Skip to content

NEWS.md

Latest commit

 

History

History
1265 lines (819 loc) · 90.6 KB

NEWS.md

File metadata and controls

1265 lines (819 loc) · 90.6 KB

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

[0.20.3] - 2026-03-23

Fixed

  • Fixed FEEC basis constructor for serendipity reffes. since PR#1260.

[0.20.2] - 2026-03-20

Fixed

  • Fixed precompilation error when extending Conformity from the FESpaces module. Since PR#1259.

[0.20.1] - 2026-03-19

Changed

  • Minor changesto incorporate the changes in Gridap 0.20, specifically the pullback machinery, into GridapDistributed. Since PR#1258.

[0.20.0] - 2026-03-18

This is a major release with breaking changes, mainly in the Polynomials and ReferenceFEs modules. The main goal of this release is to provide a more flexible and extensible framework for polynomial bases and reference finite elements based on moments, as a stepping stone for the implementation of more complicated elements (e.g. MTW, etc...).

Main PR is PR#1048.

Added

  • Documentation and refactoring of Gridap.Polynomials. Since PR#1072.
  • Two new families of polynomial bases in addition to Monomial, Legendre (former Jacobi) and ModalC0: Chebyshev and Bernstein
  • MonomialBasis and Q[Curl]GradMonomialBasis have been generalized to Legendre, Chebyshev and Bernstein using the new CartProdPolyBasis and CompWiseTensorPolyBasis respectively.
  • PCurlGradMonomialBasis has been generalized to Legendre and Chebyshev using the new RaviartThomasPolyBasis.
  • New aliases and high level constructor for CartProdPolyBasis (former MonomialBasis): MonomialBasis, LegendreBasis, ChebyshevBasis and BernsteinBasis.
  • Generalized gradient and hessian of all CartProdPolyBasis to any tensor type (even with dependant components), as long as the derivative tensor is a tensor of at most third order. Since PR#1184.
  • New high level factory FEEC_poly_basis for the bases for the scalar Lagrange, Nedelec, Raviart-Thomas, BDM spaces, and all other spaces of the Periodic Table of the Finite Elements. (for Serendipity, only scalar is supported). For example:
    • Nedelec on simplex FEEC_poly_basis(Val(D),Float64,order+1, 1,:P⁻)
    • Nedelec on n-cubes FEEC_poly_basis(Val(D),Float64,order+1, 1,:Q⁻)
    • Raviart on simplex FEEC_poly_basis(Val(D),Float64,order+1,D-1,:P⁻; rotate_90=(D==2))
    • Raviart on n-cubes FEEC_poly_basis(Val(D),Float64,order+1,D-1,:Q⁻; rotate_90=(D==2))
    • BDM on simplex FEEC_poly_basis(Val(D),Float64,order+1,D-1,:P ; rotate_90=(D==2))
  • Added BernsteinBasisOnSimplex that implements Bernstein polynomials in barycentric coordinates, since PR#1104.
  • More documentation of Gridap.ReferenceFEs. Since PR#1109.
  • Some refactoring of Gridap.TensorValues to simplify maintenance and new implementations. Since PR#1115.
    • Added SkewSymTensorValue: a new <:MultiValue 2nd order tensor type such that transpose(s)==-s.
    • Added HighOrderTensorValue: a new <:MultiValue tensors of order >=3, with all components independent. Since PR#1193.
    • congruent_prod: new operation for 2nd order tensors: a,b -> bᵀ⋅a⋅b preserving symmetry of a.
    • component_basis, representatives_of_componentbasis_dual, representatives_of_basis_dual: new APIs for ::MultiValues yielding bases of the vector space spanned by the independent components of a tensor type (1st method) and its dual space (2nd method), or the dual to a basis (3rd method).
    • contracted_product generalizes outer, dot, double_contraction and inner for tensors of arbitrary order, refactoring these operations for tensors with all components independent. Since PR#1193.
  • Refactoring of moment-based ReferenceFEs, those using face-integral linear forms for DoFs, including RaviartThomas, Nedelec, BDM and CrouzeixRaviart.
    • The mid-level MomentBasedRefFE factory function creates moment based refFEs
    • The low-level FaceMeasure implements the numerical integration of a bilinear integrand over the faces of a polytope.
    • The low-level MomentBasedDofBasis implements a discretized basis of moment DoF. It supports automatic pre-composition of the field with a differential operator. Since PR#1184.
  • Implemented moment-based scalar (H1 conform) elements for scalar elements lagrangian and serendipity under the names modal_lagrangian and modal_serendipity. They are the default elements when calling for k=0-forms in the generic FEEC reference FE constructor (P⁻/:P/:Q⁻ => lagrangian, :S=> serendipity), use the keyword nodal=true to opt-in nodal DOF based counterpart. Since PR#1173.
  • Unified the high-level constructors of ReferenceFEs
  • New high level ReferenceFEs constructor using Arnold et al FEEC notations (Periodic Table of the Finite Elements): ReferenceFE(F::Symbol, r, k, [, T::Type]; kwargs...) with F the element family, r polynomial order and k the form order.
  • Added poly_type, mom_poly_type and change_dof keyword arguments to many low and high-level reference FE constructors, they enable to control the polynomial (pre-)bases choice for the approximation space and moment test-spaces of the reference elements (See ReferenceFEs doc. page). Since PR#1173.
  • Documented the implemented ReferenceFEs with available order and other information in the ReferenceFEs section of the doc.
  • Implemented the Nedelec reference elements of the second kind nedelec2.
  • API for Geometric decomposition of polynomial bases, implemented for simplices and n-cubes. Since PR#1144. The geometric decomposition API consist in the methods has_geometric_decomposition, get_face_own_funs and apply_face_signflip.

Fixed

  • Fixed evaluation of LinearCombinationDofVector on vector of <:Fields (only impacts ModalC0 FEs and future moment based reffes). Since PR#1105.
  • Fixed get_face_dofs(::ReferenceFEs) on many moment based elements, it sometimes returned face_own_dofs.
  • Fixed hessian of Legendre (former Jacobi) and ModalC0 bases, since PR#1255.

Deprecated

  • num_terms(f::AbstractVector{<:Field}) in favor of length(f::PolynomialBasis)
  • MonomialBasis{D}(args...) in favor of MonomialBasis(Val(D), args...)
  • [P/Q][Curl]GradMonomialBasis{D}(args...) in favor of FEEC_poly_basis
  • NedelecPreBasisOnSimplex{D}(order) in favor of NedelecPolyBasisOnSimplex(Val(D), Float64, order)
  • JacobiPolynomialBasis{D}(args...) in favor of LegendreBasis(Val(D), args...)
  • return_type(::PolynomialBasis) in favor of value_type(::PolynomialBasis)

[0.19.9] - 2026-03-14

Changed

  • Split CI into separate jobs to keep test time down. Since PR#1251.

Fixed

  • Fix incorrect permutation handling in compute_field_offsets for BlockMultiFieldStyle. Since PR1247.
  • MappedGrid now returns its internal geometric map. Since PR#1226.
  • Fixed bug in skeleton transient fields. Since PR1229.
  • Fixed bug in FineToCoarseField internal indexing. Since PR1224.
  • Fixed outer(f, s::ShiftedNabla) silently ignoring the shift contribution.
  • Fixed incorrect else p == HEX control flow in EdgeBasedRefinement.jl. Since PR[1246] (#1246).
  • Fixed copy-paste bug where get_tangent_vector called get_normal_vector in CellFields.jl. Since PR[1246] (#1246).
  • Generalized change_domain_o2n so that it can now deal with triangulation portions. Since PR1249.
  • Generalized Nedelec Hexahedral elements for meshes beyond Cartesian Meshes. Since PR1250.

