attribute distinct values estimate - Mailing list pgsql-hackers

We need to estimate the number of distinct values of an attribute from
inside optimizer, for an extension we are working on.
Should we use stacommonfrac from pg_statistic, or attdisbursion, or what
else?