On Sun, 15 Aug 2004, Tom Lane wrote:
> Ole Tange <postgresql.org@tange.dk> writes:
> > As I read it the output tells me what was done during the milliseconds:
>
> No, you have a fundamental misconception here. The notation means that
> the first output row from a plan step was delivered after X
> milliseconds, and the last row after Y milliseconds.
Thanks. For a novice tuner like me it would be nice if you could see more
easily where the time was spent. However, the output is _far_ more
intuitive that MySQL's.
> It is gonna take a while to sort 175000 rows ... but possibly increasing
> sort_mem would help.
It didn't. However, I could reformulate the DISTINCT query as a GROUP BY
on all the selected fields and this uses Hash aggregate which is far
faster.
Now I am curious: Why isn't DISTINCT implemented using a Hash aggregate?
/Ole