Comments on: An Alternative way of writing Effdt clauses

By: PeopleSoft Tipster

PeopleSoft Tipster — Mon, 23 Jun 2008 14:37:27 +0000

This perhaps isn’t the place to get involved in a ‘do analytics work’ debate. They give greater database performance and developer productivity, and the gain is – in general – larger the greater the volume of data analysed.

Saying that, there will obviously be exceptions, and as I said in the article title, this is an alternative – at no point did I say ‘everyone should always do it this way from now forwards’. Before deciding what’s right for you investigate all methods and chose the fastest. Most of the time, that’ll be the analytic method due to reduced passes over the data through the use of partition result sets, and hence leading to fewer queries/subqueries (cutting out self-joins particularly), but not always.

I’ve used the functions on significantly larger data volumes than 320,000 rows and they’ve saved me much time and effort.

By: Matthew Jacobson

Matthew Jacobson — Tue, 10 Jun 2008 20:09:37 +0000

This is not complete information. The ‘faster execution’ only applies when you have a relatively low number of job rows in your PS_JOB table. The more rows you have the faster method A will be compared to Method B.

With 320,000 rows in PS_JOB –

cost for A is 24,020
cost for B is 40,912

I’d suggest sticking with the obvious approach rather than attempting to use fancy analytical clauses.

By: PeopleSoft Tipster

PeopleSoft Tipster — Fri, 09 May 2008 00:16:03 +0000

There are no issues with using this method in a multi-join query. You’d just list the next table after the alias on the inline view like this:

SELECT J.emplid, J.empl_rcd, J.effdt, J.effseq, P.NAME
FROM (
SELECT emplid, empl_rcd, effdt, effseq,
RANK () OVER
(PARTITION BY emplid, empl_rcd
ORDER BY effdt DESC
, effseq DESC) rnk
FROM ps_job
WHERE effdt <= ‘10/AUG/2007′) j,
PS_PERSONAL_DATA P
WHERE j.rnk = 1
and p.emplid = j.emplid

(This SQL hasn’t been tested, so I may have typo’ed or missed something obvious, but hopefully the method is clear)

By: fambaus

fambaus — Tue, 06 May 2008 13:29:50 +0000

Excellent tip. What about in a multi join query?

By: Noons

Noons — Mon, 08 Oct 2007 12:12:52 +0000

Excellent! Thanks for the great hint on analytics.
I always suspected they could b effectively used for Peoplesoft and their quaint timeseries format, this just confirms it.

Will definitely give it a go.