ufl team mailing list archive

[Harish Narayanan <harish.mlists@xxxxxxxxx>]

On 2/12/10 10:40 AM, Anders Logg wrote:
> On Fri, Dec 04, 2009 at 11:24:12AM +0000, Garth N. Wells wrote:
>>
>>
>> Marie Rognes wrote:
>>> Harish Narayanan wrote:
>>>> Anders Logg wrote:
>>>>
>>>>> On Fri, Dec 04, 2009 at 11:45:54AM +0100, Anders Logg wrote:
>>>>>
>>>>>> On Fri, Dec 04, 2009 at 08:31:41PM +0100, Marie Rognes wrote:
>>>>>>
>>>>>>> noreply@xxxxxxxxxxxxx wrote:
>>>>>>>
>>>>>>>> ------------------------------------------------------------
>>>>>>>> revno: 860
>>>>>>>> committer: Anders Logg <logg@xxxxxxxxx>
>>>>>>>> branch nick: ufl-dev
>>>>>>>> timestamp: Tue 2009-12-01 18:09:26 +0100
>>>>>>>> message:
>>>>>>>>  Redefine grad().T --> grad()
>>>>>>>>
>>>>>>> Have you updated div() too?
>>>>>>>
>>>>>> No, I forgot that. I'll take a look.
>>>>>>
>>>>>> Are there any other operators that do something strange and should
>>>>>> be fixed?
>>>>>>
>>>>> Strange... UFL seemed to already to compute the correct div:
>>>>>
>>>>>   def div(self, o, a):
>>>>>         i = Index()
>>>>>         g = a[i] if a.rank() == 1 else a[...,i]
>>>>>         return g.dx(i)
>>>>>
>>>>
>>>> This looks correct. Someone who uses curl should look at curl as well.
>>>>
>>>>
>>>
>>>
>>> The curl does what I expect it to do. (And the convention we used in the
>>> div/curl ffc paper.)
>>>
>>
>> The manual should be updated too.
>>
>> Garth
> 
> Should be fixed now.
> 
> I thought that the definition of div of a matrix still needed to be
> fixed (to differentiate over the last axis) but it looks like it was
> correct already.
> 
> Can someone confirm this? Harish?

I too thought that div() needed to be redefined, but I looked at [1]
it's OK.

Harish

[1] ufl/algorithms/transformations.py