Hash tables - find code

[Chris_LK_87]

Hello, 

I am using the script below to find certain codes in a very large dataset with hash-tables. Now I have to write the exact code for the dcode I would like to find (e.g: 'I412' or 'I519') for the scan-function. Instead I would like to find every code that starts with 'I41' or 'I51' with the help of hash-tables using colon (:) modifier? Is that possible?

data have;
length id $10 dcode $48;
input id$ dcode$ &;
datalines;
1 I410 J332
2 C450 I412
3 I413 R789
4 R281 I519
5 O603 C351
;
run;

data Codes;
length code $4.;
input code $ ;
datalines;
I410
I412
I413
I519
;
run;

DATA Code_scan;
length code $4.;

if _N_=1 then do;

declare hash P(dataset: 'Codes');
P.defineKey('code');
P.defineData('code');
P.defineDone();
end;

set

have

;

length Findcode $4;

do i=1 to 30 by 1;
Findcode=scan(Dcode,i,'');
if Findcode='' then leave;
if P.find(key:Findcode) = 0
then do;
output;
leave;

end;
end;

run;

[Ksharp]

I don't understand . Hash Table perform the exact match . If you want fuzz match , try other way like PRX.

data have;
length id $10 dcode $48;
input id$ dcode$ &;
datalines;
1 I410 J332
2 C450 I412
3 I413 R789
4 R281 I519
5 O603 C351
;
run;

data want;
 set have;
 if prxmatch('/\b(I41|I51)/i',dcode);
 run;

[Chris_LK_87]

I use hash-tables to save time - so I would really like to use hash-tables. 

[AMSAS]

Here's another Perl Regular Expression example, I basically hacked this example in the documentation

data have;
	length id $10 dcode $48;
	input id$ dcode$ &;
datalines;
1 I410 J332
2 C450 I412
3 I413 R789 I412 I4123
4 R281 I519 I413 I51698
5 O603 C351 
;
run;

data want ;
	retain re 0 ;
	if _n_=1 then do ;
		re=prxparse('/I41\w*|I51\w*/') ;
	end ;
	set have ;
	x=prxmatch(re,dcode) ;
	start=1 ;
	stop=length(dcode) ;
	call prxnext(re,start,stop,dcode,position,length) ;
	do while(position>0) ;
		found=substr(dcode,position,length) ;
		call prxnext(re,start,stop,dcode,position,length) ;
		put found= ;
		output ;
	end ;
run ;

	
	

[AMSAS · Accepted solution]

I checked with a colleague who handles the hash object and you can't use a wild card, still there are other ways and this one works.

data have;
	length id $10 dcode $48;
	input id$ dcode$ &;
datalines;
1 I410 J332
2 C450 I412
3 I413 R789
4 R281 I519
5 O603 C351
6 3456 I510
7 I444 2345
8 I418 6789
9 I509 1234
10 9876 I412
;
run;



data Codes;
	length code $4.;
	input code $ ;
datalines;
I410
I412
I413
I519
;
run;
data tmp;
	set codes;
	chk=substr(code,1,3);
run;


DATA Code_scan;
	length chk $3.;
	if _N_=1 then do;
		declare hash P(dataset: 'tmp');
		P.defineKey('chk');
		P.defineData('chk');
		P.defineDone();
	end;
	set have ;
	length Findcode $4;
	do i=1 to 30 by 1;
		Findcode=substr(scan(Dcode,i,''),1,3);
		if Findcode='' then leave;
		if P.find(key:Findcode) = 0
		then do;
			output;
			leave;
		end;
	end;
run;

proc print;run;