Building models with SAS Enterprise Miner, SAS Factory Miner, SAS Visual Data Mining and Machine Learning or just with programming

data step view

Reply
Occasional Contributor
Posts: 14

data step view

Can you please tell me which one of the below options is the answer for the question

Consider the following SAS log:

229 data sasuser.ranch sasuser.condo / view = sasuser.ranch;

230 set sasuser.houses;

231 if style ='RANCH' then output sasuser.ranch;

232 else if style = 'CONDO'then output sasuser.condo;

233 run;

NOTE: DATA STEP view saved on file SASUSER.RANCH

NOTE: A stored DATA STEP view cannot run under a different operating system.

235 proc print data= sasuser.condo;

ERROR: File SASUSER.CONDO.DATA does not exist.

236 run;

NOTE: The SAS System stopped processing this step because of errors.

Which one of the following explains why the PRINT procedure fails?

A. SASUSER.CONDO is a stored DATA step program.

B. A SAS data file and SAS data view cannot be created in the same DATA step.

C. A second VIEW=SASUSER.CONDO option was omitted on the DATA statement.

D. The view SASUSER.RANCH must be processed before SASUSER.CONDO is created.

However, when I tried similar code, as shown below, I didn't get an error message. CONDO.DATA  was created. Very confusing. Would you like to help me?

 

data condo_ranch;

infile datalines dsd;

input style $ sqfeet saledate : date9.;

datalines;

RANCH,1250,10MAR2004

CONDO,1400,17JUN2004

RANCH,1500,20JAN2004

;

data ranch condo /view=ranch;

set condo_ranch;

if style='CONDO' then output condo;

else if style='RANCH' then output ranch;

run;

proc print data=condo;

run;

Super User
Posts: 19,869

Re: data step view

Post your log, with above code submitted.

Respected Advisor
Posts: 4,173

Re: data step view

Hi

The answer is 'D'.

Without you asking and me doing some testing I would have got it wrong and answered 'B'. You wouldn't define anything else than "data table /view=table;" in a real life scenario and until now I wasn't aware that anything else is even possible.

Think of a view as pre-compiled code which gets executed in the moment when you use it (so for example in a SET statement).

What I never realised so far is that this also means for a data step view that you can have more than one output table. I always thought that a data step view is kept in memory and kind-of "streamed" into the data step where its used. But clearly using the code you've posted when actually executing the view using a SET statement table work.condo gets created as well.

I'm sure there will be some special case where such behaviour comes in handy - but it will need a LOT of in-program comment because such behaviour is rather exotic.

Thanks

Patrick

Here some code showing what's going on:

proc datasets lib=work kill nowarn nolist;
quit;

data condo_ranch;
  infile datalines dsd;
  input style $ sqfeet saledate : date9.;
  datalines;
RANCH,1250,10MAR2004
CONDO,1400,17JUN2004
RANCH,1500,20JAN2004
;

data ranch condo /view=ranch;
  set condo_ranch;

  if style='CONDO' then
    output condo;
  else if style='RANCH' then
    output ranch;
run;

data view=work.ranch;
  describe;
run;

title "Tables in WORK before executing the view";
proc sql;
  select memname,memtype from dictionary.tables where libname='WORK'
  ;
quit;

data _null_;
  set work.ranch;
run;

title "Tables in WORK after executing the view";
proc sql;
  select memname,memtype from dictionary.tables where libname='WORK'
  ;
quit;

Respected Advisor
Posts: 3,799

Re: data step view

Not so strange there is even an example.

Example 2: Producing Additional Output Files

Ask a Question
Discussion stats
  • 3 replies
  • 956 views
  • 0 likes
  • 4 in conversation