Thread: Still battling with Copy Control

I'm having trouble using push_back() on a vector of user defined types. Copy Control is a fascinating issue, but quiet difficult to master.

The class implements, among other things, a constructor taking a reference to a string, a copy constructor and two overloaded assignment operators.

Code:

aClass(const std::string& = "n/a");
aClass(const aClass&);
aClass& operator=(const aClass&);
aClass& operator=(const std::string&);

I then go on to create the vector in main()

Code:

vector<aClass> vec;
vec.insert(vec.end(), "Luigi"); //atempt with insert
vec.push_back("Anna"); //attempt with push_back

I was under the impression that push_back would operate by first creating a new element, then a temporary of aClass using the string constructor, and then use the copy constructor to copy this temporary to the new element.

How can I effectively do this, short of having to define a aClass object to pass as an argument to push_back()?

Hmm... some more important information...

The class only implements one data member. A string

The error states, on both push_backs, it cannot convert from const char to const aClass.

One possibility is to have a constructor that accepts a const char*
Another way would be to simply write:

Code:

vec.push_back(std::string("Anna"));

*shrugs*

Code:

class aClass {
    public:
        /* ... */
        operator std::string() { return myString; }
        /* ... */
};

I dont think that will work, Desolation. Mario F. wants to have a const char[5] or const char[6] convertible to aClass, not to have aClass convertible to std::string. From what I see, making const char* convertible to aClass by means of a constructor that accepts a const char* as an argument is one way.

hmm... perhaps more likely:

Code:

vec.push_back(aClass("Anna"));

Return value optimisation applies here, right?

It works. Thanks for the tips Citizen and Laserlight.

Converting the literal to a string was what I needed. I keep forgetting the difference between a string literal (const char) and a string. The error was telling me exactly that, but I wasn't looking. I did like the aClass() approach too. Probably easier to read since it doesn't trick the reader into thinking the vector contains strings.

But I did find one interesting thing... insert() is faster than push_back()

Actually I'm wrong. It isn't. It's slower. And also found the importance of reserve().

Every constructor, operator overload and the destructor has a cout inside letting me know when it is being called. This is a toy class I'm building as I learn about copy control.

On my code above I had insert() being called once and push_back being called next.

Code:

vector<aClass> vec;
vec.insert(vec.end(), aClass("Luigi"));
vec.push_back(aClass("Luigi"));

The output was...

*inserting an element in vector...
...inside string constructor...
...inside copy constructor...
...inside copy constructor...
...inside destructor...
...inside destructor...
*inserting another element in vector...
...inside string constructor...
...inside copy constructor...
...inside copy constructor...
...inside copy constructor...
...inside destructor...
...inside destructor...
...inside destructor...

It seemed insert() was faster. But then I swapped push_back() with insert() and now push-back was faster. It didn't make sense untill I remembered how vectors allocate memory. All those calls are happening because the vector is being moved in memory. Once I reserved some memory for it, it looked much cleaner. And also taught me insert() is slower.

*inserting an element in vector... (with insert)
...inside string constructor...
...inside copy constructor...
...inside copy constructor...
...inside destructor...
...inside destructor...
*inserting another element in vector... (with push_back)
...inside string constructor...
...inside copy constructor...
...inside destructor...