v0.9.0

This is a major release of Basix and users are advised to upgrade.

What's Changed

Features and API changes

• Expose subentity types to python for easy creation of elements on entities. by @jorgensd in #847
• dtype when creating elements by @mscroggs in #796
• dtype of enriched element by @mscroggs in #815
• Test arm64 Mac wheels by @jhale in #850
• Windows build by @jhale in #819
• Build Windows binary wheels and update wheel builder by @jhale in #849
• Remove ufcx_element_type by @mscroggs in #818
• Allow user to create symmetric quadrature element by @mscroggs in #821
• Use nanobind 2.0 typing stubgen by @garth-wells in #829
• Switch back to C++20 and use modern set_compile_features to set minimum C++ standard by @spraetor in #844
• Make degree 1 and 2 Lagrange on a pyramid not require a variant by @mscroggs in #834
• Quadrature rule on pyramids by @mscroggs in #835
• Add basix_hash method to get C++ hash of Basix elements by @mscroggs in #820
• Add permute_subentity_closure function by @mscroggs in #806
• DOF transformations: replace pull back of inverse with push forward by @mscroggs in #836

Other changes

• Remove version updater. by @jhale in #810
• Kebab case build-wheels.yml by @jhale in #813
• Remove commented code in CMake file by @garth-wells in #814
• Bump version. by @jhale in #816
• Remove caching of ufl elements by @mscroggs in #817
• Break system packages in DOLFINx integration test by @jhale in #824
• Windows: make runtime dependency install optional by @minrk in #822
• Mark all nanobind enums as nb::is_arithmetic() for compatibility with nanobind 2.x by @francesco-ballarin in #825
• Make runtime dependency install OFF by default by @jhale in #826
• DOLFINX CI updates by @garth-wells in #827
• Tweaks related to Windows DOLFINx build by @jhale in #830
• Enum and type updates by @garth-wells in #832
• Disable windows tests by @jhale in #833
• Update to std::ranges usage by @schnellerhase in #837
• Update mdspan by @schnellerhase in #840
• C++23 support by @schnellerhase in #838 (reverted)
• Update build-wheels.yml by @jhale in #843
• Re-enable Windows CI by @jhale in #845
• Correct cellname for point/vertex type for UFL by @jorgensd in #851
• Fix return type of sobolev-space in quadrature element by @jorgensd in #853

New Contributors

• @minrk made their first contribution in #822
• @schnellerhase made their first contribution in #837
• @spraetor made their first contribution in #844

