Remove the check in the Special Values : All possible values. Then add the list of values as you would like to see them ... View more

Look into using hash tables. They are much faster than the other ways of doing this.   Google it!    First read the excel sheet into a SAS table. If the example reflects the logic completely, you would want to split each row into 2 lines   customer_id1 customer_id2 combined abc cde abc/cde bbb bdc bbb/bdc   becomes a sas table called  customer_lookup like   customer_id combined abc abc/cde cde abc/cde bbb bbb/bdc bdc bbb/bdc     /* this uses a macro that simplifies the code to define a hash table In your case, maybe good to look up the syntax! */ /* If interested, I can post the macro ut_hash_define */ data want; set have; %ut_hash_define(intable=customer_lookup, keys=customer_id, datacols=combined); rc = customer_lookup.find(); run; ... View more

  I suggest that you add a line of code  options dsoptions=note2err; at the beginning of your long program. That will better highlight problems that you may have overlooked.  Then fix your code so that these new errors are avoided. ... View more

First assign timenow  to a data step character var.  The result of put is a char string.   Then you can use call symput to assign it.   . YOu also need to use call symput to assign reportname ... View more

You probably want to use format ddmmyyn8. ... View more

That is 2 GB. That's a pretty standard setting in multiuser environments. The server probably has something like 256 GB total, so 2GB each allows roughly 120 concurrent users. If the administrators gave everyone 4GB, then you would the server would be maxed with only 60 users. But maybe there is a process at the university for you to justify why you need more space and they may allocate more for a week or so. I bet that if you were a professor researching for a Nobel prize, you would easily get a lot more. I have no idea if the memory usage increase is linear (in which case you would need 6 GB) or if it increases quicker. But I would apply for 8 GB as a start. ... View more

When using _NUMERIC_ and _CHARACTER_ be aware that all columns defined before the array definition in a SET , MERGE, etc. are added to the array and columns defined after the array statement are not.  Try this: data _null_ set sashelp.cars; * This var is in the array nums ; length num_before_array 8; array nums _numeric_; * This var is not the array nums; length num_after_array 8; if _n_=1 then do; do i=1 to dim(nums); vname = vname(nums(i)); put vname; end; end; run; ... View more

If var4 is not evenly distributed, your blocks won't be of equal size, and this won't help much. You could use the firstobs and obs dataset options to simply split the big table into n blocks. Your macro then loops from 1 to n and handles each block in a loop. That means that each proc sort has input with 1/n of the records and outputs the a new temp table. You could make many more blocks so that each block can be sorted iin memory.   You are thereby only reading the big table once, albeit in 1 chunk at a time.  You really dont need a split step at all.   You still need to put the outputs from each block back together again in a following step,  So you need to output each block with the loop index as a suffix.      With this approach  you could in theory run the logic for the blocks in parallel, assuming the hardware and setup is available.    ... View more

Nice options, but I personally am moving in the opposite direction, such as using the NOTE2ERR option, since that forces me  to immediately avoid having columns created that are not spelled correctly, etc.   As a consultant , I have recently been migrating some code, and found logs that are riddled with notes about missing values, etc. which could cause problems / create missing values in some cases.   ... View more