Here is another possibility, which requires sorting the big HAVE1 table. We can then create a view (or a table, but a view is probably faster in this case) containing the possible years in the correct order (this is of course a very good solution if your actual data, unlike your demonstration data, is actually sorted already): proc sort data=have1; by name code start; run; data have1_years/view=have1_years; do until(last.code); set have1; by name code; do year=max(year+1,year(start)) to year(end); output; end; end; keep name code year; run; data want; merge have1_years(in=in1) have2(in=in2); by name code year; if in2; exist=in1; run; The solution can also be done extracting NAME, CODE and YEAR from HAVE1, and then sort the output data afterwards. This may be a better solution if you have many columns in HAVE1: data have1_years; set have1; do year=year(start) to year(end); output; end; keep name code year; run; Proc sort nodupkey; by name code year; run; ... View more

This appears to work for your example data: data want; merge have2 (in=in2) have1 (in=in1); by name code; exist = (year(start) le year le year(end)); drop start end; if in2; run; Caution: this requires the sets to be sorted by Name and Code. IF ONLY Code is to be considered for the BY match then both sets need to be sorted by CODE. If a Name/Code in the first case, or Code in the second exists on two or more observations in both sets you will need to provide some rules as that could well mean a much more complicated situation and resolution.   It is a good idea when considering comparing Year or Month values against actual dates to carefully describe the rules. It may be that you get some cases where you don't expect Year=2006 to be considered in the interval 20050712 20060531 for some reason, such as less than half of 2006 is in the interval. So if  you have such rules that were not stated then you need to add them. ... View more

data want (drop=_:); set have (keep=source target); by source target notsorted; merge have (rename=(start=_current_start)) have (firstobs=2 keep=start rename=(start=_nxt_start)); retain start ; format start yymmddn8. ; if first.target=1 or _current_start-1>lag(end) then start=_current_start; if last.target=1 or end+1<_nxt_start; run; The SET and MERGE statement set up two streams of data, read in parallel.  The first SET is accompanied by the BY statement, allowing use of first.target and last.target dummies.  The data are assumed to be grouped by source*target, and within each source*target group, to be sorted by start.  (and also sorted by END within each group).   The MERGE statement has two "substreams" of data, also read in parallel, but because of the FIRSTOBS=2 option provides a way to look ahead at the upcoming START value (in variable _nxt_start), so upcoming date gaps can easily be found.   The first IF identifies the beginning of a continuous sequence of observations. The second if identified the end of same. ... View more

MDY has nothing to do with this issue. The issue is that you don't have enough memory because the tables are so large.   Please give us the information about how many rows and columns are in each data set. ... View more