Re: Optimizing nested ConvertRowtypeExpr execution

>>>>> "Kyotaro" == Kyotaro HORIGUCHI <horiguchi(dot)kyotaro(at)lab(dot)ntt(dot)co(dot)jp> writes:

Kyotaro> My last comment was the while() loop does nothing. Ashutosh
Kyotaro> said that it is on the model of RelabelType.

Kyotaro> I examined the code for T_RelabelType again and it is
Kyotaro> intending that the ece_mutator can convert something
Kyotaro> (specifically only CollateExpr) to RelableType, then reduce
Kyotaro> the nested Relabels. So the order is not wrong in this
Kyotaro> perspective.

I'm eliminating the while loop in favour of an if, because I also think
it a good idea to ensure that the resulting convertformat is explicit if
any of the component conversions is explicit (rather than relying on the
top of the stack to be the most explicit one).

Kyotaro> So I don't object to the current patch, but it needs test like
Kyotaro> attached.

Thanks.

I'm going to pull all this together and commit it shortly.

--
Andrew (irc:RhodiumToad)