Re: EXPLAIN (plan off, rewrite off) for benchmarking

On Sat, Nov 19, 2011 at 11:47 AM, Andres Freund <andres(at)anarazel(dot)de> wrote:
> I absolutely cannot agree on the fact that the speed parse-analyze is
> irrelevant though.

Tom may be right that the speed of the parser *in isolation* is
irrelevant, in the narrow sense that if we made the parser twice as
slow but somehow by that change made up the time in the executor,
nobody would care; in fact, it would be a net win for people using
prepared statements. But I completely agree that parsing speed is
something we need to worry about. Unfortunately, I don't have a lot
of good ideas for improving it. A while back I tried ripping out most
of the parser to see whether that would speed up performance parsing
very simple statements, but the improvement was pretty small. Maybe a
more thorough job than what I did is possible, but it didn't seem
promising. Maybe we could find a way to reduce the size of the parse
tree (i.e. fewer nodes), or the number of times that it has to be
walked/copied.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company