可变参数宏的展开，宏重载

待我称王封你为后i 2022-06-12 04:41 302阅读 0赞

转自：http://www.cnblogs.com/liaokang/p/5949835.html

因某些方面需要，涉及到可变参数宏的展开及可变参数数量的确定，在网上查找了大部分资料，基本如下所述

[http://www.cnblogs.com/goooon/p/5642514.html][http_www.cnblogs.com_goooon_p_5642514.html]

**涉及到的问题点：不支持0个参数。**

经过VS2010及gcc version 4.9.1 (GCC)两个编译器编译，可以至少支持此两种编译器下的0参数问题。

以上文中的代码为例：

**定义部分**

> \#define PRIVATE\_ARGS\_GLUE(x, y) x y
> 
> \#define PRIVATE\_MACRO\_VAR\_ARGS\_IMPL\_COUNT(\_1,\_2,\_3,\_4,\_5,\_6,\_7,\_8,\_9, \_10, N, ...) N  
> \#define PRIVATE\_MACRO\_VAR\_ARGS\_IMPL(args) PRIVATE\_MACRO\_VAR\_ARGS\_IMPL\_COUNT args  
> \#define COUNT\_MACRO\_VAR\_ARGS(...) PRIVATE\_MACRO\_VAR\_ARGS\_IMPL((\_\_VA\_ARGS\_\_,10, 9,8,7,6,5,4,3,2,1,0))
> 
> \#define PRIVATE\_MACRO\_CHOOSE\_HELPER2(M,count)  M\#\#count  
> \#define PRIVATE\_MACRO\_CHOOSE\_HELPER1(M,count) PRIVATE\_MACRO\_CHOOSE\_HELPER2(M,count)  
> \#define PRIVATE\_MACRO\_CHOOSE\_HELPER(M,count)   PRIVATE\_MACRO\_CHOOSE\_HELPER1(M,count)
> 
> \#define INVOKE\_VAR\_MACRO(M,...) PRIVATE\_ARGS\_GLUE(PRIVATE\_MACRO\_CHOOSE\_HELPER(M,COUNT\_MACRO\_VAR\_ARGS(\_\_VA\_ARGS\_\_)), (\_\_VA\_ARGS\_\_))

**实现部分**

> \#define PRINT\_ARGS1(a1) std::cout<<a1<<std::endl  
> \#define PRINT\_ARGS2(a1,a2) std::cout<<a1<<","<<a2<<std::endl  
> \#define PRINT\_ARGS3(a1,a2,a3) std::cout<<a1<<","<<a2<<","<<a3<<std::endl  
> \#define PRINT\_ARGS4(a1,a2,a3,a4) std::cout<<a1<<","<<a2<<","<<a3<<","<<a4<<std::endl

**使用部分**

> INVOKE\_VAR\_MACRO(PRINT\_ARGS, 4);  
> INVOKE\_VAR\_MACRO(PRINT\_ARGS, 4, 5);  
> INVOKE\_VAR\_MACRO(PRINT\_ARGS, 4, 5, 6);  
>

如何实现?

> \#define PRINT\_ARGS0() std::cout<<"ARGS0"<<std::endl

如何使用?

> INVOKE\_VAR\_MACRO(PRINT\_ARGS);

参考  [http://en.cppreference.com/w/cpp/preprocessor/replace][http_en.cppreference.com_w_cpp_preprocessor_replace]

> some compilers offer an extension that allows \#\# to appear after a comma and before \_\_VA\_ARGS\_\_,
> 
> in which case the \#\# does nothing when \_\_VA\_ARGS\_\_ is non-empty,
> 
> but removes the comma when \_\_VA\_ARGS\_\_ is empty: this makes it possible to define macros such as `fprintf (stderr, format, ##__VA_ARGS__)`