[0.19.8] - 2026-02-19

Added

  • Added testvalue(::Type{<:Tuple}) for tuples with an arbitrary number of items. Since PR#1210.
  • Added normalize function support for MultiValue objects. Since PR#1211.
  • Added sqrt function support for TensorValue objects. Since PR#1212.
  • Added better ways to iterate over Table, following the SparseArrays model. Mainly, we introduce datarange, dataview and dataiterator functions. Since PR#1218.
  • Added a new preference to control the integer type used for indexing local quantities (e.g. faces within a cell). A new function set_local_integer_type is introduced, which sets a new type GridapLocalInt (default Int8) that is set at compile time. This allows for GeneralPolytopes with an arbitrary number of faces. Since PR#1218.
  • Added new function compute_graph returning the face-connectivity of meshes and submeshes. Since PR#1218.
  • Added a new API to change the block structure of multi-field cellfields. The MultiFieldFEBasisComponent and MultiFieldBasis have now new constructors that easily allow to recast a multi-field space with a different block structure. Since PR#1218.
  • Various improvements to the polytopal API. For GeneralPolytopes, added is_convex, signed_area, signed_volume, convexify(2D), extrude, as well as several low-level functions to manipulate polytopes. We now compute normals using Newel's algorithm, which is more stable. Since PR#1218.

Changed

  • Reworked the CellConformity API to be more flexible and extensible, while optimizing certain parts. CellConformity is now abstract, with new concrete types GenericCellConformity (stores generic arrays), CompressedCellConformity (stores compressed data optimized for low number of reffe types, old CellConformity), and DiscontinuousCellConformity (DG-type conformity, much faster for discontinuous spaces). For all new conformities, I have introduced a new API to query dof masks based on the faces dofs belong to, mainly generate_cell_dof_mask and generate_dof_mask. Since PR#1218.

Fixed

  • Fixed lack of more general tests for Nedelec ReferenceFEs. Added a test which is currently broken. Since PR#1216.
  • Small bugfix in face_labeling_from_cell_tags. Since PR#1218.
  • Fixed getindex! when the given indexes need to be converted the indexstyle of the array (cache isn't skipped anymore). Implemented the bound check elusion API in most Arrays arraysSince #1213.
  • Fixed TransientCellField skeleton evaluation for :minus side. Since PR#1229.

Changed

  • Existing Jacobi polynomial bases/spaces were renamed to Legendre (which they were).
  • Monomial is now subtype of the new abstract typePolynomial <: Field
  • MonomialBasis is now an alias for CartProdPolyBasis{...,Monomial}
  • All polynomial bases are now subtypes of the new abstract type PolynomialBasis <: AbstractVector{<:Polynomial}
  • get_order, now always returns the maximum order of the basis, the correspondence with ReferenceFEs constructors is summarized in the documentation of the module.
  • NedelecPreBasisOnSimplex is renamed NedelecPolyBasisOnSimplex
  • JacobiPolynomial is renamed Legendre and subtypes Polynomial
  • JacobiPolynomialBasis is renamed LegendreBasis
  • ModalC0BasisFunction is renamed ModalC0 and subtypes Polynomial
  • On simplices, the default polynomial bases of BDM, Nédélec and Raviart-Thomas RefFEs have changed for barycentric polynomial bases which lead to better conditioned systems for higher order.
  • Similarly, on n-cubes, the default polynomial bases of BDM, Nédélec and Raviart-Thomas RefFEs have changed for Legendre (tensor-product) polynomial bases instead of Monomials.
  • Changed Base.== for ReferenceFEs, and implemented it for LinearCombinationFieldVector and LinearCombinationDofVector. The implementation now uses AutoHashEquals.jl, so RefFEs are now only equal if they have the same prebasis, shapefun and dofs.
  • Monomial (pre)bases have been replaced with Bernstein / Barycentric bases for non-scalar finite elements on simplices.
  • ThirdOrderTensorValue{D1,D2,D3,T,L} has been redefined as alias to HighOrderTensorValue{Tuple{D1,D2,D3},T,3,L}.

[0.19.7] - 2025-12-18

Added

  • Added the symmetric quadrature rules of Witherden and Vincent, correctness tests for all quadrature rules, and minor extensions to the interfaces of Quadrature and Polytope. Since PR#1169.

    • Added the symmetric quadrature rules of Witherden and Vincent for tri/tet, quad/hex, wedge and pyramid.
    • Added maxdegree(p::Polytope, name::QuadratureName) to the interface of Quadrature.
    • Enriched the tests for quadratures by checking tensor_product, duffy, and witherden_vincent against exact integrals for monomials on the corresponding polytopes (tri/tet, quad/hex, wedge and pyramid, respectively). Other quadrature rules are compared against these tested ones.
    • Added get_measure(p::Polytope, vertex_coords) and get_diameter(p::Polytope, vertex_coords) to the interface of Polytope.

  • The default quadrature for simplices is now witherden_vincent until available, then xiao_gimbuttas until available, then duffy. Since PR#1169.

  • Added specific tags on dual numbers to allow for nested AD without perturbation confusion. Since PR#1181.

  • Updated the documentation of Polytope{D}'s type parameters D, it is correctly described as the polytope's dimension (instead of embedding dimension). Since PR#1194.

  • Added compatibilty for JSON.jl v1+ and JDL2.jl v0.6+. Since PR1198.

Fixed

  • Fixed issue #1188. Fix one() function for non-square tensors.
  • Fixed strang quadrature of order 4 for triangles. Since PR#1169.
  • Minor MuliValue bugfixes for isless and <= with scalars.

Changed

  • Added specific tags on dual numbers to allow for nested AD without perturbation confusion. Since PR#1181.
  • The default quadrature for simplices is now witherden_vincent until available, then xiao_gimbuttas until available, then duffy. Since PR#1169.

[0.19.6] - 2025-10-17

Added

  • Added support for subdividing unstructured meshes to any level, including periodic ones. Since PR#1143.

Fixed

  • Fixed issue #1174. Since PR#1175.
  • Fixes the issue that a single CellField could not be passed to TrialFESpace constructor in cases where more than one tag was being passed to dirichlet_tags kw-arg argument. Since PR#1177.

[0.19.5] - 2025-09-19

Added

  • Added extension for TikzPictures.jl. Since PR#1150.
  • Added eigen function support for TensorValue objects. Fixes issue #1156. Since PR#1157.
  • Added kwargs support for isapprox function on MultiValue objects. Since PR1158
  • Misc improvements of TensorValues APIs and indexing methods. Since PR#1162.
    • Indexing tensors using slices (:/Colon()), (static)vectors or arrays, ranges (OneTo, SOneTo), or mix of them.
    • Implemented Base.keys that relates to the array shape of the MultiValue
    • Constructors MV(::[S/M]Array) and SA(::MultiValue) for any types MV<:MultiValue and SA<:[S/M]Array for arguments for which the conversion via convert is possible.
  • Added new way of doing AD for MultiField, where partials are computed separately for each field then merged together. Since PR#1136.

[0.19.4] - 2025-08-09

Added

  • Added low-level API Arrays.invalidate_cache!(cache) that forces the re-computation of values in LazyArray caches. Since PR#1137.
  • Added support for star-patch integration, i.e patch integration with masked patch boundaries. Since PR#1138.
  • Added compatibility for DataStructures.jl v0.19.0. Since PR#1149.

Changed

  • Small modifications to PolytopalQuadrature to make it more flexible. Since PR#1149.

Fixed

  • Fixes issue #1119, allowing evaluation of CellFields on arbitrary points on periodic meshes. Since PR#1139.

[0.19.3] - 2025-07-15

Fixed

  • MultiFieldFEBasisComponent now supports addition and subtraction. Since PR#1130.
  • Fixed issues introduced in PR #1130. Closes issue #1131. Since PR #1132.
  • Removed method ambiguity in Adaptivity.setup_edge_based_rrules. Fixes issue #1133. Since PR#1135.

[0.19.2] - 2025-07-08

Added

  • Added missing operations between SkeletonPairs. Since PR#1122.
  • Added bubble elements for simplex and cube-like polytopes. Since PR#1124

Fixed

  • Update norm function to be compatible with complex vectors and tensors. Since PR#1118.
  • AdaptivityGlue can now deal with non-surjective n2o maps. Since PR#1126.

[0.19.1] - 2025-06-11

Fixed

  • Minor bugfixes from v0.19. Since PR#1111.

[0.19.0] - 2025-06-10

Added

  • Support for polytopal meshes in 2D and 3D. New structures PolytopalGridTopology, PolytopalGrid and PolytopalDiscreteModel. Since PR#1110.
  • Support for non-overlapping triangulations. Both BodyFittedTriangulation and BoundaryTriangulation have been extended to support repeated face ids. In particular, the FaceToCellGlue has a new constructor. Since PR#1110.
  • Support for patch-wise integration and local solves. New structures PatchTopology, PatchTriangulation and PatchAssembler have been added. These structures allow for the integration, assembly and solution of local problems on (potentially overlapping) patches of the mesh. Since PR#1110.
  • Support for polynomial spaces on polytopal meshes. New structure PolytopalFESpace has been added. Since PR#1110.
  • With all of the above, we have support for hybrid non-conforming methods on polytopal meshes. Examples have been added for HDG and HHO in 2D and 3D. Since PR#1110.

