Marcin 'Qrczak' Kowalczyk writes:
> Arrays can be supposed to offer constant time access to elements by
> their indices. Haskell report does not guarantee such things formally,
> but in practice it should be true.
Would it really be helpful to guarantee it formally? I can't think
of *any* implementation which will guarantee constant time access *for
arbitrarily large addresses* - up to say 2 `exp` (2 `exp` 128).
Is there really a *formal* way of writing performance specifications
that is well enough developed to be useful, in practice, for reasoning
about the performance of programs? (I'm just suspicious of your
wanting formality. Useful performance information is usually
informal, it seems to me.)
Peter
