| From: | Robins <tharakan(at)gmail(dot)com> |
|---|---|
| To: | pgsql-sql(at)postgresql(dot)org |
| Subject: | ROW_NUMBER alias |
| Date: | 2007-05-06 16:01:39 |
| Message-ID: | 36af4bed0705060901s36fed269ye5a8588eaffbd750@mail.gmail.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-sql |
Hi,
I needed ROW_NUMBER() in PostGresql and I did find the 'temporary sequence'
method as a workaround and i think it at least gets the job done relatively
well, ... so no problems there.
Its just that from a usability point of view, isn't it better that we
provide some kind of an aliasing mechanism here that allows a new user to
(unknowingly but) implicitly use a temporary sequence rather than make him
use SubQuery with a COUNT(*) and a comparison operator (with disastrous
performance) instead ??
So for a new user :
A query such as this :
SELECT ROW_NUMBER() AS row_number , a, b, c
FROM table
WHERE table_id = 973
ORDER BY record_date;
is internally interpreted by the planner as :
CREATE TEMP SEQUENCE rownum;
SELECT nextval('rownum') AS row_number , t.a, t.b, t.c
FROM (
SELECT a, b, c
FROM table
WHERE table_id = 973
ORDER BY record_date
) t;
DROP SEQUENCE rownum;
Any ideas ?
(Of what I remember, I think till recently PostgreSql internally replaced
'MAX(x)' queries with a 'ORDER BY x DESC LIMIT 1' implicitly)
--
Robins
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Stefan Becker | 2007-05-06 18:41:30 | Re: ROW_NUMBER alias |
| Previous Message | Jaime Casanova | 2007-05-06 04:37:24 | Re: Sequence vs. Index Scan |