Re: GIN improvements part2: fast scan

On Thu, Mar 13, 2014 at 8:58 PM, Heikki Linnakangas <hlinnakangas@vmware.com> wrote:

On 03/12/2014 07:52 PM, Alexander Korotkov wrote:

>
>* I just noticed that the dummy trueTriConsistentFn returns GIN_MAYBE,
>rather than GIN_TRUE. The equivalent boolean version returns 'true' without
>recheck. Is that a typo, or was there some reason for the discrepancy?
>

Actually, there is not difference in current implementation, But I
implemented it so that trueTriConsistentFn can correctly work
with shimBoolConsistentFn. In this case it should return GIN_MAYBE in case
when it have no GIN_MAYBE in the input (as analogue of setting recheck
flag). So, it could return GIN_TRUE only if it checked that input has
GIN_MAYBE. However, checking would be just wasting of cycles. So I end up
with just GIN_MAYBE :-)

I don't understand that. As it is, it's inconsistent with the boolean trueConsistent function. trueConsistent always returns TRUE with recheck=false. And in GIN_SEARCH_MODE_EVERYTHING mode, there are no regular scan keys.