Fixed

  • The tol kwarg for Interpolable was unused. It is now used for the point-to-cell search, which was breaking for very small cell sizes. Since PR#1106.

[0.18.12] - 2025-04-19

Added

  • Added functionality to allow automatic differentiation of integrals with respect to evaluation positions xh where xh is a FEFunction. Since PR#1095

Fixed

  • Fix bug for issue #927, where Nedelec FE would not work on faces (Dc < Dp). Since PR#1094.
  • Added support to evaluate polynomial bases on dualized points. Since PR#1100.

[0.18.11] - 2025-04-01

Added

  • Added definition for DiracDelta with Triangulation argument. Since PR#1087.

Fixed

  • Fixed the failure of hessian over sum/product two Fields or CellFields as reported in issue #875 by adding the relevant manual chain rules. Since PR#1053.

Changed

  • Modified the LinearStageOperator struct to align it with the NonlinearStageOperator. Added usx, ws and tx as new fields in the struct. Since PR#1089.
  • Misc TensorValues improvements since PR #1092.
    • MultiValues can now be indexed by tuples of mixed Integer and CartesianIndex
    • Indexing in MultiValues now performs bound checks, that can be disabled using @inbounds or Gridap's performance execution mode.
    • Make sure that the result of all operation (empty/zero tensor) of correctly promoted (element) type when applied to tensor(s) of length zero
    • An argument error is now thrown when calling tr(::ThirdOrderTensorValue{A,B}) with A!=B

[0.18.10] - 2025-03-04

Added

  • Added corresponding function get_tangent_vector to get_normal_vector. This method calculates the (unique up to sign) tangential unit vector to edges in 2D meshes, by rotating the normal (nx, ny) -> (ny, -nx). Since PR#1071.

Fixed

Changed

  • real and imag functions now return a real TensorValue. Since PR#1080.
  • Refactor of automatic differentiation for multi-field spaces. Now there is a more consistent implementation covering all cases. Since PR#1070.

[0.18.9] - 2025-01-13

Added

  • Added AMR-related methods mark and estimate to Adaptivity module. Implemented Dorfler marking strategy. Since PR#1063.

Fixed

  • BUG in FineToCoarseFields.jl. Since PR#1074.

Changed

  • Low level optimisations to reduce allocations. AffineMap renamed to AffineField. New AffineMap <: Map, doing the same as AffineField without struct allocation. New ConstantMap <: Map, doing the same as ConstantField without struct allocation. Since PR#1043.
  • ConstantFESpaces can now be built on triangulations. Since PR#1069

[0.18.8] - 2024-12-2

Added

  • Added get_dof_value_type for FESpacesWithLinearConstraints. Since PR#1062.
  • Added Xiao-Gimbutas quadratures for simplices. Since PR#1058.
  • Small improvements of the documentation of Gridap.TensorValues. Since PR#1051.

Fixed

  • Fixed #974, an error when weak form is real but unknown vector is complex. Since PR#1050.
  • Fixed issue where barycentric refinement rule in 3D would not produce oriented meshes. Since PR#1055.

Changed

  • Optimized MonomialBasis low-level functions. Since PR#1059.

[0.18.7] - 2024-10-8

Added

  • Added MacroFElements. These are defined as having the basis/dof-basis of a FESpace created on top of a RefinementRule. Since PR#1024.
  • Added Barycentric refinement rule in 2D and 3D. Added Simplexify refinement rule. Since PR#1024.
  • Added names to vector and tensor components in VTK exports, to avoid Paraview's automatic (sometimes wrong) guesses. See TensorValues.indep_components_names. Since PR#1038.
  • Misc improvements of the TensorValues module: See TensorValues.indep_components_names. Since PR#1040.
    • Documented all symbols exported by the module
    • Improved and added test for some API function of MultiValue (general diag of 2nd order tensors, fixed convert of 3rd order tensors to SArray, avoid unwanted fallback of num_components on MultiValue types with undefined dimensions, more autodiff tests, better double_contraction API (prevent invalid operation giving indexing errors and enable valid operations)).
    • Added a clear separation between the physical components access (getindex, num_components) and the numerical access to the stored independent components (num_indep_components, indep_comp_getindex) to enable using symmetric tensor types as unknown in FE Spaces.
    • Implemented automatic differentiation gradient and laplacian for second order tensor, and divergence for third order tensors.
    • Added AbstractSymTensorValue, an abstract type for second order symmetric tensors, and SymTracelessTensorValue (aliased to QTensorValue), a type for traceless symmetric tensors. SymTensorValue is now subtype of AbstractSymTensorValue.
    • A convergence test for Poisson problem of QTensorValue unknown field validates the implementation.
  • Added support for benchmarking, through PkgBenchmark.jl. Since PR#1039.
  • Added code quality tests, through Aqua.jl. Since PR#1039.

Fixed

  • Fixed constructor of RungeKutta with only one solver. Since PR#999.

Changed

  • Conforming FESpaces now keep track of their CellConformity info, within the metadata field. Since PR#1042.

[0.18.6] - 2024-08-29

Fixed

  • Improved performance of PR#967. Along the way, opened the door to Triangulations of different type in SkeletonTriangulation. Since PR#1026.

[0.18.5] - 2024-08-28

Changed

  • Misc changes required to support facet integration on non-conforming meshes. These changes do not involve methods of the public API. Since PR#967

[0.18.4] - 2024-08-09

Changed

  • Added WriteVTK kwargs to control the output encoding for vtk files. Since PR#1016.

Fixed

  • Passing kwargs from refine to simplexify functions in Adaptivity. Since PR#1015.
  • Fixed interpolate for ZeroMeanFESpace. Since PR#1020.
  • Fixed gather_free_and_dirichlet_values! for FESpaceWithConstantFixed. Since PR#1020.

[0.18.3] - 2024-07-11

Added

  • Added more features to Adaptivity. Notably: 3D uniform edge-based refinement for HEX and TETs. Barycentric refinement for simplices. Simplexify as a new refinement strategy. Since PR#1013.

  • Define GeneralPolytope that represents general polytopes in 2 and 3 dimensions. Since PR#1006.

Changed

  • Added local preferences for Gridap through Preferences.jl. The macro @check can now be activated/deactivated by using the local preference execution_mode. Since PR#1014.

[0.18.2] - 2024-05-02

Fixed

  • Bugfix in get_face_dofs for Nedelec GenericRefFE. Since PR#1005.
  • Ensure deterministic behavior for matrix assembly involving multiple domains. Since PR#1004.

[0.18.1] - 2024-04-12

Changed

  • Changed the sign of the residual in TransientLinearFEOperator to align with the conventions of AffineFEOperator. Since PR#996.

Fixed

  • Bugfix in restrict_to_field for BlockMultiFieldStyle. Since PR#993.

[0.18.0] - 2024-04-12

Breaking

  • ODE module extensive refactor. Breaking changes! See docs and PR for details. Since PR965.
  • Fixed name clash with Statistics.mean. Since PR#988.
  • Deprecated SubVector in favor of Julia's view. Since PR#989.

Added

  • Added some missing API methods to Assemblers and MultiField. Since PR#985.

Fixed

  • Fix when evaluating \circ operator with CellState. Since PR#987.

[0.17.23] - 2024-01-28

Changed

  • Changed how symbolic_loop_matrix_vector! loop works. Now it also takes account vector entries touched from matvecdata. Since PR#975.

[0.17.22] - 2024-01-12

Added

  • Added get_dof_to_node and get_dof_to_comp for LagrangianDofBasis. Since PR#964.

Changed

  • Changed how allocate_vector works. Now it only allocates, instead of allocating+initialising to zero. Since PR#963.

Fixed

  • Fixed issue where FineToCoarseField would not work for vector-valued fields. Since PR#970.
  • Fixed issue where BlockSparseMatrixAssembler would not work correctly when permuting variables. Since PR#971.

