There is a DATA step solution. (Hmm...maybe there's a DATA step solution to everything.)
Since this requires BY-group processing, start by loading the data into partitions:
libname mycas cas;
data mycas.have(partition=(id) orderby=(distance) replace=yes);
input id var1 var2 var3 distance;
datalines;
1 1 1 1 0.2
1 2 1 2 0.3
2 1 1 3 0.9
2 1 2 1 0.2
3 1 3 2 0.1
;
run;
Then, use the SESSREF= DATA statement argument to insist that the DATA step runs in CAS:
data mycas.want / sessref=casauto;
set mycas.have;
by id distance;
if first.id then output;
run;
Finally, check the work:
proc print data=mycas.want;
run;
There's lots of excellent DATA step in CAS documentation, such as the following:
BY-group processing in CAS
(I made an update to this post because I left off the ORDERBY= data set option on the first DATA step. With the option, the table is already organized according to the BY statement in the second DATA step--avoiding any reorganizing the data to sort it. Without the data set option, the second DATA step has to make a temporary copy of the table--partitioning and sorting it.)
