REGEXP_LIKE help

Question

Hello,
Been scouring the web looking for some help using the REGEXP_LIKE syntax. Haven't found a precise answer, but I think I wrote my own correctly and just wanted to get some validation from other users.
Running Oracle 10.2
Sample Data:
create table test(data varchar2(10));
    insert into test values('01W00012');
    insert into test values('50321070');
    insert into test values('A1391010');
    insert into test values('ZZ260010');
    insert into test values('29374C01');
    insert into test values('A938523A');
    insert into test values('23W48AC3');
    insert into test values('DKFUGHCK');
    insert into test values('09W00032');
    insert into test values('94857283');
    insert into test values('AB-29348');
    insert into test values('98W-8923');
    insert into test values('0AW00005');
    insert into test values('04300W05');
    commit;What I'm trying to do:
I have a table containing millions of work orders. They are all of length 8. I want to find all work orders where the first 2 characters are a digit only, the third character is a 'W', and the last 5 characters are digits only. Anything else I want to throw away. I think I came up with a working expression... but I'm always hesitant when running it against millions of rows. Here was what I did:
select * from test
    where regexp_like(data, '(^[0-9]{2}W{1}[[:digit:]]{5})');There are exactly 2 occurrences from up above that match the criteria I'm trying to meet.
Is this expression properly written?
Any help would be greatly appreciated.... reg expressions always make my head spin :(

Answer 1

Hi,
dvsoukup wrote:
Running Oracle 10.2
Sample Data: ...Thanks for posting that; it's really helpful!
What I'm trying to do:
I have a table containing millions of work orders. They are all of length 8. I want to find all work orders where the first 2 characters are a digit only, the third character is a 'W', and the last 5 characters are digits only. Anything else I want to throw away. I think I came up with a working expression... but I'm always hesitant when running it against millions of rows. Here was what I did:
select * from test
where regexp_like(data, '(^[0-9]{2}W{1}[[:digit:]]{5})');There are exactly 2 occurrences from up above that match the criteria I'm trying to meet.
Is this expression properly written?Yes, that's a good regular expression. However, a very important thing about regular expressions is that you should only use them when you really need to. In this case, you can get the results you want more efficiently, and at least as easily, without regular expressions. Here's one way:
SELECT  *
FROM     test
WHERE   TRANSLATE ( data
            , '012345678'
            , '999999999'
            )          = '99W99999'
;This will be more efficient than using REGEXP_LIKE, and it sounds like efficiency is important in this case.
If you really want to use a regular expression, there are a couple of small things you can do to make it clearer. (With regular expressions, every little bit helps.)
(a) If you're certain that data is always exactly 8 characters, then you don't need the ^ anchor at the beginning. If you want the regular expression to check that the length is 8 characters, then keep the ^ anchor at the beginning, but use a $ anchor at the end, too, as in the example below. As you posted it, the expression will be TRUE when data is longer than 8 characters, as long as the first 8 characters fit the pattern.
(b) [0-9] is equivalent to [[:digit:]]. Either one is fine, but I find it confusing to use both in the same expression. Use one or the other in both places.
(c) {1} is the default. I would just say 'W' instead of 'W{1}'
(d) You don't need the parentheses in this expression. If parentheses are not needed, but they improve clarity, then it's okay to include them. In this case, I don't think they add anything.
There are exactly 2 occurrences from up above that match the criteria I'm trying to meet.It helps if you post exactly what results you want. For example:
DATA
01W00012
09W00032In this case, I found your description very clear, but it doesn't hurt to post the actual results anyway.
Is this expression properly written?You might consider writing it this way, particularly if you think you'll need to debug it:
SELECT  *
FROM      test
WHERE      REGEXP_LIKE ( data
              , '^'          || -- beginning of string
                '[0-9]{2}'     || -- 2 digits
                'W'          || -- a capital W
                '[0-9]{5}'     || -- 5 more digits
                '$'             -- end of string
;As mentioned before, you only need the ^ and $ if you're not sure all the strings are exactly 8 characters.
... reg expressions always make my head spin All the more reason to use TRANSLATE.
Edited by: Frank Kulash on Jul 26, 2012 2:42 PM