Re: Normalize queries starting with SET for pg_stat_statements

On Tue, Sep 24, 2024 at 04:57:28PM +0900, Michael Paquier wrote:
> 0001 is straight-forward and that was I think a mistake to not include
> that from the start when I've expanded these tests in the v16 cycle
> (well, time..). 0002 also is quite conservative at the end, and this
> design can be used to tune easily the jumbling patterns from gram.y
> depending on the feedback we'd get.

Applied 0001 for now to expand the tests, with one tweak: the removal
of SET NAMES. It was proving tricky to use something else than UTF-8,
the CI complaining on Windows. True that this could use like unaccent
and an alternate output in a different file, but I'm not inclined to
take the cost just for this specific query pattern.

The remaining 0002 is attached for now. I am planning to wrap that
next week after a second lookup, except if there are any comments, of
course.
--
Michael