RDFLib with RDF 1.1: BNode skolemization facilities

[Ivan Herman]

These are described in

http://www.w3.org/TR/rdf11-concepts/#section-skolemization

the executive summary is that there is an (additional and, actually, optional) facility RDF environments may provide, namely to pair a recognizable URI pattern to BNodes, thereby, if necessary, change a graph replacing that BNode to a 'real' URI. The general pattern is defined in the spec, with some adaptation per library. It is to be expected, that if a graph is skolemized and then de-skolemized by the same environment, then this would constitute a round-tripping, modulo the exact BNode id-s.

To implement this, I have added a bunch of classes and methods to term.py; here they are:

class Genid(URIRef)
--------------------

all URIRef classes are assigned this class at creation time if the corresponding URI is of the form

{scheme+host}/.well-known/genid/{rest}

(This is the general form of a skolem URI per RDF 1.1.)

class RDFLibGenid(Genid)
--------------------------

all URIRef classes are assigned this class at creation time, if the corresponding URI is of the form

{scheme+host}/.well-known/genid/rdflib/{rest}

where {rest} has no fragment or query part either.

By introducing the /rdflib/ into the path, we can expect that the skolem/genid has been generated by RDFLib.

BNode.skolemize( authority = "http://rdflib.net") :
---------------------------------------------------
Generate a URIRef of the form {authority}/.well-known/genid/rdflib/{BNode identifier}, ie, an RDFLibGenid instance as above.

URIRef.de_skolemize()
---------------------
returns a BNode. If it is a RDFLibGenid, it uses the identifier part directly (ie, if only RDFLib is used, these are inverse functions), otherwise new BNodes are created. An exception is raised if the node is not a Genid instance.

Graph.skolemize(new_graph = None, bnode = None)
-----------------------------
If bnode is None, all triples in the graph are copied to new_graph, with all occurences of BNodes skolemized; if bnode is not None, then only that bnode is skolemized when copying. If new_graph is None, a new graph is created on the fly; in all cases new_graph is returned.

Graph.de_skolemize(uriref = None)
---------------------------------
like the skolemize() method, except that de_skolemization occurs on all graphs.

Note that there is no separate method for skolemization on Datasets; if effect, (de)skolemization will work on all triples regardless of graph names and a flat graph will be returned. This is because, depending on the semantics used, it may or may not be o.k. for the user to change a named graph, ie, it is not clear where the skolemized triples should be added (into the dataset with the same URI as graph name? a new dataset with the same URI as in the original? create a new URI but then what?)


Ivan


----
Ivan Herman
4, rue Beauvallon, clos St Joseph
13090 Aix-en-Provence
France
http://www.ivan-herman.net

[Markus...@tu-dresden.de]

Hello Ivan,

I am very interested in this topic. I need to skolemize some graphs with the help of rdflib and thus I discovered this topic. Is there any possibility to obtain the code you've written?

Best regards,
Markus

--
Dipl.-Ing. Markus Graube
PhD Student

Dresden University of Technology
Faculty of Electrical and Computer Engineering
Chair of Distributed Control Systems Engineering
01062 Dresden, Germany

Phone : +49 (351) 463-33387
Fax : +49 (351) 463-39681
E-Mail: markus...@tu-dresden.de
WWW: http://www.et.tu-dresden.de/ifa/?id=mg

[Ivan Herman]

Markus,

there is a separate branch

https://github.com/RDFLib/rdflib/tree/rdf11

where I uploaded my current code. Unfortunately, the discussion died out a bit which is a pity:-(

Anyway, you are free to download that!

Ivan

> --
> You received this message because you are subscribed to the Google Groups "rdflib-dev" group.
> To post to this group, send email to rdfli...@googlegroups.com.
> To unsubscribe from this group, send email to rdflib-dev+...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msg/rdflib-dev/-/Mw43ve9FQqoJ.
> For more options, visit https://groups.google.com/groups/opt_out.

>
>


----
Ivan Herman
4, rue Beauvallon, clos St Joseph
13090 Aix-en-Provence
France

http://about.me/ivan_herman

[Markus Graube]

Hello Ivan,

thank you. The code is working great. Hopefully the code can be
integrated in the main branch.

Markus

Am 17.01.2013 11:56, schrieb Ivan Herman:
> Markus,
>
> there is a separate branch
>
> https://github.com/RDFLib/rdflib/tree/rdf11
>
> where I uploaded my current code. Unfortunately, the discussion died out a bit which is a pity:-(
>
> Anyway, you are free to download that!
>
> Ivan
>

--
Dipl.-Ing. Markus Graube
Wissenschaftlicher Mitarbeiter

Technische Universität Dresden
Fakultät Elektrotechnik und Informationstechnik
Institut für Automatisierungstechnik
Professur für Prozessleittechnik
01062 Dresden

Tel. : +49 (351) 463-33387

[Graham Higgins]

On Thu, 2013-01-17 at 15:08 +0100, Markus Graube wrote:
> Hopefully the code can be
> integrated in the main branch.

Some preparatory work is necessary: applying PEP8 to the code, adding a
Python 3 port and extending test coverage so that the current level of
coverage is at least maintained.

I've pushed what I have got so far to the repos and I've added the
branch to the Travis-CI build to expose the current failures reported by
the existing tests.

https://travis-ci.org/RDFLib/rdflib/builds/4290210


--
Graham Higgins

http://bel-epa.com/gjh/

[imran.s...@gmail.com]

Hello lvan
I am working on Querying RDF document and working on the Bnode problem.Is there possible to to use Named graph and Skolmization to handel the Bnode Problem at query time.

[Ivan Herman]

I do not think so. More exactly (and if I understand your question correctly):
skolemization helps if the server/query engine plays with. The client side only
cannot do this I believe.

But I think it would be better to ask these types of questions on the
semant...@w3.org mailing list; it is not necessarily related to RDFLib.

Sincerely

Ivan