Re: Why use C++?

[Fritz Wuehler <fritz@spamexpire-201108.rodent.frell.theremailer.net>]

"Jed" <jehdiah@orbitway.net> wrote:

> I didn't say it was. I said it probably arose from that. I'll bet integer 
> literals are signed in most assembly languages (going out on a limb with 
> this, because I really don't know).

Absolutely not. Assembly language is about talking to the machine directly.
If you can't manipulate native types then your assembler isn't an assembler.

> Hence, a no-brainer translation of  HLL code to assembly (recognizing, of
> course, that compilers are free to generate machine code directly, rather
> than generating assembly).

Most (vast majority) of the compilers on IBM platforms generate object
code. It sounds like you think a goal of HLL design is to be easy to
implement. That's wrong. The goal of HLL design is to make it easy to solve
problems in the problem domain.

> It's generally considered "typeless" from the POV of HLL programmers. It 
> has relative meaning. No need to be pedantic about it.

It wasn't pedantic, it was a simple comment from an assembly coder. Assembly
is typed or it wouldn't be useful. However there's no type enforcement at
the assembler level (usually, although at the hardware level there is.)
That is no the same as not being typed, although to an HLL programmer it may
not be obvious that those two things are not one and the same.

> > especially with certain
> > systems and certain assemblers. The type in assembly language *does*
> > usually reflect the native types of the underlying machine very 
> > closely,
> > obviously.
> 
> And I'll bet, more often than not, C/C++ built-in types reflect that 
> also. It would be "silly" to specify a language to what is uncommon.

You have the tail wagging the dog. The C/C++ built-in types don't reflect
anything but the language design and they don't change their meaning from
platform to platform or from implementation to implementation. That's why
they're types. Incidentally they often map to native types on most
architectures which is why they were created in the first place. But now
that they have been created the people using C/C++ have to live with those
types or build abstract data types on top of them.

> It would be "silly" to specify a language to what is uncommon.

That's exactly the point of HLL. It must provide useful abstractions for
the problem domain. The underlying implementation should not matter at all.