- From: Steve Harris <swh@ecs.soton.ac.uk>
- Date: Mon, 17 Dec 2012 08:36:13 +0000
- To: Bob DuCharme <bob@snee.com>
- Cc: public-sparql-dev@w3.org
Bob,
I don't think that a URI is any more or less a key than a Literal.
They have very similar properties in RDF, the only significant difference is that a Literal can only appear in the object slot, and may have a datatype or language tag.
Regards,
Steve
On 17 Dec 2012, at 03:30, Bob DuCharme wrote:
> OK, that makes sense--a URI is inherently a key, so when indexed is easier to look up, while a given literal value is not necessarily, right?
>
> And congrats on the WG finishing up!
>
> Bob
>
>
> On 12/16/2012 4:41 PM, Andy Seaborne wrote:
>>
>>
>> On 16/12/12 21:24, Bob DuCharme wrote:
>>> Imagine that I have ten million triples, and these are two of them:
>>>
>>> <http://w> rdfs:label "my literal" .
>>> <http://x> <http://y> <http://z> .
>>>
>>> I got the impression somewhere that this query
>>>
>>> SELECT ?s WHERE { ?s <http://y> <http://z> }
>>>
>>> would run faster than this one:
>>>
>>> SELECT ?s WHERE { ?s rdfs:label "my literal" }
>>>
>>> Is this true, and if so is it because URIs will always be indexed and
>>> literals won't necessarily be?
>>
>> As far as I know, systems generally index literals - quite important for keys.
>>
>> Maybe there are many, many 'rdfs:label "my literal"' if it's not a key, which might make a difference, as much because there are more results.
>>
>>> Or is it all dependent on the
>>> implementation?
>>
>> Yes.
>>
>> Andy
>
>
Received on Monday, 17 December 2012 08:36:59 UTC