Oracle REGEXP_LIKE match non-numeric character

Question

I have a set of rows in my database with VARCHAR2 column values like so: tac903.2, tac903.24, tac903.2a, tac903.2b

I want to create a regex that does not match if the last character in the search expression is numeric. So when I search for "tac903.2" I want to only match: tac903.2, tac903.2a, and tac903.2b. This should be simple but I'm not getting the desired behaviour from my regex here:

REGEXP_LIKE(col, 'tac903.2[a-z]?'); //This is matching the tac903.24 record.

REGEXP_LIKE(col, 'tac903.2[^0-9]?'); //This also is matching the tac903.24 record.

I'm a beginner with regular expressions, but based on my research it seems like the above should achieve the behaviour I'm looking for. Can someone tell me what's wrong with them?

Answer 1

Set up sample data:

create table temp_matches (
  rec varchar2(10)
);

insert into temp_matches values ('tac903.2');
insert into temp_matches values ('tac903.2a');
insert into temp_matches values ('tac903.2b');
insert into temp_matches values ('tac903.24');
insert into temp_matches values ('2');
insert into temp_matches values ('a');

Query:

select *
from temp_matches
where REGEXP_LIKE(rec, '(.)*[^0-9]$');

Results:

tac903.2a
tac903.2b
a

The $ indicates the end of the line, so we just look for any values not ending in 0-9.

Answer 2

You can use word boundary [[:>:]] here.

Something like this REGEXP_LIKE(col, 'tac903.2[az]?[[:>:]]');

See demo.

http://www.sqlfiddle.com/#!9/940fd/2/0

EDIT:

`REGEXP_LIKE(col, 'tac903.2[a-z]?(\s|$|\W)');`