fenics team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Thu, Jun 17, 2010 at 11:29:28AM +0200, Kristian Oelgaard wrote:
> On 16 June 2010 22:09, Anders Logg <logg@xxxxxxxxx> wrote:
> > Looks good. I noticed a couple of things:
> >
> > 1. It would be good if class names were in another font (like courier
> > style), \texttt{VariationalProblem} etc
>
> Once the Programmer's reference is up and running, I think we should
> just use links to the class documentation (and member functions).

ok, sounds good.

> > 2. Would it be better with another variable name instead of h since
> > that will be used as mesh size in other demos. Why not g?
>
> I'll change h --> g for all Neumann BCs.

ok.

> > 3. Is the normal flux + or - grad(u).n? I think of it as -grad(u).n.
>
> The equation is:
>
> - \Delta u = f
>
> multiplying by v:
>
> - v \Delta u = vf
>
> integration by parts and switching to UFL:
>
> inner( grad(v), grad(u))*dx  - v*inner(grad(u), n)*ds = v*f*dx
>
> if we use g = - inner(grad(u), n) then
>
> inner( grad(v), grad(u))*dx  + v*g*ds = v*f*dx
>
> so
>
> a = inner( grad(v), grad(u))*dx
> L = v*f*dx - v*g*ds
>
> then we need to use g = sin(5x) for the demo to give the same solution.
>
> Of course we can move the '-' from L to g (which is what we have now)
> but then the form does not follow directly from our equations.

Perhaps g = grad u . n (with + sign) is better now when I think of
it. It looks better in two ways: we avoid the "-" sign in both the
definition of g and the definition of the form. So it should be

  g = sin(5x)
  L = v*f*dx + v*g*ds

Then one just needs to remember that g is the negative flux (if one
thinks of -grad u . n as the flux, but not everyone does).

--
Anders


> > 4. Typo in "The implemented in a single Python file"
>
> Fixed.
>
> > 5. g is used in (f, g) instead of h (but maybe h should be changed to g)
>
> yes.
>
> > The documentation seems to be coming along pretty well. It may be time
> > for you to ask people to contribute specific parts. For example,
> > Johannes and Harish could contribute with some documentation of how to
> > get Debian/Ubuntu packages and building with Dorsal. Others can
> > contribute with writing specific pages for the programmer's
> > reference, and Hans Petter can contribute with preparing his tutorial
> > in reST (and split it in Python and C++ versions).
>
> I need to figure out the framework for the Programmer's reference and
> write the styleguide for contributing, but the Tutorial and
> Installation parts can be written now.
>
> Kristian
>
> >
> >
> > On Wed, Jun 16, 2010 at 06:31:42PM +0200, Kristian Oelgaard wrote:
> >> Hello,
> >>
> >> I've finished a first iteration on documenting the Poisson demo (from
> >> dolfin/demo/pde/poisson).
> >>
> >> Check it out and let me know what you think.
> >>
> >> The Python version is here:
> >>
> >>  http://www.fenics.org/newdoc/demos/python/pde/poisson/poisson.html
> >>
> >> and the C++ version is here:
> >>
> >> http://www.fenics.org/newdoc/demos/cpp/pde/poisson/poisson.html
> >>
> >> Note that I need to finish the Programmer's reference part of the
> >> documentation before I can insert links to the relevant classes used
> >> in the demos.
> >>
> >> http://www.fenics.org/newdoc/programmers-reference/index.html
> >>
> >> Kristian
> >>
> >> _______________________________________________
> >> Mailing list: https://launchpad.net/~fenics
> >> Post to     : fenics@xxxxxxxxxxxxxxxxxxx
> >> Unsubscribe : https://launchpad.net/~fenics
> >> More help   : https://help.launchpad.net/ListHelp
> >
> > -----BEGIN PGP SIGNATURE-----
> > Version: GnuPG v1.4.10 (GNU/Linux)
> >
> > iEYEARECAAYFAkwZL2IACgkQTuwUCDsYZdHZ5QCfa7fC+zyjjYtWG2HncJVdtl/4
> > UskAn0LZx1DNof5lIRs9/+TzPhNbaErU
> > =esYb
> > -----END PGP SIGNATURE-----
> >
> >

Attachment: signature.asc
Description: Digital signature