On Thu, Jul 26, 2018 at 8:39 PM Andrey Borodin <x4...@yandex-team.ru> wrote: > I'm not sure in architectural point of view: supporting two ways (list and > heap) to store result seems may be a bit heavy, but OK. At least, it has > meaningful benefits.
It seems that Nikita have reworked it that way after my suspect that switching regular scans to pairing heap *might* cause a regression. However, I didn't insist that we need to support two ways. For instance, if we can prove that there is no regression then it's fine to use just heap... Links 1. https://www.postgresql.org/message-id/CAPpHfdu6Wm4DSAp8Pvwq0uo7fCSzsbrNy7x2v5EKK_g4Nkjx1Q%40mail.gmail.com ------ Alexander Korotkov Postgres Professional: http://www.postgrespro.com The Russian Postgres Company
