Re: [Toss-devel] F-First Normal Form

[Lukasz S. <luk...@gm...>]

On Sun, Jul 11, 2010 at 12:19 AM, Lukasz Kaiser <luk...@gm...> wrote:
>
> This looks like some nice heuristic, but you should also
> take care - in general there is a non-elementary lower
> bound for TNF, so at some points you need to copy and
> copy a lot. Moreover, I think you need conversion to DNF
> at least at some point. But optimization is always good :).
>

It is a trade-off I guess, we "don't have to" have TNF, deviation from
TNF makes things worse but so does copying. I am distributing, but
one-multiplication-at-a-time when other rules don't apply and there's
an unprotected literal inside. I'm probably reinventing the wheel. But
we can do things like doing the multiplication optionally etc. Perhaps
I'll "give up" before implementing true TNF if I'm satisfied with
results (especially if I realize I missed something).