Full Changelog: v0.8.0...v0.9.0

v0.8.0

What's Changed

• Switch from pybind11 to nanobind by @chrisrichardson in #676
• Update version to 0.8.0 by @jhale in #710
• use blocked element when creating a quadrature element with value shape by @mscroggs in #711
• Add Spack test to CI by @garth-wells in #712
• Update Spack test by @garth-wells in #713
• Update UFL element by @mscroggs in #703
• reset branches by @mscroggs in #714
• Remove setup.py by @garth-wells in #715
• Use scikit-build-core to handle minimums. by @jhale in #716
• Add CI workflow trigger by @garth-wells in #723
• Improve nanobind bindings and memory management by @garth-wells in #724
• Use absolute imports in place of relative imports by @garth-wells in #727
• Allow Spack ci to test specific Basix version by @garth-wells in #728
• Update install docs by @garth-wells in #720
• Fix hash and eq for basix.ufl elements by @conpierce8 in #718
• set gdim equal to tdim by default by @mscroggs in #729
• rename highest_degree and highest_complete_degree by @mscroggs in #731
• reset branches by @mscroggs in #732
• Throw error when trying to create BDM degree 0 by @mscroggs in #730
• implement embedded degrees for quadrature element by @mscroggs in #733
• add basix_map_type method by @mscroggs in #734
• remove unused functions by @mscroggs in #736
• Add name property to all enums by @chrisrichardson in #735
• Update wheel build action. by @jhale in #709
• Enable merge queue by @garth-wells in #737
• Update OMPI environment variables in DOLFINx integration for openmpi v5 by @francesco-ballarin in #739
• Rename dof transformation functions by @mscroggs in #738
• Reset branch name by @mscroggs in #740
• Add a pip install --no-build-isolation test by @garth-wells in #741
• Fix scikit-build-core version in pyproject.toml by @garth-wells in #745
• Use OpenBLAS in Redhat clone CI by @garth-wells in #747
• Remove Sphinx Makefile and unpin Sphinx by @jhale in #744
• More Python versions in wheels by @jhale in #742
• Update version updater script v2 by @jhale in #750
• Do not store .doctrees binary files on website by @francesco-ballarin in #751
• Fix Spack CI by @garth-wells in #753
• Update make_html for new website template by @mscroggs in #754
• Use lesstall header in web docs by @mscroggs in #755
• Use jekyll to make website header instead of hacky python code by @mscroggs in #756
• Add basix_element property to UFL element by @mscroggs in #757
• Reset branch names by @mscroggs in #758
• Add overflow: scroll to stop sidebar colliding with content in web docs by @mscroggs in #761
• Add support for float32 and float64 elements from Python by @garth-wells in #759
• Property fix by @garth-wells in #762
• Add dtype property to FiniteElement by @garth-wells in #763
• Fix equality test for different float type elements by @garth-wells in #764
• Improve type hints by @garth-wells in #765
• Fix element hash and improve types by @garth-wells in #766
• Check that dof ordering is the same in element equality check by @mscroggs in #769
• Wrap functionality in Python with nicer typing and docs by @mscroggs in #767
• Remove vtk element by @mscroggs in #770
• Rename keyword argument from lvariant to lagrange_variant by @mscroggs in #771
• Drop [email protected] in favor of [email protected] by @francesco-ballarin in #773
• Sonar CI update by @garth-wells in #775
• Switch to ruff linting by @jhale in #774
• Bump actions/{download-artifact,upload-artifact} to v4 by @francesco-ballarin in #777
• Remove packages from pyproject by @jhale in #743
• Add test_elements.py by @mscroggs in #778
• add wrap_element function to Basix.UFL by @mscroggs in #776
• Add Python 3.12 to CI by @garth-wells in #779
• Add create_tp_element function by @mscroggs in #780
• Reset branch names by @mscroggs in #781
• Use black formatting (via ruff) by @garth-wells in #783
• Improve debug build install instructions. by @jhale in #784
• Change to the Modified Gram Schmidt algorithm in orthogonalise (#679) by @ampdes in #785
• Fix Python version in python/pyproject.toml file by @garth-wells in #786
• Remove gdim input to UFL elements by @mscroggs in #772
• Reset CI branches by @mscroggs in #788
• Improve docstrings and add copyright statements by @garth-wells in #787
• Fix sonar issues by @garth-wells in #789
• Correct polyset order in demo by @mscroggs in #790
• Ruff fixes by @garth-wells in #791
• Update mdspan.hpp by @garth-wells in #792
• Reset DOLFINx CI branch by @garth-wells in #793
• Add a hash functions to finite elements by @mscroggs in #782
• Update cibuildwheel action by @jhale in #798
• Update nb::ndarray constructor in preparation of next nanobind release by @francesco-ballarin in #799
• Revise transformation names and add docs by @garth-wells in #802
• reset dolfinx branch by @mscroggs in #805
• replace lru_cache with cache by @mscroggs in #803
• DOF transformation name updates by @garth-wells in #807
• Small Doxygen config updates by @garth-wells in #808
• Revert CI to DOLFINx main by @garth-wells in #809

New Contributors

• @conpierce8 made their first contribution in #718
• @ampdes made their first contribution in #785

Full Changelog: v0.7.0...v0.8.0

v0.7.0

This version of Basix was released alongside UFL 2023.2.0 and FFCx/DOLFINx v0.7.0 tags.

User facing changes

Iso elements

• Add iso macroelement on an interval by @mscroggs in #680
• Implement iso element on quadrilateral by @mscroggs in #683
• Add iso element on hexahedron by @mscroggs in #684
• Add iso element on a tetrahedron by @mscroggs in #688

DOF ordering

• Allow permuted dof layout for Lagrange elements by @chrisrichardson in #654
• Use dof ordering when generating tensor factors by @mscroggs in #695

UFL wrapper

• Fix degree method in UFL wrapper layer for mixed elements by @garth-wells in #636
• Fix symmetric tensors by @mscroggs in #647
• Rename ufl_wrapper -> ufl, and simplify functions in UFL wrapper by @mscroggs in #655
• Fix blocked element tables by @michalhabera in #662
• Move QuadratureElement and RealElement from ffcx to Basix by @mscroggs in #682
• Correct tabulate for blocked elements by @mscroggs in #698
• Add better warning when using * by @mscroggs in #701
• Remove rank input to element and blocked_element by @mscroggs in #708

