The automatic versions of the constructor and destructor don't do anything, but they will be there if you do not build them. (The constructor you get is the "default constructor" -- no parameters -- and this is automatically built if you don't define any constructor besides the copy constructor). The automatic versions of the Copy Constructor and the Assignment operator overload are similar to each other, and their default versions are always built in a standard way.
Remember that anything that is passed by value in and out of functions causes copies to be made!
Here's an example of item #1 in the above list:
Fraction f1, f2(3,4); // declaration of two fractions Fraction f3 = f2; // declaration of f3 being initialized as a copy of f2
Note: This last line of code calls the copy constructor, since the initialization is on the same line as the declaration of f3. Contrast this with the following:
f1 = f2; // this uses the assignment operator, since f1 and f2 already exist
Format: className(const className &);
The const is not required, but it is usally a good idea, because we only want to make a copy -- we don't want to change the original. Here are some examples of copy constructor declarations for classes we have seen:
Fraction(const Fraction & f); Timer(const Timer & t); Directory(const Directory & d); Store(const Store & s);
Shallow Copy vs. Deep Copy:
The default version of the copy constructor (created by the compiler) makes what is known as a shallow copy. This simply means that the object is copied exactly as is -- each member data value is copied exactly over to the corresponding member data location in the new object. This is sufficient for many cases, but not for ALL cases.
Example:
Fraction f1(3,4);
This fraction object has a numerator of 3 and a denominator of 4. If this object is passed into a function by value, a copy will be made, and the new object's numerator will be 3, denominator 4. In this case, the shallow copy is sufficient.
Consider, however, the Directory class of the phone book example. The member data variables were currentsize and maxsize (both of type int), and a pointer, entryList (of type Entry * ), which pointed to dynamically allocated data outside the actual object. This is the situation in which a shallow copy is not sufficient. For instance, if the original object is storing the address 1024 in entryList, the copy will also get the 1024, and therefore the copy will be pointing to the original dynamic data! This will especially pose problems if, when the copy goes out of scope, it cleans up the dynamic data along with it.
When there is a pointer (inside an object) that points to dynamic data, the shallow copy is not sufficient, because it does not copy the dynamic data, only the pointer. A deep copy is needed. Here is what we might write for a copy constructor definition in the Directory class (from the phonebook database example):
Directory::Directory(const Directory & d) // copies object 'd' into the new object being created (this one) { // copy the static variables normally maxsize = d.maxsize; currentsize = d.currentsize; // create a new dynamic array for the // new object's pointer entryList = new Entry[d.maxsize]; // copy the dynamic data for (int i = 0; i < currentsize; i++) entryList[i] = d.entryList[i]; }
Fraction f1, f2; f1 = f2; // this call invokes the assignment operator
Like the copy constructor, the assignment operator has to make a copy of an object. The default version makes a shallow copy. If a deep copy is desired for assignments on a user-defined type (e.g. a class), then the assignment operator should be overloaded for the class.
The task done by the assignment operator is very similar to that of a copy constructor, but there are a couple of differences.
1) The copy constructor is initializing a brand new object as a copy of an existing one. The new object's data is being initialized for the first time. An assignment operator sets an existing object's state to that of another existing object. In situations with dynamic allocation, this may mean that old dynamic space must be cleaned up first before the copy is made.
2) Also, an assignment operator also returns the calling object (the copy constructor has no return). Consider the case of integers. In the statement:
a = b = c = 4;
The first operation is (c = 4), and this operation returns c (which now has value 4), so that the result can be used as an operand in the next assignment (b = 4). The calling object should be returned by reference when overloading = for objects. To return the object, we need to be able to refer to an object from inside the object itself.
From inside any member function, an object has access to its own address through a pointer called this, which is a keyword in C++. In an assignment operator, you must return the object itself (by reference), so you can return the target of the this pointer (which would be *this)
Like the copy constructor, the original object needs to be passed in, so there will be one parameter (of the same type as the object itself). The parameter is the same as in the copy constructor.
Declaration examples for a few classes:
Directory& operator= (const Directory &); Fraction& operator=(const Fraction &); Timer& operator=(const Timer &); Circle& operator=(const Circle &);
Here's an example of the assignment operator definition for the Directory class (in the phonebook example).
Definition in the implementation file:
Directory& Directory::operator=(const Directory & d); // copies object 'd' into the new object being created (this one) { if (this != &d) // only copy if the object passed in is not already this one { // since this is not a brand new object, we // should delete any information currently attached delete [] entryList; // similar to the copy constructor definition maxsize = d.maxsize; currentsize = d.currentsize; entryList = new Entry[d.maxsize]; for (int i = 0; i < currentsize; i++) entryList[i] = d.entryList[i]; } return *this; // return the object itself (by reference) }
Here is a link to the Phonebook example with the copy constructor and assignment operator added in.
Here is a link to what the Fraction class would look like with these functions added, although it should be noted that they are not needed in the Fraction class, because the automatic versions are sufficient. The Fraction member data does not involve pointers or dynamic allocation -- it only consists of static data, so the shallow copy is enough.