separating files for classes and thier implemenation.

Hello,

I have a class that I have written. But have a problem with a compile error. Underfined reference to 'Mammal::GetAge() const'

I have put my class into a seperate file called Mammal.hpp.
The implementation is called Mammal.cpp.
I have another file for testing the application called main.cpp

I have included the Mammal.hpp header in my Mammal.cpp file.

I am wondering if this is the correct way to do this?

Many thanks,

Mammal.hpp - file

Code:

#include <iostream>

class Mammal
{
    public:
        Mammal():itsAge(10){ std::cout << "Mammal constructor" << std::endl; }
        virtual ~Mammal(){ std::cout << "Mammal distructor" << std::endl; }

        void Move() const {std::cout << "Mammal move one step" << std::endl;}
        void Move(int) const {std::cout << "Mammal move a specified number of steps" << std::endl; }
        void GetAge() const;
        void GetWeight() const;
        virtual void Speak() const { std::cout << "Mammal speak" << std::endl; }

    protected:
        int itsAge;
        int itsWeight;
};

void Mammal::GetWeight() const
{
    std::cout << "Get Weight: " << itsWeight << std::endl;
}

Mammal.cpp - file

Code:

#include "Mammal.hpp"

void Mammal::GetAge() const
{
    std::cout << "Age: " << itsAge << std::endl;
}

Main.cpp - file

Code:

#include "Mammal.hpp"

int main()
{
    std::cout << "Hello World" << std::endl;
    Mammal myMammal;
    myMammal.Move();
    myMammal.GetAge();

    return 0;
}

I had this problem not long ago, I think it was something silly but I can't recall what it was. First try inclusion guards on Mammal.hpp

Code:

#ifndef MAMMAL_HPP
#define MAMMAL_HPP
#include <iostream>

class Mammal
{
    public:
        Mammal():itsAge(10){ std::cout << "Mammal constructor" << std::endl; }
        virtual ~Mammal(){ std::cout << "Mammal distructor" << std::endl; }

        void Move() const {std::cout << "Mammal move one step" << std::endl;}
        void Move(int) const {std::cout << "Mammal move a specified number of steps" << std::endl; }
        void GetAge() const;
        void GetWeight() const;
        virtual void Speak() const { std::cout << "Mammal speak" << std::endl; }

    protected:
        int itsAge;
        int itsWeight;
};
#endif

// and put what's below nto Mammal.cpp

void Mammal::GetWeight() const
{
    std::cout << "Get Weight: " << itsWeight << std::endl;
}

You need to do (minimally)
g++ Main.cpp Mammal.cpp

Yes, that's the way to do it.

Hello,

Thanks for the reply. I am already using guards on the header files. I didn't show them as I wanted to shorten the code displayed.

I am using codeblocks GNC complier.

I have built all the cpp files, and still having a problem.

Any more ideas?

Thanks,

That is very strange (to me at least). Doesn't work here either (GCC 4.2.3, 64-bit Linux), but with a different error -

Code:

cyberfish@cyberfish-desktop:/tmp$ g++ Main.cpp Mammal.cpp
/tmp/cc7PY0gt.o: In function `Mammal::GetWeight() const':
Mammal.cpp:(.text+0x1ca): multiple definition of `Mammal::GetWeight() const'
/tmp/cc0e9ord.o:Main.cpp:(.text+0x204): first defined here
collect2: ld returned 1 exit status

I don't see how it's defined in Main.cpp.

*edit* works if I take out the definition (implementation) of getWeight in the header (or move it to the cpp, as A10 suggests). No idea why. */edit*

Thanks,

It worked.

I moved the GetWeight out of the hpp file and put it into the mammal.cpp file.

However, I am not sure why? I can only guess that you are not allowed to have any definitions in the header file.

Thanks,

I THINK it's because Mammal.hpp is included by both Main.cpp and Mammal.cpp. That is why GetWeight gets two definitions. Inclusion guard won't help you here because the files are compiled separately.

I don't think you can ever have definitions in a header, unless it is a template function.

How about inline functions? I thought they have to be defined in the header? (otherwise the compiler won't be able to inline them)

I believe inline functions are treated differently by the compiler, and also inlining is purely up to the compiler anyway. Just because you use the 'inline' keyword doesn't necessarily mean it will be inlined.
Please do not quote me on that, as that information may be inaccurate.(AFAIK basically)

True, but the compiler cannot inline function no matter what if the caller and callee are in different files (as in separately compiled).

That is why, for example, in chess engines, which are mostly CPU-bound, people try to squeeze out the last bit of performance by not using headers at all - and have the main file include all other cpp files. That way the compiler can optimize across the whole program (inlining functions for example).

but honestly, is the (fractional) performance gain of in-lining worth it on modern hardware.

well, for a typical program running on modern hardware, about 2,000,000 positions can be searched per second. For every position, there needs to be two calls to the "inCheck()" function, which is just a few lines (so the call overhead is comparatively big).

With my program, going from -O2 to -O3 gives me about 10% increase in performance. -O3 adds things like function inlining and loop unrolling.

They do everything to get higher performance... many of the best ones have the inner loops written in assembly.

Originally Posted by cyberfish

True, but the compiler cannot inline function no matter what if the caller and callee are in different files (as in separately compiled).

Untrue. The compiler, in cooperation with the linker, can do just that.

That is why, for example, in chess engines, which are mostly CPU-bound, people try to squeeze out the last bit of performance by not using headers at all - and have the main file include all other cpp files. That way the compiler can optimize across the whole program (inlining functions for example).

That is stupid.
You should use a proper compiler instead.
A proper compiler can perform these optimizations even if they are in separate files.
One such compiler that I know if Visual C++. Microsoft's compiler. There are probably others, too.

Untrue. The compiler, in cooperation with the linker, can do just that.

GCC can't (or am I missing something). In the UNIX world they are done by two different programs entirely (compiling and linking, by gcc and ld).

It would be truly amazing of the Microsoft compiler if it can do that.

But is that to say every time a cpp file is changed, the whole project needs to be recompiled? Since that's the only way cross-file inlining can be done?

If that is the case... then what's the difference between that and including cpp files? (and keeping dummy header files for human reference, or include all headers before all cpp's?)

I thought one of the main advantages of using headers is that the project can be incrementally compiled.