[0.17.21] - 2023-12-04

Added

  • Implemented real/imag for VectorValues
  • Explicit Runge-Kutta ODE Solvers. Since PR #952
  • Improved the methods allocate_in_range and allocate_in_domain with support for BlockArrays and distributed arrays. Since PR#960.

Fixed

  • BlockMultiFieldStyle available for TransientMultiFieldFESpaces since PR #946.
  • When creating DiscreteModelPortions, some of the FaceLabeling arrays were being aliased. This caused issues when adding tags to distributed models in debug mode. Since PR #956.
  • Function add_entry! was inconsistent for AbstractMatrix and AbstractSparseMatrix. Since PR#959.

[0.17.20] - 2023-10-01

Added

  • Block assembly now generalised to work with AbstractBlockArrays, to include changes in GridapDistributed. Since PR 939.
  • Implici-Explicit Runge-Kutta ODE solvers. Since PR #919.

Fixed

  • Using Broadcasting(\circ) instead of \circ in one of the lazy_maps used to transform a coarse field into a fine field. Since PR #938.
  • Better infinite norm computation in Algebra._check_convergence. Now works for any AbstractArray type, including PVector. Since PR #940.
  • Updated Runge-Kutta solver. Since PR #919.

[0.17.19] - 2023-08-23

Fixed

  • Reimplemented DomainStyle for CellQuadrature to fix breaking low-level Poisson tutorial. Since PR #937.

[0.17.18] - 2023-08-15

Added

  • Jacobi polynomial bases. Since PR #896.
  • Replaced newest vertex bisection mesh adaptation in src/Geometry/NewestVertexBisection.jl with appropriate changes to src/Adaptivity/EdgeBasedRefinement.jl. Since PR #901.
  • When refining DiscreteModels, the FaceLabeling of the resulting AdaptedDiscreteModel will now correctly inhering the tags of the parent model. This has been made possible by the addition of the method get_d_to_face_to_parent_face. Since PR#886.
  • Added support for mixed adaptivity (i.e coarsening and refining), as well as non-conforming adaptivity. Since PR#886.
  • Added support for block assembly of FE systems. Two new types BlockMultiFieldStyle and BlockSparseMatrixAssemblers have been added. Since PR#915.

Changed

  • The API of CellQuadrature has now both data and integration domain styles as keyword arguments. Old signatures are deprecated. Since PR #885.

Fixed

  • ODE operators cache linear system at initial time or the time stored by the operator. Before, the linear system was cached at time t = 0.0, which cannot be done if the operator is not well-defined at t = 0.0. Since PR #891.
  • Fixed the method get_normal_vector for AdaptedTriangulation. The method get_facet_normal was using default, it's now using the spetialized implementation for the underlying triangulation type. Since PR #884.
  • Fixed cell_dof_ids for the case of vectorial ConstantFESpace. Since PR #888
  • Fixed generation of Modal C0 bases for Julia 1.9. Since PR #918.
  • Fixed some edge cases for change_domain between AdaptedTriangulations where inneficient coordinate transformations would be applied between physical and reference domains. Since PR#886.
  • Fixed: Domain limits can now be of any type (notably, floats) when refining CartesianDiscreteModels. Since PR#886.

[0.17.17] - 2023-02-24

Added

  • Implemented RefinementRule and AdaptivityGlue, which encode the mapping between the old and new cells in an adapted model. Since PR #838.
  • Implemented AdaptedDiscreteModel and AdaptedTriangulation, representing respectively a model and triangulation produced by adapting a parent model. This types mostly wrap around DiscreteModel and Triangulation, with some added features necessary to keep track of the adaptive hierarchy. Since PR #838.
  • Implemented tools to be able to transfer CellDatums back and forth between parent and child grids. These include changes to change_domain (which now takes the source Triangulation as argument) and a new type of Measure called CompositeMeasure, which allows the integration Triangulation to be different from the Triangulation of the resulting DomainContribution. To accommodate CompositeMeasure, Measure has been made abstract type and a GenericMeasure has been created to replace the old type. Since PR #838.
  • For the fine-to-coarse transfer of CellFields, the new FineToCoarseField has been implemented. This new structure bundles several fields defined on the fine mesh to create a single field on the coarse mesh. To enable fast interpolation of this type of field, we have also implemented FineToCoarseReferenceFE and FineToCoarseDofBasis. Since PR #838.
  • Implemented CompositeQuadrature, a quadrature for a cell that has been refined using a RefinementRule. Since PR #838.
  • Implemented simple refinement strategies for Cartesian discrete models in 2&3D as well as Unstructured discrete models in 2D. The latter is implemented by red-green refinement. Since PR #838.
  • Added optimization when calling unique for a CompressedArray. Since PR #838.
  • Added support for changing domain between adapted triangulations in cases where the target triangulation is a view, a BoundaryTriangulation or a SkeletonTriangulation. Since PR #868.

Fixed

  • Using broadcasting through in ODESolver vector operations. Since PR #860
  • Fixes to array_cache(a::Table): Now does not use the zero(T,N) function, but instead creates new empty vector using the general allocator Vector{T}(undef,N). This allows Table to work with complex composite types which don't have an easy zero(T) function defined. Since PR #838.
  • Added get_metadata to all the instances of ReferenceFE. This makes the abstract type more consistent, which is necessary for the new type FineToCoarseReferenceFE. Since PR #838.
  • ConstantFESpace is now properly exported. Since PR #872.

[0.17.16] - 2022-12-22

Fixed

  • Fixed a bug in function collect_cell_matrix_and_vector. Since PR #849

[0.17.15] - 2022-11-10

Added

  • lastindex for MultiValues for consistent usage of [end] as per length. Since PR #834
  • BDM (Brezzi-Douglas-Marini) ReferenceFEs in PR #823
  • Implemented ConstantFESpace. This space allows, e.g., to impose the mean value of the pressure via an additional unknown/equation (a Lagrange multiplier). Since PR #836
  • Methods to construct DiracDelta at generic Point(s) in the domain. Since PR #837
  • some key missing lastindex and end tests belonging to PR [#834]. Since PR #837
  • AbstractVector support to weights and Points of GenericQuadrature, in particular for FillArray usage. Since PR #839
  • Using the above generalization of GenericQuadrature, made the weights of quadrature of DiracDelta for generic points to be a Fill vector. Since PR #839
  • Optional keyword argument T to select the floating point precision for the numerical quadrature pipeline involving Measure, CellQuadrature and Quadrature functions. T defaults to Float64 when not specified, so it doesn't involve any breaking changes in public API. Since PR #840

[0.17.14] - 2022-07-29

Added

  • Functionality to take gradient of functional involving integration (DomainContribution) over Skeleton faces, with respect to the degrees-of-freedom of FEFunction. The interface remains the same - gradient(f,uh). Since PR #797
  • Extended the MultiField functional gradient (with respect to degrees-of-freedom of MultiFieldFEFunction) to functionals involving Skeleton integration. The interface remains the same gradient(f,xh). Since PR #799
  • Functionality to take jacobian of functional involving integration (DomainContribution) over Skeleton faces (obtained from testing bilinear form with the whole set of test fe_basis), with respect to the degrees-of-freedom of FEFunction. The interface remains the same - jacobian(f,uh). Since PR #803
  • A dummy tag for ForwardDiff configs being constructed in Gridap at src/Arrays/Autodiff.jl to fix issue #805. Since PR #806

Fixed

  • The behavior of gradient for functionals involving operations of CellFields inside mean and jump terms of Skeleton Integration terms. Since PR #800
  • The behavior of SkeletonCellFieldPair at the Boundary integration terms. Since PR #800
  • push_normal for rectangular Jacobians. Since PR#809
  • Nedelec FEs with triangles. Since PR #813
  • Gridap.jacobian for Skeleton integration terms, in the case of general test basis where dim not same as trial basis. Since PR #815

[0.17.13] - 2022-05-31

Added

  • KeyToValMap lazy map that dynamically creates a Dict with the outputs of a function over an array of inputs. Since PR #801
  • MappedDiscreteModel and MappedGrid, which are geometrical models with one extra geo map in the physical space. Since PR #801
  • GridWithFEMap, which has a geometrical map defined by a FE function. Since PR #801
  • Vertex bisection algorithm for refinement of triangular meshes in 2D. Since PR #733
  • Generalized-α method for 1st order ODEs. Since PR #781
  • Implemented (generalised) ModalC0 Polynomial bases and reference FEs. Since PR #777
  • Serendipity reference FEs for any dimension and order. Since PR #777

[0.17.12] - 2022-03-24

Fixed

  • Bugs related with 2nd order derivatives. Since PR #769 and #771.

