Sparse/Packed matrix assignment needs type conversion

[Marco Guazzone <marco.guazzone@…>]

In file boost/detail/matrix_assign.hpp there are two possible source of type-conversion error interesting both sparse and packed matrices.

Let:

  typedef typename M::value_type value_type;
  typedef F<typename M::..., typename E::value_type> functor_type;

Then:

1. Comparison of an E::value_type with an M::value_type

     if (v != value_type/*zero*/()) // where v is of type E::value_type
   

E.g., E::value_type is std::complex<float> and M::value_type is std::complex<double>.

2. Assignment of an M::value_type to an E::value_type

     functor_type::apply(*it, value_type/*zero*/());
   

   E.g., E::value_type is float and M::value_type is std::complex<float>.

[Marco Guazzone <marco.guazzone@…>]

A sample program for showing the problem (the program should not compile).

[Marco Guazzone <marco.guazzone@…>]

I propose a possible patch (see attachment: matrix_assign-packed_sparse_storage-type_conversion.patch).

Essentially,

1. Expressions of the first type might be changed by casting an E::value_type to a M::value_type, like in this way:

     if (static_cast<value_type>(v) != value_type/*zero*/())
   

Obviously, this does not work when E::value_type and M::value_type are not castable (e.g., std::complex and double, respectively).

2. Expressions of the second type might be changed in 2 ways:

• Option A (the one used in the proposed patch)

    typedef typename matrix_traits<E>::value_type expr_value_type;
    functor_type::apply(*it, expr_value_type/*zero*/()); // NOTE: use of E::value_type in place of M::value_type
  

• Option B

    typedef F<typename M::..., value_type> functor_type; // NOTE: use M::value_type instead of E::value_type
    functor_type::apply(*it, value_type/*zero*/()); // unchanged
  

[Marco Guazzone <marco.guazzone@…>]

Possible solution.

[Marco Guazzone <marco.guazzone@…>]

There is also a companion post un uBLAS ml:

​http://lists.boost.org/MailArchives/ublas/2010/07/4420.php

[David Bellot]

applied patch from Marco Guazzone. Should be OK, however it raises a concern about a possible security hole with static_cast<>, only when people are crazy enough to use ublas as a data storage having nothing to do with linear algebra (like I did once ;-) )

[Marco Guazzone <marco.guazzone@…>]

Test case: test copy-construction/-assignement of a sparse (symmetric) matrix. The test fails to compile if the patch is not applied.