retrieving primary key for row with MIN function

[Marcin Krol <mrkafk(at)gmail(dot)com>]

Hello everyone,

I need to retrieve PK (r.id in the query) for row with 
MIN(r.start_date), but with a twist: I need to select only one record, 
the one with minimum date.

Doing it like this does not solve the problem:

SELECT h.id AS host_id, MIN(r.start_date) AS reservation_start_date, 
r.id AS reservation_id
FROM hosts h
LEFT OUTER JOIN reservation_hosts rh ON rh.host_id = h.id
LEFT OUTER JOIN reservation r ON r.id = rh.reservation_id AND 
(r.start_date, r.end_date) OVERLAPS ('2009-04-29'::date, '2010-04-29'::date)
GROUP BY h.id, r.id
ORDER BY reservation_start_date ASC

I have to use either GROUP BY r.id or use MIN(r.id). MIN(r.id) doesn't 
select the id from the row with corresponding MIN(r.start_date), so it's 
useless, while GROUP BY r.id produces more than one row:

host_id	reservation_start_date	reservation_id
361     2009-05-11              38
361     2009-05-17              21

I need to select only row with reservation_id = 38.

I would rather not do subquery for every 'host' record, since there can 
be a lot of them...

Regards,
mk