[0.17.11] - 2022-03-16

Fixed

  • Miscellaneous bugs associated with empty meshes. Since PR #765 and #766.

[0.17.10] - 2022-03-10

Added

  • Implemented addition/subtraction between a TensorValue and a SymTensorValue. Since PR #755.
  • Nédélec elements on simplices (lowest order interpolation for the moment). Since PR #758.

Fixed

  • Restrict to the active model the evaluation of a FE function at arbitrary points. Since PR #752.
  • Renamed GridapODEs to ODEs. Since PR #756

[0.17.9] - 2022-02-28

Added

  • Extra support for dual number propagation. Since PR #653
  • Added ODE tools to Gridap. GridapODEs has been merged into Gridap. Sice PR#748

[0.17.8] - 2022-02-14

Added

  • Support for periodic conditions in CartesianDiscreteModels built from cmin, cmax. Since PR #738.
  • Function savepvd. Since PR #736.
  • Support for passing nothing as the first argument of the CartesianDiscreteModel constructor to mimic the API of GridapDistributed. Since PR #735.

Fixed

  • Bug in SymTensorValue. Since PR #740.
  • Bug in interpolation at arbitrary points. Since PR #729.
  • Bug in reference_grid(TET). Since PR #734.
  • Bug in computation of Hessian with AD for multi-field. Since PR #731.

[0.17.7]- 2021-12-03

Fixed

  • Improving compile times by removing unnecessary @inline statements. Since PR #726.

Added

  • CellFE constructor now gets optional arguments and pass them down. Since PR #728.

[0.17.6] - 2021-12-01

Added

  • Implemented Base.unaliascopy(A::SubVector). Since PR #715.

Fixed

  • Bugfix in Base.view(glue::FaceToFaceGlue,ids::AbstractArray). Since PR #724.

[0.17.5] - 2021-11-08

Added

  • Hiding the creation of FESolver and LinearFESolver from user code. Since PR #705.

[0.17.4] - 2021-10-27

Fixed

  • Using implementation of pvtk_grid provided in WriteVTK. Since PR #699.

[0.17.3] - 2021-10-27

Fixed

  • Adding a newer version of WriteVTK in the [compat] section. Since PR #698.

[0.17.2] - 2021-10-26

Fixed

  • AD with multi-field residuals with different num dofs per field. Since PR #687.

[0.17.1] - 2021-10-26

Fixed

  • Laplacian Δ operator on unstructured linear grids for quantities defined in the reference space (i.e. shape functions in standard FEM). Since PR #691.
  • Laplacian Δ operator on triangulations using GridView (e.g., when interpolating functions in a sub-domain or on the boundary). Since PR #691.
  • Fixed typo in , function solve! of LinearSolvers.jl`. Since PR #692.

[0.17.0] - 2021-10-22

Added

  • Aliases Interior, Boundary, Skeleton, and Interface for the Triangulation, BoundaryTriangulation, SkeletonTriangulation, and InterfaceTriangulation constructors. Since PR #662.
  • Function create_pvtk_file for exporting results in pvtu format. Since PR #685.

Changed

  • Major refactoring in the Triangulation interface to properly support the solution of PDEs defined on domains of different dimension. The major change from the user perspective is that Triangulation objects can be used both to integrate the weak form (as before) but also to define FE spaces (except for unfitted triangulations obviously). It is still possible to define FE spaces from DiscreteModels, but it is safer and more idiomatic (closer to the math notation) to use Triangulation objects from now on. Since PR #662.
  • Changes in assembly interface to allow optimization when assembling matrices and vectors simultaneously. Since PR #685.

Removed

  • BoundaryDiscreteModel, RestrictedDiscreteMdeol, RestrictedTriangulation, TriangulationStyle, BackgroundTriangulation, SubTriangulation, get_cell_to_bgcell, get_cell_ref_map, get_background_triangulation, and have_compatible_domains. Since PR #662.
  • Functions scale_entries! and fill_entries!. Replaced by Julia functions LinearAlgebra.rmul! and LinearAlgebra.fillstored!. Since PR #680.

[0.16.5] - 2021-09-08

Added

  • Implemented DIV operator for FE functions in RT space. Since PR #650.
  • GenericAssemblyStrategy. Since PR #655.
  • Additional high level API assembly functions. Since PR #652.

Fixed

  • Bug related with the release of ChainRulesCore version 1.3.1. Since #654.
  • Inheritance relationship for DiscreteModelPortion. Since PR #645.
  • Optimization to RT FEs. Since PR #638.
  • Bug in boundary discrete model. Since PR #651.

[0.16.4] - 2021-08-17

Added

  • Trait to CellQuadrature to support the evaluation of integrals on the reference domain. Since PR #636.
  • Type Interpolable allowing to interpolate data from two independent meshes. Since PR #632.

[0.16.3] - 2021-06-28

Fixed

  • Deactivating optimizations related with MemoArray since they are not reliable. Since PR #624.
  • Bug related with ArrayBlock. Since PR #623.

[0.16.2] - 2021-06-21

Fixed

  • Bug related with boundary integration caused by some optimization introduced in v0.16. Fixed via PR #616.

[0.16.1] - 2021-06-04

Fixed

  • Bug for 1st order FE spaces in combination of 1st order models with periodic BCs. Since PR #611.

[0.16.0] - 2021-06-04

Added

  • User API to select specific quadrature rules. Since PR #578.
  • Experimental support for mixed dimensional PDEs. Since PR #567.
  • Added get_cell_dof_basis(model,cell_reffes,::Conformity) and get_cell_shapefuns(model,cell_reffes,::Conformity). Since PR #579.
  • Implemented get_cell_dof_basis and get_cell_shapefuns for global RT FE spaces in a new file DivConformingFESpaces.jl. Since PR #579.
  • Added support to allow evaluation of FE functions at arbitrary points. Since PR #523.
  • Implemented compute_cell_points_from_vector_of_points to build CellPoint from a vector of points. Since PR #523.

Changed

  • Major refactoring in the handling of blocks (e.g. in multi-field and skeleton terms). The new code follows a much more simple approach based in the new type ArrayBlock. Since PR #583.
  • The default quadrature rule for tets has changed. Since PR #578.
  • Refactoring in SparseMatrixAssembler to make it more extensible and efficient. Since PR #568.
  • Renamed get_free_values -> get_free_dof_values. Since PR #567.
  • Renamed get_dirichlet_values -> get_dirichlet_dof_values. Since PR #606.
  • Renamed object -> value the variable in ConstantField. Since PR #606.
  • Miscellaneous changes in the FE assembly to allow the solution of mixed dimensional problems. Since PR #567.
  • Renamed get_cell_shapefuns by get_fe_basis. Since PR #579.
  • Renamed get_cell_shapefuns_trial by get_trial_fe_basis. Since PR #579.
  • Renamed get_cell_dof_basis by get_fe_dof_basis. Since PR #579.
  • Removed conformity optional keyword argument from FESpace(::DiscreteModel,::CellFE; kwargs...) constructor. Since PR #579.
  • Replaced CellFE(::AbstractArray{<:Field},::AbstractArray{<:ReferenceFE}) by CellFE(::DiscreteModel,::AbstractArray{<:ReferenceFE},::Conformity). Since PR #579.

Removed

  • All code associated with with BlockArrayCoo. Since PR #583.
  • Module Gridap.Integration has been deleted and its contents have been merged into Gridap.ReferenceFEs module.
  • Types SparseMatrixCSR and SymSparseMatrixCSR have been moved to the registered package SparseMatricesCSR. To use them simply add SparseMatricesCSR into your environment and type using SparseMatricesCSR. Since Since PR #568.
  • Removed PushForwardMap and all code depending upon it. Since PR #579.

[0.15.5] - 2021-05-25

Added

  • Differential operators (∇+k)(u), (∇+k)⋅u, (∇+k)×u, (∇+k)⊗u, and u⊗(∇+k) for some u::CellField and k::VectorValue. Since PR #597.
  • Definition of u.*v between instances of vector-valued CellField objects u and v. Also differential operators ∇.*u and (∇+k).*u. Since PR #597.

