Re: Early WIP/PoC for inlining CTEs

>>>>> "Tom" == Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> writes:

Tom> We can't inline wCTEs (those containing insert/update/delete)
Tom> without risk of semantics change.

Clearly.

Tom> I'd also not favor changing the semantics for CTEs that are read
Tom> more than once by the parent query.

This one's more debatable. There will still be cases where a CTE
referenced multiple times will be better inlined.

(It's obviously trivial to make the posted code do it that way, just by
checking cterefcount.)

Tom> However, a singly-referenced SELECT CTE could reasonably be
Tom> treated as equivalent to a sub-select-in-FROM,

In the PoC code I also excluded SELECT FOR UPDATE from inlining.

(There's already a difference between how SELECT FOR UPDATE works for
CTEs compared to subqueries and views, the comments mention it)

There might also be some merit in checking for volatility?

--
Andrew (irc:RhodiumToad)