Re: [boost] 3-state boolean type

From: Andrew J Bromage (ajb_at_[hidden])
Date: 2002-08-03 01:08:04

G'day all.

On Fri, Aug 02, 2002 at 09:43:46AM -0400, Douglas Gregor wrote:

> Potentially, yes, but then one might expect "top" as well. At that point we 
> might as well just create the boolean lattice instead of a simple 3-state 
> boolean :)

And, in fact, that would be easier to store in two bits. :-)

"Top" isn't as useful in static analysis applications as "bottom".
"Bottom" means "too little information" (i.e. undecided) and "top"
means "too much information" (i.e. inconsistent).  It's rare that
you need the inconsistent state (though it's occasionally useful).

> I could see a use of a lattice concept (with, e.g., meet and join functions 
> defined) and perhaps an adaptor to make a BGL graph into a lattice. I'm not 
> sure how one would create a lattice type -- most useful lattices tend to be 
> infinitely wide, infinitely tall, or both.

That depends.  In static analysis of programs, most machine types
have a finite number of possible values, so even their power sets
are finite.

Cheers,
Andrew Bromage