On 8/3/2013 23:09, Andy Seaborne wrote:
Please file a JIRA for this and attach the test case.
Done.
It would also be good to find out why this case arises when other
cases when FILTER/unbound evaluation is detected have not arisen.
Maybe the situations are simply less common.
I have switched off this optimizer for now so that I can continue
evaluating the new Jena version. The test case that I sent was
(obviously) artificial, just the bare minimum. The surrounding use case
is common though in our code base, e.g.:
BIND (ex:someFunction(?x) AS ?y) .
FILTER ?prebound = ?y .
It does not resemble the pattern that the optimizer is trying to
address, which I understand to be
<a> <b> ?x .
<c> <b> ?y .
FILTER ?x = ?y .
HTH - and thanks for looking into this.
Holger