[百度空间] [note] pointer to member is a POD type

C++03 3.9-10:

Arithmetic types (3.9.1), enumeration types, pointer types, and pointer to member types (3.9.2), and cv-qualified versions of these types (3.9.3) are collectively called scalar types. Scalar types, POD-struct types, POD-union types (clause 9), arrays of such types and cv-qualified versions of these types (3.9.3) are collectively called POD types.

As we all knows the pointer to member may not be cast into void* (GCC is OK but it‘s a special case anyway), according to the standrad we can store the pointer to member to a char buffer though. Unfortunately, the size of "pointer to member" is implementation defined, so we still need some hack for different compilers.

here‘s a table for sizeof(pointer to member) for some compilers

(http://www.codeproject.com/Articles/7150/Member-Function-Pointers-and-the-Fastest-Possible)

Compiler Options Single Multi Virtual Unknown

MSVC 4 8 12 16

MSVC /vmg 16# 16# 16# 16

MSVC /vmg /vmm 8# 8# -- 8

Intel_IA32 4 8 12 16

Intel_IA32 /vmg /vmm 4 8 -- 8

Intel_Itanium 8 12 16 20

G++ 8 8 8 8

Comeau 8 8 8 8

DMC 4 4 4 4

BCC32 12 12 12 12

BCC32 /Vmd 4 8 12 12

WCL386 12 12 12 12

CodeWarrior 12 12 12 12

XLC 20 20 20 20

DMC small 2 2 2 2

medium 4 4 4 4

WCL small 6 6 6 6

compact 6 6 6 6

medium 8 8 8 8

large 8 8 8 8

the original article implement each compiler‘s pointer to member mechanism, which I believe is insane, it‘s definitely not a good way to write portable codes.here‘s what I‘m doing (code snippet).

所以pointer to member 可以被保存在byte array 里面再被解出来.这对于做delegate很有帮助.这是我目前使用的方法.

class Delegate

{

public:

//per-compiler constant, or use a max enough size for all compilers

#if COMPILER == COMPILER_MSVC

static const size_t MAX_POINTER2MEMBER_SIZE = 16;

#elif COMPILER == COMPILER_GNUC

static const size_t MAX_POINTER2MEMBER_SIZE = 8;

#else

static const size_t MAX_POINTER2MEMBER_SIZE = 20;

#endif

void* mPtr;

char mMemberRaw[MAX_POINTER2MEMBER_SIZE ];

//store the pointer to member to a byte buffer

template <typename T>

explicit Delegate(T* ptr, void (T::*func)(void*) )

{

assert( sizeof(func) <= MAX_SIZE );

::memset(mMemberRaw, 0, MAX_SIZE);

::memcpy(mMemberRaw, &func, sizeof(func) );

}

};

//invoke function: extract the "pointer to member" from the byte buffer

{

...

typedef void (T::*TMemberFn)(void*);

TMemberFn TFn = NULL;

::memcpy(&TFn, mMemberRaw, sizeof(TFn) );

( ((T*)mPtr)->*TFn )( data);

}

there‘s a very efficient way to do the delegate :

http://www.codeproject.com/Articles/11015/The-Impossibly-Fast-C-Delegates

他也是用模板函数,但不同的是把函数地址作为模板参数.本来打算改用这个方法的.但是唯一的缺点就是需要显式指定模板参数<>来实例化, 而构造函数只能通过TAD来套用模板参数,不能显式指定...所以他用的是一个静态模板函数来构造一个对象

1	`delegate d = delegate::frommember<T,&T::func>( pOjb); //这样稍微有点恶心.`

而我使用的方式虽然多了一个char[], 有一点点额外的内存开销和运行时开销,但是大多数情况下几乎可以忽略不计.

主要是可以通过Template Arugment Deduction 直接创建delegate, 比如:

1	`Delegate d(pObj, &CLASS::function);`

所以考虑了一下,决定暂时不用他的那种方法(用起来稍微不方便,但是效率非常高).以后看情况再考虑要不要改.

另外,作为一个C++程序猿, 个人觉得"数据绑定", "反射","代理",这些使用时尽量谨慎, 不可滥用. 除了在event/msg 和 UI/Tools的极少数地方要用以外,其他地方还是尽量不要用为好. 最近把数据绑定框架做了简单重构, 幸好用到的地方不多(主要是编辑器在用), 要不然重构就哭了. 目前还没有反射,暂时不需要, 以后需要时再加.最好是另外封装,不影响现在的代码.

时间： 2024-10-22 21:16:46

[百度空间] [note] pointer to member is a POD type

[百度空间] [note] pointer to member is a POD type的相关文章

百度空间代码大全

Eclipse启动之四： Eclipse核心框架启动(百度空间迁移)

Eclipse启动之二：Eclipse动态库(百度空间迁移)

Eclipse启动之三：启动器插件(百度空间迁移)

“百度空间关闭”后的问题解决攻略

转自 x_x的百度空间

写给急功过利的百度：从此放弃百度空间

再见-百度空间

Android手机无法访问百度空间的解决办法