"Tom Lane" <tgl@sss.pgh.pa.us> writes:
> Gregory Stark <stark@enterprisedb.com> writes:
>> "Tom Lane" <tgl@sss.pgh.pa.us> writes:
>>> ... I'm not even sure how to fix it (the nasty case is
>>> changing directions partway through the scan); let alone how to fix it in a
>>> way that's obviously enough right to make me feel comfortable in
>>> back-patching.
>
>> It seems like the obvious fix is to just reverse the behaviour -- keep
>> reading backwards until you see the level break then return the
>> previous record from a second slot.
>
> Well, if you think it's easy, the best form of criticism is a patch.
> The change-of-direction problem seems to me to be messy --- not
> insoluble, but messy enough to need beta testing.
Hm, I must have misunderstood the bug because there's a comment in nodeUnique
which claims it already does precisely what I was suggesting:
* We return the first tuple from each group of duplicates (or the last
* tuple of each group, when moving backwards). At either end of the
* subplan, clear the result slot so that we correctly return the
* first/last tuple when reversing direction.
--
Gregory Stark
EnterpriseDB http://www.enterprisedb.com
Ask me about EnterpriseDB's RemoteDBA services!