14.03.2012 14:17, Thomas Steinmaurer wrote:

>
> Sounds good. Again. Only for those kind of statements, where an
> execution plan could differ for already known parameter values.

Sure.

> The extra prepare time might be an issue, as we have seen the difference
> in a firebird-support discussion on a largish table with a lot of
> pointer pages, which are read at prepare time, AFAIK.

This [expensive] part of the preparation happens once per the every
involved table and it doesn't depend on the computational complexity of
the chosen plan. So the time cost we're speaking about here would be
quite hard to notice.


Dmitry