>
> This seems like a good answer --- if you fix up such things during
> vacuum then the performance issue won't last too long.
>
> Rather than eating up the extra flag bit, though, would it be possible
> to change the tuple to appear to contain a NULL? AFAIK, GiST can't have
> a NULL (at least not in the first column) so this should be an unused
> state.
We would like to preserve NULL as non-special value because we hope to add
support of NULLs to GiST, although it's of low priority. In other side,
IndexTupleData contains ItemPointerData which contains block number and offset.
GiST doesn't use offset on inner pages (on leaf pages, ItemPointerData points to
heap tuple), it always equal FirstOffsetNumber. So I can use offset to indicate
correctness of tuple's key...
--
Teodor Sigaev E-mail: teodor@sigaev.ru
WWW: http://www.sigaev.ru/