Re: SHACL implementation on RDFUnit

On 27/02/2016 21:15, Dimitris Kontokostas wrote:
>
>
> On Fri, Feb 26, 2016 at 11:25 AM, Dimitris Kontokostas 
> <kontokostas@informatik.uni-leipzig.de 
> <mailto:kontokostas@informatik.uni-leipzig.de>> wrote:
> >
> > There was a question from Peter yesterday on details about my 
> (draft) implementation but I wanted to share this with the group anyway.
> >
> > the current status can be tracked here: 
> https://github.com/AKSW/RDFUnit/labels/SHACL
> > Most of the core constructs are covered and tested with my own test 
> cases but the implementation is quite fragile right now.
> > Meaning validation might break if something deviates from standard 
> (supported) definitions.
> >
> > As you already probably figured out, I use code generation to build 
> up the queries and everything that is supported so far is integrated 
> to the RDFUnit philosophy and operations, meaning
> > - can be executed on SPARQL Endpoints
> > - has different levels of result reporting (status, aggregated, 
> simplified shacl & shacl)
> > - one can mix owl, shacl, RS & DSP constraints all together and 
> execute them
> >
> > some differences with Holger's (current) implementation
> > - sh:valueShape implementation is evaluated using a bottom-up 
> approach and creates separate checks for every sub-shape/facet. IIRC 
> Holger's does the same thing but delegates this the SPARQL engine with 
> sh:hasShape while I keep the logic in RDFUnit.
> > - I plan to do the same with sh:and / sh:or / sh:not but to support 
> that I need some major refactoring on my test execution engine
>
> to make it clear for people not too technically involved, I mention 
> this only because it can affect the number of returned violations. My 
> approach may return more errors and at lower level than Holger's. This 
> is not necessarily better or worse, just a different approach.
>

Yes this sounds good, any engine should be permitted to return more 
output as long as the official minimum is covered. Are you linking the 
extra results to the parent result with sh:detail?

Holger