topic Re: MERGE equivalent in SAS in SAS Programming

MERGE equivalent in SAS

FLINT — Wed, 30 May 2018 07:57:58 GMT

Can anyone help me with the SAS equivalent of following code
MERGE TABLE A USING
(SELECT VAR1,VALUE1 FROM TABLE B) C
ON C.VAR1=A.VAR1 WHEN MATCHED THEN UPDATE
SET A.VALUE1= C.VALUE1

Re: MERGE equivalent in SAS

RW9 — Wed, 30 May 2018 08:38:59 GMT

Something like (nothing to check with):

data want (drop=val1);
  merge tablea (in=a) tableb (in=b keep=var1,value1 rename=(value1=val1));
  by var1;
  if a and b then value1=val1;
run;

Do be aware that both tablea and b both need to be sorted by var1 beforehand.

Re: MERGE equivalent in SAS

FLINT — Wed, 30 May 2018 09:43:55 GMT

Here the issue is that I cannot sort one of the tables as its a production table

Re: MERGE equivalent in SAS

Kurt_Bremser — Wed, 30 May 2018 10:04:55 GMT

Then sort it to an intermediary in WORK and use that for the merge.

@FLINT wrote:
Here the issue is that I cannot sort one of the tables as its a production table

Re: MERGE equivalent in SAS

RW9 — Wed, 30 May 2018 12:48:39 GMT

Datastep merge requires that both datasets are sorted by the by variables. This is a hard and fixed rule. There is nothing stopping you making a copy of the datasets, sorting and merging those.

Other alternatives include using a proc sql code block and rewrite the code you have to be base ANSI, maybe something like:

proc sql;
  create table want as 
  select  a.var1,
          case when b.value1 ne "" then b.value1 else a.value1 end as value1
  from    tablea a
  left join tableb b
  on       a.var1=b.var1;
quit;

This still of course does the sorts, its just implicit in the SQL parser to do this rather than you writing the code.

Another way could be hash table, not yet had a reason to use these, so look up some of the plenty of examples on the web or here.