Installation

• Test scikit-build-core and pyproject.toml for top-level install by @chrisrichardson in #673
• Enable Python 3.11 by @mscroggs in #690

Allowing multiple scalar types

• Template polyset over float type by @garth-wells in #660
• Support float and double in low-level functions by @garth-wells in #663
• Template basix::FiniteElement over scalar type by @garth-wells in #664

Performance improvements

• Input orthogonal wcoeffs rather than running orthogonalise on input by @mscroggs in #693

Other changes

• Update mdspan.hpp by @garth-wells in #632
• Run more tests on main by @jhale in #630
• Dev version bumps by @garth-wells in #635
• Fix typo in basix.sobolev_spaces module name by @francesco-ballarin in #638
• Do not run apt-get upgrade in CI by @garth-wells in #639
• Simplify constructors by @chrisrichardson in #641
• Fix Sobolev space for discontinuous elements by @garth-wells in #642
• Remove deprecated funcs by @chrisrichardson in #643
• Update python create element by @chrisrichardson in #644
• Simplify transformations by @chrisrichardson in #646
• Tweak UFL import by @mscroggs in #650
• Fix bug in N2E on tets: cells are 3D not 2D by @mscroggs in #653
• Cast scalar types for compatibility by @garth-wells in #656
• Reset branches in CI by @mscroggs in #657
• Use np.tile for blocked tabulation by @michalhabera in #659
• update docs and doc generating script by @mscroggs in #665
• Update mdspan by @garth-wells in #666
• Relax some floating point comparisons by @garth-wells in #670
• Reset FFCx and DOLFINx branches by @mscroggs in #681
• Seed random test by @mscroggs in #652
• Fix #685 by @mscroggs in #686
• Implement degree 0 to 2 macro polynomials on a triangle by @mscroggs in #687
• Fix compilation with gcc-13 (#678) by @rmsc in #691
• Consistency updates to Python docs by @garth-wells in #694
• mdspan update by @garth-wells in #699
• Reset DOLFINx branch to main by @garth-wells in #700
• Remove IrreducibleInt by @mscroggs in #702
• use gdim in default shape for an element with a rank by @mscroggs in #707

New Contributors

• @francesco-ballarin made their first contribution in #638
• @rmsc made their first contribution in #691

Full Changelog: v0.6.0...v0.7.0

v0.6.0

What's Changed

• Update version numbers to v0.5.2.dev0 by @jhale in #593
• Add Sobolev spaces by @mscroggs in #589
• reset branches to main by @mscroggs in #595
• Fix Sobolev space conversion by @mscroggs in #596
• Make "CG" work and use identity mapping for mixed elements whose subelements all use identity by @mscroggs in #597
• Add SonarCloud analysis by @jhale in #601
• use gdim as default for VectorElement (not tdim) by @mscroggs in #605
• Add deprecation warning when using "CG" by @mscroggs in #599
• Tidy up CI and packaging by @jhale in #609
• Don't run SonarCloud on external PRs. by @jhale in #613
• Try creating ARM MacOS wheels by @jhale in #390
• Bump version number due to Sobolev space API change. by @jhale in #614
• Remove unnecessary packages from pyproject by @jhale in #615
• Optimize num_sub_entities function by @tiagovla in #616
• Fix mypy (typing.Optional) by @mscroggs in #618
• Pass manifold cell information to base element of vector elements by @jorgensd in #617
• reenable mypy checks for Python 3.10 by @mscroggs in #619
• Fix CI compilers by @garth-wells in #622
• Run test in serial by @garth-wells in #623
• Include tensor element symmetry in its hash by @TTitscher in #620
• Add 'basix' string to element hash by @garth-wells in #624
• Test with Intel oneAPI and Intel Python. by @garth-wells in #625
• Fix OneAPI tests by @jhale in #629
• Bump cmake by @jorgensd in #627
• Fix oneAPI tests by @garth-wells in #631

New Contributors

• @tiagovla made their first contribution in #616
• @TTitscher made their first contribution in #620

Full Changelog: v0.5.1...v0.6.0

v0.5.1

Remove Python dependency from the CMake build for the C++ library

v0.5.0.post0

Users should prefer the tag v0.5.1.

