Question

I’ve seen this question: Class design with vector as a private/public member?, but I don’t feel that it answers the question.

I have a class called PrimeSieve that can be initialized as PrimeSieve s(10000) to perform actions such as checking whether a number under 10000 is prime or listing all the primes under 10000. In particular, I’m wondering how to perform the latter function.

Currently, I’m doing this, which I think violates the principles of OOP:

class PrimeSieve {
public:
...
     std::vector<long long> primes;

The client will never need to change the vector, but how can I still allow the client to iterate through a vector of all of the primes under some number (using things like vector.size()? I’ve thought about an accessor method that returns the vector by value, but that seems inefficient.

Answer 1

I would define the class as:

class PrimeSieve 
{
public:

    typedef std::vector<long long>::const_iterator const_iterator;

    const_iterator begin() const  { return primes.begin(); }
    const_iterator end() const  { return primes.end(); }

    long long  operator[](int index) const { return primes[i]; }

    size_t   size() const { return primes.size(); }

    //rest of the members which compute the primes!

private:
   std::vector<long long> primes ;
} ;

I think the user of the class should have read-only access to the vector, so I used const_iterator in the typedef definition. Also note that the operator[] is a const member function which means you can only access the value, but cannot modify it.

Usage:

 PrimeSieve sieve(1000); //compute

 //print to stdout
 std::copy(sieve.begin(), 
           sieve.end(), 
           std::ostream_iterator<long long>(std::cout, " "));