Pattern matching to two different patterns

[Walternate]

Hi,

I am doing some validity tests using pattern matching. 

The issue is that the valid values I want not to be flagged can be either a 4-digit string or a 5-digit string.

This is what I set up:

if _n_ = 1 

     then do;

           pattern_id3n=prxparse("/\d{5}/i");

           pattern_id3nb=prxparse("/\d{4}/i");

    end;

retain pattern:;

if (prxmatch(pattern_id3n, id3n)  ne 1) or (prxmatch(pattern_id3nb, id3n)  ne 1) then do;

           bad_id3n=id3n;

           bad_id_flag = 1;

end;

I'm not getting any bad messages, but I'm seeing values of 4-digits flagged as not valid so obviously that part of the pattern matching is not working correctly. 

[ballardw]

Where is the example values? Where are the examples of what is supposed to found as a pattern.

"The issue is that the valid values I want not to be flagged can be either a 4-digit string or a 5-digit string." Has no details of what the pattern you are looking for actually is.

---

@Walternate wrote:

Hi,

I am doing some validity tests using pattern matching. 

The issue is that the valid values I want not to be flagged can be either a 4-digit string or a 5-digit string.

This is what I set up:

if _n_ = 1 

     then do;

           pattern_id3n=prxparse("/\d{5}/i");

           pattern_id3nb=prxparse("/\d{4}/i");

    end;

retain pattern:;

if (prxmatch(pattern_id3n, id3n)  ne 1) or (prxmatch(pattern_id3nb, id3n)  ne 1) then do;

           bad_id3n=id3n;

           bad_id_flag = 1;

end;

I'm not getting any bad messages, but I'm seeing values of 4-digits flagged as not valid so obviously that part of the pattern matching is not working correctly. 

---

[Walternate]

Valid:

1234

12345

Invalid:

1

12

123

123456

1234567 

etc. 

So anything 4-digits or 5-digits is valid but nothing else is. 

[Tom · Accepted solution]

If that is the complete list of possible values just use the LENGTH() function to find the 4 and 5 digit strings.

valid = length(string) in (4:5);

If you also have values like 'FRED' then perhaps include a VERIFY() function call also.

valid = (not verify(trim(string),'0123456789')) and length(string) in (4:5) ;

[Tom]

Why not say that in the regular expression?

Use ^ to indicate it has to start at the beginning.  Add space and *$ at the end to indicate it can only be followed by the spaces used to pad the string to the full variable length.

data test;
  if _n_ = 1 then do;
    pattern_id3n=prxparse("/^\d{5} *$/");
    pattern_id3nb=prxparse("/^\d{4} *$/");
  end;
  retain pattern: ;
  drop pattern: ;
  input string $char20. ;
  test5=prxmatch(pattern_id3n,string);
  test4=prxmatch(pattern_id3nb,string);
cards;
123
1234
12345
123456
12 456
xxx 12345 yy
;

Result

Obs    string          test5    test4

 1     123               0        0
 2     1234              0        1
 3     12345             1        0
 4     123456            0        0
 5     12 456            0        0
 6     xxx 12345 yy      0        0