Thread: Initialization of primitve types

What are primitive (member) types initialized as in C++?

My integral members seem consistently to initialize to zero, float/double members often initialize the same, but can occasionally end-up with garbage, causing many headaches.

Some sources mention primitives are initialized to 0, yet elsewhere I've heard that the entire class is first memset to 0 (definitely not the same thing...)

So can anyone tell me the definitive list of initialization values for each primitive type?

Thanks in advance.

They are initialised to whatever you initialise them

Hence, initialise them in the initialisation list to avoid having them contain garbage. If you wanted to, you could value-initialise them, which means zero-initialisation, e.g.,

Code:

class X
{
public:
    X() : a_(), b_() {}
private:
    int a_;
    double b_;
};

No primitive types are automatically initialised. They're all just random junk, and as it happens random junk frequently contains many zeros.
If you need a variable to start with a particular value, you are responsible for initialising it, preferably in the declaration.

I see, thanks.
@laserlight, I thought the default constructor called all default constructors recursively on all of its members, but it seems as if your example is implying that it is only called for non-primitives (leaving the programmer to call these explicitly)?
Is this the case? It seems a bit counter-intuitive that a Foo member will be initialized with Foo(), but a int/double/primitive member won't be initialized at all...

Originally Posted by @nthony

I thought the default constructor called all default constructors recursively on all of its members, but it seems as if your example is implying that it is only called for non-primitive classes (leaving the programmer to call these explicitly)?

If the class is not a POD type, then its constructors invoke the default constructors of all members of class types that are not listed in their initialisation lists. If not, or if the members are not of class types, then the members are not initialised if they are not listed in the respective initialisation lists.

Interesting... I wonder why they chose to do this?
If I have a Foo member "bar" and am guaranteed that bar() will be called automatically if not explicitly initialized, shouldn't one naturally expect the same behavior for integer member "i" (that i() be called)?. Almost builds a case for using wrapped primitives....

Originally Posted by @nthony

My integral members seem consistently to initialize to zero, float/double members often initialize the same, but can occasionally end-up with garbage, causing many headaches.

A good compiler warns when you use a non-initialized variable.
Otherwise, just always initialize them to a safe value.