On May 24, 2007, at 8:26 PM, Tom Lane wrote:
> "Joshua D. Drake" <jd@commandprompt.com> writes:
>> Tom Lane wrote:
>>> I'm not sure I want to vote for another 10x increase by
>>> default, though.
>
>> Outside of longer analyze times, and slightly more space taken up
>> by the
>> statistics, what is the downside?
>
> Longer plan times --- several of the selfuncs.c routines grovel
> over all
> the entries in the pg_statistic row. AFAIK no one's measured the real
> impact of that, but it could easily be counterproductive for simple
> queries.
The lateness of the hour is suppressing my supposed statistics savvy,
so this may not make sense, but...
Would it be possible to look at a much larger number of samples
during analyze,
then look at the variation in those to generate a reasonable number of
pg_statistic "samples" to represent our estimate of the actual
distribution?
More datapoints for tables where the planner might benefit from it,
fewer
where it wouldn't.
Cheers,
Steve