dolfin team mailing list archive

[Thomas Fraunholz <question133595@xxxxxxxxxxxxxxxxxxxxx>]

Question #133595 on DOLFIN changed:
https://answers.launchpad.net/dolfin/+question/133595

    Status: Answered => Open

Thomas Fraunholz is still having a problem:
Ok. When I use

L = L - dt*(inner(div(grad(w)),
anisotropic_tension(c0)*div(grad(c)))*dx)

and the simplified expression

def anisotropic_tension(c0):
    return cos(c0.dx(1))

it works. But there is still a problem with the arctan I need. If I use
the ufl version, e.g.

def anisotropic_tension(c0):
    return cos(N*atan(c0.dx(1)/c0.dx(0)))

I got a native problem when c0.dx(0) equals zero. Then the solver
returns in every step 'not a number'. So  I'm looking for a possibility
like atan2 in the cmath library or the arctan2 numpy version. A first
try with an if/else expression didn't helped either. But I'm not really
sure if I am allowed to writefor example

def anisotropic_tension(c0):
    if c0.dx(0) > 0.01:
        return cos(N*atan(c0.dx(1)/c0.dx(0)))
    else:
        return cos(N*atan(c0.dx(1)/0.01))

Thanks in advance for your help,

Thomas

-- 
You received this question notification because you are a member of
DOLFIN Team, which is an answer contact for DOLFIN.