Re: New Book: The School of Niklaus Wirth

Jerry Leichter <jerrold.leichter@smarts.com> wrote:
>On the other hand, Modula-3 does not have an unsigned type - and I think
>with good reason.
>
>There are two reasons for wanting an unsigned type:
> (1) ...
> (2) To implement "modulo arithmetic". Unsigned ints typically
> (and, by standard in C/C++) the mathematical properties
> of arithmetic mod 2^k. There are times when this is
> useful.


>...you have to ask
>the question: Do we really need a *type* for this - or do we just want
>the operations? The latter is the approach Modula-3 took:


This choice bit me badly, multiple times. The problem was that
although I could write Word.T and Word.Plus all I wanted, I repeatedly
made the mistake of ``accidentally'' applying signed-integer
operations to 32-bit addresses represented as Word.T. I would have
been *very* grateful for a type system that distinguished the two
types, so as to catch these mistakes at compile time. Especially as a
lot of the 32-bit addresses got used in unsafe code :-)


N.B. I am *not* arguing for overloading +. I am happy to write
Word.Plus. What I really want is a type that won't allow me to use
signed operations by mistake.


>The *big* advantage to this approach is that it eliminates all kinds
>of issues about how signed and unsigned integers should work when they
>"meet across an operator".


Yes, this is quite useful---but it would have been even better if the
M3 designers had distinguished the two *types*, and not only the
*operations*.
--
Norman Ramsey
http://www.eecs.harvard.edu/~nr

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