Issue combining datasets and modifying data in the same step

[ChrisNZ]

@Tom Of course you are right.

Something like this would work

data combine;
  length identifier $20;
  identifier = ' ' ;
  set get_test1_nodup
      get_test2_nodup ;
  if identifier ='' then identifier=Spot_2||Spot_4||Spot_3;
run;

but is is not shorter or clearer than @mkeintz 's proposal.

I was keen to emphasize that variables could be processed at the top of the data step, before the SET statement is encountered, but my example fell flat on its face...

[mkeintz]

"There's more than one way to skin a cat".  So here's another way.  It has no explicit output and no call missing.  But it does take advantage of a DROP= parameter and conditional SET statement.  And it eliminates the unintended carry forward of stay stale identifier values.

data combine;
  set   get_test1_nodup
        get_test2_nodup  (drop=identifier in=in2);
  identifier=Spot_2||Spot_4||Spot_3;
  if in2 then set get_test2_nodup;
run;

The crucial point here is that the conditional SET is explicitly synchronized with reading data from get_test2_nodup in the unconditional SET.  

[jhealthpolicy]

Thank you everyone for the replies -- this was very educational.