Re: Any research putting c above ada?

[dweller@news.imagin.net (David Weller)]

In article <3359e813.340466234@news.pacificnet.net>,
Kevin D. Quitt <Kevin@Quitt.net> wrote:
>ways.  I've done it FORTRAN, C, C++, and assembler (not necessarily
>combined).  It's not the language or the compiler that makes it safe, it's
>the programmer.
>
Kevin,
Nobody has any doubt that you've written safe, effective software in
FORTRAN, C, or C++, or even assembler.  

I think your implication that Ada developers lean entirely on the
compiler is specious.  A careful comparison of C++ to Ada 95 would
yield these observations:

	1) There is more compiler-generated checking in Ada 95 than in
	C++.  These checks are typically things that are commonly
	overlooked in software development, particularly in large
	systems.  These compile-time checks have no effect on the
	efficiency of the generated code.

	2) Ada95 has _significantly_ more run-time checking than C++
	has.  Any run-time check obviously has an impact on
	efficiency, but when you field a system in Ada, you simply
	turn those run-time checks off (if you so desire).

	3) Because of the increased amount of compile-time checking
	(as well as a few other factors), Ada95 compilers are much
	more capable of generating efficient code compared to C++.
	Before you (or anybody else) challenges such assertions, I
	will add that there has yet to be (to my knowledge) such a
	comparison made, but it can be discerned by analysis of the
	two languages.

	4) You will generally write no more, if not less, code in
	Ada95 than you would to accomplish the same task in C++. (This
	is heresy to many, but the simple truth.  It was also
	certainly the opposite, in some cases by an order of
	magnitude, when comparing Ada_83_ to C++).  My general
	observations is that your Ada95 code can be up to around 10%
	smaller than equivalent C++ code, give or take 10%.  I don't
	think having code 10% smaller really _proves_ much, but it
	dispels the perception Bjarne Stroustrup prepetuated in his
	book, "The Design and Evolution of C++" (an excellent book, by
	the way, should be part of any computer scientists library),
	where he indicated the Booch Components in Ada83 took 125K
	lines of code, and the C++ version took ~12K of code.  The Ada
	95 version of the Booch Components will take ~10K, and that's
	from a faithfully direct translation of the C++ version.
	Those numbers I have on very good authority, since the author
	of the Ada95 components is a close personal friend of mine :-)

	5) Speaking of tasking...nah, never mind, that's a separate
	subject.

Carefully note that NONE of my points above are saying: Individually,
if you use language X, you'll be a better programmer than in language
Y.  It is quite possible to do "unsafe" things in Ada95 and to
circumvent the _support_ that the compiler and run-time environment give
you, but those "circumventions" practically stand up and scream at
others that see the code.  I've learned, over years of programming in
both Ada95 and C++, that I find MUCH more errors in my code during
execution with my C++ code than I do my Ada code.  I'm not a shoddy
programmer in either.  It doesn't really take a rocket scientist to
understand that it's easier to find and fix problems that the compiler
catches than it is to become an expert on reading stack traces from
core dumps.

Nuf sed.  I'm sure I'll start a flame war now with the crossposting
going on.

-- 
Booch Components Homepage: www.rivatech.com ||Ada Homepage: www.adahome.com
------NOTICE-----NOTICE-----NOTICE-----NOTICE-----NOTICE-----NOTICE--------
My new e-mail address is dweller@imagin.net.  The dweller@dfw.net address will 
	   terminate _totally_ (no forwarding) at the end of April