This post-release fixed some errors in the Basix wheel build process.

#586

The core Basix library is unaffected and users can use v0.5.0 or v0.5.0.post0 tags interchangeably.

Full Changelog: v0.5.0...v0.5.0.post0

v0.5.0

Users should prefer the tag v0.5.1.

What's Changed

• Fix PyPi wheel upload by @jhale in #482
• Update basix to 0.4.2.dev0 by @jhale in #487
• remove numpy version requirement by @mscroggs in #490
• Do not include cblas.h by @garth-wells in #489
• Turn on mypy checks by @mscroggs in #491
• Add CPATH to get wheel build working on RedHat by @jhale in #488
• Remove xarray completely by @IgorBaratta in #494
• Put true in wrong place in wheel build, fix. by @jhale in #497
• Add workflow trigger to test dependents against specific versions by @garth-wells in #496
• Allow UFL and FFCx version to be set when triggering FFCx integration test by @garth-wells in #499
• FFCx CI fix(handle default case) by @garth-wells in #500
• Simplify Intel/MKL CI by @garth-wells in #498
• Allow specification of version in DOLFINx integration test by @garth-wells in #501
• Bump version to 0.4.3 by @jhale in #502
• Test version tweaks by @jhale in #503
• Orthogonalise polynomial sets by @mscroggs in #495
• Install fixes for new scikit-build by @garth-wells in #510
• remove "the" from JOSS paper by @mscroggs in #505
• remove extra {}s in reference in JOSS paper by @mscroggs in #511
• Add CI to test cmake install by @mscroggs in #509
• Transpose polysets by @mscroggs in #512
• Use mdspan in polyset by @chrisrichardson in #515
• Add Hermite elements by @mscroggs in #516
• remove ; by @mscroggs in #521
• add CITATION.cff by @mscroggs in #522
• Build D to the be the correct shape rather than transposing by @mscroggs in #520
• Document shapes of xtensors by @mscroggs in #518
• fix typos in custom element demo by @mscroggs in #523
• set language to "en" by @mscroggs in #529
• Correct typing: numpy.ndarray -> numpy.typing.NDArray by @mscroggs in #527
• Remove logging by @mscroggs in #528
• Transpose polynomial::tabulate by @mscroggs in #524
• Add Bernstein variant of Lagrange for simplices by @mscroggs in #533
• run C++ demos on CI by @mscroggs in #531
• Automatically generate docs.h by @mscroggs in #530
• Update mdspan.hpp by @garth-wells in #534
• Attempt at automatic typing for basix by @jhale in #526
• Run demos on more CI workflows by @mscroggs in #537
• Remove xtensor code for Lagrange elements by @garth-wells in #539
• Remove xtensor from BDM by @garth-wells in #540
• Remove more xtensor by @garth-wells in #541
• Remove xtensor from basix::lattice by @garth-wells in #542
• make precompute return values clearer by @mscroggs in #544
• Add VectorElement, TensorElement, MixedElement, etc to UFL wrapper by @mscroggs in #535
• Reset ffcx branch by @mscroggs in #547
• Throw runtime error to avoid "may be used uninitialized" warnings by @mscroggs in #548
• Fix symmetric tensors by @mscroggs in #549
• Reset FFCx branch by @mscroggs in #550
• xtensor removal from polynomial code by @garth-wells in #546
• Reset DOLFINx branch by @garth-wells in #551
• Remove xtensor from moment code by @garth-wells in #552
• Remove xtensor from math namespace by @garth-wells in #553
• Remove xtensor from dof-transform code by @garth-wells in #554
• Final removal of xtensor by @garth-wells in #555
• Reset DOLFINx branch for CI by @garth-wells in #557
• Simplify code in polyset by @garth-wells in #558
• Fix overflow problem in matrix-matrix product heuristics by @garth-wells in #559
• Improve performance of dual matrix construction by @garth-wells in #561
• Some small optimisations when creating elements by @garth-wells in #562
• Require C++20 by @garth-wells in #563
• Revert DOLFINx CI branch back to main by @garth-wells in #564
• Skip orthogonalisation for Lagrange elements by @mscroggs in #565
• Make UFL interface consistent with UFL by @mscroggs in #566
• Only precompute matrices when they are not permutations by @mscroggs in #567
• Slightly faster prepare_permutation by @mscroggs in #569
• Fix Sphinx version by @mscroggs in #573
• Update demo by @mscroggs in #571
• The permutation from an LU decomposition can be used instead of checking the eigenvalues/condition numbers by @mscroggs in #572
• Change image for DOLFINx integration test by @jhale in #576
• Simplify prepare_matrix and apply_matrix by @mscroggs in #577
• remove brackets around asserts by @mscroggs in #579
• Add tensor product representation properties to UFL element wrappers by @IgorBaratta in #560
• Implement enriched elements by @mscroggs in #574
• Add long description by @chrisrichardson in #581

