dolfin team mailing list archive

[Johan Hake <hake@xxxxxxxxx>]

On Saturday 17 October 2009 14:23:27 Garth N. Wells wrote:
> Johan Hake wrote:
> > On Saturday 17 October 2009 11:48:57 Garth N. Wells wrote:
> >> I'd like to create a sub-class of Expression,
> >>
> >>    class my_expression(Expression):
> >>        def __init__(self, V, x):
> >>            self.x =x
> >>            Expression.__init__(V = V)
> >>
> >> but I'm having trouble with the constructor - I get the below error. Any
> >> tips?
> >
> > The following should work:
> >
> > from dolfin import *
> >
> > class MyExpr(Expression):
> >     def __init__(self, x, V=None):
> >         self.x = x
> >
> >     def eval(self, values, x):
> >         pass
> >
> > m = UnitSquare(2,2)
> > V = FunctionSpace(m,"CG",1)
> >
> > MyExpr(0.0, V=V)
> 
> Thanks. Works fine. I wouldn't have guessed the solution.

No this is a sharp corner, which has bugged me a while. The point is that the 
__init__ method you define is _not_ the real __init__ method, but it is called 
inside the generated __init__ function, which of course is not that intuitive, 
especially not now when we have forced the V/element argument to be a kwarg.

Johan

> Garth
> 
> > This is somewhat inconsistent with how sub classing in Python works. Here
> > you should _not_ call the __init__ method of the base class.
> >
> > The short reason is that this is done for you, while instantiating the
> > dynamically created class.
> >
> > This is something that could be fixed using the suggested new design for
> > CompiledExpression, so one would require an initialization as you suggest
> > above.
> >
> > Johan
> >
> >> Garth
> >>
> >>
> >> line 120, in __init__
> >>      raise TypeError, "Expression need to be initialized using either a
> >> 'cpp.FunctionSpace', using kwarg 'V', or an 'ufl.FiniteElement' using
> >> kwarg 'element'."
> >> TypeError: Expression need to be initialized using either a
> >> 'cpp.FunctionSpace', using kwarg 'V', or an 'ufl.FiniteElement' using
> >> kwarg 'element'.
> >> _______________________________________________
> >> DOLFIN-dev mailing list
> >> DOLFIN-dev@xxxxxxxxxx
> >> http://www.fenics.org/mailman/listinfo/dolfin-dev
>