[0.15.4] - 2021-03-29

Fixed

  • Bug in CartesianDiscreteModel with periodic boundary conditions that shows up in Julia 1.6 but not in Julia 1.5. Since commit da005cf.

[0.15.3] - 2021-03-16

Added

  • get_cell_map now returns array of AffineMap for linear grids of simplices. Needed to compute Laplacian operator, inverse maps etc. Since PR #553.

Fixed

  • Bug in print_op_tree. Since PR #563

[0.15.2] - 2021-03-08

Added

  • Method inverse_map for AffineMap. Since PR #552.
  • Method get_cell_points for CellDof. Since PR #551.
  • Evaluation of MonomialBasis objects at a single point. Since PR #550.
  • rand function for MultiValue objects. Since PR #530.

Fixed

  • Bug in return_value for Broadcasting(∇∇). Since PR #554.
  • Bug in dot for third order tensors. Since PR #544.

[0.15.1] - 2021-01-22

Added

  • Added support for Hessian and Laplacian operators. Only implemented for Finite Elements with an AffineMap. Since PR #514.

Fixed

  • Bug in RestrictedDiscreteModel for periodic boundary conditions. Since PR #517.
  • Bug in sum(a::LazyArray) when eltype(a) <: AbstractArray. Since PR #513.

[0.15.0] - 2020-12-14

This version is a major (backwards-incompatible) refactoring of the project which is not summarized here for the sake of brevity. Most of the functionality of v0.14.0 is available in v0.15.0, but possibly with a significantly different API. See here the changes in the sources of the Gridap Tutorials between versions 0.14.0 and 0.15.0 to effectively see the major changes in the API.

[0.14.2] - 2020-11-24

Added

  • Added additional tensor operations and new double contraction notation ⋅². Implemented a zero constructor for ThirdOrderTensorValues to allow integration of 3-tensors. Since PR #415.

Fixed

  • Bug-fix for 32-bit Julia: Replace all occurrences of Int64 by Int. Since PR #445.
  • Bug-fix for 32-bit Julia. Using inttype=Int keyword argument for JSON parsing. Since PR #456.

[0.14.1] - 2020-09-17

Added

  • Added VectorWithEntryInserted and VectorWithEntryRemoved. Since PR #401.
  • Added missing get_constant_approach() getter to FESpaceWithConstantFixed. Since PR #409.

Deprecated

  • The name FESpaceWithLastDofRemoved has been deprecated in favor of its generalization FESpaceWithConstantFixed. Since PR #396 and PR #404.

[0.14.0] - 2020-08-27

Removed

  • Support for Julia v1.0. Now, the minimum supported is Julia v1.3. Since PR #376.

Changed

  • Major refactoring associated with the handling of elemental matrices and vectors in multi-field computations and also on the skeleton. Since PR #376.
  • First and second argument switch in update_state_variables! in order to have function-first style. Since PR #376.
  • Table struct has been generalized such that data and ptrs arrays can be of an arbitrary type extending AbstractArray. Since PR #310
  • interpolate, interpolate!, interpolate_dirichlet... switched argument order to function first style. For instance interpolate(u, V) instead of interpolate(V, u)

Added

  • Allowing the construction of an HomogeneousTrialFESpace from a TrialFESpace. Since PR #384.
  • Support for automatic differentiation of residuals and Jacobians in multi-field computations since PR #383.
  • New FilterKernel since PR #379.

Fixed

  • Bug associated with boundary triangulation in 1D discrete models. Since PR #393.

[0.13.4] - 2020-08-23

Added

  • New FilteredCellArray since PR #372.

[0.13.3] - 2020-08-12

Added

  • Visualization.visualization_data function that makes it easier to bring fields into visualization library friendly formats. Since PR #354.
  • Gradient of a product binary operation (*) between a scalar and a field. Since PR #340.

[0.13.2] - 2020-07-31

Added

  • Automatic differentiation of the Jacobian from a given residual and the Jacobian and the residual from a given energy. Not working at this moment on the Skeleton nor for multi-field (WIP), but yes for other cases. Now, the user can omit jac from FETerm(res,jac,trian,quad), i.e. FETerm(res,trian,quad) and the Jacobian will be automatically generated. In addition, the user can write FEEnergy(ener,trian,quad) for a given ener(uh) function and the residual and the Jacobian will be automatically generated. Since PR #338.

[0.13.1] - 2020-07-24

Fixed

  • Bugs associated with the degenerated case of 0-length arrays. Since PR #331 and #332.

[0.13.0] - 2020-07-23

Added

  • Automatic differentiation for symmetric gradient, i.e. ε(u) for a given vector-valued function u. Since PR #327.
  • Added missing SparseMatrixAssembler constructor for MultiFieldFESpaces. Since PR #320.
  • kw-argument space to LagrangianRefFE constructor in order to select the type of underlying polynomial space, i.e., :Q, :S, or :P. Since PR #321.

Changed

  • The meaning of inward/outward has slightly changed for SkeletonCellBasis objects. Now, by accessing to these properties a ReducedSkeletonCellBasis is returned, which allows to use the result in a more flexible way (in particular, the result can be used in a similar way than the result of jump or mean). Since PR #317.
  • Major refactoring in ReferenceFEs module. Since PR #319 and #321. In particular:
    • NodalReferenceFE has been replaced by a new abstract type LagrangianRefFE.
    • GenericNodalCartesianRefFE has been replaced by GenericLagrangianRefFE.

Removed

  • Removals associated with the ReferenceFEs refactoring in PR #319:
    • Removed QDiscRefFE constructor. Use a standard LagrangianRefFE and L2Conformity instead.
    • Removed PDiscRefFE constructor. Use LagrangianRefFE constructor with the kw-argument space=:P.
    • Removed CDLagrangianRefFE constructor. Use a standard LagrangianRefFE and CDConformity instead.
    • Removed fields face_own_dofs and face_own_dof_permutations from GenericRefFE.
    • Removed struct DiscRefFE.

Fixed

  • Better handling of FE terms defined on empty triangulations. Since PR #329.
  • Replaced += by add_entry!. Since PR #316.
  • Minor fix to let Vtk.jl support changes in Vtk 1.7.X versus 1.6.X. Since PR #324.

[0.12.0] - 2020-07-07

Added

  • Added SkeletonTriangulation constructor in order to integrate, where a given interpolation is discontinuous. Since PR #304.
  • New ConformingFESpace constructor. Since PR #293.
  • Added QDiscRefFE constructor for DiscRefFE. Since PR #293.
  • New FESpace constructor that takes an instance of ReferenceFE. Since PR #294.
  • New FESpace constructor that takes an instance of Conformity. Since PR #311.
  • New CDLagrangianRefFE struct, that provides a Lagrangian reference FE with different conformity per direction. Since PR #299.
  • New FESpace method that takes a model and a RefFE. Since PR #299.
  • Possibility to have 0 order in DISC directions of a CDLagrangianRefFE. Since PR #308.
  • Added setindex! method for Reindexed. Since PR #309.

Changed

  • Changed the interfaces of ReferenceFE and NodalReferenceFE in relation of DOF ownership. Now function get_face_own_dofs and related ones are parametrized by a Conformity object. Since PR #311.
  • The constructors GenericRefFE, GenericNodalCartesianRefFE, and compute_conforming_cell_dofs take an extra argument of type Conformity. Since PR #311.
  • Renamed PDiscRefFE -> DiscRefFE struct keeping the name for constructor. Since PR #293.
  • One of the GradConformingFESpace methods now more general ConformingFESpace. Since PR #293.
  • DivConformingFESpace and CurlConformingFESpace constructors eliminated. Since PR #293.
  • Extend table to support arbitrary vector types. Since PR #310.

Fixed

  • Construction of VectorValue, TensorValue, et al. objects from non-homogeneous arguments. This solves some problems associated with automatic differentiation. Since PR #298.
  • CDLagrangianRefFE node ordering. Since PR #305.

[0.11.2] - 2020-06-22

Added

  • Method solve!(x,ls,op::AffineOperator,cache::Nothing,newmatrix). Since PR #288.

