On Tue, 2008-01-08 at 02:12 +0000, Gregory Stark wrote:
> I also don't understand how this proposal deals with the more common use case
> of unloading and loading data. Normally in partitioned tables we build the
> data in a side table until the data is all correct then load it as a
> partition. If you treat it as a lower-level object then I don't see that
> working. The layout of the new table won't often match the layout of the
> target partitioned table.
We optimised for that in 8.2, but I would say that not many people
noticed and that it isn't normal.
The problem with that approach, and the reason many people don't use it
is that it requires all data for a partition to be available at the time
you add the partition. That necessarily implies a time delay into the
process of loading data, which is no long acceptable in the world of
straight-through-processing or whatever you call the need for zero
processing delay in an/your industry. So people choose to load data
directly into the main table, allowing it to be immediately available,
though at the cost of some performance.
-- Simon Riggs 2ndQuadrant http://www.2ndQuadrant.com