Re: [pooma-dev] [RFA/PATCH] FieldEngine FIXMEs anyone?

[Richard Guenther]

On Thu, 6 Nov 2003, John H.Hall wrote:

> Richard:
> I think we kinda got mixed up in the latest version of field and I am
> right in the middle of a proposal about how to correct our design. The
> concept of multi-material fields really should probably have been
> multi-material engines with different views for a multi- and
> single-material engine both pointing to the same data. We are getting
> pretty far along in our sparse mixed cell engine treatment and with
> some small modifications we would be able to interchange between the
> 2.4 multi-material model and our new sparse multi-material model simply
> by choosing new engines.
>
> The confusion you are pointing out is a strong indicator that we didn't
> quite get it right. I will follow up on this after I study it some more
> and discuss it with more of the old pooma team.

Yes, I also think the current design of Fields has some flaws. In the end
all that should distinguish Fields from Arrays is that the former has
meta-data such as
 - a centering
 - multiplicity (aka materials)
 - a mesh
 - relations

So in the end we need sort of a MultiBrick engine that provides a
collection of identical Brick engines and the Field maintaining the
mapping from (material, centering) -> MultiBrick[i].  So basically
FieldEngine should be a _real_ engine and the metadata part of FieldEngine
fold into Field.  This would also allow for multi-component Arrays, if we
want to allow this.

Also of course the Field interface should be a superset of the Array
interface, so we may just inherit from Array and save some code
duplication in evaluators?  Or we can at least allow Array views of Fields
and vice-versa.

Ah and while we're at it - mesh and coordinate system support should be
changed a lot, too.

Maybe we can discuss the needs of different groups some further here. My
main use of Fields is representing a staggered grid, so I use (canonical)
centerings a lot, and also (in future) handling of multi-component fluids
which require multiple materials.

Thanks for your input,

Richard.