Question

I have seen lots of codes where the coders define an init() function for classes and call it the first thing after creating the instance.

Is there any harm or limitation of doing all the initializations in Constructor?

Answer 1

It is a design pattern that has to do with exceptions thrown from inside an object constructor.

In C++ if an exception is thrown from inside an object costructor then that object is considered as not-constructed at all, by the language runtime.
As a consequence the object destructor won’t be called when the object goes out of scope.

This means that if you had code like this inside your constructor:

int *p1 = new int;
int *p2 = new int;

and code like this in your destructor:

delete p1;
delete p2;

and the initialization of p2 inside the constructor fails due to no more memory available, then a bad_alloc exception is thrown by the new operator.
At that point your object is not fully constructred, even if the memory for p1 has been allocated correctly.
If this happens the destructor won’t be called and you are leaking p1.

So the more code you place inside the constructor, the more likely an error will occur leading to potential memory leaks.

That’s the main reason for that design choice, which isn’t too mad after all.

More on this on Herb Sutter’s blog: Constructors exceptions in C++