Watch out the parsing mechanism of C++ compiler

From the book named <<Effective STL 50 Specific Ways to Improve Your Use of the Standard Template Library>>, I knew the suggestion of " Watch out the parsing mechanism of C++ compiler", which is the sixth suggestion in that book. This fact is because of C++ using the strategy of "interpreted as being function as possible".

     There is a classic and typical example about reading data from a file and using the data as parameters to construct a list.

The code is below:

   

#include <iostream>
#include <fstream>
#include <list>
#include <iterator>
using namespace std;

int main(int argc, char* argv[])
{

	std::ifstream dataFile( "Test.txt" ); 

	std::list< int > intData( std::istream_iterator< int >( dataFile ), 

                        std::istream_iterator< int >() ); 

	std::list< int >::iterator iter = intData.begin();
	
	for(iter; iter != intData.end(); ++iter)
	{
		std::cout<<*iter<<std::endl;
	}
	return 0;
}

 

 

If you use this code compiled with gcc 4.4.7, you will get below output:

parseMech.cc:19: error: request for member ‘begin’ in ‘intData’, which is of non-class type ‘std::list<int, std::allocator<int> >(std::istream_iterator<int, char, std::char_traits<char>, int>, std::istream_iterator<int, char, std::char_traits<char>, int> (*)())’

 It says we are using a non-class type, which means what we want to construct a list above is not valid.

So It's important for us to explore how the compiler deals with the statement of definition of the variable iniData. Here C++ Compiler uses the strategy which introduced above to interpret the definition as a declaration of a function. It treats the dataFile as a parameter with type of std::istream_iterator<int> and treats the std::istream_iterator<int>() as a function pointer which points to a function having no parameter and return type of std::istream_iteratior<int>.

 

Now we know what the C++ Compiler actually do , then what we can do to avoid this mechanism of C++ Compiler?

 

There are two solutions to solve this problem:

1. wrapped the std::istream_iterator<int>(dataFile) with pair of parentheses. the result is

#include <iostream>
#include <fstream>
#include <list>
#include <iterator>
using namespace std;

int main(int argc, char* argv[])
{

	std::ifstream dataFile( "Test.txt" ); 

  

	std::list< int > intData(( std::istream_iterator< int >( dataFile )), 

                        std::istream_iterator< int >() ); 


	std::list< int >::iterator iter = intData.begin();
	
	for(iter; iter != intData.end(); ++iter)
	{
		std::cout<<*iter<<std::endl;
	}
	return 0;
}

 

Then we re-compile this code and the damn error would be gone.  But unfortunately, this solution is not well supported by all of compilers. So We will introduce the other solution.

 

2. The better solution is to avoid using anonymous object, the following code would always work correctly:

#include <fstream>
#include <list>
#include <iterator>
using namespace std;

int main(int argc, char* argv[])
{

	std::ifstream dataFile( "Test.txt" ); 

  	std::istream_iterator< int > dataBegin(dataFile);
  	 
  	std::istream_iterator< int > dataEnd;

	std::list< int > intData(dataBegin, dataEnd ); 

	std::list< int >::iterator iter = intData.begin();
	
	for(iter; iter != intData.end(); ++iter)
	{
		std::cout<<*iter<<std::endl;
	}
	return 0;
}

 Here we explicitly declare the dataBegin and dataEnd.