On 2017-03-22 13:15:41 -0400, Tom Lane wrote:
> BTW, I'm fairly concerned by what you did in nodeTidscan.c, ie delaying
> compile of the TID expressions until TidListCreate. I think that probably
> fails for cases involving, eg, subplans in the expressions; we need
> subplans to get linked to the parent node, and this way won't do it till
> (maybe) too late. Barring objection I'm going to rearrange that so that
> we still do the compile part at executor start.
No objection here.
