迭代器与 STL 容器的关系

Question

I understand the general idea behind iterators and I can use them at a basic level, but I don't get what happens under the hood and the relationship they have with STL containers.

From reading http://www.cplusplus.com/reference/iterator/iterator/ :

• Iterators are objects defined with a class template

• "Iterators are classified into five categories depending on the functionality they implement"

• There are different types of iterators (input, output, forward, bidirectional, random access)

My assumptions are (I will be using vector throughout as an example)

• In STL containers, for example vector , the iterator is a nested class template and is created for each unique vector type.
• An iterator being considered to be one particular type of iterators is just a concept because it ultimately depends on what is implemented as member functions in the vector class.
  • My reasoning is that for example: std::list<int>::iterator iterator; which is a bidirectional iterator vs std::vector<int>::iterator iterator; which is a random access iterator are both declared with ::iterator and there is no distinction made.
• Functions like begin() and end() are overloaded in the vector class

I hope I am making sense, please correct me.

Answer 1

Let's go point by point.

Iterators are objects defined with a class template

Not necessarily. Iterators are objects that have certain operations . Pointers are such objects, as are objects of various class types.

Each container defines a member type container::iterator , and another member type container::const_iterator .

These can be directly (a nested class), or it could be with a type alias referring to some other typename.

Iterators are classified into five categories depending on the functionality they implement

As of C++14, yes. C++17 and C++20 each introduce another category.

There are different types of iterators (input, output, forward, bidirectional, random access)

Those are the (C++14) categories, but there are an infinity of types within each category. std::vector<int>::iterator is a random access iterator, and so is double * , but they are distinct types. These categories overlap, the definitions are in terms of previous ones in the hierachy. RandomAccess is defined as an extension to Bidirectional , which is defined as an extension to Forward , which is defined as an extension to Input .

In STL containers, for example vector , the iterator is a nested class template and is created for each unique vector type.

Mostly. In many implementations it is, but in the particular case of std::vector<T> , no rule stops an implementation from using T * as the iterator type.

An iterator being considered to be one particular type of iterators is just a concept because it ultimately depends on what is implemented as member functions in the vector class.

Yes. C++ has a notion Concept , which are just labels for things that behave in similar ways. Because of how templates work, there does not need to be any relationship between the types that satisfy a particular Concept. Contrast Java and C#, where interface s have to be explicitly mentioned in the definition of a type.

Functions like begin() and end() are overloaded in the vector class

This is literally true, but probably not what you mean. Each container does have two member functions named begin . container::iterator container::begin() and container::const_iterator container::begin() const .

There is also the free function template std::begin , which is specialised for each container and also (C style) arrays.