Re: DFA complement/intersection problem

"Paul J. Lucas" <pjl@mac.com> wrote in message


> If I have two languages:
>
> L = a*
> M = (a|b)*
>
> it's obvious that L <= M (L is a subset of M) because all
> sequences of A* are also sequences of (a|b)*. However, when I
> write code for this, I get L <= M and M <= L which is wrong.
> From:
>
> L <= M === intersect(L,~M) = 0
>
> I first have two minimal DFA for both languages:
>
> L: (s0) --{a}-> (s0)
> M: (t0) --{a|b}-> (t0)
....


> However, if you do this for M <= L, you get the same result
> because ~L has no accepting states so N' = intersect(M,~L)
> doesn't either; therefore N' is empty and M <= L. But this
> isn't right!
>
> What am I missing? Where is the mistake.


The "flipping accepting/nonaccepting states" method works only if L is
specified as a *complete* DFA (over the alphabet {a,b}), namely


L:(s0)--(a)->(s0)
      (s0)--(b)->(s1)
      (s1)--(a|b)->(s1)


where (s1) is non-final. Now it's obvious that ~L=a*b(a|b)* (i.e.
strings that contain at least one b).

Return to the comp.compilers page.
Search the comp.compilers archives again.