Full Changelog: v0.4.2...v0.5.0

v0.4.2

Fixes build issues related to including cblas.h. Otherwise identical to v0.4.1.

Full Changelog: v0.4.1...v0.4.2

v0.4.1

What's Changed

• Update version numbers by @chrisrichardson in #450
• Add double contravariant map test by @mscroggs in #451
• Speed up sympy tests by @mscroggs in #448
• correct serendipity value shape by @mscroggs in #458
• Avoid manipulating version.h in source directory by @garth-wells in #459
• Use blas for large matrix-matrix multiplications by @IgorBaratta in #453
• Remove TODO that I've already done by @mscroggs in #460
• Add Red Hat CI by @mscroggs in #463
• Fix redhat by @mscroggs in #465
• Add MacOS CI by @mscroggs in #466
• Use Accelerate/Accelerate.h include directory for APPLE by @IgorBaratta in #467
• Remove assumption that degree == highest_degree by @mscroggs in #457
• FFCx and DOLFINx branches by @mscroggs in #470
• remove slow (48s) test by @mscroggs in #472
• add version updating utility by @mscroggs in #468
• Overload FiniteElement by @mscroggs in #473
• only push from one workflow by @mscroggs in #475
• Add == for custom element by @mscroggs in #474
• Fix installation error on mac by @mscroggs in #478
• Remove xarray in favor of xt::xtensor<double, 2> by @jorgensd in #476
• Add maybe unused to variables in assert by @jorgensd in #479
• Include summary from README in website docs by @mscroggs in #484
• Update joss paper by @mscroggs in #481
• Add functions to UFL wrapper by @mscroggs in #464
• Tweak JOSS paper and docs by @mscroggs in #486

