search

w3c / data-shapes

RDF Data Shapes WG repo
87 stars 33 forks source link

requirements on sh:message not checked in test cases process #60

Closed pfps closed 7 years ago

pfps commented 7 years ago

There are requirements on sh:resultMessage that are not checked by the process in the test suite document.

(non-normative) 2.1.5 Declaring Messages for a Shape If a shape has at least one value for sh:message in the shapes graph, then all validation results produced as a result of the shape will have exactly these messages as their value of sh:resultMessage,

(normative) 5.3.2 Mapping of Solution Bindings to Result Properties sh:resultMessage

  1. The binding for the variable message
  2. For SPARQL-based constraints: The values of sh:message of the SPARQL-based constraint. For SPARQL-based constraint components: The values of sh:message of the validator of the SPARQL-based constraint component.
  3. For SPARQL-based constraint components: The values of sh:message of the SPARQL-based constraint component.
HolgerKnublauch commented 7 years ago

Not sure what you are referring to. Are you asking for specific test cases to cover sh:message better, or do you think the test framework needs to change. If the latter, what do you suggest?

pfps commented 7 years ago

The test framework needs to change to not remove sh:resultMessage triples.

HolgerKnublauch commented 7 years ago

I have addressed this by adding a special rule for handling sh:resultMessage, and added a corresponding test case. See commit above. This makes it possible for certain tests to specifically test for the presence of certain messages without requiring specific messages elsewhere.

pfps commented 7 years ago

I think that the logic is the wrong way around - it should be the subject that is checked, but I don't think that even that is correct.

HolgerKnublauch commented 7 years ago

The approach seems to have been clear enough for those who have submitted implementation reports. I see no need to complicate this any further. The test suite never claimed to be complete or replace what's already stated in the spec.

pfps commented 7 years ago

No longer worth worrying about at this time.