An Interview With C++ Creator Bjarne Stroustrup

DevTool writes "Bjarne Stroustrup talks about the imminent C++0x standard and the forthcoming features it brings, the difficulties of standardizing programming languages in general, the calculated risks that the standards committee can afford to take with new features, and even his own New Year's resolutions."

Comments

Is C++ ever the right tool for the job?

[ron_ivi]

It seems to me that most tasks that seem good for C++ would be better handled using a mix of an easier-to-program language (Ruby, Python, heck even lisp or smalltalk or anything else) with C extensions.

IMHO C++ seems not very good at very low-level programming; since with C++ it's not always obvious what a compiler might want to do with '+' thanks to operator overloading and rather convoluted implicit casting rules. In C you're using a pretty good tool for low-level programmings (especially with a dialect where you can sneak in a few assembly calls where you need to). In Ruby you're using a reasonably nice and efficient to develop in OO language. With the incredible ease of writing C extensions for Ruby, it's easy to use the best tool for each part of the job you're doing. The only compelling reason to use C++ I can think of is if some political policy forces you to use a single language for an entire project; and then I guess C++ not quite as clunky as java or c#.

( though I'm kinda repeating myself - a longer rant I made on slashdot about the pains of C++ years ago is here http://slashdot.org/comments.pl?sid=100202&cid=8540772 . An even more condemning annoyance about C++ is that thanks to so many convoluted tricks in the language, most people who claim C++ knowledge don't actually know it, as evidenced by the comments in that old thread )

Re:Is C++ ever the right tool for the job?

[EsbenMoseHansen]

I tried to use ruby for a while, but ended up back with C++. I couldn't live without static typing, as it turned out, and also C++ had more libraries leading to faster development (for me, at least).

C++ is the worst language, excepting every other language I have tried. And I am happy about the new features, they should make coding much nicer.

Re:Is C++ ever the right tool for the job?

[godrik]

I aggree that people tend to do too many things in C++ (me first). However, there are quite some place where C++ should be used. Mainly gaming engines and high performance computing. You NEED low level programming to deal with that. They could be written in C. But numerous C++ programming techniques (mainly template mechanism) makes it much easier to program. You could not get reusable algorithm accross multiple data type with efficient compile-time type checking at no runtime cost without templates. You could do it with macros but you will have terrible compilation problem if you screw up. Or you could use inheritance but will pay the cost at runtime.

Re:Is C++ ever the right tool for the job?

[mswhippingboy]

It seems to me that most tasks that seem good for C++ would be better handled using a mix of an easier-to-program language (Ruby, Python, heck even lisp or smalltalk or anything else) with C extensions.

IMHO C++ seems not very good at very low-level programming; since with C++ it's not always obvious what a compiler might want to do with '+' thanks to operator overloading and rather convoluted implicit casting rules. In C you're using a pretty good tool for low-level programmings (especially with a dialect where you can sneak in a few assembly calls where you need to). In Ruby you're using a reasonably nice and efficient to develop in OO language. With the incredible ease of writing C extensions for Ruby, it's easy to use the best tool for each part of the job you're doing. The only compelling reason to use C++ I can think of is if some political policy forces you to use a single language for an entire project; and then I guess C++ not quite as clunky as java or c#.

( though I'm kinda repeating myself - a longer rant I made on slashdot about the pains of C++ years ago is here http://slashdot.org/comments.pl?sid=100202&cid=8540772 . An even more condemning annoyance about C++ is that thanks to so many convoluted tricks in the language, most people who claim C++ knowledge don't actually know it, as evidenced by the comments in that old thread )

I disagree on several points. C is great for low level programming, but many times you want to use OO rather than procedural programming. C++ allows you to do exactly the same low level programming you can do in C (including inline-assembler if needed), but also offers the ability to design in OO. I can't think of a good reason not to use C++ rather than C unless it is a simple monolithic (preferably small) project. Mixing languages brings a whole other set of headaches (including staffing issues). Ruby, Python or other languages are fine and have their place, but I can't imagine using them for systems level programming anymore than I would use C++ to build a web application. Languages like Python, Ruby, (errr.. LISP? - there is considerable debate over whether LISP is even a true "programming language" but rather an alternate implementation of a Turing machine, but I digress) are orders of a magnitude too slow to be used as systems programming languages.

Re:Is C++ ever the right tool for the job?

[EvanED]

I used to be a huge C++ fan, though that has waned over the years as I've used better-designed languages and have also seen other people struggle with C++'s testier features.

