Fix wrong range and jacobian type in  hermitebasistest

Closes #88

See merge request !492

Feature/transformed finite element mixin

See merge request !490

There are many Hermite-type elements. The one implemented here
is commonly known as cubic Hermite element.

This commit is not ment to be merged but should
be removed once the test is in. It only serves
to disable it in a minimal invasive way such that
the CI can be run.

Maik Porrmann authored Oct 02, 2024 and Carsten Gräser committed Oct 10, 2024

Layout cleanup:

* Unify indentation.
* Unify placement of opening/closing braces.
* Unify placement of `static_assert`s and add missing conditions.
* Surround each method by empty lines.
* Put initializer and base class specifier in new line.
* Fix a few typos.
* For my inner Monk: Unify indentation of multiline doxygen comments.

Other cleanup:

* Cleanup includes.
* Drop `coeffsize` and use `constexpr` function instead of nested ternary operators.
* Rename `MorleyLocalBasis` to `MorleyReferenceLocalBasis`.

Co-authored-by: Carsten Gräser <graeser@math.fau.de>

Maik Porrmann authored Sep 30, 2024 and Carsten Gräser committed Oct 10, 2024

Layout cleanup:

* Unify indentation.
* Unify placement of opening/closing braces.
* Unify placement of `static_assert`s and add missing conditions.
* Surround each method by empty lines.
* Put initializer and base class specifier in new line.
* Fix a few typos.
* Order some `using` statements.
* For my inner Monk: Unify indentation of multiline doxygen comments.

Other cleanup:

* Remove redundant members.
* Cleanup includes.
* Drop `coeffsize` and use `constexpr` function instead of nester ternary operators.
* Rename `HermiteLocalBasis` to `HermiteReferenceLocalBasis`.
* Improve documentation.
* Extract generic mapper-based utilities into separate header:
  * A simple range over all subindices associated to an element within a `MCMGMapper`.
  * A function computing average mesh sizes for all DOFs addressed by a `MCMGMapper`.
* Cleanup internal members:
  * Incorporate `LFE::fillMatrix()` into `LFE::bind()` to avoid useless indirection.
  * Incorporate `PreBasis::updateState()` into `PreBasis::update()` to avoid useless indirection.
  * Use meaningfull names for member variables instead of generic ones like `localState` and `data`.
  * Use `std::array<T,n>` instead of `std::vector<T>` if `n` is a small compile time constant.

Interface improvements:

* Add global basis.

Further changes:

* Test on a non-uniform grid.
* Fix a bug (and memory leak) when binding to an element.
* Order includes lexicographically.

Co-authored-by: Carsten Gräser <graeser@math.fau.de>

* Extend documentation.
* Move non-working 'example' into doxygen documenation.
* Unify coding style.
* Remove redundant calls to `resize()`.

Co-authored-by: Maik Porrmann <maik.porrmann@tu-dresden.de>

This is needed to specify the nature of the DOFs
of a Hermite-basis in order to identify those
that need to be constrained for a certain boundary
condition.

Maik Porrmann authored Sep 26, 2024 and Carsten Gräser committed Oct 06, 2024

This implements the global basis for the
nonconforming Morley-element. The implementation
is based on `TransformedFiniteElementMixin`.
This only provides the classical quadratic
element in 2d and no higher-order or 3d extensions.

Notice that the element is H^1- and H^2-nonconforming.

Maik Porrmann authored Sep 26, 2024 and Carsten Gräser committed Oct 06, 2024

This uses the `TransformedFiniteElementMixin` helper
class to imeplement the crucial local transformation
needed to ensure global continuity of derivatives.
The implementation provides:

* The 1d cubic Hermite basis (aka cubic C1 splines)
* The 2d cubic Hermite basis (aka Hermite triangle)
* A version of the 2d cubic Hermite basis where
  the cubic element bubble is eliminated. This
  is known as 'reduced Hermite triangle'
  and provides one part of the mixed DKT-element.
* The 3d cubic Hermite basis

Notice that, except for the 1d case, these FE-spaces
are one H^1 but not H^2 conforming.

Maik Porrmann authored Sep 26, 2024 and Carsten Gräser committed Oct 06, 2024

This CRTP-mixin class implements the `localBasis()`
part of a `LocalFiniteElement` using a transformation
of another basis. The implementation class is required
to implement member functions `referenceLocalBasis()`
and `transform()`. The former allows to define the
local ansatz-space in terms of _some_ local basis
based on the corresponding interface in dune-localfunctions.
The target basis is then given by a set of linear
combinations of those reference basis functions.
The `transform()` methods implements the linear
transformation indiced by those linear combinations.

Since the same transformation is used for function
values as well as any order of derivative, this
__cannot__ be used to implement range transformation
as e.g. Piola-transformations.

[doc][cleanup] Fix typo

See merge request !493

Feature/add differentiability checks

See merge request !480

Maik Porrmann authored Sep 27, 2024 and Carsten Gräser committed Sep 27, 2024

This adds optional checks for continuity at vertices and continuity of normal derivatives at edge midpoints. The latter is implemented by means of a general framework for differentiability checks.

Add new function class MonomialSet<K,dim,n>

See merge request !470

Maik Porrmann authored Jun 18, 2024 and Carsten Gräser committed Sep 25, 2024

Co-authored-by: Simon Praetorius <simon.praetorius@tu-dresden.de>

This implements all `dim`-variate monomials up to order `n`
as components of a vector-valued differentiable functions.
Currently this is only specialized for order `dim=1` and `dim=2`.

The new `MonomialSet` is used and tested in `CubicHermiteBasis`.

Update versions release 2 10

See merge request !488

Use Hybrid::equal_to instead of Hybrid::equals

See merge request !486

Add container descriptor for blocked-interleaved index-merging strategy

See merge request !467

Update COPYING

See merge request !484

Add SPDX license information

See merge request !475

[examples][doc] Fix sign of pressure in Stokes example

See merge request !482

This has been changed to the physically correct version in the
example code, too.

While the exact sign is not important mathematically,
the example used the physically wrong sign. Thus this
is also uncommon in literature. The physically correct form
of the Stokes equation is `-\Delta u + \grad p = f` which,
after partial integration leads to a `-div(v)p` term in the
weak form.

This issues exists since the example was added. Maybe the
initial version of the example was implemented based on the
book by Braess. There, indeed the uncommon sign is used for
the Stokes as well as for the Navier-Stokes equation.

No surprisinly, the uncommon convention leads to confusing
results. E.g. for the lid driven cavity flow, the pressure
decreases towards the corner where the fluid is driven to.
Similarly, for Poiseuille flow in a pipe the fluid is driven
towards the high pressure side.