On Sun, 20 Jun 2004, Tom Lane wrote:
> Gavin Sherry <swm@linuxworld.com.au> writes:
> > On Mon, 21 Jun 2004, Tatsuo Ishii wrote:
> >> Also I think we need to enhance ALTER INDEX to assign new table spaces
> >> for indexes. Assigning different tables spaces for tables and indexes
> >> are essential to gain more I/O speed IMO.
>
> > I thought about this. ALTER INDEX doesn't exist yet and I figured that,
> > unlike the case of tables, its easy to drop and recreate indexes in new
> > tablespaces.
>
> The precedents we already have (ALTER OWNER, RENAME, SET STATISTICS)
> are that ALTER TABLE applies to any relation type for which it makes
> sense. So I'd expect ALTER TABLE SET TABLESPACE to just work on
> indexes, not that we'd go and invent an ALTER INDEX ... command.
Yes, of course.
>
> Given that you implement the data transfer as a straight block-by-block
> copy and not some kind of tuple-at-a-time thing, I would think that
> it would be trivial to consider them the same case from an
> implementation point of view, too.
But I did implement it as a tuple at a time thing. I reused the code from
rebuild_relation()...
What did you have in mind?
Gavin
