> Are you happy with these names? Is `P Lambda` the best choice?
> It seems a bit odd with a space in the name but I can't think of
> anything better.
Actually my choice would be
("P-", cell, r, k)
("P", cell, r, k)
("Q-", cell, r, k)
("S", cell, r, k)
The Lambda is always present so conveys no information.
> Not that I'm aware of, but we could easily add a function `simplex`
> which would allow this:
>
> FiniteElement("P Lambda", simplex(n), r, k)
Seems like a good idea to me.
On 03/03/2014 11:54 AM, Anders Logg wrote:
On Mon, Mar 03, 2014 at 09:49:39AM -0600, Douglas N Arnold wrote:
Let me recall that there is one way to specify
elements that is truly consistent and leads to
dimension free code:
(name, cell, r, k)
where name is 'P- Lambda' or 'P Lambda' for cell
= 'interval', 'triangle', or 'tetrahedron',
and name is 'Q- Lambda' or 'S Lambda' for
cell = 'interval', 'quadrilateral', or
'hexahedron'. From one point of view
(the one conveyed by the periodic table), everything
else is a convenience alias.
Are you happy with these names? Is `P Lambda` the best choice?
It seems a bit odd with a space in the name but I can't think of
anything better.
BTW, is there a dimension-independent way
to refer to a simplex or cube, (e.g., (simplex, 2)
instead of triangle) in UFL? If you want code
that changes dimension by just changing the one
line "n = 2", I guess you need this.
Not that I'm aware of, but we could easily add a function `simplex`
which would allow this:
FiniteElement("P Lambda", simplex(n), r, k)
--
Anders
On Mon, Mar 03, 2014 at 02:18:36PM +0100, Marie E. Rognes wrote:
On 03/03/2014 02:08 PM, Anders Logg wrote:
On Mon, Mar 03, 2014 at 01:15:47PM +0100, Marie E. Rognes wrote:
For clarification: will these names be aliases in addition to the
existing ones? If so, I'm fine with either.
I hope that we (after some time) can cleanup some of the existing
aliases.
For example, I would rather have `N1F` instead of `RT` for the div
spaces on tetrahedra.
Well, keeping RT for the div spaces on triangles and tets allows for
dimension-independent code, which I would be strongly in favor of. (Even
if the term may be historically dubious.) I find the N[1|2][e|f] are
simpler and cleaner than the current Nedelec combinations though.
But, as you say, that could be discussed in the future.
_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics
_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics
