How to remove the prefix and suffix at one step?

[leehsin]

Hi, I wanted to extract pact of the characters from a string which has delimiter '_'. The string is like this:

'abcd_ggg_fff_1234'

My question is: is there a single step to get 'ggg_fff" from that string, or have to do it in two steps?

Thanks!

[Jagadishkatam]

please try perl regular expression

data have;
x='abcd_ggg_fff_1234';
y=prxchange('s/(\w+)(ggg_fff)(.\d+)/$2/',-1,x);
run;

[leehsin]

Thanks, Jagadishkatam! But I have more strings with different middle parts:

'abcd_ggg_fff_1234'

'abcd_ttt_www_1234'

'qadc_hhh_lll_4321'

'dret_eee_1278'

I just want to do the general removal of prefix and suffix for all strings, not for a specific one.

Thank you!

[novinosrin]

HI @leehsin 

Call scan 

data have;
input str $30.;
cards;
abcd_ggg_fff_1234
abcd_ttt_www_1234
qadc_hhh_lll_4321
dret_eee_1278
;


data want;
set have; 
call scan(str, 1, position, length,'_');
substr(str,1,length+1)=' ';
call scan(str, -1, position, length,'_');
substr(str,position-1)=' ';
drop position length;
run;

[leehsin]

Hi, novinosrin, thanks for the solution. So, one step is not feasible. How about I want to keep both the original string and the new string?

[novinosrin]

Hi @leehsin  My second version gives you both. Please review

[novinosrin · Accepted solution]

Or 

data have;
input str $30.;
cards;
abcd_ggg_fff_1234
abcd_ttt_www_1234
qadc_hhh_lll_4321
dret_eee_1278
;


data want;
set have; 
call scan(str, 1, p, l,'_');
call scan(str, -1, position, length,'_');
want=substr(str,l+2,position-(l+3));
drop p: l:;
run;

[leehsin]

Great! This solved my problem! Thank you so much!

[novinosrin]

Hello @leehsin  Just to accept your challenge, here is a one step solution. It's not any faster or better but I loved your question.

data have;
input str $30.;
cards;
abcd_ggg_fff_1234
abcd_ttt_www_1234
qadc_hhh_lll_4321
dret_eee_1278
;


data want;
set have; 
want=substr(str,index(str,'_')+1, findc(str,'_','B')- (index(str,'_')+1));
run;

Also, Call scan is by far much faster than a regular expression for a simple problem like this unless your pattern really needs a regex

[novinosrin]

@leehsin   This linear approach is pretty quick

data have;
input str $30.;
cards;
abcd_ggg_fff_1234
abcd_ttt_www_1234
qadc_hhh_lll_4321
dret_eee_1278
;

data want;
set have;
length want $30;
do _n_=2 to countw(str,'_')-1;
want=catx('_',want,scan(str,_n_,'_'));
end;
run;

[leehsin]

Hi, novinosrin, 

Thank you for providing more solutions! These versatile ways will serve many kinds of scenarios. Thank you!

[Jagadishkatam]

data have;
input text :$100.;
if countw(text,'_')=4 then y=prxchange('s/(\w+\_)(\w+\_\w+)(\_\d+)/$2/',-1,text);
else if countw(text,'_')=3 then y=prxchange('s/(\w+\_)(\w+)(\_\d+)/$2/',-1,text);
cards;
abcd_ggg_fff_1234
abcd_ttt_www_1234
qadc_hhh_lll_4321
dret_eee_1278
;

[leehsin]

Jagadishkatam, yours is another good solution. Maybe more codes needed if I want to automatically detect the numbers for countw to use.

Thank you!