Shai,
I think it would be enough to add to the javadocs of BooleanQuery that
the minimum number of SHOULD clauses is ignored when no such clauses
are added.
Regards,
Paul Elschot
On Tuesday 15 January 2008 13:04:51 Shai Erera wrote:
> Hi
>
> I'm not sure if this is a bug or working as designed. In BooleanQuery, one
> can set the minimum number of SHOULD match terms. So if I send the query
> "abc +def" and set it to 1, only if "abc" has matches in the index I should
> get back results.
> However, if I send the query "+def" and set the minimum to 1, I get back
> results, even though I don't have any SHOULD terms in the query. However, if
> I send the query "+abc +def" and set the minimum to 1, I get back 0 results,
> as expected.
> I think this is actually a bug, mostly because the inconsistencies of "+abc
> +def" and "+def" with the minimum set to 1.
>
> The reason lies in BooleanQuery.rewrite, followed by Query.scorer.
> BooleanQuery.rewrite has an optimization that if the number of clauses is 1
> (as is in "+def"), the query returned is clauses(0).getQuery().rewrite(),
> which in this case is TermQuery. However, if the number of clauses is
> greater than 1, the query returned is the BooleanQuery with all of its
> clauses rewritten.
> Then, the searcher calls query.scorer.
> In the first case ("+def"), TermQuery.scorer is not aware of the minimum
> number should match constraint.
> In the second ("+abc +def"), BooleanQuery.scorer returns a BooleanScorer2,
> which does handle the minimum number should match constraint.
>
> I think that a patch is pretty straight-forward, and I'll open an issue and
> provide the patch once you confirm this is indeed a bug.
>
> Cheers,
>
> Shai Erera
>
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
