Bugs with 1x0 and 0x1 matrices

[lqy16]

I find that there are several bugs with 1x0 and 0x1 matrices (Matrix.Dense).

• The method toString() cannot work normally with 1x0 and 0x1 matrices (althogh it does work with 0x0 matrices)
• When computing AxB (A is 1x0 matrix and B is 0x1 matrix) with MKL library, there will be a ERROR: "Intel MKL ERROR: Parameter 9 was incorrect on entry to cblas_dgemm", which severely affects the computation speed.

BTW, such matrices are of practical application. For example, in state space model, we need 0x0, 0x1 and 1x0 matrices to describe a gain-only process.

[lqy16]

Guys, is there any progress on this bug?

[cdrnet]

Hm yes, we should address this before the v5 release.

[jkalias]

Has this been addressed already?