Re: Postgresqlism & Vacuum?

At 01:23 PM 14-04-2000 -0400, Bruce Momjian wrote:
>
>> Perhaps its time for the two functions to be separated - controlled by an
>> option?
>> Perhaps VACUUM STATONLY could collect stats, not lock table and not reclaim
>> space.
>
>Makes sense.

Actually it may be more logical to have
VACUUM        reclaim
VACUUM ANALYZE     reclaim then analyze (for downward compatibility)
ANALYZE     just analyze

But I still think that some sort of stats could be used dynamically. Even
if the stats aren't optimal (e.g. don't know which MVCC copy to use), using
them should be better than nothing in most cases. In the cases which don't
work then just do a manual analyze - people are already used to doing it
regularly as it's a postgresqlism ;). There are probably only extremely
few/unusual cases where such automatic stats would make things slower.

Well, OK I could be totally wrong - I'm far from a database expert and
don't have a wide experience of databases.

Cheerio,

Link.