W3C home > Mailing lists > Public > semantic-web@w3.org > December 2010

Re: A path language for RDF tighly integrated with Python

From: Martin Hepp <martin.hepp@ebusiness-unibw.org>
Date: Thu, 2 Dec 2010 11:37:29 +0100
Cc: Semantic Web <semantic-web@w3.org>
Message-Id: <A33898E8-1AD1-4815-BB03-E1C52BFFF6D3@ebusiness-unibw.org>
To: Pierre-Antoine Champin <swlists-040405@champin.net>
Hi Pierre:
Nice approach - two comments:

1. There is support for path expressions in SPARQL in ARQ and  
Virtuoso, see e.g.

http://virtuoso.openlinksw.com/presentations/OpenLink_Data_Spaces_SIOC_and_SPARQL_Guide/OpenLink_Data_Spaces_SIOC_and_SPARQL_Guide_Part_II/OpenLink_Data_Spaces_SIOC_and_SPARQL_Guide_Part_II.html#%2816%29

2. I have doubts that a dedicated path language is needed, because I  
expect that typically
- you will only need short path expressions from nodes that you first  
find via SPARQL
- you will often need to combine information from multiple nodes not  
closely related, so you will have to use a SPARQL query anyway.

So my personal opinion is that support for short path expressions in  
SPARQL is much more important than a general path language. The  
Semantic Web is huge, messy, and often sparsely populated, so  
traversing a rigid path pattern in the graph will often not work,  
because it requires that you have and can rely on structural  
assumptions over the data ex ante.

Martin


On 02.12.2010, at 11:15, Pierre-Antoine Champin wrote:

> Hi all,
>
> Jeni Tenisson's recent post [1] suggesting that RDF should have a  
> standard path language encouraged me to experiment with some ideas  
> that I have been having for a while.
>
> The main idea is to implement an XPath-like language for RDF,  
> directly *inside* the programming language, meaning that the path is  
> not a string, but build using the actual "/" and "[]" operators of  
> the language.
>
> It works pretty well in python, and does not involve a lot of code,  
> as demonstrated by [2]. I guess this could easily be ported to other  
> languages supporting operator overloading.
>
> This is only a proof of concept, but I think it has some potential  
> (I already have several ideas to extend this...), so I put this on  
> googlecode. If anybody also thinks it is a good idea, please let me  
> know.
>
>  pa
>
>
> [1] http://www.jenitennison.com/blog/node/149
> [2] http://code.google.com/p/rdf-pypath/source/browse/trunk/rdfpypath.py
>
Received on Thursday, 2 December 2010 10:38:01 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 21:45:40 GMT