Re: compute_query_id and pg_stat_statements

On Tue, May 11, 2021 at 09:43:25AM +0200, Magnus Hagander wrote:
> 
> 30 second review -- wouldn't it be cleaner to keep a separate boolean
> telling the backend "include it or not", which is set to true/false in
> the guc assign hook and can then be flipped from false->true in
> queryIdWanted()? (I'd suggest a more verbose name for that function
> btw, something like requestQueryIdGeneration() or so).
> 
> (Again, just the 30 second review between meetings, so maybe I'm completely off)

It it surely would, but then that variable would need to be explicitly handled
as it wouldn't be automatically inherited on Windows and EXEC_BACKEND right?