Efficient way to extract words from text and place in new variable

[Caetreviop543]

I'm trying to figure out the best, most succinct way to extract words from a text variable, and place them in a new variable. I have a text variable:

TXT

Some text

More text

List text

Read text

And I want to extract certain words from TXT and create a new variable:

TXT                   Extract

text Some          Some

More text           More

text word List     List

Read text          Read

So far this is the only code I came up with:

data new;
set old;

if find(TXT, "Some")>0 then Extract = "Some";
else if find(TXT, "More")>0 then Extract = "More";
else if find(TXT, "List")>0 then Extract = "List";
else if find(TXT, "Read")>0 the Extract = "Read";

run;

It works, but is pretty clunky. Does anyone know  a better way?

[novinosrin]

Extract=scan(txt,1,' ');

[Caetreviop543]

I made the example too simple; the word I want to extract isn't always the first. 

[novinosrin]

Oops sorry I overlooked completely. my apologies

[Caetreviop543]

That's ok! I should have made the example better.

[novinosrin]

HI @Caetreviop543  Would this come close?

Basically, eliminating the word 'text' in the string

data have;
input TXT & :$10. ;
cards;

Some text

More text

List text

Read text
text red
;

data want;
 set have;
 temp=strip(tranwrd(txt,'text',' '));
run;

[Caetreviop543]

Thanks for your response. Is there a way to extract just the word if the text isn't all the same, i.e., the actual word text doesn't appear throughout? I just used text to symbolize a string of words. 

[novinosrin]

Is there a way to extract just the word if the text isn't all the same,

By the above, do you mean you only want to extract a set of known words? Like as though you have a list of words to extract from the string???Please clarify

---

@Caetreviop543 wrote:

Thanks for your response. Is there a way to extract just the word if the text isn't all the same, i.e., the actual word text doesn't appear throughout? I just used text to symbolize a string of words. 

---

[ChrisNZ]

You can write this in a more compact manner, but it is less legible.

EXTRACT = prxchange('s/.*(Some|More|List|Read).*/\1/',-1,TXT);

This means: find anything then one of the strings then anything else and replace with the string found.

[Caetreviop543]

It told me the prxchange argument doesn't have enough arguments. How does it know which variable to change/extract words from?

[ChrisNZ]

Try again, I corrected within seconds . I clicked too fast and you are too fast. 🙂

[ChrisNZ]

1 is probably fine rather than -1 actually. Sorry I can't test.

[Caetreviop543]

Thanks, but that didn't work. It returned a much larger list of lots of text.

[ChrisNZ]

Mmm that's odd. I probably forgot something obvious. I can't test sadly. 

[novinosrin · Accepted solution]

Hi @Caetreviop543  For hard coded char constants, you could consider a fast approach which is merely tweak of your original. See if this helps

data have;
input TXT & :$10. ;
cards;

Some text

More text

List text

Read text
text red
;

data want;
 set have;
 array t(5) $32 _temporary_ ('Some' 'More' 'List' 'Read' 'red') ;
 length extract $32;
 do _n_=1 to dim(t) until(not missing(extract));
  if index(txt,strip(t(_n_))) then extract=t(_n_);
 end;
run;