topic Re: Searching a string variable for various words in SAS Procedures

Searching a string variable for various words

Dani08 — Tue, 28 Feb 2017 14:19:54 GMT

I have a lookup table of product names (around 200)

For example,

BARKAT
CALOGEN
CALSHAKE
etc.

and I would like to search each of these products through all of my observations.

I am not sure how to go about this as I am still new to SAS...

I need to do some sort of a scan as the product names do not exactly match up.

For example, when I search for these names in the productname variable, I would like to flag things like...

Barkat brown rice
Calogen extra
Calogen extra shots
Calshake
etc.

I made a cntlin list of the product names, and tried to use indexw/findw, but these seem to require individual words. My list is up to 200, so a manual search is not an option. I wonder if there is a way to do this?

Re: Searching a string variable for various words

RW9 — Tue, 28 Feb 2017 14:28:44 GMT

This isn't the best topic to be starting to learn with if your new to SAS.

As for how to do it, there are a few ways, the way I would go with (and you have not provided any test data in the form of datasteps so this is only an example) - assumes listofwords contains the words, and have contains one string variable called sentance:

data _null_;
  set listofwords end=last;
  if _n_=1 then call execute('data want; set have; length wordsfound $2000;');
  call execute('if findw(sentance,"'||strip(word)||'") >0 then wordsfound=catx(",",wordsfound,"'||strip(word)||'");');
  if last then call execute('run;');
run;

What this does is create the datastep code after this is executed, which has an if statement for each word in listofwords, with a catx() function call to add any words found to wordsfound variable.

As mentioned this is advanced, not a good area for learning as the above probably doesn't make any sense to you.

Re: Searching a string variable for various words

Dani08 — Tue, 28 Feb 2017 14:46:28 GMT

Thank you this absolutely worked, but as you say I will need to look into the code to understand it fully. Thank you for your time.