Hi KaiGai-san,
Sorry about the delay in answering. I have been swamped with another thing
lately.
> BTW, your patch does not make sense in my environment that is just
> after initdb without any parameter customizing. Could you give us
> the step to reproduce the Nested-Loop plan from Hash-Join?
I examined both the Nested-Loop and Hash-Join plans by using the enable_xxxxx
options, and found that their total costs are almost equal. The result might
depend on the execution environment (My environment is Linux 2.6.18 on x86_64.).
I think it is preferable to use the enable_xxxxx options for this regression
test like src/test/regress/sql/join.sql.
Thanks,
Best regards,
Etsuro Fujita