Enable sh:pattern on IRIs about pyshacl

[ashleysommer]

I don't believe there is need for separate 'relaxed' mode for this change. Changing sh:pattern to work on IRIs as well as strings does not cause any of tests in the SHACL test suite to fail.

[dr-shorthair]

FWIW it looks like the TQ SHACL implementation is here: https://github.com/TopQuadrant/shacl

[HolgerKnublauch]

The spec is IMHO clear that IRIs can be checked too.

[ashleysommer]

@HolgerKnublauch
Yes you're right. Re-reading the relevant sections I see it is pretty obvious. I think it was this part which threw me off:

The values of sh:pattern in a shape are literals with datatype xsd:string

I originally read that as "the focusnodes of sh:pattern constraint are literals with datatype xsd:string" so thought it excluded IRIs.

However, just looking at implementing this feature in pySHACL I see it is already implemented. All string-based constraint operations already work on IRIs, and not by accident, there is even a method in there for all string-based constraints to turn an IRI to a string before evaluating. So the spec must've been clear to me back when I wrote that code too.

So can close this as not a bug.

[dr-shorthair]

@HolgerKnublauch It is explicit on sh:minLength and sh:maxLength with the phrase "This can be applied to any literals and IRIs, but not to blank nodes.". Not explicit on sh:pattern (this briefly threw us off the trail), but not excluded either.