[mikeash]

For me, and perhaps others, this is because C++ enshrines idiotic counter-intuitive operator overloading in its standard library.

[tragomaskhalos]

Examples ?

[mikeash]

The use of << and >> for streams is the one that always comes to mind. So minor, and yet so awful.

[tragomaskhalos]

Well this is clearly a matter of taste; the visual implication of movement (http://images.google.com/images?q=diversion+sign+chevrons) is quite elegant IMHO, and the precedence seems natural also. But others' mileage obviously varies ...

[mikeash]

I disagree that the precedence is natural. Without looking it up, and without compiling it, what do the following (IMO quite reasonable in general) lines of code do?

    cout << x & y;
    cout << (x & mask) != 0;
    cout << boolVar ? "Yes" : "No";

As for the visual implication of movement, I have no objection whatsoever to using the << operator for stream operations in general. However, I strenuously object to doing this in a language which already defines them as bitshift operators. Repurposing an operator to do something completely different (and especially in this case, where a functional operator suddenly turns into one that's almost entirely about causing a side effect) is bad. Operator overloading can be useful and result in great, comprehensible code, but only if operators only mean the same thing everywhere.

[tragomaskhalos]

Point taken, but small beer IMO; I use parens in boolean expressions because I can never remember the precedence of && vs ||, so would naturally use them here too.

And I don't buy the fact that << is somehow off limits because it's used for bitshift; the domain of usage is usually disjoint, although one would have to squint to grok the outputting of a bitshifted value to a stream - but how frequent is that?

Yes the << notation isn't perfect, but it is much nicer than named functions (and certainly superior to < (which Stroustrup initially considered)!). Adding arbitrary operators to the language would have required mechanisms to define associativity and precedence, which would have added a weight of complexity disproportionate to the reward.

[mikeash]

I don't really see the big advantage to << over named functions here. Compare:

    cout << string << " and " << otherstring << endl;

    cout.write(string).write(" and ").write(otherstring).write(endl);

It's about as readable to me. To me, operator overloading is good when the meaning is already obvious because of shared context that everybody has. That's not the case when co-opting << for stream operators.