Ok.
I can push this, but we also need a test. What with:
def testCoefficient():
v = Constant(triangle)
assert round(expand_derivatives(diff(v,v))-1.0, 7) == 0
in test_diff.py
It fails with present master but not with the suggested fix.
Johan
On Tue, Oct 14, 2014 at 11:07 AM, Martin Sandve Alnæs <[email protected]>
wrote:
> Yes.
>
> On 14 October 2014 10:59, Johan Hake <[email protected]> wrote:
>
>> Yes, it looks like you are correct Miroslav.
>>
>> I suppose the line you link to should add a self check too. Changing that
>> line to
>>
>> if f._ufl_is_terminal_ and f != v:
>>
>> fixes the problem.
>>
>> Is this fix something we just can add Martin?
>>
>> Johan
>>
>> On Tue, Oct 14, 2014 at 10:14 AM, Miroslav Kuchta <[email protected]>
>> wrote:
>>
>>> Hi, this seems to be due the simplification made here
>>> <https://github.com/FEniCS/ufl/blob/master/ufl/differentiation.py#L87>
>>>
>>> from dolfin import *
>>>
>>> mesh = UnitSquareMesh(10, 10)
>>> V = FunctionSpace(mesh, "CG", 1)
>>> u = Function(V)
>>>
>>> # 0
>>> print u._ufl_is_terminal_
>>> print diff(u, u)
>>>
>>> # 1
>>> print (Constant(1)*u)._ufl_is_terminal_
>>> plot(diff(Constant(1)*u, u), mesh=mesh)
>>>
>>> # 1
>>> u._ufl_is_terminal_ = False
>>> plot(diff(u, u), mesh=mesh)
>>> interactive()
>>>
>>> Regards, Miro
>>>
>>> On 10/13/2014 10:32 PM, Johan Hake wrote:
>>>
>>> from dolfin import *
>>> mesh = UnitSquareMesh(10, 10)
>>> V = FunctionSpace(mesh, "CG", 1)
>>> u = Function(V)
>>> print diff(u, u)
>>>
>>>
>>>
>>
>> _______________________________________________
>> fenics mailing list
>> [email protected]
>> http://fenicsproject.org/mailman/listinfo/fenics
>>
>>
>
_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics
