Re: Selecting distinct records

"Dave [Hawk-Systems]" <dave@hawk-systems.com> writes:
>> What I am looking to do is
>> - grab every record for $user
>> - remove any records that have identical ipaddress+sessionid+refid
>> - then sort the results by date_time or something else

> this last requirement is where the problem is...

I think you need two levels of select.  To do the DISTINCT ON you must
sort by the columns to be distinct'd.  You can re-sort the result in
an outer select.  For example:

select * from
  (select distinct on (ten) * from tenk1 order by ten) ss
order by hundred;

In practice the inner select's order by will need additional sort
columns, with which you control which row gets chosen out of any set of
duplicates.

            regards, tom lane