Full Changelog: v0.4.0...v0.4.1

Version 0.4.0

What's Changed

• Support single install from top-level directory by @garth-wells in #263
• Remove Intel 'Classic' compiler from CI by @garth-wells in #264
• Allow discontinuous elements by @mscroggs in #260
• Reset FFCx and DOLFINx branches by @mscroggs in #265
• Add Issac's GLL points on triangle and tetrahedron by @mscroggs in #267
• Disable GLL points on pyramid by @mscroggs in #271
• Round data_size up, not down by @mscroggs in #272
• Make numba requirement optional by @mscroggs in #273
• Generalise lattice creation by @mscroggs in #278
• Remove type_to_str and str_to_type functions from cpp by @mscroggs in #270
• Add Chebyshev point lattice by @mscroggs in #279
• Reset branches by @mscroggs in #280
• Update API to use vector by @chrisrichardson in #281
• Remove include set by @mscroggs in #282
• Add centroid simplex method by @mscroggs in #283
• Add GL points for Lagrange by @mscroggs in #284
• Use lib/ when installing the C++ library using Python by @garth-wells in #287
• Use CMake relative path when installing with top level pip by @garth-wells in #288
• Do not install xtensor-blas via CMake by @garth-wells in #289
• Remove xtensor-blas dependency by @garth-wells in #291
• Add interpolation between elements by @mscroggs in #292
• Improve documentation of implementation by @mscroggs in #294
• Update docs by @mscroggs in #295
• Builds and tests manylinux wheels for various x86-64 Pythons by @jhale in #286
• Tweak wheel action. by @jhale in #297
• Jhale/wheel action tweaks by @jhale in #298
• Improve wheel creation by @jhale in #299
• Missing ref in build wheel action by @jhale in #300
• More RPATH adjustments by @jhale in #301
• add BDMCE and BDMCF to map of finite element names by @mscroggs in #302
• Improve documentation by @mscroggs in #303
• Add helper function that returns shape of the array used in FiniteElement::tabulate by @garth-wells in #308
• Remove duplicate tabulate by @mscroggs in #309
• reset FFCx branch by @mscroggs in #312
• Write some demos by @mscroggs in #307
• Allow creation of low order Lagrange without a variant by @mscroggs in #315
• Use an enum instead of a string for quadrature rules by @mscroggs in #317
• revert FFCx and DOLFINx branches to main by @mscroggs in #318
• Fix small typo by @IgorBaratta in #319
• Allow DG0 on a point by @mscroggs in #316
• Tighten up scoping by @garth-wells in #320
• Add JOSS paper by @mscroggs in #321
• Add VTK variant of Lagrange by @mscroggs in #293
• Add simple cpp demo by @IgorBaratta in #322
• Fix for 'maybe unused' variable in release mode by @jorgensd in #323
• Update JOSS paper by @mscroggs in #324
• Apply "iwyu" suggestions by @IgorBaratta in #326
• Use MODULE not SHARED in CMakeLists.txt by @chrisrichardson in #327
• Implement blocked interpolation by @mscroggs in #328
• Remove redundant function FiniteElement::num_points by @garth-wells in #329
• Added documentation of interpolation_matrix by @mscroggs in #330
• Correct shape in interpolation matrix docs by @mscroggs in #331
• Remove compute_expansion_coefficients and store dual matrix by @mscroggs in #333
• Remove interpolate() by @mscroggs in #334
• Revert to DOLFINx main by @mscroggs in #335
• Restructure implementation of Nedelec spaces by @mscroggs in #336
• Use fixed-width xtensor arrays now that we don't support the legacy I… by @garth-wells in #337
• Add AAE, AAF as alternative names by @mscroggs in #339
• Reduce quadrature orders used in element creation by @mscroggs in #338
• Add Xiao-Gimbutas quadrature on tetrahedra by @mscroggs in #340
• Return map function for an element by @garth-wells in #342
• Remove public map_type by @jorgensd in #344
• Improved CMake/pybind11/python3 interaction by @jhale in #343
• Add information about highest degree of polynomials in a finite element by @mscroggs in #346
• Change quadrature numbering to ijk by @IgorBaratta in #350
• Add tensor product representation of Lagrange on quads and hexes by @mscroggs in #349
• Add z direction for tensor product test by @adeebkor in #353
• Reordering of GLL quad points and weights by @adeebkor in #355
• Create FUNDING.yml by @jhale in #358
• Remove xtensor, make download via git explicit option by @jhale in #357
• Improve polyset by @mscroggs in #360
• Add contributing and support info to README by @mscroggs in #362
• Minor suggestions and typo fixes by @IgorBaratta in #364
• a -> an by @mscroggs in #365
• Allow integral moments to be taken against polynomials by @mscroggs in #361
• Set demo CMake language to CXX by @garth-wells in #367
• Store the Lagrange variant when creating a Lagrange element by @mscroggs in #368
• contributing: Fix typo by @wence- in #371
• Python packaging fixes by @jhale in #372
• Minor housekeeping by @garth-wells in #373
• Check if two elements are equal by @mscroggs in #374
• Add more Python versions to CI, including Python 3.10. by @garth-wells in #378
• Fix value shape for Lagrange elements by @garth-wells in #377
• Correct JOSS paper: functionals must be linear by @mscroggs in #380
• Add variants of serendipity in 1D by @mscroggs in #382
• Reset FFCx branch by @mscroggs in #383
• Add variants of DPC spaces by @mscroggs in #384
• Reset branches by @mscroggs in #385
• Variants of 2D and 3D serendipity by @mscroggs in #386
• reset branches by @mscroggs in #387
• Update syntax for wheel build by @jhale in #388
• Remove -DDEBUG definition that doesn't do anything. by @garth-wells in #389
• Store variant for discontinuous Lagrange by @mscroggs in #391
• Convert variant names to lowercase to allow eg "GLL" to be used by @mscroggs in #392
• Test if interpolation matrix is an identity matrix by @mscroggs in #396
• Make serendipity family polysets more stable by @mscroggs in #397
• Remove memory assignment in polyset by @mscroggs in #393
• Add & and const by @mscroggs in #398
• Avoid C++ errors by @mscroggs in #399
• Throw RuntimeError if user tries to make element with negative degree by @mscroggs in #400
• Implement discontinuous Regge by @mscroggs in #401
• Allow user to create a custom element by @mscroggs in #402
• Compute entity transformations algorithmically for all elements rather than implementing them for each element by @mscroggs in https://github.com/FEniCS/basi...