Date: Wed, 13 Jul 2005 16:06:35 +0000 (GMT)
From: David Blankley <david@blankley.com>
To: pgsql-sql@postgresql.org
Subject: Re: Dynamic Offset Determination 
Message-ID: <20050713160327.H33783-100000@hallsofworlds.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

I came up with a clever solution to this myself...
I know the determined value for my CriteriaCol, so I can simply get the
count of all the rows with criteria <= my value...

SELECT count(*) FROM foo WHERE CriteriaCol<=constraint ORDER BY OrderCol;

Thanks,
Dave


-----------------------------------
Problem Statement:
I want to return n rows from a table.
These n rows are relative to an offset.

The part I can't figure out:
The offset needs to be determined from values in the rows.

Example:
Given a table foo:
CriteriaCol 	integer
OrderedCol	integer

I can figure out the number of rows
SELECT COUNT(CriteriaCol) FROM foo;

How do I determine the offset?
SELECT {rownum?} FROM foo WHERE CriteriaCol = SomeValidValue;

With the rownum I can then determine the rows I want.
SELECT * FROM foo ORDER BY OrderedCol LIMIT 21 OFFSET (rownum-10);

Don't worry about boundary conditions, functions, etc, I can sort that out
once I know how to get the offset.
--------------------------

Searching the forums, the one potential hack I've found is:
Quote from: http://archives.postgresql.org/pgsql-sql/2003-04/msg00287.php
create temp sequence foo;
	select nextval('foo'), * from
	(select ... whatever ... order by something) ss;
drop sequence foo;

Presumably I would get my ordinal value from the sequence before dropping
it.

This seems like it would be a potentially slow process.  Also, this
solution dates to 2003, so I thought it worth asking in case a better
solution has come along.

Thanks for suggestions/solutions
Dave