Tom Lane wrote:
>I believe the proposed patch breaks many nontrivial cases (though it's
>quite hard to be sure exactly what it does, given a non-contextual diff
>:-().
>
Hi Tom,
it was a cvs diff, should I upload the whole source?
> Have you tried running its output back into the system to see
>whether the querytree is reconstructed exactly? Don't forget to try
>cases where parentheses were used in the original source to force a
>non-default evaluation order.
>
>
I'm quite sure about JOINS, T_BoolExpr and T_OpExpr. I had a lot of view
definitions fed in and out, to see if things work as expected.
I wasn't able to force a T_CoerceToDomain, and only got T_RelabelType
with simple variables. If you can tell me how to force complex args for
these cases, I'll check this. As far as I understand, both types always
have only one arg. If this is wrong, there definitely has to be an
additional check for a simple variable arg.
Regards,
Andreas