On Mar 19, 2008, at 11:25 AM, Michael Waters wrote:
I'm running into a new error however, and was hoping someone could
explain to me what is happening. I am trying to create a simple
solidification model.
The problem comes up as this:
TypeError: unsupported operand type(s) for +: 'binOp' and 'unOp'
I am new to fipy and python so im having trouble conceptualizing
binOp vs. unOp. The error comes up around line 120.
thanks!
I believe that the immediate problem is that
dxi = (A*phase).getGrad()._take((1, 0), axis = 1) * (-1, 1)
returns a VectorCellVariable (because of getGrad()), such that
anisotropySource =(dxi).getDivergence()
returns a FaceVariable. Then, when you do
thetaMag + anisotropySource
you're trying to add a CellVariable to a FaceVariable, which is not
allowed. The "unsupported operand type(s) for +: 'binOp' and 'unOp'"
or "'unOp' and 'unOp'" is generated by Python, and we don't have a
great deal of control over it (but we recognize that it's not very
helpful). A 'unOp' is a unary operator (like sin(x)) and a 'binOp' is
a binary operator (like a + b). That's not what's important; FiPy is
quite capable of adding upOp's and binOp's (they're designed for
that); the problem is that an OperatorVariable can be either a
FaceVariable or a CellVariable, and it's those that are incompatible.
I'll will try to figure out a way to make Python return a more
informative error message, but I'm not optimistic.
I couldn't even swear that VectorCellVariable.getDivergence() is even
doing the right thing (we don't test it, and I don't believe we ever
use it). It's not very useful, even if it does work. If we don't get
rid of it, we should probably make it return a CellVariable.
phi.getGrad().getDivergence() would be a 1st order Laplacian, but at
least it you'd be able to use it where probably everybody would
expect to; I don't know what you'd want a divergence as a
FaceVariable for (and I think I wrote it), but I'll see if Daniel has
any thoughts on keeping the status quo.
Instead, if you do this
dxi = (A*phase).getFaceGrad()._take((1, 0), axis = 1) * (-1, 1)
you'll get a VectorFaceVariable, and then
anisotropySource =(dxi).getDivergence()
will return a CellVariable (desirably 2nd order), which is a proper
source to use in your equation.
