On 29 April 2010 15:43, Anders Logg <logg@xxxxxxxxx> wrote:
>On Thu, Apr 29, 2010 at 02:21:49PM +0100, Garth N. Wells wrote:
>>
>>
>>On 28/04/10 17:46, Kristian Oelgaard wrote:
>>>
>>>
>>>On 28 April 2010 18:41, Anders Logg <logg@xxxxxxxxx> wrote:
>>>>On Wed, Apr 28, 2010 at 06:33:36PM +0200, Kristian Oelgaard wrote:
>>>>>
>>>>>
>>>>>On 28 April 2010 18:13, Anders Logg <logg@xxxxxxxxx> wrote:
>>>>>>Sounds good, I hope Kristian can give you some instructions on what
>>>>>>you can help out with.
>>>>>
>>>>>Sure, we're still in the process of figuring out exactly how the
>>>>>format should be, but I think the demos might be the easiest place
>>>>>to start. Do we want to recategorize the demos as suggested in the
>>>>>blueprint, or should we just start adding all the demos that we have
>>>>>in DOLFIN and keep the directory structure?
>>>>
>>>>I think some recategorization is necessary, but I don't know which is
>>>>best: to first add them and then move them around of first decide on
>>>>the categories and then add the demos.
>>>
>>>Maybe we don't need to add all of them before we have a better idea
>>>about the categories.
>>>I don't think we'll get it right the first time anyway though.
>>>
>>>>Have you planned which categories should go into the programmer's
>>>>reference? We should probably try to match those categories with the
>>>>demos and link to them.
>>>
>>>No, not yet, but I think the structure in the DOLFIN source tree is
>>>pretty logical.
>>>The question is if we can match everything up with a demo, although it
>>>would be nice if we could.
>>>
>>
>>I'm not so sure that this is possible/sensible. The demos
>>'demonstrate' how to solve various problems and how to use various
>>features. Shouldn't the programmers reference document the
>>interface?
>
>I think there's also room for simple demos that illustrate the use
>of basic classes like Mesh, MeshFunction, input/output, linear
>algebra, parameters etc like we have now, without necessarily solving
>a PDE.
Yes, sure. The example code in the programmer's reference should just cover one function and the immediate usage, but then we can link to a demo that puts the use in a more elaborate context.
Kristian
