Comments (6)

1. Patrick Wolf wrote:

   Solution 3:

   Reconsider your data model if you are really going to compare two columns which obviously have the same content and meaning but are of different data type…

   
   Patrick

   December 10th, 2007, at 12:11 pm #
2. Eddie Awad wrote:

   @Patrick: Sure, if you can change the data model, that will be a very good solution. But, sometimes reality hurts and you’re stuck with a model that you have no control over. I sometimes do find myself in such a situation, unfortunately, especially when I deal with FlexFields (varchar2 columns) in Oracle EBS.

   December 10th, 2007, at 12:21 pm #
3. Patrick Wolf wrote:

   Yeah, reality sometimes is so far way from the “ideal” world

   December 10th, 2007, at 12:25 pm #
4. Justin Cave wrote:

   When trying to filter out rows, be aware that there is no guarantee about what order Oracle executes the various conditions. So you might well find that the invalid number error returns down the road when the query plan changes.

   Even if you nest the subquery, i.e.

   SELECT t1.col1, t2.col2
     FROM (SELECT t.col1
             FROM table1 t
            WHERE filter_out_non_numbers) t1,
          table2 t2
    WHERE t1.col1 = t2.col1
   

   there is no guarantee that the inner subquery is logically materialized before the join (and the implicit conversion) happens.

   December 10th, 2007, at 6:59 pm #
5. Eddie Awad wrote:

   @Justin: Good points.

   One more thing, if the to_char is used in the WHERE clause, that will break the use of indexes which may impact performance. So, a function based index may be a good option here.

   December 10th, 2007, at 9:29 pm #
6. Pythian Group Blog » Blog Archive » Log Buffer #75: a Carnival of the Vanities for DBAs wrote:

   [...] also has the second episode of his Oracle in 3 Minutes webcast, called A Ticking Bomb, or A Sleeper [...]

   December 14th, 2007, at 10:00 am #