initialize a row vector

Can I initialize a constant row vector?

I have tried Constant((1,1)), and Expression(('1.','1.')). They both return a column vector. When I tried to transpose it, it gives an error that, transpose is only defined for the Matrix.

Thanks!

asked Dec 17, 2015 by truemerlin FEniCS Novice (410 points)

Can you give more details? E.g. code?

commented Dec 18, 2015 by Yair Daon FEniCS User (1,350 points)

just normal setup.

from dolfin import *
mesh = UnitSquareMesh(2,2)
TFS = TensorFunctionSpace(mesh, 'CG', 1)
F = Function(TFS)
N = Constant((1.,1.))

N*F

this won't work

commented Dec 18, 2015 by truemerlin FEniCS Novice (410 points)

Is this what you're looking for?
As you said, a transpose operation could only be done on a matrix.

from dolfin import *
mesh = UnitSquareMesh(2,2)
TFS = TensorFunctionSpace(mesh, 'CG', 1)
F = Function(TFS)

# as tensor
K = as_tensor([\
    [1.0, 0.0],\
    [0.0, 1.0]\
    ])
K.T
K*F

# as vector
K = Constant((1.,1.))
dot(K, F)

commented Dec 19, 2015 by Chao Zhang FEniCS User (1,180 points)
moved Dec 19, 2015 by Chao Zhang

thanks, chao!

it is not what I am looking for. I want to left multiply a row vector to a matrix. I have to initialize row vector first to achieve that.

But your answer inspired me to do the following. And it works!! The trick is to use as_vector

from dolfin import *
mesh = UnitSquareMesh(2,2)
TFS = TensorFunctionSpace(mesh, 'CG', 1)
F = Function(TFS)

# row vector
N = as_vector([[1. , 1.])

N.shape() # output (1,2)

NF = N*F

NF.shape() # output (1,2)

commented Dec 19, 2015 by truemerlin FEniCS Novice (410 points)

...