Regex pattern inside SQL Replace function?

You can use PATINDEX to find the first index of the pattern (string's) occurrence. Then use STUFF to stuff another string into the pattern(string) matched.

Loop through each row. Replace each illegal characters with what you want. In your case replace non numeric with blank. The inner loop is if you have more than one illegal character in a current cell that of the loop.

DECLARE @counter int

SET @counter = 0

WHILE(@counter < (SELECT MAX(ID_COLUMN) FROM Table))
BEGIN  

    WHILE 1 = 1
    BEGIN
        DECLARE @RetVal varchar(50)

        SET @RetVal =  (SELECT Column = STUFF(Column, PATINDEX('%[^0-9.]%', Column),1, '')
        FROM Table
        WHERE ID_COLUMN = @counter)

        IF(@RetVal IS NOT NULL)       
          UPDATE Table SET
          Column = @RetVal
          WHERE ID_COLUMN = @counter
        ELSE
            break
    END

    SET @counter = @counter + 1
END

Caution: This is slow though! Having a varchar column may impact. So using LTRIM RTRIM may help a bit. Regardless, it is slow.

Credit goes to this StackOverFlow answer.

EDIT Credit also goes to @srutzky

Edit (by @Tmdean) Instead of doing one row at a time, this answer can be adapted to a more set-based solution. It still iterates the max of the number of non-numeric characters in a single row, so it's not ideal, but I think it should be acceptable in most situations.

WHILE 1 = 1 BEGIN
    WITH q AS
        (SELECT ID_Column, PATINDEX('%[^0-9.]%', Column) AS n
        FROM Table)
    UPDATE Table
    SET Column = STUFF(Column, q.n, 1, '')
    FROM q
    WHERE Table.ID_Column = q.ID_Column AND q.n != 0;

    IF @@ROWCOUNT = 0 BREAK;
END;

You can also improve efficiency quite a lot if you maintain a bit column in the table that indicates whether the field has been scrubbed yet. (NULL represents "Unknown" in my example and should be the column default.)

DECLARE @done bit = 0;
WHILE @done = 0 BEGIN
    WITH q AS
        (SELECT ID_Column, PATINDEX('%[^0-9.]%', Column) AS n
        FROM Table
        WHERE COALESCE(Scrubbed_Column, 0) = 0)
    UPDATE Table
    SET Column = STUFF(Column, q.n, 1, ''),
        Scrubbed_Column = 0
    FROM q
    WHERE Table.ID_Column = q.ID_Column AND q.n != 0;

    IF @@ROWCOUNT = 0 SET @done = 1;

    -- if Scrubbed_Column is still NULL, then the PATINDEX
    -- must have given 0
    UPDATE table
    SET Scrubbed_Column = CASE
        WHEN Scrubbed_Column IS NULL THEN 1
        ELSE NULLIF(Scrubbed_Column, 0)
    END;
END;

If you don't want to change your schema, this is easy to adapt to store intermediate results in a table valued variable which gets applied to the actual table at the end.

Instead of stripping out the found character by its sole position, using Replace(Column, BadFoundCharacter, '') could be substantially faster. Additionally, instead of just replacing the one bad character found next in each column, this replaces all those found.

WHILE 1 = 1 BEGIN
    UPDATE dbo.YourTable
    SET Column = Replace(Column, Substring(Column, PatIndex('%[^0-9.-]%', Column), 1), '')
    WHERE Column LIKE '%[^0-9.-]%'
    If @@RowCount = 0 BREAK;
END;

I am convinced this will work better than the accepted answer, if only because it does fewer operations. There are other ways that might also be faster, but I don't have time to explore those right now.