> > This (simple) patch creates (if the -T flag is set) a transaction of every
> > insert block, so de import is mostly 10 times faster.
> ... kinda seems like a kluge ...
I haven't looked at the patch to see if, for example, the -T flag takes
a tuple count to determine the size of the transaction blocks. But istm
that this is actually a very useful feature, even though it should not
be required as a direct result of Ferdinand's problems with the copy
functionality.
Why not have a feature which can increase the speed of an insert-style
import? We're not demanding that everyone use it, and we have someone
who went to the trouble to implement it.
- Thomas