Find Strings matching a certain pattern and remove the exceptions.

[david27]

Hello,

Please advise.

%let exception=b_n1h1 c_n1h1 f_n1h1 m_n1h1;
%let have=(n1h2_output*((n1h1_n1h2*((a_n1h1*a[i]-b[1]) + (b_n1h1*b[i]-b[1]) + (c_n1h1*c[i]-b[1]) + (d_n1h1*d[i]-b[1]) + 
(e_n1h1*e[i]-b[1]) + (f_n1h1*f[i]-b[1]) + (g_n1h1*g[i]-b[1]))-b[2]);

I want a list of all values(anything that has an underscore) from 'have' that are NOT in 'exception'

want==n1h2_output n1h1_n1h2 a_n1h1 d_n1h1 e_n1h1 g_n1h1

Thank You

[novinosrin · Accepted solution]

Hi @david27  Please see if this works. I just took a quick stab at your string with some easy assumptions-

%let exception=b_n1h1 c_n1h1 f_n1h1 m_n1h1;
%let have=(n1h2_output*((n1h1_n1h2*((a_n1h1*a[i]-b[1]) + (b_n1h1*b[i]-b[1]) + (c_n1h1*c[i]-b[1]) + (d_n1h1*d[i]-b[1]) + 
(e_n1h1*e[i]-b[1]) + (f_n1h1*f[i]-b[1]) + (g_n1h1*g[i]-b[1]))-b[2]);

data have;
  have="&have";
run;
data want;
 set have;
 array ex(4)$30 _temporary_('b_n1h1' 'c_n1h1' 'f_n1h1' 'm_n1h1');
 do _n_=1 to countw(have,'*');
  w=scan(have,_n_,'*');
  want=scan(w,-1);
  if index(want,'_') and want not in ex then output;
 end;
 drop w;
run;

have	want
(n1h2_output*((n1h1_n1h2*((a_n1h1*a[i]-b[1]) + (b_n1h1*b[i]-b[1]) + (c_n1h1*c[i]-b[1]) + (d_n1h1*d[i]-b[1]) + (e_n1h1*e[i]-b[1]) + (f_n1h1*f[i]-b[1]) + (g_n1h1*g[i]-b[1]))-b[2])	n1h2_output
(n1h2_output*((n1h1_n1h2*((a_n1h1*a[i]-b[1]) + (b_n1h1*b[i]-b[1]) + (c_n1h1*c[i]-b[1]) + (d_n1h1*d[i]-b[1]) + (e_n1h1*e[i]-b[1]) + (f_n1h1*f[i]-b[1]) + (g_n1h1*g[i]-b[1]))-b[2])	n1h1_n1h2
(n1h2_output*((n1h1_n1h2*((a_n1h1*a[i]-b[1]) + (b_n1h1*b[i]-b[1]) + (c_n1h1*c[i]-b[1]) + (d_n1h1*d[i]-b[1]) + (e_n1h1*e[i]-b[1]) + (f_n1h1*f[i]-b[1]) + (g_n1h1*g[i]-b[1]))-b[2])	a_n1h1
(n1h2_output*((n1h1_n1h2*((a_n1h1*a[i]-b[1]) + (b_n1h1*b[i]-b[1]) + (c_n1h1*c[i]-b[1]) + (d_n1h1*d[i]-b[1]) + (e_n1h1*e[i]-b[1]) + (f_n1h1*f[i]-b[1]) + (g_n1h1*g[i]-b[1]))-b[2])	d_n1h1
(n1h2_output*((n1h1_n1h2*((a_n1h1*a[i]-b[1]) + (b_n1h1*b[i]-b[1]) + (c_n1h1*c[i]-b[1]) + (d_n1h1*d[i]-b[1]) + (e_n1h1*e[i]-b[1]) + (f_n1h1*f[i]-b[1]) + (g_n1h1*g[i]-b[1]))-b[2])	e_n1h1
(n1h2_output*((n1h1_n1h2*((a_n1h1*a[i]-b[1]) + (b_n1h1*b[i]-b[1]) + (c_n1h1*c[i]-b[1]) + (d_n1h1*d[i]-b[1]) + (e_n1h1*e[i]-b[1]) + (f_n1h1*f[i]-b[1]) + (g_n1h1*g[i]-b[1]))-b[2])	g_n1h1

[david27]

Thank You for the code @novinosrin 

Can this be done without a datastep?

Also not sure if this helps but,

the string that i am looking for

• has an underscore
• begins with ( and ends with *
• is not in exception list.

[novinosrin]

Hi @david27  "Without a datastep makes it more a heavier syntax in my opinion-

 options minoperator mindelimiter=' ';
%macro t;
 %global want;
 %let want=;
 %let exception=b_n1h1 c_n1h1 f_n1h1 m_n1h1;
 %let have=(n1h2_output*((n1h1_n1h2*((a_n1h1*a[i]-b[1]) + (b_n1h1*b[i]-b[1]) + (c_n1h1*c[i]-b[1]) + (d_n1h1*d[i]-b[1]) + 
(e_n1h1*e[i]-b[1]) + (f_n1h1*f[i]-b[1]) + (g_n1h1*g[i]-b[1]))-b[2]);
 %do i=1 %to %sysfunc(countw(%bquote(&have),%str(*)));
  %let w=%scan(%bquote(&have),&i,%str(*));
  %let need=%scan(%bquote(&w),-1);
  %put &=need;
  %if %index(&need,%str(_))>0 and %eval(&need  in (&exception))=0  %then %let want=&want &need;
 %end;
%mend t;
%t

%put &=want;

LOG:

       %put &=want;
WANT=n1h2_output n1h1_n1h2 a_n1h1 d_n1h1 e_n1h1 g_n1h1

[david27]

You were correct. Data step is better. Also it’s more flexible.
Thank You very much.