When I write code I don't only want to write code that is correct. I also want to write code that is understandable, and maintainable. I want to deliver code that is easy to read not only for the compiler but also for other human beings. After all, humans will read my code more frequently than compilers.
I have been thinking what are the single most important keywords that help us write readable code. Probably this question doesn't make much sense, but const
and using
are definitely among these. We already discussed const
a lot, this time it's time to see how using using
can improve our code.
We are going to review the 4 ways we can use it:
- type aliasing with
using
- introducing complete namespaces with
using
-directive - introducing members of another namespace with
using
-declaration - importing class members with
using
-declaration
Aliasing
In old C++ we could use typedef
to give another name, to give an alias for our types. Sometimes you might want to use it instead of strong typing, just to benefit from more meaningful names like int
.
typedef int Horsepower;
Other times you want to shorten long types for easier usage:
typedef std::vector<std::string>::iterator Iterator;
Since C++11 we can use using
instead of typedef
to achieve the same results.
using Horsepower = int;
using Iterator = std::vector<std::string>::iterator;
Why would you use using
over the good old typedef
? Just read the above statements! Exactly like the T.43 core guideline says, it's more readable! The keyword has a very clear meaining, then the name comes first and the old comes after a =
.
Besides, using
can be used more generally. It can be used for template aliases where typedef
would lead to a compilation error.
template<typename T>
typedef std::map<int, T> MapT; // error
template<typename T>
using MapT = std::map<int, T>; // OK
Using-directive in namespace and block scope
You've probably seen many code examples that right after the #include
statements contain the line using namespace std
.
You've probably seen lots of such application code.
You've probably been told that it's bad.
It's particularly bad if you do in at the global scope in a header file, just like [SF.7 from the Core Guidelines says]:
// bad.h
#include <iostream>
using namespace std; // bad
// user.cpp
#include "bad.h"
// some function that happens to be named copy
bool copy(/*... some parameters ...*/);
int main()
{
// now overloads local ::copy and std::copy, could be ambiguous
copy(/*...*/);
}
In my opinion, even the fact that as a reader you cannot be sure where a function comes from is bad. This is a simplistic example, but when you use using namespace
in a long .cpp
file it's hard to keep track of where certain objects come from. I prefer having using
-declarations instead and I also often introduce alias namespaces.
//some.h
#include <other.h>
using mcs = mynamespace::component::subcomponent;
msc::Class foo();
//some.cpp
msc::Class foo() {
using msc::AnotherClass;
AnotherClass bar;
// ...
}
As such, I don't pollute the global namespace. What you have to keep in mind is that when you introduce a using
-directive into a header file at the global namespace header, you don't just mess things up in the current scope.
If you include the header file in other files, you'll also bring the inclusion of all those introduced symbols. If you introduce different header files with different global levels using
-directives, the situation becomes even worse and the results of name lookup might depend on the order of inclusion.
To avoid all such problems, just follow SF.7 and don’t write using namespace at global scope in a header file.
Using-declaration in namespace and block scope
While the using
-directive brings all the symbols of a namespace into the current scope, a using
-declaration will bring only one item!
using std::string;
string foo{"bar"};
In the above example, we just demonstrated how it works. After using std::string
, we can refer to std::string
without mentioning the std
namespace.
It's still something not to overuse! A using
-declaration may also expand an overload set. It's less dangerous to use it at a file scope than having a using
-directive at the same scope, but risks still remain.
Starting from C++20, you can also introduce scoped enumerators into a namespace of block scope!
enum class Color { red, green, blue };
class MyClass {
using Color::red;
Color c = red; // This is OK from C++20
};
In fact, it would also work with the old-style unscoped enum
, but why would we do that?
Importing base class members with using
-declaration
With using
-declaration, you can introduce base class members - including constructors - into derived classes. It's an easy way of exposing protected
base class members as public
in the derived class. It can be used both for functions and variables.
#include <iostream>
class Base {
protected:
void foo() {
std::cout << "Base::foo()\n";
}
int m_i = 42;
};
class Derived : public Base {
public:
using Base::foo;
using Base::m_i;
};
int main() {
Derived d;
d.foo();
std::cout << d.m_i << '\n';
}
/*
Base::foo()
42
*/
If you try to modify the above example and remove any of the two using
-declarations, you'll see the compilation failing.
If the derived class already has a member with the same name, the compilation will not. The imported symbol from the base class will be hidden.
#include <iostream>
class Base {
protected:
void foo() {
std::cout << "Base::foo()\n";
}
};
class Derived : public Base {
public:
using Base::foo;
void foo() {
std::cout << "Derived::foo()\n";
}
};
int main() {
Derived d;
d.foo();
}
/*
Derived::foo()
*/
I find this technique really useful for unit testing. When you're writing a mock by hand, you often have to expose protected member functions from the base class, from the class that you are about to mock.
One way of doing it is forwarding the call.
Hopefully, the function's name in the mock is not changed, but I've seen it a couple of times. It really puts an extra burden on the maintainers when they realize that there is a better option.
class ClassUnderTest {
public:
virtual void testMe() {
}
virtual void testMeToo() {
}
};
class MockClassUnderTest : public ClassUnderTest {
public:
void testMe() override {
ClassUnderTest::testMe();
}
void mockedTestMeToo() {
ClassUnderTest::testMeToo();
}
};
Apart from tying a lot of unnecessary code, the problem above is that if the parameter list of testMe
or testMeToo
changes, you'll also have to update MockClassUnderTest
. You can get rid of that need by using using
.
class MockClassUnderTest : public ClassUnderTest {
public:
using ClassUnderTest::testMe;
using ClassUnderTest::testMeToo;
};
Now we have less code and it's more understandable what's happening. As a bonus, even the maintenance is simplified.
Conclusion
In this article, we discussed the 4 different ways that we can use the using
keyword. It's the right way to create aliases and import base class members in derived classes. At the same time, they can be also used to introduce whole namespaces into the current scope which can be particularly dangerous. Last but not least, using
can also introduce single types to the current scope which is a less dangerous option than introducing whole namespaces, still, it should be used with care.
Connect deeper
If you liked this article, please
- hit on the like button,
- subscribe to my newsletter
- and let's connect on Twitter!
Top comments (3)
Formatting has gone out of control in your message!
Oh, thanks. I added a backtick to fix it!
IMHO, the C++ committee should have allowed
template
withtypedef
also for completeness and symmetry.