That said -- I'd still take C++ over plain C for essentially anything in a heartbeat. (Basically the only exception would be stuff like microcontroller programming or other environments where there isn't really a good C++ compiler.) RAII alone is enough to seal that deal.

I don't buy most of the arguments of C over C++. For instance, take your statement that "IMHO C++ seems not very good at very low-level programming; since with C++ it's not always obvious what a compiler might want to do with '+' thanks to operator overloading and rather convoluted implicit casting rules."

But with modern compilers, I feel it's already very not obvious what the compiler is going to do with your code. What function calls are inlined? What loops are unrolled? For what I think is a reasonably dramatic example of this, take the following snippet:

int main(int argc, char** argv) {
    int y = 50;
    if (argc > 1) {
        y = 100;
    }
    return y;
}

and compile with optimization on. Look at the resulting assembly. There's no branch! (I'm assuming a variant of x86 or x64 here.)

If you're on a 64-bit system with GCC you'll probably see a cmov (conditional move) instruction, which kind of makes sense. But you don't even need that instruction to be available for it to omit an actual control-flow jump. Recompile with -m32 and look at the assembly again. Much longer, but still no branch. Instead, it uses one of the setxx instructions (setg in my case) and a bit of bit twiddling.

In my opinion, today's optimizers make the generated code so far removed from what you type into your editor that saying "+ can do anything!" is a drop in the bucket.

Re:Is C++ ever the right tool for the job?

[Spykk]

If you don't like operator overloading then don't use it. I don't see how that has any effect on C++'s usefulness as a low level language. C++ can use inline assembly the same way C can. C++ has all the power of C with the convenience of classes.

There is no mandate that you must use templates or operator overloading in C++. If you are going to complain that it gives you the option then why aren't you complaining about C having goto?

In my experience when someone wants to use C over C++ for a new project it is generally because they don't know C++.

Re:Is C++ ever the right tool for the job?

[derGoldstein]

You moved from portable, scalable, flexible C++ to an architecture-specific assembler? Ah, I see you were modded up. Good, Slashdot is working perfectly today.

Re:Is C++ ever the right tool for the job?

[EvanED]

So you have to be paranoid and check the class definition just in case.

So I have two responses to your general comment, though they are related.

You don't need to check the class definition at all -- you just need to know that one of the operands is an object. That's all. IMO, that's the only* difference between an overloaded operator call and a normal function call.

Once you're at that point, I don't see any difference between calling the function "+" and calling the function "doSomethingSimple". If you're paranoid enough about the class defining "+" in a way that will cause it to open sockets and thus you have to check the class definition to see if it does, you should be equally paranoid that "doSomethingSimple" will do the same thing and you have to check the class definition anyway.

The second part of my response is this: why do you CARE what "+" does internally? I can think of two reasons.

First, because you're concerned about efficiency. I've already said why I think this is a red herring, especially because I suspect most overloaded operator functions will be inlined to something efficient in general other way.

The second reason is because you want to know whether "+" is a good name for that function. But again I don't see how this differs from normal functions. If I call a function add and write a.add(b) instead of a + b, why are you more suspicious about the latter? I can come up with bad names without operator overloading plenty easily enough. Hell, I come up with bad function names for my non-overloaded functions far more often than bad function "names" for my operators. (And that even neglects the personal opinion that the normal convention of a.add(b) for a + b -- e.g. like what's used in Java's complex class -- feels wrong to me. a.add(b) feels way more like a += b than a + b to me. But again -- this isn't the fault of Java not supporting operator overloading per se, but rather the fault of Sun's function naming.)

* Okay, there is one other difference. Operator overloading is cool and tempting to use. It invites abuse way more than function naming does. But some high-profile cases aside (in particular, the proposal to overload the comma operator or whatever it was to load containers, a la Vector v = 4, 5, 6, 7, 8; or something), it certainly seems to me that, in practice, programmers are actually able to hold their tongues pretty well. Strangely enough, the biggest exception to this, I feel, is the bitshift overloading in iostreams. And IMO, that's easy enough to learn and brings about enough benefits (I/O type safety, non-repetition of type information, and extensibility vs a printf-style interface, and way less syntactic overhead vs repeated function calls) that it's probably the best solution even though it's not entirely satisfactory.

Re:Is C++ ever the right tool for the job?

[firewrought]

If you don't like $FEATURE then don't use it.

