topic How to set a user-defined pattern to be checked by like/contains condition in SAS Programming

How to set a user-defined pattern to be checked by like/contains condition

Dontik — Thu, 23 Oct 2014 09:27:17 GMT

I have a small problem, I have a large data set and I want to take these values that have certain characters (in my case - prefixes). There are dozens of these prefixes so doing LIKE/CONTAINS each time does not only take space but is inefficient.. Prefixes usually look like that - NLLL - so 1 number and 3 letters. Is there a way to check for that pattern in like/contains coditions or just a way to have these conditions contain multiple values i.e. IN option with prefixes or a list of prefixes that could be checked instead of only one?

Thanks in advance

Re: How to set a user-defined pattern to be checked by like/contains condition

RW9 — Thu, 23 Oct 2014 09:41:54 GMT

Yes, have a look at perl regular expressions - prxparse/prxmatch. You could also break it out a bit:

data want;

set have;

where (substr(PREFIX,1,1) in ('1','2','3'))

and (substr(PREFIX(2,3) in ('ABC','CBD'));

run;

If you have lots, then make a dataset of the combinations and use that to generate your code from:

data _null_;

set list_of_combinations end=last;

if _n_=1 then call execute('data want; set have; where prefix in ("'||strip(item)||'");

else call execute(',"'||strip(item)||'"');

if last then call execute('); run;');

run;

Re: How to set a user-defined pattern to be checked by like/contains condition

Patrick — Thu, 23 Oct 2014 10:46:34 GMT

If you don't have too many different patterns then a Regular Expression could be quite efficient for coding. Eg. for your pattern - NLLL - it could be something like:

where prxmatch('/^\d[[:alpha:]]{3}/o',<your variable>)