On Mon, Nov 27, 2023 at 5:07 PM Peter Geoghegan <pg@bowt.ie> wrote:
> One of the reasons why we shouldn't do this during parse analysis is
> because query rewriting might matter. But that doesn't mean that the
> transformation/normalization process must fundamentally be the
> responsibility of the optimizer, through process of elimination.
>
> Maybe it should be the responsibility of some other phase of query
> processing, invented solely to make life easier for the optimizer, but
> not formally part of query planning per se.
Support for SEARCH and CYCLE clauses for recursive CTEs (added by
commit 3696a600e2) works by literally rewriting a parse node into a
form involving RowExpr and ScalarArrayOpExpr during rewriting. See
rewriteSearchAndCycle(). These implementation details are even
mentioned in user-facing docs.
Separately, the planner has long relied on certain generic
normalization steps from rewriteHandler.c. For example, it reorders
the targetlist from INSERT and UPDATE statements into what it knows to
be standard order within the planner, for the planner's convenience.
I'm not suggesting that these are any kind of precedent to follow now.
Just that they hint that rewriting/transformation prior to query
planning proper could be the right general approach. AFAICT that
really is what is needed. That, plus the work of fixing any
undesirable/unintended side effects that the transformations lead to,
which might be a difficult task in its own right (it likely requires
work in the planner).
--
Peter Geoghegan