Date: Tue, 26 Aug 2003 18:36:49 +0000
Reply-To: sashole@bellsouth.net
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Paul Dorfman <paul_dorfman@HOTMAIL.COM>
Subject: Re: Ways to code (if Var contains the value of an Obs in another
SAS dataset).
Content-Type: text/plain; format=flowed
Ya,
Nope, it is a very old trick I bet you have already seen. Sometimes it
referred to as macro array, but I think the term is confusing. The simple
truth is that since the ability to look macro variables by name in the
symbol table is embedded in the macro processor, you can store key values as
macro variable names - most often, via CALL SYMPUT() or SQL with INTO: - and
the satellite(s) - as those macro variables' values. Then at the time of
search, supply a search key to SYMGET(), and extract the satellite info. Of
course the key values themselves cannot be just anything: They have to
conform to the rules of a valid SAS names. This leaves us with 31-byte
alphanumeric keys because to avoid the situation when it begins with a
digit, it makes sense to precede it with an underscore. As an illustration,
try, say,
data lookup ;
input state: & $char31. nickname: & $char30. ;
call symput ('_' || translate (state, '_', ' ') , nickname) ;
cards ;
Alabama Yellowhammer State
Alaska The Last Frontier
Arizona Grand Canyon State
Arkansas The Natural State
California Golden State
Colorado Centennial State
Connecticut Constitution State
Delaware First State
Florida Sunshine State
Georgia Peach State
Hawaii Aloha State
Idaho Gem State
Illinois Prairie State
Indiana Hoosier State
Iowa Hawkeye State
Kansas Sunflower State
Kentucky Bluegrass State
Louisiana Pelican State
Maine Pine Tree State
Maryland Free State
Massachusetts Bay State
Michigan Wolverine State
Minnesota North Star State
Mississippi Magnolia State
Missouri Show-me State
Montana Treasure State
Nebraska Cornhusker State
Nevada Sagebrush State
New Hampshire Granite State
New Jersey Garden State
New Mexico Land of Enchantment
New York Empire State
North Carolina Tar Heel State
North Dakota Sioux State
Ohio Buckeye State
Oklahoma Sooner State
Oregon Beaver State
Pennsylvania Keystone State
Rhode Island The Ocean State
South Carolina Palmetto State
South Dakota Mount Rushmore State
Tennessee Volunteer State
Texas Lone Star State
Utah Beehive State
Vermont Green Mountain State
Virginia The Old Dominion
Washington Evergreen State
West Virginia Mountain State
Wisconsin Badger State
Wyoming Equality State
;
run ;
%put _user_ ;
option error=0 ;
data _null_ ;
input state: & $31. ;
nickname = symget ('_' || translate (state, '_', ' ')) ;
put state @ ;
if not _error_ then put @20 nickname ;
else put @23 '*BAD KEY!' ;
_error_ = 0 ;
cards ;
abraca dabra
Alabama
tasmania
Florida
ukraine
Georgia
Indiana
Maryland
South Carolina
south pole
South Dakota
Washington
franklin
West Virginia
east virginia
;
run ;
As to the advantages and disadvantages of this approach, I bet you have
already realized what they are. Frankly, given the roster of other methods
having no drawbacks of the above, I am having a difficult time envisaging
circumstances under which it could be the method of choice. Job security,
maybe... However, it is not completely devoid of interest. First, this is
the only method except the newly-fangled V9 hash that allows for dynamic
O(1) storage and search in the same data step, the macro memory serving as
the repository. Secondly, the table is available without the need to reload
it, to any subseqent process, capable of accessing the macro memory, in the
same SAS session.
Kind regards,
=================
Paul M. Dorfman
Jacksonville, FL
=================
>From: "Huang, Ya" <yhuang@AMYLIN.COM>
>Reply-To: "Huang, Ya" <yhuang@AMYLIN.COM>
>To: SAS-L@LISTSERV.UGA.EDU
>Subject: Re: Ways to code (if Var contains the value of an Obs in another
> SAS dataset).
>Date: Tue, 26 Aug 2003 09:09:54 -0700
>
>Paul,
>
>I'm not sure what you mean by "macro memory search",
>can you give a sample code. Is this a new trick that
>has never been revealed yet.
>
>Kind regards,
>
>Ya
>-----Original Message-----
>From: Paul Dorfman [mailto:paul_dorfman@HOTMAIL.COM]
>Sent: Tuesday, August 26, 2003 8:43 AM
>To: SAS-L@LISTSERV.UGA.EDU
>Subject: Re: Ways to code (if Var contains the value of an Obs in
>another SAS dataset).
>
>
> >From: "Richard A. DeVenezia" <radevenz@IX.NETCOM.COM>
> >
> >>"Haynes, Zac" <ZHaynes@BCBSM.COM> wrote in message
> >>What are some efficient ways to compute whether a variable ($char80.)
> >>contains the value of one of the observations of another SAS dataset?
>
> >Depends on the nature of the variables being looked up. If lookup is
> >equality (col-n = v in table X is equal to col-m in table Y) than all
>these
> >can work well:
> >array, hash, format, merge, SQL left join, SQL subquery lookup
>
>...bitmaps, conditional IN/OR/AND lists, SAS index lookup, INDEX() against
>a
>concatenated string, binary/interpolation search with POINT=..., macro
>memory search, on...
>
>Kind regards,
>=================
>Paul M. Dorfman
>Jacksonville, FL
>=================
>
>_________________________________________________________________
>Get MSN 8 and help protect your children with advanced parental controls.
>http://join.msn.com/?page=features/parental
_________________________________________________________________
Help protect your PC: Get a free online virus scan at McAfee.com.
http://clinic.mcafee.com/clinic/ibuy/campaign.asp?cid=3963
| 
