Re: Types and "join compatibility"

"Marshall Spight" <marshall.spight_at_gmail.com> wrote in message news:1123182042.699165.231320_at_z14g2000cwz.googlegroups.com...
> vc wrote:
>> Marshall Spight wrote:
>> ...
>>
>> >
>> > Alternatively, if we add two ints, we could come up with an
>> > equivalent expression that involves complex numbers.
>>
>> Interesting... Could you, like, give an example of two integer
>> addition resulting in a complex number ?
>>
>> > So the
>> > type of integer addition is complex!
>>
>> Integer addition type is (+):int*int->int unless you are using your
>> private vocabulary for "integer addition".
>
> I think you need to read the whole thread to understand the
> context. My statement that (+):int, int -> complex was
> the end point of a reducto ad absurdum argument; it is not
> a statement I support. Specifically, I was disproving the
> TTM claim that the type of an expression should be
> the same type as any equivalent expression. Substitutability
> means that we can always construct an equivalent expression
> that "passes through" an intermediate state involving a
> supertype. So if we buy into that line of reasoning, the
> result type of any function could be considered Top. (Or
> Alpha, as TTM would have it.) That's the "absurdum" part.
>
>

OK. If you are interested in possible approaches to relation type systems, you may find this useful:

Polymorphism and Type Inference in Database Programming
PETER BUNEMAN
ATSUSHI OHORI
Kyoto University

Type Inference in the Polymorphic Relational Algebra Jan Van den Bussche, Emmanuel Waller

> Marshall
>
Received on Fri Aug 05 2005 - 02:44:03 CEST