C++ Matrix Class - Suggestions

Question

I am trying to build a templated Matrix class in C++. Below is its implementation. I implemented two operators +,+= as of now just to give an idea on how it will look and I thought it would be best to ask for feedback before I proceed any further.

The whole implementation is public and also there are no explicit bound/error checks, this is because its not meant to be a full fledged matrix library and hence avoided unnecessary code.

It would be very helpful if someone could comment on it and may be suggest a few improvements or suggestions.

Thank you.

template<class T>
class Matrix
{
    public:
    int r,c;
    vector< vector< T > >mat;

    Matrix() {}

    // Constructor to set the size of the matrix
    Matrix(int _r,int _c)
    {
        r=_r;c=_c;
        mat.resize(r);
        for(int i=0;i<r;i++)
            mat[i].resize(c);
    }
    // Constructor to build a matrix from a C 2d array
    // Pointer to the first element is passed (&arr[0][0])
    Matrix(T *arr,int _r,int _c)
    {
        r=_r;c=_c;
        mat.resize(r);
        for(int i=0;i<r;i++)
            for(int j=0;j<c;j++)
                mat[i].push_back(arr[i*c+j]);
    }
    template<typename U>
    Matrix<T>& operator +=(const Matrix<U>&M)
    {
        for(int i=0;i<r;i++)
            for(int j=0;j<c;j++)
                mat[i][j]+=static_cast<T>(M.mat[i][j]);
        return *this;
    }
    template<typename U>
    Matrix<T> operator +(const Matrix<U>&M)
    {
        Matrix<T>tmp=*this;
        return tmp+=M;
    }
};

template<typename T>
istream& operator >>(istream &in,Matrix<T>&M)
{
    in>>M.r>>M.c;
    Matrix<T>tmp(M.r,M.c);
    for(int i=0;i<M.r;i++)
        for(int j=0;j<M.c;j++)
            in>>tmp.mat[i][j];
    M=tmp;
    return in;
}
template<typename T>
ostream& operator <<(ostream &out,Matrix<T>M)
{
    for(int i=0;i<M.r;i++)
    {
        for(int j=0;j<M.c;j++)
            cout<<M.mat[i][j]<<" ";
        cout<<endl;
    }
    return out;
}

EDIT: Thank you all for the suggestions.

I just have one small question, say I do want to implement error checking (ex: checking for bounds,valid arguments etc..) however I do want the provide user with an option to disable error checking completely, is there any good way to implement this ? What I need is something like example:`ios_base::sync_with_stdio(0);. Thanks again.

Answer 1

Use initialization lists when initializing contructor args:

 Matrix(int _r,int _c)
    {
        r=_r;c=_c;

..

should be:

 Matrix(int _r,int _c) : r(_r), c(_c)
    {
        ....

Answer 2

A couple of points:

• Use a single std::vector<T> instead of std::vector<std::vector<T>>. Index it with y*r+x - use and operator overload to make this easier (see next point). This will be more memory efficient and slightly faster (and your init will be a lot easier: resize(r*c)).
• Make your data members private to protect your matrix against unintentional size changes. For example, user code can currently resize the vectors but leave the old r and c.
• Overload operator() to access the matrix (both const and non-const). If you really must use the matrix[r][c] syntax instead of matrix(r,c), consider overloading operator[] and returning an iterator to the correct row (vector iterators are random access, so they will provide operator[]).
• Implement operator+ as a non-friend non-member function instead - Improves encapsulation!
• Use initialization lists as others suggested.
• Let the constructor that currently takes a T* take an iterator instead. That way you automagically get the pointers support along with a lot of other cool things, such as range checking for debug-iterators, automatic type conversion for compatible value types and support for all other random access iterators. Also consider filling your matrix row-wise, so you can use forward iterators too.

Answer 3

1. You might want to make r, c and mat members private.
2. Your streaming operators are not symmetric (operator<< doesn't write out r and c).