Removing nonnumerical data out of a number + SQL

Question

I'm trying find the best way to remove nonnumerical data from a varchar in SQL e.g.

'(082) 000-0000' to '0820000000' or
'+2782 000 0000' to '0820000000'

The difficulty is i'm not always sure what number formats are coming in, as shown above, so I'd like like everything that is not a number removed essentially.

Update:
From what you guys have said this is a little spike done:

declare @Num varchar(20)

set @Num = ' + (82) 468 6152 '

--strip nonnumrical data out of @num

print @Num

set @Num = replace(@Num, ' ', '')
set @Num = replace(@Num, '+', '')
set @Num = replace(@Num, '-', '')
set @Num = replace(@Num, '(', '')
set @Num = replace(@Num, ')', '')

print @Num

Couldn't get the replace [^0-9] expression right though.

Answer 1

If you're using SQL Server 2005 or newer then your best option is to create a user-defined CLR function and use a regular expression to remove all non-numeric characters.

If you don't want to use a CLR function then you could create a standard user-defined function. This will do the job although it won't be as efficient:

CREATE FUNCTION dbo.RemoveNonNumerics(@in VARCHAR(255))
RETURNS VARCHAR(255)
AS
BEGIN
    DECLARE @out VARCHAR(255)

    IF (@in IS NOT NULL)
    BEGIN
        SET @out = ''

        WHILE (@in <> '')
        BEGIN
            IF (@in LIKE '[0-9]%')
                SET @out = @out + SUBSTRING(@in, 1, 1)

            SET @in = SUBSTRING(@in, 2, LEN(@in) - 1)
        END
    END

    RETURN(@out)
END

And then select from your table like so:

SELECT dbo.RemoveNonNumerics(your_column) AS your_tidy_column
FROM your_table

Answer 2

Have a look at this post (it's the 8th post down - the first LONG one) which details how to use regular expressions in SQL Server. It's not the fastest (that would be do it before you get to SQL) but it provides a decent way to do it.

Answer 3

There are several different options for this, depending on your DBMS.

For T-SQL, this link has some good approaches.

Answer 4

It is much easier to handle string parsing in your business layer. However, baring that use the T-SQL REPLACE() function (assuming MS SQL).

You could do a loop with that function on the parameter that was passed in to strip all non-numeric letters out of it.

Answer 5

What flavour of SQL backend are you using? If there's a regexp_replace kind of function, you could use that to replace [^0-9] with nothing.

Answer 6

The most effective and flexible that I have found is using the numbers/tally table method as shown in mwigdahl's answer on 10 Mar 2009

i.e. an allow list is much safer than the do-not-allow list that you have put in the bottom of your question above.

What you have not stated is how you handle non-integers ... what do you do with decimal points?