- From: Gregory Williams <greg@evilfunhouse.com>
- Date: Fri, 7 Sep 2012 11:55:22 -0400
- To: Lee Feigenbaum <lee@thefigtrees.net>
- Cc: Steve Harris <steve.harris@garlik.com>, Barry Bishop <barry.bishop@ontotext.com>, "Polleres, Axel" <axel.polleres@siemens.com>, "public-sparql-dev@w3.org" <public-sparql-dev@w3.org>
On Sep 7, 2012, at 11:38 AM, Lee Feigenbaum wrote:
> On 9/7/2012 11:30 AM, Steve Harris wrote:
>> 4store and 5store - by default the default graph is the union of the named graphs, but if you do e.g.
>>
>> INSERT DATA { <s> <p> <o> }
>>
>> then <s> <p> <o> will get written into a "special" named graph.
>>
>> There's a flag you can use (in both I think) to get the "special" graph to behave like the unnamed graph.
>>
>> - Steve
>>
>> PS for the record, I still think the whole unnamed/default graph thing is silly.
>
> :-) Me too, that's why I've been surprised to learn that most named graph stores include one :-)
As the author of one of these systems I can say that I certainly wouldn't do it this way again. In my (RDF::Query's) case, it's a result of the API growing and changing over ~10 years from something rather trivial with early support for RDQL to full SPARQL 1.1 support today. In new projects, I've been going with the more obvious approach of having a pure quadstore with one graph identified as the default graph during dataset construction.
.greg
Received on Friday, 7 September 2012 15:55:47 UTC