ufl team mailing list archive

[Martin Sandve Alnæs <martinal@xxxxxxxxx>]

On Fri, Mar 6, 2009 at 2:29 PM, Anders Logg <logg@xxxxxxxxx> wrote:
> On Fri, Mar 06, 2009 at 02:16:06PM +0100, Martin Sandve Alnæs wrote:
>> Hi,
>> say that we have a form with some defined options and some undefined:
>>
>>   a = withopt*dx(0, myoptions) + noopt*dx(0)
>>
>> A natural interpretation of this is that the form compiler
>> uses its default options on noopt, and overrides its defaults
>> with myoptions for withopt.
>
> What does withopt mean for the integrand? Is it possible to attach
> metadata also to integrands (not only measures)?

No. It's just "an expression".


>> Now what if the form compiler gets nondefault global options?
>>
>>   form = jit(a, someoptions)
>>
>> Naturally, someoptions will overrule default options.
>> But should myoptions still overrule someoptions?
>> I'm leaning towards this, but I want your opinion.
>>
>> In other words, is the precedence order
>>   myoptions > someoptions > defaultoptions
>> or
>>   someoptions > myoptions > defaultoptions
>> ?
>
> I think:
>
>  myoptions > someoptions > defaultoptions
>
> Motivation: myoptions allows more fine-grained control over
> someoptions

Agree.

Martin