Date:         Mon, 6 May 1996 23:22:05 -0700
Reply-To:     Karsten Self <kmself@IX.NETCOM.COM>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Karsten Self <kmself@IX.NETCOM.COM>
Subject:      Re: Compressed SAS datasets
Comments: To: Christopher Keane <keane@GLUE.UMD.EDU>

You can always specify your libref as '/dev/null'....  But seriously...
 
SAS compression in my experience generally results in 10 - 30% savings, but
your mileage will vary considerably with the type of data and its
composition -- long text does better than short numbers.  Compression *can*
result in *larger* datasets -- SAS does not determine this ahead of time,
nor does it warn you, you must look for the 'compressing dataset...'
reduced/increased size message.
 
Unless something's changed, you can zcat SAS datasets, but cannot access
them except after decompressing them -- SAS will not read compressed data
directly.  SAS *can* read compressed raw data (ascii/ebcdic) through a
'pipe' fileref.  I've found this unreliable.  zcat's comression method is
*much* more effective than SAS's.
 
At 06:36 PM 5/6/96 GMT, Christopher Keane wrote:
>I have a large number of 8M+ record datasets.  As text files they are
>about 30M in size.  Likewise, when I save the file, even with setting
>LENGTH _NUMERIC_ 4 is 44MB.
>
>I desparately need to reduce the disk foot print of these datasets.
>
>I vaguely remember seeing something about being able to zcat a file (I am
>using SAS 6.07 on a Sparc).  I also see that one can set COMPRESS=YES in
>the data step.  I have a few questions:
>
>1. Which provides better compression (I am getting 90% compression with
>unix compress)?
>2. What is the exact syntax to do this?
>
>Thanks in advance.
>
>As a note, thank you to all who answered my last question.  It is awfully
>nice to find a newsgroup with so many helpful and courteous people!
>
>
>--
>Cheers,
>------------------------------------------------------------------------------
>Christopher M. Keane                                    keane@jei.umd.edu
>Joint Education Initiative                              keane@glue.umd.edu
>Department of Geology
>University of Maryland at College Park
>*************************************************************************
>* Steamrolling another dataset.                                         *
>*************************************************************************
>
>
----------------------------------------
Karsten Self / kmself@ix.netcom.com
 
What part of gestalt don't you understand?