Holger
This is down to a new optimizer in 2.10.2, however that optimizer is based
off an existing optimizer that has been present in ARQ for a long time
which makes the same assumption.
So the following query would give the exact same behavior when used with
initial bindings:
ASK
{
FILTER (?a = <http://constant>)
}
Both optimizers are making appropriate decisions based on the information
available to them.
I'm going to start a discussion on the dev list about initial bindings
which you are welcome to participate in because there are a bunch of
issues here around interpretation of SPARQL and API intention.
Rob
On 8/2/13 3:34 AM, "Holger Knublauch" <[email protected]> wrote:
>Hi,
>
>I believe there is a regression bug somewhere between 2.7.2 and 2.10.2.
>See test case which works green in the old version but fails on the
>latest snapshot. I believe the cause is that
>TransformFilterImplicitJoin.testSpecialCaseUnused() incorrectly assumes
>it can ignore the whole query if it encounters unbound variables.
>However, in my case the variables do have values, coming from the
>initial binding of the QueryExecution.
>
>Thanks,
>Holger
>
