← Back to team overview

dolfin team mailing list archive

  1. dolfin team
  2. Mailing list archive
  3. Message #24006

Re: Adding form names F and J

  Thread Previous • Date Previous • • Thread Next 
On 07/01/11 13:13, Anders Logg wrote:

On Fri, Jul 01, 2011 at 11:07:15AM +0200, Martin Sandve Alnæs wrote:

On 1 July 2011 09:05, Anders Logg<logg@xxxxxxxxx>  wrote:

They are always generated as it is now, but I assume it would be
simple to make it so that they are only generated if they are used.

I'm not sure there's a big chance they will be misunderstood. Users
will not know that the classes ResidualForm and JacobianForm are
defined if they don't read through the generated code. Users will only
know about them if they look at a nonlinear example and then it will
be clear how they should be used.

Users rarely follow the train of thought they're supposed to :)

I don't have a strong opinion either, do as you like.

ok. :-)
Always interpreting F and J as a linear and bilinear form without adding more checks is slightly error-prone cf the failing FFC buildbot. 

--
Marie


--
Anders



Martin


But I don't have a firm opinion on this other than it's a natural
thing to have so that one can work with F and J for nonlinear problems
instead of a and L.



On Fri, Jul 01, 2011 at 08:31:13AM +0200, Martin Sandve Alnæs wrote:

Are the typedefs only generated if the forms are named F and J? My worry is
that it could easily be misunderstood and misused.
Martin
Den 30. juni 2011 21.50 skrev&quot;Anders Logg&quot;<logg@xxxxxxxxx>
følgende:

On Thu, Jun 30, 2011 at 06:35:19PM +0200, Anders Logg wrote:

Could we add the additional form names&quot;F&quot; and&quot;J&quot; to

UFL to be used

for nonlinear problems? These would then be mapped (typedeffed) to
&quot;ResidualForm&quot; and&quot;JacobianForm&quot; by the DOLFIN C++

wrappers.

Then one can use F and J in the .ufl file and then initialize the
forms in C++ by

NonlinearPoisson::ResidualForm F(V);
NonlinearPoisson::JacobianForm J(V, V);

instead of using L and a and then initializing by

NonlinearPoisson::LinearForm F(V);
NonlinearPoisson::BilinearForm J(V, V);

Any objections?

I didn't hear any objections and it sounded like a good idea so I have
added this now (will merge into main when the buildbot is green).

In short, I've made a small change to UFL to allow 'F' and 'J' to be
used in form files and a small change in the DOLFIN wrapper code
generators to generate typedef aliases:

BilinearForm = JacobianForm
LinearForm = ResidualForm

So nothing should break for old code. One can still use BilinearForm
and LinearForm for nonlinear problems, but no in addition the more
logical ResidualForm and JacobianForm.


_______________________________________________
Mailing list: https://launchpad.net/~dolfin
Post to : dolfin@xxxxxxxxxxxxxxxxxxx
Unsubscribe : https://launchpad.net/~dolfin
More help : https://help.launchpad.net/ListHelp

_______________________________________________
Mailing list: https://launchpad.net/~dolfin
Post to     : dolfin@xxxxxxxxxxxxxxxxxxx
Unsubscribe : https://launchpad.net/~dolfin
More help   : https://help.launchpad.net/ListHelp

Follow ups

References

  Thread Previous • Date Previous • • Thread Next