dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Thu, Nov 26, 2009 at 02:20:56PM +0100, Anders Logg wrote:
> We are currently able to automatically detect the value shape of an
> Expression defined like this:
>
>   f = Expression(("x[0]", "sin(x[1])"))
>
> But not if it is defined like this:
>
>   class MyExpression(Expression):
>       def eval(self, values, x):
>           values[0] = x[0]
>           values[1] = sin(x[1])
>
> How should the value shape be specified for a subclass of an
> expression?
>
> We had value_rank, value_dim before. We might either require
>
>   def value_shape(self):
>       return (2,)
>
> or sending value shape to the constructor:
>
>   f = MyExpression(shape=(2,))
>
> Or there might be some other trick we can do.

I have added a dim() function to the interface of subclassed
Expressions.

This can return either an integer or a tuple:

 1      <--> scalar
 2      <--> vector of dimension 2
 3      <--> vector of dimension 3
 (1,)   <--> scalar
 (2,)   <--> vector of dimension 2
 (2, 2) <--> matrix of dimension 2 x 2

It should work now but the elasticity demo is failing because FFC is
not correctly adjusting the cells for mixed elements (I think).

--
Anders

Attachment: signature.asc
Description: Digital signature