To separate letters in a word

[Ken_nuga]

I have a patient unique ID, say "CT00001A" and I want to separate the prefix "CT" from the serial number "00001" and the suffix "A". Here is a program I used

DATA TEST;

SET sasuser.CIHP_RADET;

B = COMPRESS (A,"","A");

C = COMPRESS (A,B); RUN;

PROC PRINT DATA = TEST;

RUN;

and it returned 00001 in one column and CTA in another. Now I want to separate CT from A in 2 separate columns

[Reeza]

What are the rules? 
Why can't you just substring the first two characters using SUBSTR()?

Obviously this will need to be used on multiple observations so is it always the first two characters? Or first set of letters? How do you know A needs to be removed? What if it was "CT000B1A" what would be kept or is that not possible?

[Ken_nuga]

Alright, here is a better explanation.

I have attached 2 SAS OUTPUT pictures and 2 Excel pictures of the dataset, the variables "prefix", " serial number" and "suffix" were meant to concatenate as "Patient Unique ID", but a data clerk erroneously copied the wrong thing.

The A or B after "CT00001A" is the suffix.

Thank you.

[Ken_nuga]

The prefixes could be any of CT, TI, TII, PM, OPI, AMOPI, TB, OPEI, PNE. Therefore, it is not possible to substring the first two characters, but it is the first set of letters that needs to be moved into a separate column as "prefixes."
The alphabet that comes after the numbers always comes after the numbers because they were meant to be entered into a separate column as suffixes and then concatenated.
Presently, my challenge is how to separate the last alphabet from the following examples CTA, AMOPIB, OPEIC as the COMPRESS function groups all the alphabets into one column.

[mkeintz]

Your rules apparently expect the original value to have a alphabetic prefix, a numerical middle and an alphabetic suffix, in which each  component can have a varying length.  You can use SCAN to get the prefix and suffix, when you declare all digits as delimitrs.  And the middle is nothing but the original string with all alphabetic characters removed (i.e. compressed out):

data _null_;
  string='CTX00001A';
  str_prefix=scan(string,1,,'d');
  str_middle=compress(string,,'A');
  str_suffix=scan(string,2,,'d'); 
  put (str:) (=);
run;

[Ken_nuga]

Many thanks, but I am actually trying this on a variable with several observations. Here is a short copy of it:

The SAS System

Obs	A
10	CT00010A
54	AMOPI00066
55	CT00069DV
56	CTPM00069B
67	OPEI00079DV
68	PM00080A
69	PM00080C
70	TI00080A
71	PM00080b
72	CT00081B
73	CT00081A
75	PM00084A
76	TI00084B
77	CT00085A
78	CT00085B
79	PM00086A
80	PM00086B
81	PM00088A
82	PM00088B
88	PM00092A
89	PM00092B
101	TI00117B
102	TI00117A
179	PM02414
200	PI03049A
201	PI03049B

[Reeza]

---

@Ken_nuga wrote:

Many thanks, but I am actually trying this on a variable with several observations. Here is a short copy of it:

---

It doesn't matter how many observations, the code is the same. You need to reference your data set and your variables, but the idea is correct. 

[chaatak · Accepted solution]

Please try this:

data new;
infile datalines;
input x $15.;

k1 = anydigit (x);
A = substr (x, 1, k1-1);
temp = substr(x,k1);
k2 = anyalpha (temp);
Num = substr(temp, 1, k2-1);
B = substr(temp, k2);
keep x A Num B;
datalines;
CT00010A
AMOPI00066
CT00069DV
CTPM00069B
OPEI00079DV
;

[Ken_nuga]

Alright, it is done! Here is what I did and a partial print out.

proc import out = sasuser.CIHP_RADET datafile = 'C:\Users\ONUGHA\Desktop\A.xlsx' dbms=xlsx replace;

run;

DATA TEST;

SET sasuser.CIHP_RADET;

k1 = anydigit (A);

PUID_PREFIX = substr (A, 1, k1-1);

temp = substr(A,k1);

k2 = anyalpha (temp);

PUID = substr(temp, 1, k2-1);

PUID_SUFFIX = substr(temp, k2);

keep A PUID_PREFIX PUID PUID_SUFFIX;

RUN;

PROC PRINT DATA = TEST; RUN;

The SAS System

Obs	A	PUID_PREFIX	PUID	PUID_SUFFIX
1	PM00003DV	PM	00003	DV
2	CT00004MSG	CT	00004	MSG
3	OPI00004AMOPI	OPI	00004	AMOPI
4	CT/PI00005	CT/PI	00005
5	CCP/PM00006	CCP/PM	00006
6	CT00007	CT	00007
7	PM00008	PM	00008
8	OPI00008	OPI	00008
9	PM00009	PM	00009
10	CT00010A	CT	00010	A
11	PM00010	PM	00010

Great, thank you!

[Reeza]

@Ken_nuga in this case you shouldn't have marked your answer as the solution, @chaatak provided the correct solution and it's worth a few seconds to acknowledge that.