ffc team mailing list archive

[Martin Sandve Alnæs <martinal@xxxxxxxxx>]

On Fri, Mar 6, 2009 at 10:23 AM, Kristian Oelgaard
<k.b.oelgaard@xxxxxxxxxx> wrote:
> Quoting FFC <ffc@xxxxxxxxxx>:
>
>> One or more new changesets pushed to the primary ffc repository.
>> A short summary of the last three changesets is included below.
>>
>> changeset:   1372:19f68f217564f18384dd29d30081268e1d583163
>> tag:         tip
>> user:        Anders Logg <logg@xxxxxxxxx>
>> date:        Thu Mar 05 22:19:10 2009 +0100
>> files:
>> ffc/compiler/representation/quadrature/uflquadraturerepresentation.py
>> ffc/compiler/representation/tensor/tensorrepresentation.py
>> ffc/compiler/representation/tensor/ufltensorrepresentation.py
>> ffc/compiler/uflcompiler.py
>> description:
>> Use form instead of form_data in compiler stage 3
>
> This caused me quite a bit of grief.
>
> Try:
> print form
> print form.form_data().form
>
> on any form involving functions. In the first instance the coefficients has
> arbitrary indices, in the second the index corresponds to the correct function
> number in the form.
>
> Maybe I'm the only one not knowing this but it did break the quadrature
> generation.
>
> Kristian

Also, in the second one, derivatives are expanded
and indices renumbered from 0.

Without expanded derivatives, the functions and
basis functions are not correct for all forms.

Without renumbered indices and arguments,
the repr-based signature will often break.

I'm always using form_data().form in the form compiler.

For the record, try this:
  print form.form_data().original_form

Which gives back the untransformed form.

Any reason why you're not using load_forms?

Martin