How to assign unique character variable to each observation

[Luke_D]

Hi,

I'm trying to assign a unique 6 character variable to each observation in a dataset.

eg.

1="AAAAAA"

2="AAAAAB"

3="AAAAAC"

.

.

26="AAAAAZ"

27="AAAABA"

etc.

It needs to be dynamic based on the first observation which may not always be "AAAAAA".

What would be the best way of doing this?

Thanks

[Reeza]

Please explain further what you mean by dynamic? Feel free to add more examples to illustrate the issue.

Here's a method to create an anonymized key data set if that's what you're looking for.
https://gist.github.com/statgeek/fd94b0b6e78815430c1340e8c19f8644

[Luke_D]

I mean that obs 1 won't always start "AAAAAA" so the next observation would need to add the next letter in the sequence based on the values of the previous observation.

eg.
1=""ADBBAB"
2="ADBBAC"
3="ADBBAD"

Thanks for the link but I'm trying to get each observation to follow the sequence rather than just an anonymized key.

[Reeza]

So you want each letter to be an alphabetic sequence?
Separate each character into an array. Then increment each level as needed.
The BYTE() function will convert a number to a character, 65 to 90 are the capitalized values, but you'll have to manually do the increment for Z to A to get the value.

Essentially you're using a numbering system with base 26 instead of 10 so you have to do all the math calculations yourself. It would be much easier to use numbers......

[Luke_D]

---

@Reeza wrote:
Essentially you're using a numbering system with base 26 instead of 10 so you have to do all the math calculations yourself. It would be much easier to use numbers......

---

I agree, but the variable is limited to 6 bytes and using letters gives us 300m+ more combinations than base 10.

[Reeza]

6 bytes for a numeric variable supports values up to 137,438,953,472 which is 444 times more values than using 6 characters.
1 byte is not a character when it comes to numeric values.

[FreelanceReinh]

Combining @Reeza's idea

Essentially you're using a numbering system with base 26 instead of 10 so you have to do all the math calculations yourself.

---

with @ChrisNZ's magic %ConvertBase macro one can calculate the n-th value of the sequence (for an almost arbitrary positive integer n) directly from the first without looping through the intermediate values.

Example:

%let value1=ADBZZU;
%let n=19;
%let c=%sysfunc(collate(65,90)); /* assuming the ASCII collating sequence */
%let d=0123456789%sysfunc(collate(65,80));
%let value&n=%sysfunc(translate(%substr(%ConvertBase(10,26,%eval(%ConvertBase(26,10,%sysfunc(translate(&value1,&d,&c)))+&n-1+26**6)),2),&c,&d));
%put &&value&n;

The code looks a bit cumbersome because the mathematical base-26 system uses digits 0, ..., 9, A, ..., P rather than A, ..., Z. That's where the TRANSLATE and COLLATE functions come into play.

[FreelanceReinh · Accepted solution]

Hi @Luke_D and welcome to the SAS Support Communities!

Try this:

/* Create sample data for demonstration */

data have;
set sashelp.class;
if _n_=1 then string='ADBBAB';
run;

/* Continue the alphabetic series of strings */

data want(drop=_:);
set have;
if _n_=1 then _s=string;
else do;
  _i=6-findc(_s,'Z',-6,'k');
  if _i>0 then substr(_s,7-_i)=repeat('A',6);
  if _i<6 then substr(_s,6-_i,1)=byte(rank(char(_s,6-_i))+1);
  else put 'CAUTION: Overflow "ZZZZZZ" --> "AAAAAA" occurred in obs. ' _n_;
  string=_s;
end;
retain _s;
run;

[novinosrin]

Sir @FreelanceReinh  Brilliant!!! I had to read your idea twice to get the hang of it.  Yes, got it. Very neat indeed. 🙂

[Luke_D]

Thanks for the welcome and the fantastic solution!

[Ksharp]

Just for fun.

%let dsn=sashelp.heart;
%let string= ADBZYZ ;







%let dsid=%sysfunc(open(&dsn));
%let nobs=%sysfunc(attrn(&dsid,nlobs));
%let dsid=%sysfunc(close(&dsid));
%let n1=%sysfunc(rank(%substr(&string,1,1)));
%let n2=%sysfunc(rank(%substr(&string,2,1)));
%let n3=%sysfunc(rank(%substr(&string,3,1)));
%let n4=%sysfunc(rank(%substr(&string,4,1)));
%let n5=%sysfunc(rank(%substr(&string,5,1)));
%let n6=%sysfunc(rank(%substr(&string,6,1)));


%put _user_;

data key;
 do n1=&n1 to rank('Z');

  loop2+1;
   _n2=ifn(loop2=1,&n2,rank('A'));
 do n2=_n2 to rank('Z');

   loop3+1;
   _n3=ifn(loop3=1,&n3,rank('A'));
 do n3=_n3 to rank('Z');

   loop4+1;
   _n4=ifn(loop4=1,&n4,rank('A'));
 do n4=_n4 to rank('Z');

   loop5+1;
   _n5=ifn(loop5=1,&n5,rank('A'));
 do n5=_n5 to rank('Z');

   loop6+1;
   _n6=ifn(loop6=1,&n6,rank('A'));
 do n6=_n6 to rank('Z');
   n+1;string=cats(byte(n1),byte(n2),byte(n3),byte(n4),byte(n5),byte(n6));output;
   if n = &nobs then stop;
 end;end;end;end;end;end;
 keep string;
run;


data want;
 merge &dsn key;
run;