在某些元素上提供带有迭代器的整数矩阵

Question

I'm trying to find a solution for the following problem:

Suppose we have implemented an abstract data type "integer matrix" (assume that integers are stored in rows). We need to define an iterator to iterate over the matrix by columns and only over the even elements of the matrix.

Executing the folling code:

 Matrix M; ... Matrix::iterator it; for(it = M.begin(); it != M.end();++it) cout << *it 

on the matrix:

 5 4 3 2 1 2 9 0 2 8 9 1 

should produce 2 8 4 0 2 2

I face several problems to do it. First, I would like to avoid providing the class iterator with much information about the matrix it is iterating over. Is it inevitable to provide the number of rows and columns of the matrix? Second, the solution depends a lot on the internal representation of the matrix so I'd say I should implement a basic iterator to iterate row by row...

What do you think is the best solution to this problem? I would just need advice about the representation of the matrix and the iterator and the functions begin,end,operator*,operator++ and operator ==

Answer 1

Without guessing at which of the operations that you haven't specified need to be especially efficient, it appears a flat linear storage of the elements would be best. By calling it Matrix::iterator (and using begin() and end() as you have shown) you have implied this is the ordinary iterator used for this Matrix class (not some weird extra iterator). So it is worth a little extra content in the matrix to keep the iterator light weight. You can make the iterator's data content be only a pointer, so in an optimized compile the iterator itself is only a pointer. It doesn't need to know row count or column count or limit.

To do that, you need to allocate and initialize a guard position at the limit of the matrix data and it must hold an even number and end() must be a pointer to it (wrapped in the iterator class). Then begin() can simply find the first even number and the iterator's ++ can simply find the next even number.

To support other operations that we assume must exist (but you haven't asked about) the Matrix itself must know number of columns. For end() and other operations, the Matrix must know the total size (represented as number of rows or number of elements or limit address are anything else easily translated to limit address). But that extra info is better kept in the Matrix and not duplicated into the iterator.