No opinions on the topic below?

I would like to hear how to change syntax of fields and variables in short syntax.

The current way is:

fields = { 'displacement': ('real', 'vector', 'Omega', 2), # 'vector' or 3 in 3D 'pressure' : ('real', 'scalar', 'Omega', 1), # 'scalar' or 1 }

variables = { 'u' : ('unknown field', 'displacement', 0), 'v' : ('test field', 'displacement', 'u'), 'p' : ('unknown field', 'pressure', 1), 'q' : ('test field', 'pressure', 'p'), }

I can imagine, for example (fields are easy...):

fields = { 'displacement': ('real', 'Omega', 2), 'pressure' : ('real', , 'Omega', 1), }

variables = { 'u' : ('unknown field', 'displacement', 3, 0), 'v' : ('test field', 'displacement', 'vector', 'u'), 'p' : ('unknown field', 'pressure', 'scalar', 1), 'q' : ('test field', 'pressure', 'scalar', 'p'), }

In long syntax, the field would be called 'n_components' or 'shape' or?

r.

On 12/13/10 19:27, Robert Cimrman wrote:

Hi,

I am thinking about decoupling the shape of a Field from the number of components of a Variable.

The current state of affairs is that the field shape defines the components of related variables, but in reality the FE base is always scalar, applied to each component of a vector variable separately. This is in contrast with truely vector FE spaces like the edge elements used for the Maxwell equations.

To fix that I propose:

1. For scalar FE approximations the field shape will ve 1 - this is for all examples we have.

2. Variable syntax will be changed to allow for the number of components.

Apart from the clean-up reasons, it will be IMHO easier to implement the edge elements then.

Opinions/ideas?

r.