Hi,   A good approach here is to first create the target table without any data but with the definitions as you want them. Then you run your extract and do a force append of that to the dataset you created. This will make certain you get what you want instead of letting the access engine decide. For example:   /** * Step 1: create the target table. */ data target; attrib col1 length=$10; attrib col2 length=$40; attrib col3 length=8; call missing(col1, col2, col3); stop; run; /** * Step 2: define the source, preferrably as a view. */ proc sql; create view source as (select col1, col2, col3 from ....); quit; /** * Step 3: Load the source data into the target table */ proc append base=target data=source force; run; This is a pattern that is typically implemented in the data loader transform in SAS DI Studio.   Hope this helps, -- Jan. ... View more

How about this?   data have; input id count result; cards; 1 1 0 1 2 0 1 3 1 1 4 0 1 5 0 1 6 1 2 1 0 2 2 0 3 1 1 4 1 0 4 2 0 4 3 1 5 1 1 5 2 0 5 3 1 ; proc sql; create table res1 as select id, min(count) as count1 from t where result=1 group by id; quit; data want; merge have res1; by id; if count<count1 then test_before = 'yes'; else if count>count1 then test_after = 'yes'; run; I create an intermediate set with al id's and the count at which the result was 1 for the first time. Then I left-join that with the original set to determine if a row was before or after the first result=1.   This all assumes the data was sorted by ID and Count. If not you need to add an additional sort step.   And yes, with some effort this can be rewritten in a single SQL. As usual, many roads lead to Rome.   Hope this helps, - Jan. ... View more

What you're looking at is primarily a presentation issue. The default format for numeric does not cater for so many decimals. That's why SAS reverts to the scientific ("E") notation. Always keep in mind that what you see is not exactly always what you have.   Use the format statement to handle this. Also, read up on numeric precision in SAS. This is a very interesting topic and can be very helpful in your case. As you are actually trying fuzzy comparison, also have a look at the FUZZ() function.   in your case this should help:   data comb1; merge ds1 ds2; by Index; format a b d comma12.10; D=a-b; if A = B then A_eq_B = "Yes"; if round(A,0.01) = round(B,0.01) then A_eq_B_1 = "Yes"; run;   Hope this helps, - Jan. ... View more

Looks like you based your code on another Communities answer where If_True and If_False wereactually part of the data. They are not part of any regex syntax. Here they are not so you will always see found=0. I have simplified your regex and this works for me. I hope it also work for you.   data dataout; set datain; if prxmatch("m/.*?5648/i",name) > 0 then found=1; else if prxmatch("m/.*?9463/i",name) > 0 then found=2; else if prxmatch("m/.*?4721/i",name) > 0 then found=3; else found=0; run;   Regards, - Jan.     ... View more

I am assuming you have LSF as the scheduler and looking at Flowmanager for the color coding. This is the most common scenario. LSF does not know of a warning condition; it only distinguishes between success (rc=0) and failure (rc>0). This is why many admins decide to reduce a warning completion code (rc=2) to rc=0. Otherwise a warning would potentially stall the progress of a flow. The color yellow in Flowmanager does not indicate a warning but merely the "initializing" or "waiting" state. In that common scenario a warning will  implies success. Otherwise it will be interpreted as an error and the job will show up red.   The code to accomplish this is already present in the sasbatch.sh script as comments at the bottom of the file: # Add this code to capture exit=1 (SAS warning) and make it exit=0 # Comment out exec line above and uncomment the code below to allow the return code to be captured #"$SAS_COMMAND" -noxcmd -lrecl 32767 "$@" "${USERMODS_OPTIONS[@]}" #rc=$? #if [ $rc -eq 1 ]; then # rc=0 #fi #exit $rc Follow the instruction to turn rc=2 into rc=0.   The difference between "completes successfully" and "ends with any exit code" is what it says: a dependent job will either run only when its predecessor end with rc=0 or will always run when the predecessor ends regardless of its return code.   The condition "Ends with exit code" allows for user defined return codes (using the SAS abort statement) to control specific paths through the flow.   The value for "Runs more than" will trigger a dependent job only when the predecessor runs longer than a specific amount of time. This could help for example when you want to be alerted for unusually long running jobs.   SAS and scheduling docs are here. LSF is extensively documented in this IBM Knowledge Center doc . Keep in mind that the LSF software that comes with SAS is somewhat limited compared to the full functionality you would buy from IBM.   Hope this helps, - Jan. ... View more