Reference documentation for deal.II version 9.4.1

This is the list of changes made between the deal.II releases listed above. made to the three sublibraries base, lac, and deal.II, as well as changes to the general infrastructure, documentation, etc.
All entries are signed with the names of the author. Regular contributor's names are abbreviated by WB (Wolfgang Bangerth), GK (Guido Kanschat), RH (Ralf Hartmann).
Following are a few modifications to the library that unfortunately are incompatible with previous versions of the library, but which we deem necessary for the future maintainability of the library. Unfortunately, some of these changes will require modifications to application programs. We apologize for the inconvenience this causes.
Removed: The FullMatrix::add_diag
function was removed. It offered functionality of questionable use in the first place, and its implementation was correct only for matrices of size \(3 \times 3\), \(4 \times 4\) and \(8 \times 8\) without anyone noticing for 10 years. Consequently, it can't have been used very frequently.
(WB 2007/12/09)
Changed: The namespace deal_II_numbers has been renamed numbers. The old name stemmed from a time when not everything was already in namespace dealii
. The old name is retained via a namespace alias but is deprecated and will eventually be removed.
(WB 2007/11/02)
Changed: When writing output files in UCD format using either the DataOutBase or the GridOut class, we used to write a preamble at the beginning of the file that included the date and time the file was created. However, several visualization programs get confused when confronted with comments at the beginning of UCD files. Rather than printing a sensible error message, they usually simply refuse to show any output, making it very hard to track down the actual cause.
The classes mentioned above previously allowed to suppress writing a preamble by setting a flag in the DataOutBase::UcdFlags or GridOutFlags::Ucd structures. Given how complicated it is to find the actual source of trouble, the default for these flags has been changed to not produce this preamble any more. If desired, the flag can be used to still produce a preamble.
(WB 2007/10/30)
Changed: The version number of the deal.II intermediate format written by DataOutBase::write_deal_II_intermediate has been increased to 3 to accommodate the fact that we now support writing vectorvalued data to output files in at least some output formats. (Previously, vectorvalued date was written as a collection of scalar fields.) Since we can only read files written in intermediate format that has the same number as the files we generate, this means that files written with previous format numbers can now no longer be read.
(WB 2007/10/11)
Changed: FilteredMatrix now can be applied to any matrix having the standard set of vmult
functions. In order to achieve this, its interface had to be overhauled. Only the VECTOR
template argument remains. Furthermore, instead of PreconditionJacobi being applied to FilteredMatrix, FilteredMatrix can now be applied to any preconditioner.
(GK 2007/09/25)
Changed: The deprecated typedefs internal::Triangulation::Line
, internal::Triangulation::Quad
, and internal::Triangulation::Hexahedron
have been removed.
(WB 2007/09/07)
New: step33, written mainly by David Neckels, illustrates how the Euler equations of compressible inviscid gas dynamics can be solved. More generally, the program also provides a framework for the solution of arbitrary hyperbolic, nonlinear conservation laws.
(WB, 2008/05/23)
Improved: step16 has been updated to show the use of Multigrid with Dirichlet boundary conditions.
(GK, 2007/05/13)
New: step22, written by Martin Kronbichler and Wolfgang Bangerth, shows how to solve, and in particular how to solve efficiently, the Stokes equations that describe viscous incompressible fluid flow.
(WB, 2008/03/28)
New: step29 demonstrates how problems involving complex numbers can be implemented by viewing real and imaginary parts of a complexvalued solution as the two components of a vectorvalued function.
(Moritz Allmaras, 2007/10/31)
New: A significantly simpler way to code the assembly of linear systems for problems with more than one solution variable has been implemented. This is explained in detail in the report on Handling vector valued problems and tutorial programs step20 and step21 have been converted as well.
(WB, 2008/02/23)
Improved: On Mac OS X, the operating system provides for "frameworks", which are essentially collections of shared libraries. We now link with the "Accelerate" framework (from Mac OS X 10.4 onwards) or the "vecLib" framework (for previous versions) instead of the individual BLAS and LAPACK libraries if they are needed. This insulates us from having to use the actual name of these libraries, which may be subject to change, and it may also link with optimized or vectorized libraries if they are available.
(Eh Tan, WB 2007/10/22)
New: The FunctionParser::default_variable_names() function returns default names for variables in the given space dimension. For example, in 2d, this would be "x,y".
(WB 2008/05/12)
New: ConvergenceTable::evaluate_convergence_rates has been incomplete for a long time. An implementation for reduction_rate and reduction_rate_log2 has been added in the case where one specifies a reference column that allows to compute error rates also for local refinement.
(Luca Heltai 2008/04/16)
Fixed: The Utilities::int_to_string function did not work for 5 or 6 digits, while its counterpart Utilities::needed_digits does. This is now fixed, and the two are coherent.
(Luca Heltai 2008/04/16)
New: The SymmetricTensor class now has a constructor that creates an object from an array consisting its independent components.
