I realise that alot of work has gone into this so please don't take
this the wrong way: after reading for a bit it seems to me like a long
rant against C++. I think many of the criticisms are valid but are
phrased in an aggressive way. It makes it  appear the people
responsible for C++ are stupid or irresponsible, which IMO is not the
case.

Would you be ammenable to suggestions from this NG that would tone
some of the entries down a bit?

-Andrew Marlow

"Within C++, there is a much smaller and cleaner language struggling
to get out," which he says "would ... have been an unimportant cult
language."

.../and/ of course because of quotes like this (http://
www.research.att.com/~bs/bs_faq.html#C++success):

"the chance of success was affected by marketing clout, which I did
not have."

P.S. I'll try to answer the rest of the meaningful criticism in the
evening (I've quit all online discussions in non-moderated newsgroups
though). I'm publishing factual corrections at this page (which also
has my definition of "factual" in this context):

http://yosefk.com/c++fqa/web-vs-fqa.html

If you want to get a reply from me, e-mail is the most effective way:
http://yosefk.com

-- Yossi

But I'm not quite sure that I understand the point you're trying to make.

The FQA contains at least one FAQ (in the literal sense, rather than the
sense that FAQ is used for Answers to Frequently Asked Questions), in
FQA 23.6 http://yosefk.com/c++fqa/inheritance-mother.html#fqa-23.6
"How about trying another programming language?"

Could this be what you're driving at?  If so, what do you have in mind?

LR

http://yosefk.com/c++fqa/disclaimers.html

One particular thing that AFAIK applies almost always: low-level work
is better done in C than C++. I speak from lots of experience (but it
doesn't matter - what matters is whether my points in the FQA convince
you or not). By "low-level", I mean "when your resources are really
scarce" (compared to the "size" of the problem).

For example where I used POSIX mutexes it would have made for shorter
simpler C++ if I used RAII. Instead in C I had to be more careful and
tended to use goto's to jump to cleanup code on errors. This made the
code longer and of course, meant was using the dreaded goto. This lost
opportunity to use RAII is probably what David Abrahams was referring
to when he said his code was typically shorter and simpler in C++.

So I wonder why you would chose C. I did because of special project
needs, what are your reasons please? Why is the scarcity of resources
an issue here?

Regards,

Andrew Marlow

I could prefer C because it would compile in a snap (I never use
Boost.Python since it takes forever to compile - python.org cites this
close to the front page), and the verbose code (wrapping "tricky"
things like mutexes) would at least be very clear, and in C++ this
isn't really the case (I don't think I can convince you here and vice
versa, it's largely a matter of taste and experience of each person).

And I wouldn't introduce a third language, especially a tricky one
like C++ (which could throw exceptions that C couldn't deal with and/
or require initialization before main and finalization after main that
wouldn't necessarily work in the already multilingual environment and/
or there might be dynamic libraries involved and then C++ type_info
objects may end up being instantiated twice, etc.) I saw what happens
when you have Ada and C++ in the same process (didn't manage to figure
out why it crashed in _exit).

But then again, I don't know what I'd do in your case, since I don't
really know it. It doesn't seem to deal with "low-level work" by
itself though, it's a very special issue, "linguistic" at the very
core (mixing C and Java).

on Tue Nov 06 2007, apm35-AT-student.open.ac.uk wrote:

     * High-performance math: expression templates allow optimal
       implementations of complex matrix calculations while maintaining
       readable syntax.  The simplest example is summing an arbitrary
       set of matrices (known at compile-time) without creating any
       temporary matrices.

          Matrix sum = a + b + c ... ;

       The optimal code walks the summed matrices in parallel, placing
       the sums of each corresponding element into the result.

       The equivalent 'C' is bigger, uglier, and even if you manage to
       write a library routine that does it just as efficiently for
       this case, when you add one more matrix to your sum, you have to
       rewrite the library to maintain performance (so naturally you
       sacrifice performance).  The operations involved in a realistic
       case aren't limited to +, and even introducing '-' creates a
       combinatorial explosion of possible computations so there's no
       way to hardcode a C library that can do the computation
       optimally.

     * The Python interpreter is written in 'C' and uses
       reference-counting as part of its garbage-collection scheme.
       Not having the smart pointer abstraction available to manage the
       reference counting is a nightmare, and frequently leads to
       mistakes that can be hard to detect and debug.  Furthermore, all
       the manual reference-counting operations account for a great
       deal of syntactic overhead.  The 'C' code is bigger, uglier, and
       more error-prone than the equivalent C++, and no more efficient.

I could go on.  Abstraction without efficiency penalty is key to
writing robust, correct, efficient, sustainable software, and 'C'
doesn't provide nearly the capabilities that C++ does in that
department.

--
Dave Abrahams
Boost Consulting
http://www.boost-consulting.com

On Nov 5, 9:08 am, "Alf P. Steinbach" <al...@start.no> wrote:

In the short term I can understand the decision, but from a
longer term perspective I'm not convinced it was a good decision.
Now someone down the road is going to have a bigger pile to
clean up than before.  (I'm also impressed with the accomplish-
ment, but I agree with what I quoted above -- things can only
be organized so much with C.)  Biting the bullet seems the only
alternative.  What C++ compilers are developed with older
versions of C++ compilers?

Brian Wood
Ebenezer Enterprises
www.webebenezer.net

--
Dave Abrahams
Boost Consulting
http://www.boost-consulting.com

Brian Wood
Ebenezer Enterprises
www.webebenezer.net