Damian,
Thanks, that clarified how to use BINDINGS, which I'd been wondering about!
I've now tried that approach, but it turns out to be very slow...
My benchmarks so far (milliseconds):
Query per resource: 735
Filter using IN(), in batches of 50: 109
Bindings, in batches of 50: 4047
One query for all resources: 2000
(This is for querying 200 resources from a model containing about 100K
triples; the query is repeated 10 times to get significant durations).
Cheers,
David.
Damian Steer wrote:
>>Is the performance of the IN() function O(n), and if so can anything be
easily done to improve this?
>I suspect the answers are: yes, and maybe.
>and, is there a better way to do this kind of query?
>I've never used them, but how about using bindings? [1]
>SELECT ?res ?lat ?lon { ?res <http://example.com#lat> ?lat; <
http://example.com#lon> ?lon . } BINDINGS ?res {
(<http://example.com#myresource123>)
(<http://example.com#myresource124>) }
>My naïve understanding is that this ought to operate like a regular join.
