Re: more closed-world chatter
From: Marshall <marshall.spight_at_gmail.com>
Date: 7 May 2007 13:12:03 -0700
Message-ID: <1178568723.706223.284750_at_q75g2000hsh.googlegroups.com>
> TTM has it the disjunction, which is sound
> but overbroad.
Date: 7 May 2007 13:12:03 -0700
Message-ID: <1178568723.706223.284750_at_q75g2000hsh.googlegroups.com>
On May 7, 11:27 am, Marshall <marshall.spi..._at_gmail.com> wrote:
>
> TTM has it the disjunction, which is sound
> but overbroad.
Correction: TTM can express type constraints only as a single node of the type lattice, (which is what the vast majority of type systems do) so in the case of A \ B, the type of the single attribute must be the least upper bound of the type of the attribute in A and in B. Which means that to preserve the equivalence
type(A & B) = type(A \ (A \ B))
it must be the case that the type of attributes in A&B must be the least upper bound of the type of the attributes in A and in B, which is sound but not the most specific type it could be.
Marshall Received on Mon May 07 2007 - 22:12:03 CEST