Decibel! wrote:
On Apr 3, 2008, at 4:51 PM, Andrew Dunstan wrote:
Several years ago Bruce and I discussed the then theoretical use of a
SELECT query as the source for COPY TO, and we agreed that the sane
analog would be to have an INSERT query as the target of COPY FROM.
This idea seems to take that rather further. If doable I think it
would be cool, as long as people don't try using it as an alternative
storage engine. I can just imagine people creating views over such
SELECT statements ...
Why not? There's certainly cases where doing just that could be very
valuable. Storing older information that you're less likely to query
comes to mind... in those cases you're going to be seqscanning anyway,
so being able to read off a compact on-disk form is likely to be a win
performance-wise. It could certainly be a win storage-wise.
If someone wants to look at syntax options, I'm pretty certain that
Oracle supports this. IIRC you actually create what appears to the
database to be a real table, except for restrictions on what you can
actually do with it (for example, IIRC it's read-only).
You're serious aren't you?
Quite apart from any other reason why not, this would be a horrid hack
and is just the sort of "feature" we rightly eschew, IMNSHO. COPY is
designed as a bulk load/unload facility. It's fragile enough in that role.
If we really want to support an alternative storage engine then we
should tackle that front on and not via a back door like this.
cheers
andrew
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
