Re: shape programs [was Re: shapes as classes]

"Peter F. Patel-Schneider" <pfpschneider@gmail.com> wrote on 12/15/2014 
04:18:41 PM:

> I think that you mean that a shape program is something that contains 
> shapes/constraints, i.e., the top-level construct of a constraint 
language, 
> except that some proposals don't really have a top-level construct.

I think of any proposed shape language as implicitly defining a virtual 
machine. The code for the VM comes from the "shape program", whatever that 
is. The VM+code takes as input an RDF graph (or maybe an RDF dataset) and 
creates some output, maybe just YES/NO, or maybe a, possibly empty, list 
of error messages.

> SPIN constraints are always (I think) associated with classes. 
> However, there 
> is no need for a SPIN document to contain an ontology.  I don't think 
that 
> SPIN can handle OWL ontologies anyway, only RDF classes.

Perhaps I am using the term "ontology" imprecisely. The SPIN overview 
states "The SPIN Modeling Vocabulary defines a collection of properties 
and classes that can be used to link RDFS and OWL classes with SPARQL 
queries." [1]

> In sum, one has to be prepared to have constraints/shapes not connected 
to an 
> ontology and not even connected to classes.

I agree completely. That is the design of OSLC shapes.

> PS:  Is it really true that OSLC shapes define sets of graohs?  Where is 
this 
> stated?

This is implicit in how programmers use constraints. A constraint is 
either satisfied or violated. A constraint defines the set of graphs that 
satisfy it.

[1] http://www.w3.org/Submission/spin-overview/


-- Arthur Ryman

Received on Thursday, 18 December 2014 17:24:52 UTC