Re: RDF parsing without the building the SiRPAC model

uche.ogbuji@fourthought.com wrote:
> 
> I too looked to the SiRPAC effort, in my case to try to re-use an existing
> interface for RDF, but I disliked their interface so much that we're rolling
> our own for our Python RDF processor.  As soon as I have time, I'll post the
> whole catalog of problems I have with the SiRPAC folks' proposed standard RDF
> interface.

Stefan Decker pointed me to two other implementations... 

It should be possible that all parser implementors agree on a very 
simple callback interface that abstracts from the rdf syntax parsed, 
but does NOT generate a complete in-memory model: Something like SAX 
for RDF, maybe 

org.w3c.rdf.parser.StatementHandler, 

with a few method declarations, e.g.

literalStatement (String predicate, String subject, String
literalObject);
ressourceStatement (String predicate, String subject, String
ressourceObject);

The corresponding parser interface (org.w3c.rdf.parser.Parser) would
need 

setStatementHandler (StatementHandler statementHandler);
parse (InputSource inputSource);

The inputSource and some more methods could siply be borrowed from SAX.

If most rdf parsers would implement that interface, we would have all
the advantages of SAX for RDF: 

- We could combine the model builders/processors (=handler
  implementations) and parsers from different vendors,

- I could e.g. use Sergey Melnik's strawman parser with my own 
  graph model.

- Parsers for for other RDF serializations could be added 
  simply, reusing an existing processor/model.

- Pseudo parsers could walk through a memory model and generates
  the events, thus an existing rdf processor could simply be feed from
  an existing memory model.

- The simple interface would show how simple usage of RDF is. That 
  is not obvious when looking at a RDF document or at SiRPAC.

Best regards

Stefan



-- 
Stefan Haustein
University of Dortmund
Computer Science VIII
www-ai.cs.uni-dortmund.de

Received on Tuesday, 4 January 2000 06:04:47 UTC