Re: new correlation metric

Jeff Davis <pgsql(at)j-davis(dot)com> writes:
> On Sun, 2008-10-26 at 12:44 -0400, Tom Lane wrote:
>> We might need to invent some
>> other catalog besides pg_statistic if we want to represent per-index
>> properties like correlation.

> Why can't we just use pg_statistic with the starelid set to the index
> oid?

Well, because pg_statistic is built for per-column stats. You'd have to
invent some value for staattnum, which would be problematic for views
like pg_stats that expect it to join to a valid pg_attribute row;
and you'd have useless columns like stanullfrac and stadistinct.

There's no problem with using pg_statistic for stats that correspond to
individual index columns (and in fact we do that already); but ISTM
the point here is that correlation/ordering is about the index as a
whole, not any one column of it.

regards, tom lane