As you read a set of records, you need to compare the current-record LEAVE_DATE to the next record's ENTER_DATE. You can preserve the first ENTER_DATE in an extra variable (_ENTER1). If the gap is greater than 30 days, or you are at the end of an ID group, then you can re-read the same records (without re-reading LEAVE_DATE or ENTER_DATE), and output those records.
You don't need to re-read ENTER_DATE because you can replace it with _ENTER1. And you don't need to re-read LEAVE_DATE because the LEAVE_DATE at the end of the first pass is the one to keep:
data have;
infile datalines ;
input id (enter_date leave_date service_date) ( :mmddyy9.);
format enter_date -- service_date date9.;
datalines;
1 1/21/19 1/27/19 1/21/19
1 2/25/19 3/9/19 2/25/19
1 4/1/19 4/2/19 4/1/19
1 5/1/19 5/5/19 5/1/19
2 3/1/18 3/4/18 3/1/18
2 3/10/18 6/10/18 3/7/18
3 6/7/18 7/5/18 6/2/18
3 7/3/19 7/9/19 7/3/19
3 8/11/19 8/30/19 8/11/19
3 9/1/19 9/5/19 9/1/19
4 1/31/18 2/14/18 1/31/18
4 2/14/18 5/14/18 2/14/18
4 6/1/18 6/7/18 6/1/18
5 11/12/18 12/12/18 11/11/18
5 12/28/18 2/1/19 12/28/18
6 1/1/19 2/1/19 1/1/19
7 2/21/18 2/24/18 2/21/18
7 2/27/18 3/3/18 2/26/18
7 10/10/18 10/23/18 10/10/18
7 11/1/18 12/1/18 11/1/18
run;
data want (drop=_:);
do _n=1 by 1 until (last.id=1 or _nxt_enter > leave_date + 30);
set have (keep=id);
by id;
merge have
have (firstobs=2 keep=enter_date rename=(enter_date=_nxt_enter));
if _n=1 then _enter1=enter_date;
end;
enter_date=_enter1;
do _i=1 to _n;
set have (drop=enter_date leave_date);
output;
end;
run;
- If the first do group there is both a SET statement and a MERGE statement. These are parallel streams of data. The SET statement keeps only the ID variable and is accompanied by a "BY ID" statement, which produces the dummy variable LAST.ID (indicating that the record in hand is the last one for the current ID.
- The MERGE statement reads the identical record as the SET statement and simulaneously one variable (ENTER_DATE renamed to _NXT_ENTER) from the next record (due to the "FIRSTOBS=2" parameter). The _NXT_ENTER is compared to the current LEAVE_DATE+30 at the end of every do-loop iteration, because of the "UNTIL …" condition.
- Note the "KEEP=ID" in the first SET statement is not absolutely necessary (all the variables are re-read in the MERGE statement). But I use it as a technique to remind me of why I use the SET statement (just to implement the "BY ID" statement).
- Once the UNTIL condition is satisfied copy _ENTER1 to ENTER_DATE,
- The re-read exactly _N records, keeping all the variables except ENTER_DATE and LEAVE_DATE, and output each of those re-read records.
