string-->wstring-->string

	std::string src("三毛三毛三毛三毛三三三三流浪记");

 	size_t size = mbstowcs(NULL,src.c_str(),0);
 	std::wstring dst;
 	dst.resize(size);
 	setlocale(LC_CTYPE,"chs");//setlocale(LC_CTYPE,"UTF-8") on mac
 	mbstowcs(&dst[0],src.c_str(),size); 
	wstring::size_type pos = dst.find_first_of(L'\0'); // L'\0' or wchar_t()
 	dst = dst.substr(0,pos);
 	stable_sort(dst.begin(),dst.end());
 	wstring::iterator it = unique(dst.begin(),dst.end());
 	dst.erase(it,dst.end());
 	size_t newSize = wcstombs(NULL,dst.c_str(),0);
 	string newStr;
 	newStr.resize(newSize);
 	wcstombs(&newStr[0],dst.c_str(),newSize);
 	setlocale(LC_ALL,"C");

 

wcstombs:

  http://msdn.microsoft.com/zh-cn/subscriptions/downloads/5d7tc9zw(v=vs.71).aspx

     mbstowcs(NULL,src.c_str(),0) 返回的wstring需要的长度,比实际需要的多

    导致wstring.resize的时候,多了许多'\0' ,经过排序去重后,'\0'在最前面

         wcstombs 看到'\0' 立马结束,返回string的长度需要0,导致错误。

  If wcstombs encounters the wide-character null character (L'\0') either before or when count occurs, it converts it to an 8-bit 0 and stops. Thus, the multibyte character string at mbstr is null-terminated only if wcstombs encounters a wide-character null character during conversion. If the sequences pointed to by wcstrand mbstr overlap, the behavior of wcstombs is undefined.

 

 

 

你可能感兴趣的:(String)