Thread: Getting information from one class to another

Here is the situation

there is

Code:

class A{
  bool managment; // this variable is initialized later after construction

  struct UserInfo
  {
     std list::groups;
     std list::rights;
  }

  struct GroupInfo
  {
    std list::rights;
  }

  UserInfo users[];
  std::map<string,GroupInfo> groups;
}

Now in case managment is set to true then the way i am suppose to return rights for a single user is to check every group he is a member of and sum up all the rights. In case management is set to false, that means that the rights member of the UserInfo struct has been filled with values, and working with group is redundant.

My idea was to turn the UserInfo struct into a class and set the rights member to private and expose instead a getRights() method which would decide how to return the rights. My question is how to do that? In this case I need somehow to pass the information about the state of the management variable to the UserInfo class and also a reference to the groups map? What is the most "objective" way to do this?

Any member function of a class can see any other variable in the class, private or not. getRights() would seem to me to be a member of the A class, which means it would see the managment variable without you doing any work.

Originally Posted by farmer

Here is the situation

Code:

class A{
  bool managment; // this variable is initialized later after construction

  struct UserInfo
  {
     std list::groups;
     std list::rights;
  }

  struct GroupInfo
  {
    std list::rights;
  }

  UserInfo users[];
  std::map<string,GroupInfo> groups;
}

this is incorrect C++ syntax.

the correct way to declare a list is:

Code:

std::list rights;
std::list groups;

etc.

if you don't want the caller of getRights() to directly modify the contents of the list (it's generally considered good practice not to allow this), you should return a copy of the list. the easiest way to do this is to declare the getRights() function as follows:

Code:

std::list getRights();

std::list's copy constructor and assignment operator will automatically copy the elements properly, assuming the template type is trivially copyable. if a shallow copy is not appropriate, you will need to define a custom copy constructor and assignment operator for your user-defined type.

a better way might be to pass a std::list to the getRights() function by reference or pointer, and use std::copy or some other method to fill the list. it could avoid an unnecessary additional copy of the data.

> if you don't want the caller of getRights() to directly modify the contents of the list (it's generally considered good practice not to allow this), you should return a copy of the list.

I have to disagree with this one. Any container that can contain an unspecified (but possibly) large amount of objects (such as strings, lists, vectors and so on) I would return a const reference instead.

Originally Posted by Shakti

I have to disagree with this one. Any container that can contain an unspecified (but possibly) large amount of objects (such as strings, lists, vectors and so on) I would return a const reference instead.

but a const reference can still be cast to a normal reference. it shouldn't be done, but it can be done, and some people still do it. the safest way is to copy the entire list and return the copy. it's true that a large list of objects could be expensive, in terms of clock cycles, to copy, but it all depends on how much isolation you want between the class code and the calling code.

Originally Posted by Elkvis

but a const reference can still be cast to a normal reference. it shouldn't be done, but it can be done, and some people still do it. the safest way is to copy the entire list and return the copy. it's true that a large list of objects could be expensive, in terms of clock cycles, to copy, but it all depends on how much isolation you want between the class code and the calling code.

Absurd! Design a library under the assumption that people will use it according to the rules you design it under (do not cast away const for instance) and keep to good design practices (yes i consider pass by const reference rather than pass by value a good design practice, so does the author of Effective C++, a widely cited book when it comes to C++ coding practices). Plus the "isolated" data can still be modified (not sure it's even 100% valid C++ code but hey, if the people you design for break the rules of the library why assume they adhere to the C++ standard...):

Code:

#include <iostream>
#include <vector>

class Foo
{
private:
	std::vector<int> veryLongVector;
public:
	Foo()
	{
		for(int i=0; i<100000; i++)
			veryLongVector.push_back(i);
	}

	std::vector<int> getVector() { return veryLongVector; }
};

int main(int argc, char** argv)
{
	Foo f;
	std::vector<int> *vec = (std::vector<int>*)&f;
	(*vec)[0] = 2;
	std::cout << f.getVector()[0] << std::endl;
}

And I am saying you can't guarantee this so why even bother. Do things right (for example do not invoke copy contructor unnessecary, especially for large data-structures) and tell the people who use the library wrong to bugger off. Really, everything else is just poor design and creates badly written libraries. And yes, I agree my example is ridiculous but as grumpy said: either the person casting away const knows what he is doing (which is fine) or he can only blame any problems on his own coding (also fine) so why delibirately write inferior code?