🚀 go-pugleaf

Hallvard B Furuseth <h.b.furuseth@usit.uio.no> wrote:
>In article <1995Aug13.144107.21353@dcs.warwick.ac.uk> A.Main@dcs.warwick.ac.uk (Zefram) writes:
>> atleast: certainly *is* needed.
>
>halasz and I had an e-mail conversation after that article:
>He is only talking about *syntax* here.  He is saying that the
>"atleast:" keyword is not necessary, a shorter syntax would be
>sufficient.  Which is of course true.

Ah.  If one will only ever want atleast semantics, this is true.  In
that case, I would suggest that something similar to the existing
bitfield notation could be used.  But exact semantics are sometimes
required, so for clarity it would be much better to have an exact:
specification in addition to atleast:, in which case the atleast
keyword (or some confusing puctuation, or an overload of an existing
keyword (return would be unambiguous :-))) is required.

I think exact semantics are used sufficiently that language support is
warranted.  At the moment, there is a great deal of code that assumes
that int (or sometimes long) is exactly 32 bits, and that short is
exactly 16 bits.  Much code relies on variables declared as unsigned
int having a range of 0 to 0xffffffff.  If no form of exact
specification is provided, people *will* write code that relies on int
:64 (or int atleast:64) having exactly 64 bits (which it will in most
implementations).  The SBEIR paper explains the need for the exact:
specification quite well.

-zefram