Wrong solution result for non square matrix.

[Gluttton]

Based on test: MunkresTest.solve_3x2_NonObviousSolutionCase001_Success solution result for non square matrix is wrong.

Possibly reason (in my opinion):

1. Wrong test case.
2. Algorithm designed only for square matrix (so special assertion is required).
3. Error in algorithm.

[saebyn]

This test passes when I disable the row/column minimization functions, so I'm going to start by looking in there to see if that's the problem.

[saebyn]

Nope, I never bothered to add support for non-square input matrices (and now I've discovered that fact), so that seems to have been the problem.

[saebyn]

Commit ae77d7e adds support for rectangular inputs.

[Gluttton]

The problem have not solved completely.

Here is the related issue.
Also commits
0acab86
5b66cd7
were added and related to the issue.

The problem is required more deeper research.

P.S. If input into the Google search field "the assignment problem no" the Google will prompt "n square matrix". So it looks like that such problem is really exists ;) .

[Gluttton]

I've simplified test data with preserving their behaviour.
Instead of:

  Matrix<double> etalon_matrix{
    {-1.0, -1.0, -1.0, -1.0},
    { 0.0, -1.0, -1.0, -1.0},
    {-1.0,  0.0, -1.0, -1.0},
    {-1.0, -1.0, -1.0, -1.0},
    {-1.0, -1.0, -1.0,  0.0},
    {-1.0, -1.0,  0.0, -1.0}
  };
  Matrix<double> test_matrix{
    {1.79769e+308, 7.33184e+08,  9.41561e+08,  2.79247e+08},
    {3.06449e+08,  1.79769e+308, 3.3464e+08,   7.06878e+08},
    {9.93296e+08,  1.9414e+08,   1.79769e+308, 1.14174e+08},
    {3.51623e+08,  2.48635e+08,  7.81242e+08,  1.79769e+308},
    {7.02639e+08,  8.51663e+08,  9.37382e+08,  4.96945e+07},
    {7.58851e+08,  8.58445e+08,  8.7235e+07,   5.47076e+08}
  };

I use:

  Matrix<double> etalon_matrix{
    {-1.0, -1.0},
    { 0.0, -1.0},
    {-1.0,  0.0}
  };
  Matrix<double> test_matrix{
    {1.0e+17, 3},
    {2,       1.0e+17},
    {4,       1}
  };

etc...

After playing with simple tests I found out that my test cases for 7x5 and 5x7 matrices was wrong.

So it looks like the problem was fixed with commit 0acab86 .