Date:         Mon, 26 Apr 2004 13:58:51 -0400
Reply-To:     "Fehd, Ronald J. (PHPPO)" <rjf2@CDC.GOV>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Fehd, Ronald J. (PHPPO)" <rjf2@CDC.GOV>
Subject:      Re: Reverse PROC CONTENTS
Comments: cc: jim.groeneveld@VITATRON.COM
Content-Type: text/plain; charset="us-ascii"

Perhaps the following idea is what is wanted:

*this step is relatively easy to read from a text file;
DATA Structure;
attrib VarA length = <$> 8 label = 'VarA label'
       ...;
stop;

or
infile Structure '...';
DATA _Null_;
attrib Name   length = $ 32
       Type   length = $  1
       Length length =    4
       Format length = $ 32
       Label  length = $ 40
       ...
       ;
call execute('DATA <Library.>Structure ...;attrib ');
infile Structure
       pad lrecl = 72;
input @1 Name ...;
if Type eq some interesting value
then TypeStr = '$'
else TypeStr = ' ';
attribute = trim(Name)
          !! ' length = ' !! TypeStr !! Length
          !! ' format = ...;
call execute(Attribute);
*at end of file read:
call execute(';stop;run;');
stop;
run;

DATA Child;
if 0 then
set  <Library.>Structure;
do until(EndoFile);
*set SomeOtherData
     end = EndoFile
or
input etc.;
end;%*read loop;
stop;

there is an SQL sentence something like:
create table X like Y
meaning
create table X with same data structure as Y

Ron Fehd  the macro maven  CDC Atlanta GA USA RJF2@cdc.gov

--> cheerful provider of UNTESTED SAS code from the Clue?Gee!Wrx <--

If you want creative workers,
   give them enough time to play.
   -- John Cleese, comic actor (1939- )
... or write one data structure for many data sets

> From: Groeneveld, Jim [mailto:jim.groeneveld@VITATRON.COM]
> Maybe I'm missing something, but I am looking for some kind
> of reverse PROC CONTENTS. I read ascii data and don't want to
> enter the metadata (variable labels and many other
> attributes) by hand by writing a piece of data step code. The
> information is avaliable in another ascii file, which I would
> like to read using the standardized variable names from PROC
> CONTENTS (into a
> dataset) and feed back to a (still empty) dataset, just like
> the CNTLIN option with PROC FORMAT allows for importing
> formats from a dataset. If that is not possible I would could
> read the metadata to generate normal data step code to
> %include, but that is more work.
>