dividing a sparse matrix by a scalar gives a dense matrix, but multiplying gives a sparse one

[williamstein]

This is a bug:

sage: A = matrix(ZZ, 2, [1..4], sparse=True)
sage: type(A*1)
<type 'sage.matrix.matrix_integer_sparse.Matrix_integer_sparse'>
sage: type(A/1)
<type 'sage.matrix.matrix_rational_dense.Matrix_rational_dense'>

Component: linear algebra

Issue created by migration from https://trac.sagemath.org/ticket/5032

[robertwb]

comment:1

This is because it creates the matrix space over the fraction field when division is done, and I agree that it is a bug that the sparce flag doesn't get passed on. In contrast

sage: A = matrix(QQ, 2, [1..4], sparse=True)
sage: type(A)
<type 'sage.matrix.matrix_rational_sparse.Matrix_rational_sparse'>
sage: type(A/1)
<type 'sage.matrix.matrix_rational_sparse.Matrix_rational_sparse'>

What needs to change is

File:           /Users/robert/sage/sage-3.1.3/local/lib/python2.5/site-packages/sage/matrix/matrix_space.py
Definition:     MS.construction(self)
Source:
    def construction(self):
        from sage.categories.pushout import MatrixFunctor
        return MatrixFunctor(self.__nrows, self.__ncols), self.base_ring()

It should read

        return MatrixFunctor(self.__nrows, self.__ncols, is_sparse=self.is_sparse), self.base_ring()

[robertwb]

comment:2

Attachment: 5032-sparce-matrix.patch.gz

[rlmill]

comment:3

I was right in the middle of tracking this down when you posted the patch, so I figured I'd review it right away. ;)

[sagetrac-mabshoff]

comment:4

Merged in Sage 3.3.alpha2