Fixed

  • Bug related with WriteVTK version 1.7. Fixed via PR #287.
  • Bug in outer constructor of Table{...} for input arrays of abstract type. Fixed via PR #285.

[0.11.1] - 2020-06-19

Fixed

  • Bug in the handling of caches in NLSolver. Fixed via PR #283.
  • Bug that showed up when interpolating a FE function defined on an ExtendedFESpace onto a non-extended FESpace. Fixed via PR #282.

[0.11.0] - 2020-06-16

Added

  • Operator ⊙ (\odot) as an alias of inner. Since PR #239.
  • Operator ⊗ (\otimes) as an alias of outer. Since PR #239.
  • Support for (symmetric) 4th order tensors. Since PR #239.
  • Optimizations for symmetric 2nd order tensors. Since PR #239.
  • Methods for cross function (aka × (\times)) to operate with VectorValues. Since PR #280.
  • Interpolation is now supported also for multifield spaces. Since PR #279.

Changed

  • Major refactoring in the module Gridap.TensorValues. Since PR #239. The following changes are likely to affect all users:
    • The operator * is not allowed for expressing the dot product anymore. Use LinearAlgebra.dot function aka ⋅ (\cdot).
    • The syntax ∇*u is not allowed anymore. Use ∇⋅u instead.
    • Gridap re-exports dot, ⋅, and other names from LinearAlgebra that are used often in Gridap code.
    • Function n_components is renamed to num_components.
  • The SingleFieldFESpace interface has changed. The function gather_free_and_dirichlet_values! has been added as mandatory for all FE space implementations and the old function gather_free_and_dirichlet_values is now optional. Since PR #279.

[0.10.4] - 2020-06-8

Added

  • Functions create_vtk_file and createvtk. Since PR #273.

[0.10.3] - 2020-05-29

Added

  • Function print_op_tree to visualize lazy operation trees. Since PR #270.
  • Exported apply and reindex from Gridap top level. Since PR #270.
  • Extended support of CartesianDiscreteModel to models with periodic boundary conditions. PR #266.

Deprecated

  • Optional argument map for CartesianDescriptor converted to a key-word argument. Since PR #266.

Fixed

  • Fixed some methods of the sparsecsr generic function. Since PR #262.
  • Fixed BUG in findnz function for SparseMatrixCSR. Since PR #264.
  • Fixed restrict(::AbstractArray,::TriangulationPortion) for portions of triangulations extending BoundaryTriangulation. Since PR #267.

[0.10.2] - 2020-05-21

Added

  • New key-word arguments zeromean_trian and zeromean_quad in the FESpace constructor. Since PR #257.
  • New method reindex(::Triangulation,indices). Since PR #257.
  • New functions get_face_to_face(::BoundaryTriangulation) and get_cell_around(::BoundaryTriangulation). Since PR #256.

[0.10.1] - 2020-05-19

Fixed

  • Added missing implementation of simplexify(SEGMENT) and simplexify(VERTEX). Since PR #252.

[0.10.0] - 2020-05-14

Added

  • Extended support of TriangulationPortion to boundary and skeleton triangulations. Since PR #249.
  • Added FESpaceWithLinearConstraints. Since PR #247.
  • Added inner constructor to CartesianDiscreteModel allowing to build a model that represents a subgrid of a larger grid. Since PR #245.

Changed

  • The part associated with the imposition of constraints in the FESpace interface has changed slightly. Since PR #247.
  • Simplified the signature of zero_free_values(::FESpace). Since PR #249.
  • Simplified the signature of zero_initial_guess(op::NonlinearOperator). Since PR #249.
  • Major refactoring in the Assembler interface. Important change: Now, assembly-related functions take the data returned by functions like collect_cell_matrix as it is. Example: the old user code assemble_matrix(assembler,collect_cell_matrix(du,dv,terms)...) now is written simply as assemble_matrix(assembler,collect_cell_matrix(du,dv,terms)), i.e., the unpack of the last argument is not used anymore. In addition, with the new assembler interface, it is possible to customize the assembly process via a so-called AssemblerStrategy object. Since PR #249.
  • Change the types of the sizes and partition fields of CartesianDescriptor to tuples instead of points. Since PR #246.

[0.9.2] - 2020-04-26

Added

  • Automatic differentiation of manufactured solutions. Since PR #236.

[0.9.1] - 2020-04-20

Added

  • Function cell_measure. Since PR #234.

Fixed

  • Several bugs associated with ExtendedFESpace. In particular, we have fixed a bug that showed up when combining ZeroMeanFESpace and ExtendedFESpace. Since PR #234.

[0.9.0] - 2020-04-18

Added

  • Function HomogeneousTrialFESpace. Since PR #226.
  • Function lazy_append in order to lazily append two objects (implemented for AbstractVector, Triangulation, and CellQuadrature). Since PR #220.
  • Support for FE spaces with DOFs defined in the physical space. Since PR #216 and #218.

Changed

  • Replaced non_linear -> nonlinear and NonLinear -> Nonlinear. Since PR #223.
  • The FESpace interface has slightly changed, mainly the return type of functions get_cell_basis and get_cell_dof_basis.. Since PR #216 and #218.

Fixed

  • Bug that showed up in multi-field computations when some field had no contribution to the rhs vector. Since #229.
  • Bug in gradient operator in the void part of ExtendedFESpace objects. Since PR #219.
  • Bug in jumps of quantities restricted to InterfaceTriangulation objects. Since PR #215.

[0.8.0] - 2020-03-17

Added

  • Support for surface-coupled multi-physics. See SurfaceCouplingTests.jl for further details. Since PR #209.
  • Support for constitutive laws with state / historical variables. See IsotropicDamageTests.jl for further details. Since PR #208.
  • Curl-conforming reference FE NedelecRefFE and corresponding FE space constructor since PR #199.
  • New constructors AffineFETermFromCellMatVec and FETermFromCellJacRes that provides full control in the definition of cell matrices and vectors. Since PR #191.
  • Support for simultaneous integration of matrices and vectors. Since PR #191.

Changed

  • Renaming NonLinear to Nonlinear since it is one word and it is not consistent with style

  • The definition of interpolation order in Raviart-Thomas and Nédélec reference FEs has changed. Now, the divergence of functions in the Raviart-Thomas space of order k belongs to P_k or Q_k depending on the underlying polytope. Idem for Nédelec, but using the curl instead of the divergence. Since PR #212.

  • The order in which test and trial spaces are written in the code has changed and also the other in the arguments of functions defining bi-linear and linear forms, and weak residuals and Jacobians. This affects everybody that is using Gridap, even the most basic users. Now, we write the trial space before the test one in all methods taking two spaces in their arguments. E.g., we have changed AffineFEOperator(V,U,terms...) to AffineFEOperator(U,V,terms...), where U is the trial and V is the test space. For functions defining weak forms, now we have: The new signatures for bi-linear and a linear forms are a(u,v), l(v), where u is a trial function and v is a test one. For weak Jacobians and residuals jac(u,du,v) and res(u,v), where u is the (trial) function in which we evaluate these quantities, du is the direction in which we evaluate the Jacobian and v is a test function. Since PR #195 and PR #197.

  • The part related with the application of constraints in the FESpace interface has changed. Since PR #191.

Fixed

  • Bug in 1d Cartesian grids. Since PR #192.

[0.7.1] - 2020-02-18

Added

  • New DirichletFESpace that can be used to compute matrices and vectors associated with the Dirichlet DOFs. Since commit 972afcc

[0.7.0] - 2020-02-13

This version is a major refactoring of the project which is not summarized here for the sake of brevity. Most of the functionality of v0.6.0 is available in v0.7.0, but with a possibly slightly different API. See here the changes in the sources of the Gridap Tutorials between versions 0.6.0 and 0.7.0 to effectively see the major changes in the API.

[0.6.0] - 2020-01-24

Added

  • New GenericRefFE. Since commit 876ef1e
  • New NedelecRefFE constructor that generates Nedelec FEs of arbitrary order in 2D and 3D on hex. Since commit 876ef1e
  • New keyword argument map in the constructor of CartesianModel, which allows one to transform the original domain, by default [0,1]^d to a new domain through a homeomorphic map. Since commit 30cc4bc
  • New keyword argument map in the constructor of CartesianGrid and a new map attribute in this structure, since commit 30cc4bc
  • CartesianGridPoints has new attribute map since commit 30cc4bc
  • Added SparseMatricesCSR support to SparseMatrixAssembler and MultiSparseMatrixAssembler in PR #118.

