🚀 go-pugleaf

In article <1992Dec24.172333.7339@grebyn.com> malak@grebyn.com (Michael Malak) writes:
> In my opinion, there is no fast track to either C or C++.  I feel the
> following must be learned in order:
>   1) Pascal, with two semesters experience

IMHO, Pascal has nothing useful that C doesn't have, except for array
bounds checking.  You you use C++ you can use a vector class that
gives you that.  I would suggest learning Scheme for a semester or two
and then moving to C++.

>   2) Assembly, learned anywhere, with tinkering on your own

I'm not sure it's so important to really program in assembly.  You
should understand basic machine architecture, and have some idea of
what an instruction set looks like, but writing real code, especially
for a RISC, doesn't teach you all that much.

>   3) C, 2-3 years min. anywhere (college, real world, or tinkering)
> After that you can learn and appreciate C++.  You need to first learn
> structured programming and algorithms, most importantly, linked lists
> and trees.  Since C is tightly coupled to assembly in many ways (C is a
> shorthand for assembly, and C integrates easily with assembly modules),

Two point here.  First, you can start with a subset of C++ that is
close to C, but includes important things such as mandatory function
prototypes and flexible declaration positioning, and add classes and
other OO features gradually.  You don't have to jump into multiple
inheritance the first day.  There are plenty of ways that C++ improved
on C that don't add much conceptual complexity.

Second, it's not right to call C "shorthand for assembly".  One thing
you can say about C and C++ is that it discourages language features
that can't be implemented at compile time, but that doesn't make it
any closer to assembly than any other compiled language.  And most
languages "integrate easily with assembly modules", simply because
that's what compilers produce.

> One reason that FORTRAN is associated with good number crunching
> capabilities is that there exist out there optimized, proven math
> libraries.

The other reason (I think) is that the semantics of fortran array
parameters make it easier for vectorizers and optimizers to handle
loops.  I don't remember the details, but this was behind the
"noalias" controversy with ANSI C a while back.

	Wayne