Your earlier answers indicate that your dates are true SAS dates. As you have seen, it's not necessary for them to have the same formats in order to merge them I justed wanted to be sure that if I added 1 to a date value, that it would get me to the next day (regardless of crossing month boundaries or year boundaries).
Here's a mildly intuitive approach that should be flexible enough. It turns the second data set into a format, and then runs through the first data set using the format to find matching values.
Starting with the second data set:
data fmt;
retain fmtname '$returns';
set B end=done;
start = catx(' ', permo, date);
label = ret;
output;
if done;
hlo='O';
label=' ';
output;
run;
proc format cntlin=fmt;
run;
This gives you a format that translates the combination of PERMO and DATE into a value for RET. The last piece of the format translates unrecognized values into a blank. Depending on what is actually contained in RET, you may have to deal with commas later. That's a relatively small puzzle piece. The bigger piece is that the results from a format always return a character string. So we will need to convert those to numeric. So ....
data want;
set A;
text = put( catx(' ', permo, date), $returns.);
ret_today = input(text, 8.); /* might need to change 8. if there are commas! */
* Get return from next available date, limited to 4 days of search;
do k = 1 to 4 until (text > ' ');
text = put( catx(' ', permo, date + k), $returns.);
end;
if text > ' ' then ret_tomorrow = input(text, 8.);
* Get return from previous available date, limited to 4 days of search;
do k = 1 to 4 until (text > ' ');
text = put( catx(' ', permo, date - k), $returns.);
end;
if text > ' ' then ret_yesterday = input(text, 8.);
drop k;
run;
This is untested code, so it might need some tweaking. It's probably a good strategy to get just the first four lines of this DATA step working first (getting RET_TODAY), before adding the rest. Also note, there's probably a way to do this using hashing, but my hashing skills aren't good enough! Good luck.
