🚀 go-pugleaf

In article <1de603INN5qv@early-bird.think.com>, barmar@think.com (Barry Margolin) writes:
|> [...]
|> And it's possible to have aliasing without pointer variables.  Fortran has
|> this problem because parameters are passed by reference -- their solution
|> was simply to legislate that programs are not permitted to pass array
|> parameters (or is it all parameters) in such a fashion that such aliases
|> would occur.  It's not possible to detect this statically in the general
|> case.

Quite true.  Fortran commits the same error as C, but in the reverse
direction.  What I want is for the language to allow aliasing, allow
the user to explicitly state whether things are *really* aliased or
not (with *NOT ALIASED* and the default).  The fact that the non-alias
legislation is not statically determinable is not relevant to me: I'm
perfectly willing to take that responsibility for my use of the code.

The point is that aliasing is an *inherent* part of what pointers do,
it is merely a rarely used possibility with arrays (one which I'm
perfectly willing to do completely without).  The distinction is 
therefore naturally divided - if you *call* something an array, the
compiler assumes it's not aliased to anything; if you *call* something
a pointer the compiler assumes that it might be aliased.  In short, 
part of the distinction you make between pointers and arrays is the
explicit control over aliasing I mentioned in the preceding paragraph.

Thus, if you coerce one to the other in every procedure call, you have
lost a useful distinction.  It's a bad idea to coerce arrays to pointers.

-- 
J. Giles