覆盖自定义数据结构的迭代器的仅运算符*（）

Question

I have an array class MyArray with MyArray .
For simplicity, this is the whole code. It works OK.

Current code

#include <iostream>
using namespace std;
template<class T> class MyArray;
template<class T>class MyIterator{
    public: int index=0;
    public: MyArray<T>* myArray;
    public: MyIterator<T> operator++(){
        index++;
        return *this;
    }
    public: T& operator*(){
        return myArray->database[index];
    }
    public: friend bool operator!=(MyIterator<T> b,MyIterator<T> c){
        return b.index!=c.index;
    }
};
template<class T>class MyArray{
    public: T database[5];
    public: MyArray(){
        database[2]=3; //just fill with something to test
    }
    public: MyIterator<T> begin(){
        MyIterator<T> r;  r.index=0; r.myArray=this;
        return r;
    }
    public: MyIterator<T> end(){
        MyIterator<T> r;  r.index=5; r.myArray=this;
        return r;
    }
};

Here is its usage :-

int main() {
    MyArray<int> test;
    for(int ele:test){
        std::cout<<ele<<std::endl;
    }
    return 0;
}

Problem / requirement

I have specific classes, let's say their names are B and C .
I have a converter from B to C named convertBToC(B) .

Now, I want a new datastructure (named MyArray2 ) that :-

• act mostly like MyArray<B> ....
• except that function operator*() of MyArray2 's iterator return C instead of B
• the C that is returned, convert from B using convertBToC(B) .

This is the usage that I wished for ( #1 ) :-

MyArray2 test;
//test.push_back(B());  //used like "MyArray<B>"
for(C c:test){          //"C", not "B"
    .... logic about "c" ....  
}  

The above code would work as if I call it like :-

MyArray<B> arr;   
for(B& b: arr){
    C c= convertBToC(b);   //<-- I want to hide this line
    .... logic about "c" .... 
}

Question: How to code MyArray2 ?

Criteria

I want a solution that :- (sorted by priority)

• efficient (not use std::function and its family)
• not refer to MyIterator directly (because MyIterator is an internal class)
• cute (low amounts of statement/line, readable)
• minimum change to MyArray<T> (if any)

The most related question is here , but it mentions about std::vector .

My poor solutions

Solution 1 (2x inheritance)

Create 2 classes :-

• MyArray2 derived from MyArray<B>
  Override: begin() or end() - return MyIterator2 .

• MyIterator2 derived from MyIterator<B>
  Override: operator*() - return C (using convertBToC() ).

Disadvantage:

• Dirty - I have to create 2 classes just to override only 1 function.
• MyArray2.h code contains the word MyIterator<T> .

Solution 2 (lambda)

Create only 1 class :-

• MyArray2 derived from MyArray<B>
  new function: iterate() :-

Here is the draft of iterate() :-

template<typename F> MyArray2::iterate( F lamdbaFunction ){
    for(B b: MyArray<B>){
         C c= convertBToC(b);
         lamdbaFunction(c);
    }
}

The usage have to be changed from #1 to be ... ( #2 )

MyArray2 arr;   
auto lambdaF=[&](C c){     
    .... logic about "c" .... 
}
arr.iterateElement(lambdaF);

Disadvantage

• It sometimes destroy readability. (Lambda sometimes harder to read)
• It limits coding style. (Sometimes, I prefer range-based loop.)

Answer 1

Another approach would be to use Boost range adaptors, specifically boost::adaptors::transformed .

#include <boost/range/adaptor/transformed.hpp>
...

MyArray<B> test;
for(C ele : test | boost::adaptor::transformed(convertBToC)) {
    // something...
}

Judge how much of an improvement it is yourself. Maybe better put the operator| call inside a new MyArray 's member function.