I'll patch locally and try it out... I'm running a head that's older
than yesterday:)

Thanks,
--tim

On Wed, Dec 4, 2013 at 12:17 PM, Aaron McCurry <[email protected]> wrote:
> Hmm, I thought I fixed that problem here.
>
> https://git-wip-us.apache.org/repos/asf?p=incubator-blur.git;a=commit;h=de596916af31c8f563cf1d16f6fbc4969684f189
>
> The issue is because the SUPER score SCORE_TYPE is used, and when the
> double produced from the score is cast to float.  Sometimes the float is a
> NAN if the double was very small, and when the Lucene collector collects
> the hit and the score NAN the docId is switched to Integer.MAX_VALUE for
> some reason.  Did you get this update?
>
> Aaron
>
>
>
> On Wed, Dec 4, 2013 at 10:56 AM, Tim Williams <[email protected]> wrote:
>
>> On certain queries, when converting to blur results
>> (convertToBlurResults), I get a BlurException down in the fetchRow
>> suggesting that the "Location id ... with docId... is not valid".
>> Curiously, the docid happens to equal Integer.MAX_VALUE.  I notice the
>> sentinel object's have MAX_VALUE, but don't immediately see how one
>> could stay intact.  Any ideas what's going on?
>>
>> Thanks,
>> --tim
>>

Reply via email to