sh:pattern not evaluated correctly

[berezovskyi]

shaclex 0.1.103
jena 4.4.0
jdk 11 temurin
eclipse lyo 5.0.0-snapshot

We disabled our validation module build during the JDK 11 migration in Eclipse Lyo and now after enabling it and updating to latest ShaclEx we discovered that a number of old passing test cases now fail. For example:

ValidationResult [resultPath=http://www.sampledomain.org/sam#aStringProperty, focusNode=null, message=pattern violation. Expected "Between" to match '^B', resultSeverity=http://www.w3.org/ns/shacl#Violation]

The spec says to follow the SPARQL patterns: https://www.w3.org/TR/shacl/#PatternConstraintComponent Which in turn gives us almost the same regex as an example: https://www.w3.org/TR/sparql11-query/#func-regex

When we change the annotation on the property from @ShaclPattern("^B") to @ShaclPattern("^B.*"), all tests pass.

I have two thoughts: either it's due to Jena 4 migration or cough someone may have introduced a regression in weso/utils#148. However, I don't think that going back to Xerces is a good idea, so we may want to replicate whatever Jena does to evaluate regexps in a SPARQL-compliant way.

[ericprud]

Relevant spec section: 5.4.4 XML Schema String Facet Constraints, which references XPath 5.6.1 Regular expression syntax. The examples in 5.6.3 fn:matches indicate that your reading is correct.

The expression fn:matches("abracadabra", "bra") returns true().

This also suggests that we need a test.

[berezovskyi]

Hello, I haven't had time to follow the SHaclEX developments and today I tried to upgrade to SHaclEX 0.2.2 and all tests passed without having to modify the regex.

It seems like ShEx module re-introduced the Xerces dependency, which we tried to exclude and it works so far (we only call the SHACL). Also, we found that there is a fair number of dependencies wihin SHaclEX with mismatched version numbers, which we were forced to duct tape due to our maven-enforcer-plugin settings: https://github.com/eclipse/lyo/pull/248/files#diff-f5751dbd520a10ddcb49fe2ab71c084c779c3e2cf88765b9cc00593b5e50208cR105

Cautiously, I could say that this issue could be closed given that the other two issues don't impact sh:pattern evaluation directly.

[labra]

Thanks for the update. I think we can close this issue.

About the other topics, it would be great if add specific issues for them.

One possibility to clean the dependencies a bit more if you are only using SHACL is that you consider to depend on SHACL-S which is specific about SHACL...at this moment, SHaclEX is an umbrella project which invokes either ShEx-s o SHACL-s depending on which type of RDF validation you use, so depending on them may avoid to use some extra dependencies which you don't need.