Date: Sat, 5 Oct 1996 14:01:07 EDT
Reply-To: whitloi1@WESTATPO.WESTAT.COM
Sender: "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From: Ian Whitlock <whitloi1@WESTATPO.WESTAT.COM>
Subject: Re: Format for numeric representation of PIB character varia
Subject: Format for numeric representation of PIB character variables
Summary: Deja vu
Respondent: Ian Whitlock <whitloi1@westat.com>
Karsten Self <kmself@PMSQUARED.COM> wants to store huge numeric values
in 4 bytes of character data.
>I'm exploring the possibility of saving storage space by coding
>numeric integer values into a character field. This corresponds to
>unsigned integers in other programming languages. These are serially
>incremented counters, no negative values, and I expect a large
>quantity of data.
As he points out, he can get 4,294,967,296 different values in 4
character bytes, but only 2,097,152 different numbers in 4 byte numeric
variables.
When I began programming, I worked for a firm, that stored counts in
each half byte to keep the record small. The consequence was that
every program was significantly harder to write and took much longer to
execute. At the time the data base had 600 reels of tape which needed
updating twice a week, so storage space was at a premium. Each record
was 4096 bytes and contained many small counts. They managed to make
the system work (with an army of 150 programmers), so who can knock it.
Unlike the above problem, it is far from clear what is to be gained by
using 4 character bytes instead of 8 numeric bytes. But the conversion
routines are trivial to code in SAS
store = input ( put ( numin , hex8. ) , $hex8. ) ;
numout = input ( put ( store , $hex8. ) , hex8. ) ;
Hence the view
data v ( drop = temp ) / view = v ;
set &data ( rename = ( store = temp ) ) ;
store = input ( put ( store , $hex8. ) , hex8. ) ;
format store comma14. ;
run ;
would appear to solve the problem wherever needed. (In case anyone has
deja vu, this is just another twist on converting from character to
numeric and has been discussed all too often on SAS-L.)
As for a format the SAS/TOOLKIT will allow one to construct such an
algorithm at a lot more expense, but I doubt whether much is too be
gained in doing so.
Ian Whitlock
