On Wed, 2009-02-04 at 14:42 +0800, Don Marvick wrote:
> 4. any other issue needs consideration?
Most attempts to improve sorting further have fallen to nothing because
of the lack of repeatable test results. In particular, coming up with
test cases *after* writing the code is a good way to get lost in
discussions and have your work passed over.
The best starting point is to collect a number of test cases, both
typical and extreme cases. Do some research to show that "typical" cases
really are that and be prepared for some expressions of reasonable
doubt. Publish that data so test results can be objectively verified and
then measure those cases on existing code, with various settings of
tunable parameters.
Then it will be a simple matter to prove your changes are effective in
target cases without damaging other cases. We would also want the
changes to work automatically without additional tunable parameters.
-- Simon Riggs www.2ndQuadrant.comPostgreSQL Training, Services and Support