This is borne out, IMHO, by the fact that even though everyone supposedly "knows" it's broken and how to fix it, it is by far still the most popular choice when you need very high performance with reasonable abstraction power in industry. It's been around for decades, and there's nothing worthwhile to replace it that fits in the particular domain it fits into.
I'm not advocating here -- I suspect what will happen is that as the low-level architecture of computers change, something may take it's place that maps better, but for now, that statement is no lie.
What do you find controversial about that statement?
You can certainly argue over which languages are most powerful, but there's really only one other general purpose language that rivals C++ in terms of performance and that's C. C++ being nearly a strict superset of C is almost by definition more powerful.
This whole thing implicitly assumes there's some acceptable "balance" you're making when you compare abstraction power and performance. Ignoring the fact that "choosing speed" or "choosing abstraction" is a bit of a false dichotomy, that's okay. But according to this page, C++ is the perfect harmony between these two areas. I'm calling bullshit on how arbitrary that is.
WTF are you on about? The page doesn't say any of that. They acknowledge that C++ is not perfect but give specific reasons why it's a good fit for this domain: performance, access to cutting edge tech (e.g. GPUs, iPhone, iPad) and access to native OS facilities. The only other language that fits that description is C, and you're damn right that "speed" and "abstraction" do not sum to zero: C++ is C with more abstractive power.
> C++ being nearly a strict superset of C is almost by definition more powerful.
Not necessarily. In the general case, restrictions can make a language more powerful [1], by allowing the user and the compiler to make more assumptions. E.g. removing gotos and mutations seem good ideas.
In the case of C vs C++, removing operator overloading and exceptions could turn out to make the language more powerful.
[1] If you use the right definition of `powerful'. You need a sensible definition that avoids "They are all Turing complete and thus equally powerful."
Removing operator overloading makes vector and matrix math deeply unpleasant - something that this kind of "creative coding" tends to use rather heavily.
I like how Haskell handles operator overloading: You need to conform to some kind of type class, _and_ you can make up your own operators instead of having to reuse bit-shifts for I/O.
Oh, I do not really have an opinion on operator overloading in C++. I just conceded the possibility that "removing operator overloading [...] could turn out to make the language more powerful [...]", because I have heard the opinion being advanced with some good arguments.
This is borne out, IMHO, by the fact that even though everyone supposedly "knows" it's broken and how to fix it, it is by far still the most popular choice when you need very high performance with reasonable abstraction power in industry. It's been around for decades, and there's nothing worthwhile to replace it that fits in the particular domain it fits into.
I'm not advocating here -- I suspect what will happen is that as the low-level architecture of computers change, something may take it's place that maps better, but for now, that statement is no lie.