Date:         Thu, 18 Nov 2004 10:30:09 -0500
Reply-To:     "Richard A. DeVenezia" <radevenz@IX.NETCOM.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Richard A. DeVenezia" <radevenz@IX.NETCOM.COM>
Subject:      Re: proc format vs. merge

Jack Hamilton wrote:
> Perhaps this problem is the source of the suggestion on the SASWare
> ballot to have an option on the SET, MERGE, and UPDATE statements to
> tell you when any of the input data sets have been exhausted?
>
> I have a suggestion in (which did not make it on to the ballot) to
> allow a partial-composite-key lookup in the data step.  What I wanted
> was a lookup on all parts of some of the keys (that is, if the index
> is on a, b, and c, I wanted to get all the values matching a
> particular A, in a, b, c order).  I'm not sure whether that would
> give you what you want - did you want to match part of the value of
> some of the keys, not all of the values of some of the keys?

Sounds like VSAM and GENKEY and a few other IBM mainframe goodies I can't
recall.

Another balloty idea is to allow a variable parts to be aliased to other
identifiers and intelligently used in the case of indexing.
I.e., suppose a factory has some part identifying convention
partId is a constructed value <topic1><topic2><topic3><topic4>, and everyone
in a plant is familiar with partId in its longest form.  Analysts on the
other hand may only be interested in certain components or sections of the
partId value.

alias topic1 partId:1:2;
alias topic2 partId:3:7;
alias topic3 partId:8:10;
alias topic4 partId:11;
alias topic5 partId:12;
alias shazam partId:8:11;

An alias' would be useable in any context a variable could be.

If partId is a variable common in some database schema, it would be far
easier to have aliases than tens or hundreds of views that break the partid
up (or combine five discrete variables into the partid everyone is familiar
with)

--
Richard A. DeVenezia
http://www.devenezia.com/