How do C++ class members get initialized if I don't do it explicitly?

Question

In lieu of explicit initialization, initialization of members in classes works identically to initialization of local variables in functions.

For objects, their default constructor is called. For example, for std::string, the default constructor sets it to an empty string. If the object's class does not have a default constructor, it will be a compile error if you do not explicitly initialize it.

For primitive types (pointers, ints, etc), they are not initialized -- they contain whatever arbitrary junk happened to be at that memory location previously.

For references (e.g. std::string&), it is illegal not to initialize them, and your compiler will complain and refuse to compile such code. References must always be initialized.

So, in your specific case, if they are not explicitly initialized:

    int *ptr;  // Contains junk
    string name;  // Empty string
    string *pname;  // Contains junk
    string &rname;  // Compile error
    const string &crname;  // Compile error
    int age;  // Contains junk

First, let me explain what a mem-initializer-list is. A mem-initializer-list is a comma-separated list of mem-initializers, where each mem-initializer is a member name followed by (, followed by an expression-list, followed by a ). The expression-list is how the member is constructed. For example, in

static const char s_str[] = "bodacydo";
class Example
{
private:
    int *ptr;
    string name;
    string *pname;
    string &rname;
    const string &crname;
    int age;

public:
    Example()
        : name(s_str, s_str + 8), rname(name), crname(name), age(-4)
    {
    }
};

the mem-initializer-list of the user-supplied, no-arguments constructor is name(s_str, s_str + 8), rname(name), crname(name), age(-4). This mem-initializer-list means that the name member is initialized by the std::string constructor that takes two input iterators, the rname member is initialized with a reference to name, the crname member is initialized with a const-reference to name, and the age member is initialized with the value -4.

Each constructor has its own mem-initializer-list, and members can only be initialized in a prescribed order (basically the order in which the members are declared in the class). Thus, the members of Example can only be initialized in the order: ptr, name, pname, rname, crname, and age.

When you do not specify a mem-initializer of a member, the C++ standard says:

If the entity is a nonstatic data member ... of class type ..., the entity is default-initialized (8.5). ... Otherwise, the entity is not initialized.

Here, because name is a nonstatic data member of class type, it is default-initialized if no initializer for name was specified in the mem-initializer-list. All other members of Example do not have class type, so they are not initialized.

When the standard says that they are not initialized, this means that they can have any value. Thus, because the above code did not initialize pname, it could be anything.

Note that you still have to follow other rules, such as the rule that references must always be initialized. It is a compiler error to not initialize references.

You can also initialize data members at the point you declare them:

class another_example{
public:
    another_example();
    ~another_example();
private:
    int m_iInteger=10;
    double m_dDouble=10.765;
};

I use this form pretty much exclusively, although I have read some people consider it 'bad form', perhaps because it was only recently introduced - I think in C++11. To me it is more logical.

Another useful facet to the new rules is how to initialize data-members that are themselves classes. For instance suppose that CDynamicString is a class that encapsulates string handling. It has a constructor that allows you specify its initial value CDynamicString(wchat_t* pstrInitialString). You might very well use this class as a data member inside another class - say a class that encapsulates a windows registry value which in this case stores a postal address. To 'hard code' the registry key name to which this writes you use braces:

class Registry_Entry{
public:
    Registry_Entry();
    ~Registry_Entry();
    Commit();//Writes data to registry.
    Retrieve();//Reads data from registry;
private:
    CDynamicString m_cKeyName{L"Postal Address"};
    CDynamicString m_cAddress;
};

Note the second string class which holds the actual postal address does not have an initializer so its default constructor will be called on creation - perhaps automatically setting it to a blank string.

If you example class is instantiated on the stack, the contents of uninitialized scalar members is random and undefined.

For a global instance, uninitialized scalar members will be zeroed.

For members which are themselves instances of classes, their default constructors will be called, so your string object will get initialized.

• int *ptr; //uninitialized pointer (or zeroed if global)
• string name; //constructor called, initialized with empty string
• string *pname; //uninitialized pointer (or zeroed if global)
• string &rname; //compilation error if you fail to initialize this
• const string &crname; //compilation error if you fail to initialize this
• int age; //scalar value, uninitialized and random (or zeroed if global)

Uninitialized non-static members will contain random data. Actually, they will just have the value of the memory location they are assigned to.

Of course for object parameters (like string) the object's constructor could do a default initialization.

In your example:

int *ptr; // will point to a random memory location
string name; // empty string (due to string's default costructor)
string *pname; // will point to a random memory location
string &rname; // it would't compile
const string &crname; // it would't compile
int age; // random value

It depends on how the class is constructed

Answering this question comes understanding a huge switch case statement in the C++ language standard, and one which is hard for mere mortals to get intuition about.

As a simple example of how difficult thing are:

main.cpp

#include <cassert>

int main() {
    struct C { int i; };

    // This syntax is called "default initialization"
    C a;
    // i undefined

    // This syntax is called "value initialization"
    C b{};
    assert(b.i == 0);
}

In default initialization you would start from: https://en.cppreference.com/w/cpp/language/default_initialization we go to the part "The effects of default initialization are" and start the case statement:

• "if T is a non-POD": no (the definition of POD is in itself a huge switch statement)
• "if T is an array type": no
• "otherwise, nothing is done": therefore it is left with an undefined value

Then, if someone decides to value initialize we go to https://en.cppreference.com/w/cpp/language/value_initialization "The effects of value initialization are" and start the case statement:

• "if T is a class type with no default constructor or with a user-provided or deleted default constructor": not the case. You will now spend 20 minutes Googling those terms:
  • we have an implicitly defined default constructor (in particular because no other constructor was defined)
  • it is not user-provided (implicitly defined)
  • it is not deleted (= delete)
• "if T is a class type with a default constructor that is neither user-provided nor deleted": yes
  • "the object is zero-initialized and then it is default-initialized if it has a non-trivial default constructor": no non-trivial constructor, just zero-initialize. The definition of "zero-initialize" at least is simple and does what you expect: https://en.cppreference.com/w/cpp/language/zero_initialization

This is why I strongly recommend that you just never rely on "implicit" zero initialization. Unless there are strong performance reasons, explicitly initialize everything, either on the constructor if you defined one, or using aggregate initialization. Otherwise you make things very very risky for future developers.