(WB 2008/02/21)
New: There are now output operators (i.e. operator<<
) for the SymmetricTensor class.
(WB 2008/02/12)
Improved: LogStream is now thread safe and output lines from different threads are separated now. Additionally, a function LogStream::log_thread_id() has been added to log the id of the thread printing the message. The semantics has slightly changed in that the header is now printed at the time when the line is finished.
(GK 2008/01/22)
New: There is a function Threads::this_thread_id() now returning an integer id of the current thread.
(GK 2008/01/22)
Improved: Quadrature now has an assignment operator.
(GK 2007/12/27)
Fixed: MultithreadInfo::n_cpus now gives the correct result also on my Intel MacBook Pro.
(Luca Heltai 2007/11/12)
New: There is now a template numbers::NumberTraits that provides the means to implement linear algebra and other algorithms for both real and complex data types.
(WB 2007/11/03)
New: The DataOutBase class and derived classes can now deal with data that is logically vectorvalued, i.e. derived classes can pass down information that some of the components of the output should be grouped together as vectors, instead of being treated as a number of separate scalar fields. This allows visualization programs to display these components as vectorfields right away, without the need to manually group them together into a vector.
While all output format writers receive the information necessary to do this, currently only the VTK reader in DataOutBase::write_vtk as well as the deal.II intermediate format writer make use of it.
The use of this ability is explained in the step22 and step33 tutorial programs.
(WB 2007/10/11)
New: Macro AssertDimension introduced for easier handling of ExcDimensionMismatch.
(GK 2007/09/13)
Changed: A few simple PETSc wrappers for helper functions PETScWrappers::MatrixBase::transpose, PETScWrappers::MatrixBase::is_symmetric, PETScWrappers::MatrixBase::is_hermitian, and PETScWrappers::MatrixBase::write_ascii have been added.
(Toby D. Young 2008/04/30)
Fixed: The SolverQMRS class could not deal with block vectors and block matrices. This is now fixed.
(WB 2008/3/20)
Fixed: The implementation of SparseILU::decompose was rather inefficient in that it accessed random elements of the matrix in its inner loop. It has been replaced by the algorithm given in the book by Yves Saad: "Iterative methods for sparse linear systems", second edition, in section 10.3.2. The code is now about 8 times faster than before.
(WB 2008/2/29)
Fixed: The implementation of SparseILU::vmult very needlessly called SparseMatrix::vmult just to throw away the (nonsensical) result away immediately. This is now fixed.
(WB 2008/2/27)
New: The functionality print_gnuplot is now also available for sparsity patterns derived from the class BlockSparsityPatternBase, e.g. for BlockSparsityPattern in the same way as for general sparsity patterns.
(Martin Kronbichler 2008/2/26)
Fixed: The PETScWrappers::MatrixBase::clear_row and PETScWrappers::MatrixBase::clear_rows functions had a memory leak that made PETSc become progressively slower over time. This is now fixed.
(Daniel Goldberg 2008/2/20)
Improved: All GrowingVectorMemory objects that have the same template argument will access the same memory pool. Therefore, it is now not a crime any more to just create a memory pool and discard later. Furthermore, logging of statistics has been switched off by default, such that linear solvers remain silent.
(GK 2007/12/16)
Fixed: Vector::operator/= can't work when the scaling factor is zero, but it happened to check whether the factor was positive. That's of course bogus, the check should have been whether it is nonzero. This has now been fixed.
(WB 2007/11/03)
New: A class ScaledMatrix was introduced which combines the vector operations of an underlying matrix with a scaling.
(GK 2007/10/30)
Improved: FilteredMatrix has an iterator now that allows users to access the constraints individually.
(GK 2007/10/30)
Improved: GridOutFlags::Gnuplot now takes an additional curved_inner_cells parameter. Similar to curved_inner_cells in DataOut this allows the output of interior cells with curved boundaries. This is useful for MappingQEulerian or similar mappings.
(RH 2008/05/12)
Fixed: Functions::FEFieldFunction<dim>.set_active_cell was declared inline in the declaration but not in the implementation. This is now fixed.
(Luca Heltai 2008/04/16)
New: The FE_RaviartThomasNodal now supports hp functionality.
(Zhu Liang, 2008/04/04)
Improved: The StraightBoundary object Triangulation::straight_boundary is now made public. This allows to use Triangulation::straight_boundary instead of recreating a StraightBoundary object when needed.
(RH, 2008/04/04)
Extendend: SolutionTransfer has an additional template parameters DH for the dof handler class to use, which defaults to DoFHandler. However, an instantiation is also provided for hp::DoFHandler.
(Tobias Leicht, 2008/04/03)
Extended: DataOut has an extended interface now which enables a user decision, which cells shall be written as curved cells using the provided mapping: none, only cells at the boundary, or all cells. The last choice can be useful when a mapping like MappingQEulerian is employed.
