Jim C. Nasby wrote:
>On Wed, May 17, 2006 at 05:48:52PM -0400, Emi Lu wrote:
>
>
>>I think it is a very useful feature for postgresql to support it.
>>
>>If we have this feature supported, I do not have to recreate the table
>>and resetup all foreign key constraints, views, triggers, etc that are
>>based on the table.
>>
>>
>
>Uh, why do you have to do that? Please don't tell me you're using SELECT
>*...
>
>
>
No. It is not for select.
I have tens of tables with very clean structure. For example, username,
application_code, last_modified_by, etc in specific orders.
Since the business model is changed, I have to add some columns to
serveral tables.
I prefer columns orders following my other tables.
>In any case, there's extensive discussion about this in the -hackers
>archives. IIRC, there is consensus that this would be nice to have but
>no one has cared enough to actually make it happen. There are some
>non-trivial issues since this would mean either completely re-writing
>the table when you do an ALTER or you'd have to be able to divorce the
>catalog representation of a table with the on-disk representation.
>Though there are other advantages to doing the later, it's non-trivial.
>
>
If it does not support, I will recreate my tables.
Thanks.