Date:         Thu, 9 May 2002 13:28:28 -0400
Reply-To:     Dianne Rhodes <RHODESD1@WESTAT.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Dianne Rhodes <RHODESD1@WESTAT.COM>
Subject:      Re: Revised stupid question
Comments: To: "Kevin F. Spratt" <kevin-spratt@UIOWA.EDU>
Content-Type: text/plain; charset="iso-8859-1"

Kevin F. Spratt [mailto:kevin-spratt@UIOWA.EDU] wrote :

> In olden times, (Version 5 and before),
  < stuff about how SAS used to work snipped >

Well in versions before version 5 (and maybe in version 5, I skipped that
iteration)
SAS did not know if your data was sorted unless you sorted it in SAS.
IE if you read in a flat file say written by the Cobol coven that was
presorted by IDVAR
and then tried to use by IDVAR group processing, SAS would complain that the
data wasn't sorted and stop.
As a result, I learned and taught that you should use SAS to explicitly sort
your data (we were on MVS and had SYNCSORT option)

You can see how this has led to lots of ways to tell SAS "It is too sorted,
trust me"
NOTSORTED SORTEDBY

And I like that SAS now knows its sorted and tells me, thanks but it's all
ready sorted, so I'll skip that.

And then there are the things SAS complains about when you delete or subset
observations that
were in sort order.

So my feeling is that I would like the programmer to know their data well
enough to know if
it is in the correct sort order.  IMO good programming practices would make
me be
leary of any more "automagically sort my data" options.  I'm still worrying
about SQL, although
Sig and Ian tell me "just trust me".  I don't mind having SAS tell me
"Dummy, you forgot to sort this"
or maybe "Ain't sorted in my book"

Dianne Louise Rhodes
Sr. Systems Analyst
Westat