Re: Equivalent keys/elements in Ada.Containers Maps and Sets

["Adam Beneschan" <adam@irvine.com>]

On Jan 24, 2:50 pm, "Matthew Heaney" <mhea...@on2.com> wrote:

> This is a silly argument.  The Ada container library is modeled on the
> C++ STL, and in that library the requirement for "strict weak ordering"
> for the binary predicate for ordered associative containers is clear.
>
> Whatever putative difficulties Ada programmers have defining a
> less-than operator for instantiating a generic container, that is no
> different from the same problems C++ programmers would have defining a
> binary predicate for instantiating a container template.  Yet somehow,
> some way, thousands of C++ programmers have managed to instantiate
> their container templates without any problems.  I hope you're not
> suggesting that C++ programmers are smarter than Ada programmers!

No, of course not.  But I guess I am suggesting that the Ada
programming community expects the RM to be correct.  The idea that we
should let the RM say something is correct when it isn't, because
nobody will ever do it wrong anyway, isn't the Ada way.  That sounds
more like the C way.  The same "C way" that says "We don't need to
check our input for sanity because nobody will enter insane input
anyway."  "We don't need to check for buffer overflow because no data
is going to be that large anyway."  Or maybe that's the Microsoft
way---I don't know.  But I think our standards are higher than that.

> Ambiguities in the Ada RM happen all the time, that's why we have an
> ARG.   There's nothing special about (assumed) ambiguity in the
> requirement for strict weak ordering.

But there's no ambiguity in the RM.  The RM's requirements for the "<"
function are unambiguous.  Wrong, but unambiguous.

                     -- Adam