STL list_iterator代码问题(STL 4.0.0)

Question

有人能解释为什么_List_const_iterator会使用_List_node_base，并在需要的时候将其转换为_List_node吗？--我认为这背后一定有什么原因。

谢谢

struct _List_node_base
{

    _List_node_base* _M_next;   ///< Self-explanatory
    _List_node_base* _M_prev;   ///< Self-explanatory
    // ...
};

template<typename _Tp> 
struct _List_node : public _List_node_base

{
    _Tp _M_data;                ///< User's data.
};


template<typename _Tp>
struct _List_const_iterator {

    // Must downcast from List_node_base to _List_node to get to
    // _M_data.
    reference operator*() const
    { return static_cast<_Node*>(_M_node)->_M_data; }

    // ...
    // The only member points to the %list element.
    const _List_node_base* _M_node;  ///WHY NOT USING _List_node here?
};

Answer 1

我猜测_M_node是_List_node_base*类型的，因此可以使用_M_next和/或_M_prev (如您所示，它们都是_List_node_base*类型)来分配/初始化它。

我想知道为什么会有一个_List_node_base类，而不是将_M_next和_M_prev声明为_List_node类的成员。一个原因可能是减少生成的代码量:如果_List_node类有许多不同的专门化，则将其大部分(如果不是全部)代码/实现放在非泛型基类中可以减少生成的代码量。

Answer 2

这来自对EASTL list.h实现的评论-

https://github.com/paulhodge/EASTL/blob/d77d94b0d75399ac957d683ef84a8557b0f93df2/include/EASTL/list.h

/// ListNodeBase
///
/// We define a ListNodeBase separately from ListNode (below), because it allows
/// us to have non-templated operations such as insert, remove (below), and it
/// makes it so that the list anchor node doesn't carry a T with it, which would
/// waste space and possibly lead to surprising the user due to extra Ts existing
/// that the user didn't explicitly create. The downside to all of this is that
/// it makes debug viewing of a list harder, given that the node pointers are of
/// type ListNodeBase and not ListNode. However, see ListNodeBaseProxy below.
///