
	sh:targetClass ex:Person ;
	sh:property [   # _:b1
		sh:path ex:address ;
		sh:minCount 1 ;
		sh:nodeByExpression ex:AddressShape ;
	] .

This is not a good example because instead of sh:nodeByExpression, we can use a plain sh:node (as in https://raw.githack.com/w3c/data-shapes/master/shacl12-core/index.html#example-48).

@mgberg @afs @HolgerKnublauch
Can we come up with a better example?

I see in one of the commits of #408

Only using an IRI Expression here because Core doesn't define interesting node expressions

So I guess you want to keep Core self-contained and not use expressions except the trivial ones in https://raw.githack.com/w3c/data-shapes/master/shacl12-core/index.html#node-expressions. I think there are 2 better alternatives:

1. That example could use a more complex expression, refer to the Node Expressions spec, and mention that supporting such expressions by Core-conforming engines is not required
2. Or I could argue that sh:nodeByExpression itself should not be part of Core but of Node Expressions.
   • @mgberg can you confirm that the evaluation result of sh:node and sh:nodeByExpression on IRIs/literals is the same?

Now I think that 2 is the better option: If an engine doesn't support non-trivial Node Expressions, why should it support sh:nodeByExpression?