I always cringe when I see this statement in a discussion about programming languages. The presence or absence of a particular feature will impact the third-party libraries you use, the code samples and tutorials you come across, and the legacy apps you inherit. Even if you develop in a vacuum, some features can impact you by accident [operator overloading isn't a good example, but implict-conversion-of-numeric-types-to-boolean is].

In $FEATURE desirable or not for a particular programming language? Most always, that's going to be a complex usability/design question answerable by some combination of analysis, research, experience, testing, etc. Read the programming languages weblog or the discussion forums for D, Haskell, etc., to see the level of thought that goes into make these tradeoffs.

Interview

[Mongoose Disciple]

I always preferred this interview with Bjarne Stroustrup.

(Yes, I know it's not real, but...)

Re:Interview

[Prune]

I wrote C for about 5 years and then switched to C++ with which I've stuck for the last dozen years. With that perspective, the article you linked to is such a lousy attempt at humor that it made me cringe. C++ has, like many other large entities with significant history, become a bit messy due to the need for legacy support given its huge installed base. This does not mean that you cannot write neat programs with it! It's easy to blame the language instead of the software developer. It is just a matter of a bit of attention and discipline to do great stuff with C++ even while using what some may refer to its more esoteric features like template metaprogramming etc. I could never give up things like multiple inheritance, which may require a bit of care during usage, but I've found to be the way to go in many situations. The majority of the C++0x additions are very welcome as well. What's really needed is a completely modern reference text that minimizes time spent on legacy issues. It's basically the same situation with OpenGL, which has become much larger in the 3.x and 4.x versions, while still supporting all the legacy code--and there is no proper reference so one usually sees a messy mix of 2.x and 3/4.x in most current projects, rather than a clean, organized design following the most recent variant. A proper text (other than the standards document itself) would make a huge difference.

Re:Too little and too much, way too late

[JBMcB]

> Games developers, a few corporate app maintainers and...

Most Mozilla project applications including Firefox.. Pretty much all of KDE and some of GNOME. WebKit. Google Chrome. Opera. A good chunk of OpenOffice. Most Adobe applications. Most Microsoft desktop applications including Internet Explorer. CUPS. The Qt toolkit and pretty much everything that uses it. MySQL. Autodesk Maya. Winamp.

I wouldn't say a *few* corporate apps are written in C++, I'd say pretty much every major desktop app that's undergone a major re-write within the last two decades is probably C++.

Slashdotted, here is article text

[Anonymous Coward]

Interviewer: Well, it's been a few years since you changed the
world of software design, how does it feel, looking back?

Stroustrup: Actually, I was thinking about those days, just before
you arrived. Do you remember? Everyone was writing 'C'
and, the trouble was, they were pretty damn good at it.
Universities got pretty good at teaching it, too. They were
turning out competent - I stress the word 'competent' -
graduates at a phenomenal rate. That's what caused the
problem.

Interviewer: Problem?

Stroustrup: Yes, problem. Remember when everyone wrote Cobol?

Interviewer: Of course, I did too

Stroustrup: Well, in the beginning, these guys were like demi-gods.
Their salaries were high, and they were treated like
royalty.

Interviewer: Those were the days, eh?

Stroustrup: Right. So what happened? IBM got sick of it, and
invested millions in training programmers, till they were a
dime a dozen.

Interviewer: That's why I got out. Salaries dropped within a year,
to the point where being a journalist actually paid better.

Stroustrup: Exactly. Well, the same happened with 'C' programmers.

Interviewer: I see, but what's the point?

Stroustrup: Well, one day, when I was sitting in my office, I
thought of this little scheme, which would redress the
balance a little. I thought 'I wonder what would happen, if
there were a language so complicated, so difficult to learn,
that nobody would ever be able to swamp the market with
programmers? Actually, I got some of the ideas from X10,
you know, X windows. That was such a bitch of a graphics
system, that it only just ran on those Sun 3/60 things.
They had all the ingredients for what I wanted. A really
ridiculously complex syntax, obscure functions, and
pseudo-OO structure. Even now, nobody writes raw X-windows
code. Motif is the only way to go if you want to retain
your sanity.

Interviewer: You're kidding...?

Stroustrup: Not a bit of it. In fact, there was another problem.
Unix was written in 'C', which meant that any 'C' programmer
could very easily become a systems programmer. Remember
what a mainframe systems programmer used to earn?

Interviewer: You bet I do, that's what I used to do.

