Restriction to vector valued shape functions of size <= 3

[bzfschie]

Trying to implement a mixed formulation for a shell I have as fields the displacement, which is vector valued (from a domain in \R² into \R³) and the gradient of the displacement (from a domain in \R² into \R^{3 x 2]). To create shape-functions for the latter field either requires to define tensor valued shape functions (order 2, dimensions 2 and 3) or to define a vector valued shape function in \R⁶. The second option seems straightforward, but is currently not possible, because Tensors.jl does not support tensors with dimensions larger than 3.
Is theré a way to circumvent this problem, or might it be possible to extend the functionality of Ferrite to such cases? Many thanks in advance for your consideration.

[KnutAM]

This is already possible - we call this embedded in the docs.

You can do CellValues(qr, ip^3, ipg^2). The outputs will soon become MixedTensor (see #1286), but are currently SArrays.

[bzfschie]

... but how does the DofHandler know, how many degrees of freedom there are? It only has Interpolations, but no CellValues...