Hello,

I have had troubles getting matrix multiplication to always work. It seems to rely on the way one initializes matrices as well as the size of the matrices involved (square or rectangular).

The script below works when I use

M1 = matrix((1,2,3),(4,5,6),(7,8,9))

But fails with the complaint "Operator cannot be expanded by the element" when I use

M1 = matrix[3,3]

repeat i = 1 to 3 repeat j = 1 to 3 M1[i,j] = (i-1)*3+j endrepeat endrepeat

In addition, I have had trouble initializing vector matrices. I have begun to use vector matrices because flexpde doesn't seem to like me adding a matrix and an array despite the fact that they are the same size. However it seems I cannot initialize such vector matrices in the way I want to.

The script below works when I use

M4 = matrix((1),(2),(3))

But fails with the error "Empty list" when I use

M4 = matrix for i (1 by 1 to 3) for j (1 by 1 to 1): i

While I can use "for j (1)", this is inconvenient as it makes it difficult for me to switch back and forth between 1 and more than 1 objects since the columns of the matrix correspond to different objects.

The full working script is below with the variations that cause the error commented out:

DEFINITIONS { parameter definitions }

! M1 = matrix[3,3]

! repeat i = 1 to 3 repeat j = 1 to 3 M1[i,j] = (i-1)*3+j endrepeat endrepeat

M1 = matrix((1,2,3),(4,5,6),(7,8,9))

M2 = matrix((1,2),(3,4),(5,6))

M3 = M1**M2

M4 = matrix((1),(2),(3))

! M4 = matrix for i (1 by 1 to 3) for j (1 by 1 to 1): i

M5 = M1**M4

BOUNDARIES

REGION START(0,0) LINE TO (1,0) TO (1,1) TO (0,1) TO CLOSE

PLOTS

summary

repeat i = 1 to 3 repeat j = 1 to 3 report(M1[i,j]) endrepeat endrepeat

repeat i = 1 to 3 repeat j = 1 to 2 report(M2[i,j]) endrepeat endrepeat

repeat i = 1 to 3 repeat j = 1 to 2 report(M3[i,j]) endrepeat endrepeat

repeat i = 1 to 3 repeat j = 1 to 1 report(M4[i,j]) endrepeat endrepeat

repeat i = 1 to 3 repeat j = 1 to 1 report(M5[i,j]) endrepeat endrepeat

END

Hopefully there can be some way to get around these issues other than explicitly writing out every component of relatively large matrices. If anyone has any ideas or if there are fixes available, it would be nice to know about. Thanks,

Jared Barber