Stroustrup: OK, so this new language had to divorce itself from
Unix, by hiding all the system calls that bound the two
together so nicely. This would enable guys who only knew
about DOS to earn a decent living too.

Interviewer: I don't believe you said that...

Stroustrup: Well, it's been long enough, now, and I believe most
people have figured out for themselves that C++ is a waste
of time but, I must say, it's taken them a lot longer than I
thought it would.

Interviewer: So how exactly did you do it?

Stroustrup: It was only supposed to be a joke, I never thought
people would take the book seriously. Anyone with half a
brain can see that object-oriented programming is
counter-intuitive, illogical and inefficient.

Interviewer: What?

Stroustrup: And as for 're-useable code' - when did you ever hear
of a company re-using its code?

Interviewer: Well, never, actually, but...

Stroustrup: There you are then. Mind you, a few tried, in the
early days. There was this Oregon company - Mentor
Graphics, I think they were called - really caught a cold
trying to rewrite everything in C++ in about '90 or '91. I
felt sorry for them really, but I thought people would learn
from their mistakes.

Interviewer: Obviously, they didn't?

Stroustrup: Not in the slightest. Trouble is, most companies
hush-up all their major blunders, and explaining a $30
million loss to the shareholders would have been difficult.
Give them their due, though, they made it work in the end.

Interviewer: They did? Well, there you are then, it proves O-O
works.

Stroustrup: Well, almost. The executable was so huge, it took
five minutes to load, on an HP works

Re:Too little and too much, way too late

[mswhippingboy]

There's still a market for low level C programmers, and for Java, C#, Perl, Python, VB, and just about everything else, but who uses C++ these days? Games developers, a few corporate app maintainers and... uh... hang on...

Not sure if you're being facetious or not here, but C++ is consistently in the top 3 languages (the others being C and Java) in terms of popularity and usage. It's well above PHP, C#, Perl, Python, VB, Obj-C, and so on. I realize that stats from Tiobe index, Sourceforge, Langpop and other sources are merely an educated guess as to what people are really using, but the fact that nearly all agree on these three languages as being the top contenders seems to add credibility to their accuracy. It certainly beats the opinion of some individuals who are biased toward their favorite (or least favorite) languages.

kept alive by beardy wierdies and a few caffeine soaked kids who'll burn out and end up writing SQL for banks in a couple of years

Hey wait a minute! I resemble that remark!

C++0x compiled!

[hackingbear]

According to Wikipedia, "C++0x" is pronounced "see plus plus oh ex". After three rounds of macro preprocessing, four expansions of template substitutions, and reversing five levels of dynamic cast operator overloads, the name is eventually compiled to something readable: C plus plus? Oh, my ex-programming language!.

Re:Linking

[shutdown -p now]

It can't be standardized in the language spec itself, because there's no notion of "object files" or even "linking" in general in the spec - it's not tied to a particular implementation technique. That's why you can actually write a conformant C++ interpreter.

Now a separate standard for C++ ABI is quite possible, and they do, in fact, exist. The trick is getting the compilers to subscribe to them, and the problem is that they already have their existing ABI (usually incompatible with other vendors), and moving to standardized ABI would break compatibility with libraries compiled with older compiler versions. It's not a big deal in Unix world where source code is generally available these days, which is why Unix compilers (e.g. g++ and Intel) converged on a common one; but Windows stuff is still broadly incompatible.

Re:C++0when?

[c 0xFF]

My user name seems appropriate here.

Re:Multi-processor Extensions

[scorp1us]

So, I am an unapologetic Qt freak.

It does everything pretty damn easily and extends your C++ compiler by using a MetaObjectCompiler (MOC pre-processor) and gives you most, not all, of what C++0xForever is promising. Platform independently.

Of you don't like Qt, then there is Boost. Both are just C++ libraries.

Re:Multi-processor Extensions

[lgw]

The core Boost classes were written by guys on the C++0x committee - they were written with the intent of becoming part of the standard library eventually. It's nice to have commonly used libraries incorporated into the language standard, especially to standardize cross-platform interfaces to platform-dependent stuff like threads.

But some stuff like Lamba just gets better syntax sugar if you change the language instead of using a library. While I'm not all that happy with the new C++ lambda syntax (what's wrong with making lamba a keyword, guys? same for you C# guys?), is still way easier than the Boost lamba syntax.

And some stuff like move semantics are just an outright fix to the language. Vector will stop requiring that Foo be copyable - finally!