Thread: C++: (ab)using?

Originally Posted by EVOEx

Myself, I don't use "using" in any case (with the exception of programming contests, where speed is more important than good code). I find that it entirely kills the entire use of namespaces. I barely use "using", and yet the few times I did use it I ran into trouble, the only one I can remember now was "int map[25][25];". And somehow I find it tells the reader of the code a lot more on the object to specify, each time, where this object is located.

In C++ Coding Standards, Sutter and Alexandrescu argue the opposite:
"But the vast majority of the time there is no ambiguity: And that is why namespace using declarations and directives are what make namespaces usable, because they greatly reduce code clutter by freeing you from having to tediously qualify every name every time (which would be onerous, and frankly people just won't put up with it) and still letting you qualify names only in those rare cases when you need to resolve an actual ambiguity."

Personally, I would rather only use using declarations and using directives when in some restricted scope as I find that it is easier to spot those "rare cases" that way.

Originally Posted by EVOEx

Honestly, I'm quite surprised how many people use using, and the fact that that keyword even exists.

There are uses for using declarations that are beyond any argument of convenience, e.g., it is used to bring in names from a base class that are otherwise hidden by an overload in a derived class.

I usually use using namespace std; since I know better than to use variable names that are identical to STL names. I very rarely run into a conflict; but when I do, I just explicitly qualify the name with std::

But of course when I write a header file, I never put a using declaration there because it's just an evil thing to do to other people that use your header file...

I think there are cases when you'd like to use "using namespace X", and other times when using "X::".

As to the cases where it causes conflicts, "int map[...]" is not causing a big problem - either rename the variable [you are compiling often, aren't you? So changing the variable name would be little work] - or prefix it to the global namespace: "int ::map[...]" whenever it is used.

The worst kind of problem is when there is a silent "failure":

Code:

int cin;
... 
int x = 0;

cin >> x;

This would compile and link fine, and even run without any problems at all - it just wouldn't do what you expected...

--
Mats

>should one use the keyword "using", or should one preface
>a namespace every time something in it is referenced?
That's a matter of personal preference. I tend to prefer the prefix except when my lines start to get longer than I'm comfortable with, and then I restrict the scope of the using directive as much as possible.

>do you use "using namespace" or just the referenced item, eg. "using std::cout"?
I've never needed the added control of the using declaration. In my opinion, it just adds clutter compared to intelligent use of the using directive.