How to convert string to numbers (if numeric), otherwise leave as string in SQL - sql

Question

my column has customer numbers that are both numeric and combinations of text and numbers. I only want to convert them if they are all numeric otherwise leave them as a string
ex.
91036
ab321
10001
Only convert 91036 and 10001 as numbers (using cast or convert) but leave ab321 as string. I tried isnumeric but this isn't a recognized function in the query builder I'm using. And if I convert using cast or convert the strings are disappearing.

Answer 1

This is an example based on Oracle, but I believe that the same principle would work elsewhere (at least, in databases that support regular expressions).
I don't quite understand what you meant by saying that you want to "convert" those values to numbers; convert where? I presume that they are currently stored in a VARCHAR2 (or some other "character" datatype column). There's no use in converting them to number an leave them in the same column, as that wouldn't do absolutely anything - they would still be treated as strings (stored in the character datatype column, remember?).
So, I've created another - NUMBER datatype column and will populate it with a number whenever possible.
Sample data:
SQL> create table test (col_char varchar2(10),
2 col_num number);
Table created.
SQL> insert into test (col_char)
2 (select '91036' from dual union
3 select 'ab321' from dual union
4 select '10001' from dual
5 );
3 rows created.
Update uses REGEXP_LIKE function which checks whether COL_CHAR looks like a number:
^ anchors you to the beginning of the string
\d+ takes as many consecutive digits as it can find
$ anchors you to the end of the string
which, all together, means that string must begin with a digit, must end with a digit, and everything in between have to be digits.
SQL> update test set
2 col_num = to_number(col_char)
3 where regexp_like(col_char, '^\d+$');
2 rows updated.
SQL> select * From test;
COL_CHAR COL_NUM
---------- ----------
10001 10001
91036 91036
ab321
SQL>
Things are getting somewhat more complex when there are decimals and/or thousands separators in a string, so you might need to alter both TO_NUMBER function and regular expression pattern. For example:
SQL> update test set
2 col_num = to_number(col_char, '999990D00000', 'nls_numeric_characters = '',.''')
3 where regexp_like(col_char, '^[0-9]+|(\,)[0-9]+$');
2 rows updated.
SQL> select * from test;
COL_CHAR COL_NUM
---------- ----------
10001 10001
91,036 91,036
ab321
SQL>