(Tobias Leicht, 2008/03/20)
Changed: From now on the SolutionTransfer class does not use any user pointers of the underlying Triangulation object. Thus several SolutionTransfer instances with different DoFHandler objects can be used simultaneously. It is no longer necessary to assemble all solution vectors into a new one with a combined DoFHandler.
(Tobias Leicht, 2008/03/18)
New: There is now a namespace DoFRenumbering::boost that contains the implementation of interfaces to three reordering strategies provided by the Boost Graph Library, namely DoFRenumbering::boost::Cuthill_McKee, DoFRenumbering::boost::king_ordering, and DoFRenumbering::boost::minimum_degree.
(WB 2008/03/07)
Changed: The DoFRenumbering class has been converted into a namespace. Since all its members were static, this hasn't changed anything fundamental and use of the functions therein should work in exactly the same way as before.
(WB 2008/03/07)
New: The new TriaObjectAccessor::set_all_boundary_indicators function does not only set the boundary indicator of the current object but of all that bound it as well. For example, in 3d, if TriaObjectAccessor::set_boundary_indicator is called on a face, then the boundary indicator of the 4 edges that bound the face remain unchanged. On the other hand, the boundary indicators of face and edges are all set at the same time using the new function.
(WB 2008/02/26)
Improved: The tria pointer and the GridIn::debug_output_grid function are changed from private to protected to be accessible in derived classes.
(RH 2008/02/24)
New: The new Boundary::get_new_point_on_face and get_intermediate_points_on_face functions offer an dimension independent interface to the Boundary class.
(RH 2008/02/24)
New: There is now a FEFaceValuesBase::get_face_index function which returns the index of the face selected the last time the reinit() function was called.
(RH 2008/02/15)
New: The MappingQEulerian class provides an arbitrary order Eulerian mapping where the nodes of a mesh are displaced by a previously computed vector field.
(Joshua White 2008/02/05)
New: The MappingQ class constructor now takes a parameter that determines whether a higher order mapping should also be used for interior cells. By default, the higher order mapping is only used for cells on the boundary; cells in the interior are bounded by straight edges described by a (bi, tri)linear mapping.
(WB 2008/02/05)
New: The function VectorTools::compute_no_normal_flux_constraints computes the constraints that correspond to boundary conditions of the form \(\vec u \cdot \vec n = 0\).
(WB 2008/01/23)
Fixed: Neither ConstraintMatrix::print nor ConstraintMatrix::write_dot produced any output for constraints of the form \(x_i=0\), i.e. where the right hand side is a trivial linear combination of other degrees of freedom. This is now fixed.
(WB 2008/01/23)
Improved: GridGenerator::subdivided_hyper_rectangle now also colorizes cells according to the octant they are in.
(GK 2007/12/20)
New: The DataOut, DataOutRotation, DataOutStack, and DataOutFaces can now deal with vectorvalued data if the functions in DataOutBase that write in a particular graphical output format can deal with it. Previously, if a finite element field had more than one component, they were all output as logically independent scalar components; most visualization programs then allowed to display vector fields by composing them of individual scalar fields for each vector component.
With the new scheme, the DataOut_DoFData::add_data_vector() functions inherited by the classes listed above take an additional parameter that may be used to indicate that certain components of the data logically form a vector field. The output formats for which this is presently implemented then indicate this fact in the output file. The mechanism is shown in action in step22 and step33.
(WB 2007/10/14)
Improved: UpdateFlags::update_q_points has been renamed to UpdateFlags::update_quadrature_points. Additional update flags for support points have been added without functionality, yet.
(GK 2007/10/12)
Improved: The number of blocks of an FESystem was properly defined and the constructors changed accordingly. At least non of the test programs noticed the change.
(GK 2007/10/03)
Improved: In an effort to make names more consistent, second derivatives in FEValuesBase and UpdateFlags have been renamed to Hessians. Thus, the clash between the forms 2nd
and second
has been removed. Old function and enum names are available for compatibility but have been marked deprecated.
(GK 2007/09/12)
Fixed+New: The GridOutFlags::Ucd and GridOutFlags::Msh structures now take a new parameter write_lines
to output lines with boundary id different from 0 in threedimensional meshes. This fixes an annoying bug for which meshes with ids different from zero where not written in a compatible way, and if reread with the corresponding GridIn functions, would not yield the same mesh upon refinement.
(Luca Heltai 2007/09/10)
Extended: The possibilities of graphical output via the DataOut, DataOutFaces and DataOutRotation classes have been extended by the ability to perform a postprocessing step before a given data vector is written to the output. This way, derived variables can be written instead of the original data. To this end, there is a new version of DataOut_DoFData::add_data_vector taking a data vector and a DataPostprocessor, which performs the actual calculation of new data based on the values and possibly derivatives of the original data at each point of the patch.
(Tobias Leicht 2007/09/10)