W3C home > Mailing lists > Public > public-rdf-dawg@w3.org > April to June 2012

Property paths

From: Paul Gearon <pgearon@revelytix.com>
Date: Tue, 19 Jun 2012 15:45:58 -0400
Message-ID: <CAOQ8B2HuatDkbCH-uJ6u4MjGdMbFdNmM2oDC0DtVC_kCq6xC6A@mail.gmail.com>
To: SPARQL Working Group <public-rdf-dawg@w3.org>
Apologies that this hasn't come through before now...

This review is restricted to section 9 of the query document, and the
property path sections of the algebra.

9.1 Property Path Syntax

Is the description of "zero or more occurrences" and "one or more
occurrences" clear enough to a user who is unfamiliar with the
concept. We understand that a path involves walking from one node to
another using connecting predicates, and the "occurrences" refers to
how many steps are taken, but will that always be understood? If not,
then perhaps it could be mentioned in 9.4.


9.2 Examples

"Inverse Path Sequence"

This is fine, but given the prior examples with equivalent queries
then I'd have expected a similar example here, something like:

{
    ?x foaf:knows ?k .
    ?y foaf:knows ?k .
    FILTER(?x != ?y)
}


9.3 Property Paths and Equivalent Patterns

All good.


9.4 Arbitrary Length Path Matching

As I said earlier, an understanding of "zero or more" and "one or
more" property paths is presumed. Given the level of understanding of
some of the audience, this may be worthwhile explaining. If there is
agreement with me, then a simple explanatory sentence might be:

"The length of a property path is the number of steps taken to
traverse a graph from one node to another using a given set of
properties, where a step from one node to another is described by a
triple in the graph with the two nodes as subject and object, and
having one of the desired predicates."

... or words to that effect.


Algebra

This first one was posted to Andy offline and fixed, but I'll mention
it here so that it's documented.

The definition for "Evaluation of ZeroOrOnePath" was incorrect. The
first part of the definition previously read:

  eval(Path(X:term, ZeroOrOnePath(P), Y:var)) = { (Y, yn) | yn = X or
(yn,X) in eval(Path(X,P,Y)) }
  eval(Path(X:var, ZeroOrOnePath(P), Y:term)) = { (X, xn) | xn = Y or
(xn,Y) in eval(Path(X,P,Y)) }

It now correctly reads:

  eval(Path(X:term, ZeroOrOnePath(P), Y:var)) = { (Y, yn) | yn = X or
{(Y, yn)} in eval(Path(X,P,Y)) }
  eval(Path(X:var, ZeroOrOnePath(P), Y:term)) = { (X, xn) | xn = Y or
{(X, xn)} in eval(Path(X,P,Y)) }

*fixed*


Definition: Function ALP

  "Let eval(x:term, path) be the evaluation of 'path', returning a
multiset of RDF terms reached by path."

This does not describe the element "x:term". Suggest adding the following:

  "Let eval(x:term, path) be the evaluation of 'path' starting at the
term x, returning a multiset of RDF terms reached by path."


Regards,
Paul Gearon
Received on Tuesday, 19 June 2012 19:46:29 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 16:15:48 GMT