Recently, I was browsing through my开发者_StackOverflow社区 copy of the C++ Pocket Reference from O'Reilly Media, and I was surprised when I came across a brief section and example regarding user-defined conversion for user-defined types:
#include <iostream>
class account {
private:
double balance;
public:
account (double b) { balance = b; }
operator double (void) { return balance; }
};
int main (void) {
account acc(100.0);
double balance = acc;
std::cout << balance << std::endl;
return 0;
}
I've been programming in C++ for awhile, and this is the first time I've ever seen this sort of operator overloading. The book's description of this subject is somewhat brief, leaving me with a few unanswered questions about this feature:
- Is this a particularly obscure feature? As I said, I've been programming in C++ for awhile and this is the first time I've ever come across this. I haven't had much luck finding more in-depth material regarding this.
- Is this relatively portable? (I'm compiling on GCC 4.1)
Can user-defined conversions to user defined types be done? e.g.
operator std::string () { /* code */ }
Is this a particularly obscure feature?
Yes, conversion operators aren't used very often. The places I've seen them are for user-defined types that can degrade to built-in ones. Things like a fixed-precision number class that supports converting to/from atomic number types.
Is this relatively portable?
As far as I know, it is. They've been in the standard forever.
Can user-defined conversions to user defined types be done?
Yes, that's one of the features of constructors. A constructor that takes a single argument effectively creates a conversion operator from the argument type to your class's type. For example, a class like this:
class Foo {
public:
Foo(int n) {
// do stuff...
}
}
Let's you do:
Foo f = 123;
If you've used std::string
before, odds are you've used this feature without realizing it. (As an aside, if you want to prevent this behavior, declare any single-argument constructors using explicit
.)
It's not particularly obscure; it is very portable (it is part of the language after all), and conversion to user-defined types is possible.
One word of caution, having a lot of possible implicit conversion paths can lead to unexpected conversion being invoked and surprising bugs. Also, having non-explicit converting constructors and conversion functions between several user-defined types can lead to more ambigious conversion sequeunces which can be a pain to resolve.
It was about one of the first things I stumbled across when I was learning C++, so I'd say that no, it is not all that obscure.
One thing I would caution on though: Use the explicit
keyword with them unless you know exactly what you are doing. Implicit conversions can cause code to behave in unexpected ways, so you should avoid using them in most cases. Frankly, I'd be happier if the language didn't have them.
This is a particulary useful standard C++ feature and not a bit obscure :) You can use fundamental and user defined types alike for conversion operators.
精彩评论