@SM1 wrote: @s_lassen   I ran the suggested syntax; no errors were reported.   The PRXMATCH syntax properly flagged the observations (n=14) that met the criteria.   The FINDW syntax didn't flag any of the observations (n=0), although the new variable FLAG was created.   It seems like there should be a way to specify more than 1 parameter to match without having to specify the universe of acceptable temporal orders for the parameters.     If you do not get any matches with the FINDW method, that is probably because FINDW looks for whole words, and for some reason your keywords are not properly delimited. When I test against the test data supplied by @ballardw, the results look OK. If you are OK with not checking for whole words (as in the PRX example), you can try the FIND function instead. Or you can check the syntax of the FINDW function, and find out how you should specify the correct delimiters. It is true that you have to specify the "universe of acceptable temporal orders", as you call it, in order to find the matches with a single match string. Unless you have a way of specifying "look for NORMANDY or OVERBOARD, and then look for the one of the two words that was not found the first time". Which would take search string syntax to a whole new level of complexity, I think. ... View more

This is not a very good case for using a single PRXMATCH call. But of course it can be done: data want; set have; array findtxt (5) narrative1 narrative2 narrative3 narrative4 narrative5; do i = 1 to 5; if prxmatch ("/(?:NORMANDY.*OVERBOARD|OVERBOARD.*NORMANDY)/i",findtxt(i)) then Flag = 1; end; drop i; run; But I would prefer the FINDW function: data want; set have; array findtxt (5) narrative1 narrative2 narrative3 narrative4 narrative5; do i = 1 to 5 until(flag=1); if findw(findtxt(i),'NORMANDY',,'i') and findw(findtxt(i),'OVERBOARD',,'i') then Flag = 1; end; drop i; run; And then I put in an UNTIL, as there is no reason to keep looking once a match is found. And actually, you should put in some \b markers in the PRXMATCH string, so that you only get whole words in the result; FINDW does that for you. ... View more

Sounds like what you basically want to do is to change the value of the MASS variable to the number of records having that last name and DOB? Something like this may work: proc sort data=have; by last_name DOB; run; data counts; do mass=1 by 1 until(last.DOB); set have(keep=last_name DOB); by last_name DOB; end; run; data want; merge have(drop=mass) counts; by last_name DOB; run;   ... View more

I assume that you got as far as to allocate an XML library and write to that. So what you have is a source path, which is the path of your XML library. And a destination path, which is the archive folder. And a member name, the name of the dataset that you wrote in the source path library. You now have 3 options. 1: to allocate the destination path with an XML libname as well, and use proc copy to copy it to the archive library;   libname dest xml '<destination folder>'; proc copy in=<source lib> out=dest; select <dataset name>; run; proc delete data=<source lib>.<dataset name>; run; This is probably the slowest, as SAS may do some processing of variables etc. along the way. As shown here, it also lacks a check for a successful copy, so that the source table only gets deleted if the copy goes well.     2: to copy the generated XML file as a flat file, using the FCOPY function:   filename in '<source folder>\<dataset name>.XML'; filename out '<destination folder>\<dataset name>.XML'; data _null_; if fcopy('in','out') then do; msg=sysmsg(); put msg; end; else rc=fdelete('in'); run; This is also portable, and probably as fast as it gets. And as you can see, it was quite simpe to put in a check for successful copy before deleting the input.   (3: Except if the source file is on the same physical disk as the archive folder, then you may get better performance using a native command to move the file, as it is only the reference to the file (the name of the directory) that needs to be changed, the actual file needs not be moved, But as others have already pointed out, you may not be allowed to use OS commands from your EG session anyway.)   ... View more

That's right, in SAS SQL you have to do it as SELECT table1.ID, table1.FirstName, (SELECT City FROM table2 WHERE table2.ID=table1.ID) as Hometown FROM table1; which I think is more correct SQL anyway. ... View more

One possibility is to create a macro which only lists the datasets that actually exist: %macro ifexist(datasets); %local i w; %do i=1 %to %sysfunc(countw(&datasets,%str( ))); %let w=%scan(&datasets,&i,%str( )); %if %sysfunc(exist(&w)) %then %do; &w%end; %end; %mend; data work.appended; set %ifexist( work.monday_em_t work.monday_da_t work.monday_ef_t work.monday_pr_t work.monday_lf_t work.monday_ov_t work.tuesday_em_t work.tuesday_da_t work.tuesday_ef_t work.tuesday_pr_t work.tuesday_lf_t work.tuesday_ov_t work.wednesday_em_t work.wednesday_da_t work.wednesday_ef_t work.wednesday_pr_t work.wednesday_lf_t work.wednesday_ov_t work.thursday_em_t work.thursday_da_t work.thursday_ef_t work.thursday_pr_t work.thursday_lf_t work.thursday_ov_t work.friday_em_t work.friday_da_t work.friday_ef_t work.friday_pr_t work.friday_lf_t work.friday_ov_t work.saturday_em_t work.saturday_da_t work.saturday_ef_t work.saturday_pr_t work.saturday_lf_t work.saturday_ov_t work.sunday_em_t work.sunday_da_t work.sunday_ef_t work.sunday_pr_t work.sunday_lf_t work.sunday_ov_t ); run; ... View more

Here is a case where Pearl Regular Expressions shine: data want; set have; if prxmatch('/\b(ab|bd|ge|mf|ac)\/\d{3,3}\/\d{3,3}\//',path); run; A short explanation of the match string: \b is a word boundary (ab|bd|ge|mf|ac) means either of the letter combinations you are looking for \/ means a slash ("/", which has to be escaped with a backslash) \d{3,3} means exactly 3 digits ... View more

The fastest and easiest way to do it is probably to modify the dataset in place: data x; if 0 then modify x; /* modify, but do not read any rows */ output; output; stop; run; ... View more

When doing rolling sums/means, the easiest is often to put the values in a temporary array. Only here we have to use LAG because we do not want the newest values in: data want; array values (0:10) 8 _temporary_; call missing(of values(*)); do _N_=0 by 1 until(last.permno); set have; by permno; sum=sum(of values(*)); output; values(mod(_N_,11))=ifn(_N_=0,.,lag(Ret)); /* use IFN so that LAG is always called */ end; run; If you do not want the sum before you have all 11 values, use "if _N_>=10 then sum=sum(of values(*));" ... View more