I see two ways that might work, but both are a bit annoying: - Wrap the points that are required into a (Pseudo)Quadrature. As these points are different for each cell, this requires a new FEValues object for each cell. A variation might be to use the hp variant of FEValues and use a collection of quadratures, one for each cell. This would allow to at least recycle the FEValues object.
That's the canonical way of doing it.
- Calculate the Jacobian on my own from the information I have available (probably Mapping::support_point_value is everything I need). But I do not want to reinvent the wheel, Mapping can already calculate that.
That might work. The Mapping classes can definitely compute all of this information, but their real purpose is to interact with the FEValues class and not be something you'd use from user programs. That doesn't mean that you can't, but the interface isn't made for that, it's strictly made for use by FEValues.
So, if you can find a way to work with just the Mapping, then that would avoid the overhead of the FEValues class, otherwise I'd go with your first suggestion.
Best W. _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