Changed

  • The RaviartThomasRefFE has now been replaced by GenericRefFE, and the constructor for Raviart-Thomas FEs is called RTRefFE. Since commit 876ef1e
  • The default map in the CartesianModel constructor is [0,1]^d instead of [-1,1]^d, since commit 30cc4bc
  • CartesianGrid has attribute map since commit 30cc4bc

[0.5.2] - 2019-10-22

Fixed

  • Incompatibility problem with TensorValues version 0.3.5. Via commit 3c0682a.

[0.5.1] - 2019-10-03

Added

  • Pretty printing for the types most exposed to users. Since PR #109.

Fixed

  • Bug related to ZeroMeanFESpace. Via PR #111.

[0.5.0] - 2019-09-27

Added

  • Added a high level constructor, namely FESpace, to create different types of FE spaces. See issue #100 for more details. Since PR #102.
  • Added ZeroMeanFESpace to construct FE spaces whose functions have zero mean value. Since PR #102.
  • Added Hdiv FE space using Raviart-Thomas reference FEs in 34bfa34
  • Added the corresponding DOF basis for Raviart-Thomas reference FEs for interpolation of fields 60b9021
  • Added an arbitrary order div-conforming Raviart-Thomas reference FE of arbitrary order on quads in commit 60b9021
  • Now, the tags argument is optional when constructing SkeletonTriangulation and BoundaryTriangulation objects from a DiscreteModel. Since commit e6424a3.
  • Added mean operator for quantities restricted to a SkeletonTriangulation. Since commit 83798b4.
  • Extended NormalVector to SkeletonTriangulations. Since commit 5fb8487.
  • Now, TrialFESpaces can be constructed from values instead of functions if the corresponding Dirichlet conditions are constant. Since commit bae237e.
  • Added the possibility of adding new tags to a FaceLabels object via the function add_tag_from_tags! and using it to construct FE spaces. Since commit e9dfac4.
  • Added BackslashSolver to facilitate the usage in Gridap of the build-in Julia backslash linear solver. Since commit 8e3a9b7.
  • Added NLSolver to facilitate the usage in Gridap of the non-linear solvers available in the official Julia package NLsolve. Introduced in commit e5a933f as JuliaNLSolver. Renamed to NLSolver in PR #108.

Changed

  • The Signature of solve! for NumericalSetup objects. The argument for the system matrix has been removed. The information about the matrix is already in the NumericalSetup object. Since commit ac212d3.
  • The signature of solve!(::FEFunction,::FESolver,::FEOperator). Before it was used as cache = solve!(uh,solver,op), now it is used as uh, cache = solve!(uh,solver,op). Since PR #102.
  • Previous ConformingFESpace constructor is H1ConformingFESpace since 34bfa34

Deprecated

  • JuliaNLSolver. Renamed to NLSolver. Since PR #108.
  • Key-word argument order in CellQuadrature constructor. Renamed to degree. Since PR #108.

Fixed

  • Bug in @law macro for more than one FEBasis arguments. Solved via PR #104.
  • Bug in NonlinearFEOperator constructor with default assembler in multi-field computations. Solved via PR #104.
  • Bug in NormalVector for non-Cartesian grids. Solved via PR #98.

[0.4.0] - 2019-09-07

Added

  • Added support to high order simplicial Lagrangian finite elements. Since commit cbefe9b.
  • Now the built-in simplicial grids are oriented. Since commit cbefe9b.
  • Added binary operations between FEFunction and Number, and FEBasis and Number. Since PR #88.
  • Added PDiscRefFE, DiscFESpace, and ConstrainedFESpace. Since PR #88.
  • Now its possible to pass a CellNumber or an Array of numbers into a constitutive law. Useful to identify which is the material of the current Gauss point in multi-material problems. Since commit 62cb2c3.
  • LinearFESolver is now optional for solving a LinearFEOperator. Since commit 5c1caa8.
  • Assembler is now optional to build FEOperator objects. Since commit b1bf517.
  • Binary operations between Function and FEFunction. Since commit a7f22f5.
  • Extended constructions of CLagrangianFESpace and DLagrangianFESpace. diritags and dirimasks are now optional. diritags can now be also a vector of String. Since commit 776b402.
  • Added div, curl, and trace operators. Since commit 5a0f322.
  • Macro @law to facilitate the definition of constitutive laws. Since commit 30b67f2.
  • Definition of linear forms b(v) = inner(v, f) directly from a function f. Since commit bb42847
  • Serialization and de-serialization of DiscreteModel objects into and from json format. Since PR #76.
  • Support for boundary integration (e.g., Neumann BCs) for multi-field computations. Since PR #75.

Changed

  • Signature of LagrangianRefFE constructor. Since commit 529c764.

Fixed

  • Bug in LinearFETerm for multi-field computations. Fixed via commit 2b957d1.
  • Bug in MultiCellArray constructor. Fixed via commit bbc3b1c.
  • Bug in binary operations between FEFunction and FEBasis. Fixed via commit aa49689.

[0.3.0] - 2019-08-06

Added

  • CurlGradMonomialBasis spanning the polynomial space needed for RT elements on n-cubes.
  • CLagrangianFESpace and DLagrangianFESpace types providing an efficient implementation for continuous and discontinuous Lagrangian FE spaces respectively. In contrast to ConfirmingFESpace, the new types allow to select which are the components that are actually prescribed on the Dirichlet boundary. Since PR #64.
  • simplexify function to convert Grid and DiscreteModel objects made of n-cubes to the corresponding counterparts made of n-simplices. Since PR #62.
  • Duffy transformation based integration for n-simplices of arbitrary dimension. Since PR #61.
  • NormalVector to construct the outward normal vector to a given BoundaryTriangulation. Since PR #60.
  • Support for tensor-valued FE computations. Since PR #57.
  • Support for integration on the skeleton of the mesh. This includes SkeletonTriangulation, an integration mesh for the skeleton, restrict function is extended to restrict to the skeleton, jump function to compute jumps of CellFields and CellBasis restricted to the skeleton, extension of FETerms to allow integration on the skeleton. See PR #47
  • Support for Robin boundary conditions. Since commit 946054a
  • Support for Neumann boundary conditions. Since commit 4dcd16f
  • FETerm and AffineFETerm abstract types and several concrete implementations. They allow to deal with problems whose weak form has terms integrated over different geometrical entities. NonlinearFEOperator and LinearFEOperator can be constructed using several terms. Since commit 0f99234
  • Extended Assembler and MultiAssembler to deal with several terms. See issue #42 and PR #43.
  • IdentityCellNumber, an indexable cell number that simply returns the given index. Also efficient implementation of reindex for this type (i.e. do nothing). Available since commit b6b4c32
  • Function restrict for restricting CellField and CellBasis objects to surfaces. Available since commit e981f3c
  • BoundaryTriangulation an integration mesh used to integrate CellField and CellBasis objects restricted on a surface. Available since commit e981f3c
  • NonIterableCellMap, a cell map that has iteration intentionally disabled. Available since commit 956a537.
  • CompressedCellValue, CompressedCellArray, and CompressedCellMap, as well as efficient versions of apply, evaluate, and reindex for these types. See PR #41 for more details.
  • NEWS.md file (a changelog file)

Changed

  • Domains are now in [0,1] instead of [-1,1]. Quadratures and nodes arrays modified accordingly. Since commit 268dfe1.
  • Changed the signature of assemble, apply_constraints, apply_constraints_rows, and apply_constraints_cols to support FE assembly of several terms, which are integrated in different domains. The old API of asseble is still functional, but not for the apply_constraints et al. Since PR #43. Further changed in commit a335aed.

Fixed

  • Bug in generation of the cellwise local to global DOF map for high order interpolations. Fixed via PR #56.
  • Bug in numerical integration. There was a bug for computations where the number of cell DOFs was different from the number of integration points. Fixed via commit 0b3d4bf

[0.2.0] - 2019-06-29

A changelog is not maintained for this version.

This version introduces the core finite element machinery for linear and non-linear problems, single field and multi-field problems with terms integrated over the interior of the computational domain.

[0.1.0] - 2019-05-20

A changelog is not maintained for this version.

This version is non functional. It is just a tag for registering the package.