`看不懂？那看看http://blog.csdn.net/sakaue/article/details/29591631`

`那是不是可以改成这样呢？`

> \#define PRIVATE\_MACRO\_VAR\_ARGS\_IMPL\_COUNT(\_0,\_1,\_2,\_3,\_4,\_5,\_6,\_7,\_8,\_9, \_10, N, ...) N  
> \#define PRIVATE\_MACRO\_VAR\_ARGS\_IMPL(args) PRIVATE\_MACRO\_VAR\_ARGS\_IMPL\_COUNT args  
> \#define COUNT\_MACRO\_VAR\_ARGS(...) PRIVATE\_MACRO\_VAR\_ARGS\_IMPL((0,\#\#\_\_VA\_ARGS\_\_,10, 9,8,7,6,5,4,3,2,1,0))

在gcc version 4.9.1 (GCC)下是没有问题的，

可以输出

> ARGS0  
> 4  
> 4,5  
> 4,5,6

但是在vs2010中就出现了编译错误

> warning C4003: “PRINT\_ARGS1”宏的实参不足  
> error C2059: 语法错误:“<<”  
>

那是不是意味着vs2010不支持 **”如果可变参数被忽略或为空，‘\#\#’操作将使预处理器（preprocessor）去除掉它前面的那个逗号。”**

在vs2010中定义

> \#define debug(format, ...) fprintf (stderr, format, \#\#\_\_VA\_ARGS\_\_)

vs2010中实现

> debug("abc");

结果编译没有问题，输出也正常。

因此，应该是vs2010和gcc在预编译中，对于宏的处理存在差异。最明显的一个例子就是参考google chromium中logging，

> typedef int LogSeverity;  
> const LogSeverity LOG\_VERBOSE = -1;  // This is level 1 verbosity  
> // Note: the log severities are used to index into the array of names,  
> // see log\_severity\_names.  
> const LogSeverity LOG\_INFO = 0;  
> const LogSeverity LOG\_WARNING = 1;  
> const LogSeverity LOG\_ERROR = 2;  
> const LogSeverity LOG\_FATAL = 3;  
> 
> 
>  
> 
> LOG(INFO) << "Found " << num\_cookies << " cookies";
> 
> 其中涉及使用 ::logging::LOG\_ \#\# severity  组成一个 ::logging::LOG\_INFO

个人修改时，添加了一个LOG\_DEBUG，结果vs2010上编译使用都没有问题，gcc version 4.9.1 (GCC)下编译不过，说找不到LOG\_1，修改为LOG\_DEBUGGING，编译及输出均OK。

那么如何在此两种编译环境下解决不支持0个参数问题呢？

测试代码

> \#define PP\_ARG\_N(\_1,\_2,\_3,\_4,\_5,\_6,\_7,\_8,\_9, \_10, N, ...) ((\#\_1 == "") ? 0 : N)  
> \#define PP\_NARG\_(args)  PP\_ARG\_N args  
> \#define PP\_NARG(...) PP\_NARG\_((\_\_VA\_ARGS\_\_,10,9,8,7,6,5,4,3,2,1,0))
> 
> \#define TESTINVOKE\_VAR\_MACRO(M,...) std::cout<<PP\_NARG(\_\_VA\_ARGS\_\_)<<std::endl
> 
> TESTINVOKE\_VAR\_MACRO(PRINT\_ARGS);  
> TESTINVOKE\_VAR\_MACRO(PRINT\_ARGS, 4);  
> TESTINVOKE\_VAR\_MACRO(PRINT\_ARGS, 4, 5);  
> TESTINVOKE\_VAR\_MACRO(PRINT\_ARGS, 4, 5, 6);

输出

> 0  
> 1  
> 2  
> 3

但要针对\#define PRIVATE\_MACRO\_VAR\_ARGS\_IMPL\_COUNT(\_1,\_2,\_3,\_4,\_5,\_6,\_7,\_8,\_9, \_10, N, ...) N 的修改，那就要需要慢慢修改，为什么呢？

参考宏替换规则，如果将N改为一个表达式，那还能找到对应的 PRINT\_ARGSX 吗？显然是不能的。但如果作为函数传参，这个方法还是能够胜任的，比如上述的测试代码。

[http_www.cnblogs.com_goooon_p_5642514.html]: http://www.cnblogs.com/goooon/p/5642514.html
[http_en.cppreference.com_w_cpp_preprocessor_replace]: http://en.cppreference.com/w/cpp/preprocessor/replace