Thank you, Roy.
Code runs but sets the Dirichlet b.c to zero, instead to the value I want.
This is what I have done:
class BcFunction : public FunctionBase<Number>
{
virtual Number component (unsigned int i, const Point &p, Real time=0.)
{
if (i == 0) return 2.0;
if (i == 1) return 2.0;
if (i == 2) return 300.0;
if (i == 3) return 300.0;
}
virtual Number operator() (const Point &p, const Real time=0.)
{
libmesh_error();
}
virtual void operator() (const Point &p, const Real time,
DenseVector< Number > &output)
{
libmesh_error();
}
virtual UniquePtr<FunctionBase<Number> > clone() const
{
return UniquePtr<FunctionBase<Number> >(new BcFunction());
}
};
BcFunction func;
DirichletBoundary dirichlet_bc_v(boundary_ids, variables,
func,LOCAL_VARIABLE_ORDER);
system.get_dof_map().add_dirichlet_boundary(dirichlet_bc_v);
Does it look correct? May be I should implement something else in
addition to component()?
Thank you,
Michael.
On 6/15/2017 5:32 PM, Roy Stogner wrote:
On Thu, 15 Jun 2017, Michael Povolotskyi wrote:
I'm fine with creating a child of FunctionBase.
Will it be enough to override only the component() method?
You'll have to override operator() with *something*, even if only
libmesh_error(), since IIRC it's pure virtual. If you do go the
libmesh_error() route your code should work fine for now, but be
forewarned that it may then break with some libMesh release in the
future.
---
Roy
On 6/15/2017 5:08 PM, Roy Stogner wrote:
> > On Thu, 15 Jun 2017, Michael Povolotskyi wrote:
> > > AnalyticFunction<> dirichlet_function_object(zero1);
> > > void zero1(DenseVector<Number> & output,
> > const Point & p,
> > const Real)
> > {
> > output(0) = 0; output(1) = 0;
> > output(2) = 10; output(3) = 10;
> > > > }
> > > > Unfortunately, this does not work. Somehow argument output
in the > > function zero1 has 2 components. Most likely I'm doing
something wrong.
> > You're suffering from the interaction of a couple questionable
design
> decisions, is all. Sorry!
> > AnalyticFunction::component(i) only resizes output to be large
enough
> to fill i, because it doesn't actually know how large an output
vector
> your C function is ready to handle.
> > The internal dirichlet boundary constraint code iterates one
variable
> at a time, and uses component() to query values of just that
variable,
> because it's complicated enough code even without trying to project
> multiple variables at once.
> > > Could you, please, tell me how to achieve what I want
> > Most easily? Just test output.size() and only fill indices
below that
> size.
> > Most efficiently? Subclass FunctionBase yourself and override
> component() too.
> > > or point me to a similar example?
> > I don't think there is one! introduction_ex4 and ex5 use
> AnalyticFunction with DirichletBoundary, but only for a single
> solution variable. All the multi-variable DirichletBoundary
examples
> appear to be using subclassing.
> ---
> Roy
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Libmesh-users mailing list
Libmesh-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libmesh-users
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Libmesh-users mailing list
Libmesh-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libmesh-users
