Date:         Sun, 20 Oct 2002 22:24:22 -0400
Reply-To:     Sigurd Hermansen <hermans1@WESTAT.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Sigurd Hermansen <hermans1@WESTAT.COM>
Subject:      Re: MERGE (was "safe" merging?)
Comments: To: "Shane (SASL) Hornibrook" <shornibr@IS.DAL.CA>

While I agree that Pud's eloquent and civil reply to my deliberately
provocative statements certainly deserve wider circulation, I do hope that
you will include one caveat: in the timing test of SQL vs. MERGE, the test
creates data that have the sort order required by the MERGE. The SQL
compiler takes the precaution of sorting the test datasets before joining
them. This special case does not reflect accurately the best case
performance of SQL vs. best case performance of the data step MERGE. A
number of tests posted earler on SAS-L show almost equal performance of the
SQL and MERGE in best-case tests (except for a parallel MERGE of three or
more datasets on the same BY group, which the MERGE wins hands down).

I'll respond to the other arguments directly.
Sig

On Sun, 20 Oct 2002 21:11:54 -0300, Shane Hornibrook <shornibr@IS.DAL.CA>
wrote:

>Pud,
>
>This one is going into my sasl-archive folder, to be forwarded to
>oh-so-many people.
>
>Thanks,
>--Shane
>Shane Hornibrook
>GIS Analyst
>Web: http://www.geologist.net
>Mobile: 902.441.4158
>
<snip>