Bruce Momjian wrote:
>
> Sorry to be commenting so late.
>
> We could do that CLUSTER way of making a new heap file, but we rejected
> that for DROP COLUMN, so I am not sure why we would use that for ALTER
> COLUMN. Can anyone think of a good reason?
>
> Clearly if the new data type is binary compatible and it is just a
> catalog change, we can do that in place.
...with the caveat that any indexes that refer to the column must be
rebuilt, since the value interpretation could be quite different.
Guess that's obvious enough, though...
Kevin, a.k.a. "Mr. Obvious" :-)
--
Kevin Brown kevin@sysexperts.com
