← Back to team overview

dolfin team mailing list archive

Re: [noreply@xxxxxxxxxxxxx: [Branch ~dolfin-core/dolfin/main] Rev 6190: Return shared_ptr for Mesh from FunctionSpace and regenerate form files.]

 

On 13 September 2011 14:47, Anders Logg <logg@xxxxxxxxx> wrote:
> On Tue, Sep 13, 2011 at 02:40:04PM +0100, Garth N. Wells wrote:
>> On 13 September 2011 14:31, Anders Logg <logg@xxxxxxxxx> wrote:
>> > Why was this changed? I thought we had agreed that (in C++), functions
>> > that return data should be
>> >
>> >  [const] Foo& foo() [const];
>> >
>> > and in addition we offer
>> >
>> >  boost::shared_ptr<[const] Foo> foo_shared_ptr() [const];
>> >
>>
>> I thought that we agreed to use a single function in place of two for
>> low-level library functions to:
>>
>>  - reduce the amount of code
>>  - avoid possible memory errors
>>  - avoid the need to ignore functions in the swig interface (which has
>> been a source of bugs).
>
> Agree, in all places except a few that are most often exposed to
> users.
>
>> I advocate the shared_ptr interface over extra work and complicated
>> memory issues. I don't see an advantage to a reference version in
>> parts of the code that only advanced users will see.
>
> Me neither, but isn't FunctionSpace::mesh() one of the more basic
> calls that many users will see, like Function::vector(),
> Function::function_space() and the like?
>

Keeping a reference version for Function::vector() I agree with. I
don't believe that many users are exposed to FunctionSpace::mesh(). In
simple usage cases the user has the original mesh that was used to
create a FunctionSpace in scope and readily accessible. If they don't
have it in scope, then they're in trouble anyway with the simple
interface because the FunctionSpace will have a dangling pointer to a
Mesh.

FunctionSpace::mesh(), unlike Function::vector(), is not exposed in
any of the C++ demos.

Garth

> --
> Anders
>
>
>> Garth
>>
>> > And in the Python interface, we ignore foo and map foo_shared_ptr to
>> > foo.
>> >
>> > Accessing the mesh is very handy from C++ and this change will force
>> > users of the C++ interface to use shared pointers which I think
>> > should be optional.
>> >
>> >
>> >
>> > ---------- Forwarded message ----------
>> > From: noreply@xxxxxxxxxxxxx
>> > To: Anders Logg <logg@xxxxxxxxx>
>> > Date: Tue, 13 Sep 2011 12:05:16 -0000
>> > Subject: [Branch ~dolfin-core/dolfin/main] Rev 6190: Return shared_ptr for Mesh from FunctionSpace and regenerate form files.
>> > ------------------------------------------------------------
>> > revno: 6190
>> > committer: Garth N. Wells <gnw20@xxxxxxxxx>
>> > branch nick: dolfin-main
>> > timestamp: Tue 2011-09-13 12:34:53 +0100
>> > message:
>> >  Return shared_ptr for Mesh from FunctionSpace and regenerate form files.
>> > modified:
>> >  bench/fem/assembly/cpp/forms/Elasticity3D.h
>> >  bench/fem/assembly/cpp/forms/NSEMomentum3D.h
>> >  bench/fem/assembly/cpp/forms/Poisson2DP1.h
>> >  bench/fem/assembly/cpp/forms/Poisson2DP2.h
>> >  bench/fem/assembly/cpp/forms/Poisson2DP3.h
>> >  bench/fem/assembly/cpp/forms/StabStokes2D.h
>> >  bench/fem/assembly/cpp/forms/THStokes2D.h
>> >  bench/fem/convergence/cpp/forms/Poisson2D_1.h
>> >  bench/fem/convergence/cpp/forms/Poisson2D_2.h
>> >  bench/fem/convergence/cpp/forms/Poisson2D_3.h
>> >  bench/fem/convergence/cpp/forms/Poisson2D_4.h
>> >  bench/fem/convergence/cpp/forms/Poisson2D_5.h
>> >  bench/fem/convergence/cpp/forms/Poisson3D_1.h
>> >  bench/fem/convergence/cpp/forms/Poisson3D_2.h
>> >  bench/fem/convergence/cpp/forms/Poisson3D_3.h
>> >  bench/fem/convergence/cpp/forms/Poisson3D_4.h
>> >  bench/fem/convergence/cpp/forms/Poisson3D_5.h
>> >  bench/fem/multicore/cpp/NavierStokes.h
>> >  bench/fem/multicore/cpp/Poisson.h
>> >  bench/fem/speedup/cpp/Poisson.h
>> >  bench/function/evaluation/cpp/P1.h
>> >  demo/la/eigenvalue/cpp/StiffnessMatrix.h
>> >  demo/pde/biharmonic/cpp/Biharmonic.h
>> >  demo/pde/cahn-hilliard/cpp/CahnHilliard2D.cpp
>> >  demo/pde/cahn-hilliard/cpp/CahnHilliard2D.h
>> >  demo/pde/cahn-hilliard/cpp/CahnHilliard3D.cpp
>> >  demo/pde/cahn-hilliard/cpp/CahnHilliard3D.h
>> >  demo/pde/hyperelasticity/cpp/HyperElasticity.h
>> >  demo/pde/mixed-poisson/cpp/MixedPoisson.h
>> >  demo/pde/navier-stokes/cpp/PressureUpdate.h
>> >  demo/pde/navier-stokes/cpp/TentativeVelocity.h
>> >  demo/pde/navier-stokes/cpp/VelocityUpdate.h
>> >  demo/pde/poisson/cpp/Poisson.h
>> >  demo/pde/stokes-iterative/cpp/Stokes.h
>> >  demo/pde/stokes-iterative/cpp/StokesPreconditioner.h
>> >  demo/undocumented/advection-diffusion/cpp/AdvectionDiffusion.h
>> >  demo/undocumented/advection-diffusion/cpp/Velocity.h
>> >  demo/undocumented/assembly/cpp/ReactionDiffusion.h
>> >  demo/undocumented/auto-adaptive-navier-stokes/cpp/AdaptiveNavierStokes.h
>> >  demo/undocumented/auto-adaptive-poisson/cpp/AdaptivePoisson.h
>> >  demo/undocumented/bcs/cpp/Poisson.h
>> >  demo/undocumented/block-matrix/cpp/StiffnessMatrix.h
>> >  demo/undocumented/conditional/cpp/Conditional.h
>> >  demo/undocumented/curl-curl/cpp/CurrentDensity.h
>> >  demo/undocumented/curl-curl/cpp/EddyCurrents.h
>> >  demo/undocumented/dg-advection-diffusion/cpp/AdvectionDiffusion.h
>> >  demo/undocumented/dg-advection-diffusion/cpp/Projection.h
>> >  demo/undocumented/dg-advection-diffusion/cpp/Velocity.h
>> >  demo/undocumented/dg-poisson/cpp/Poisson.h
>> >  demo/undocumented/elasticity/cpp/Elasticity.h
>> >  demo/undocumented/elastodynamics/cpp/DG0_eps_xx.h
>> >  demo/undocumented/elastodynamics/cpp/ElastoDynamics.h
>> >  demo/undocumented/eval/cpp/Projection.h
>> >  demo/undocumented/extrapolation/cpp/P1.h
>> >  demo/undocumented/extrapolation/cpp/P2.h
>> >  demo/undocumented/functional/cpp/EnergyNorm.h
>> >  demo/undocumented/lift-drag/cpp/Drag.h
>> >  demo/undocumented/lift-drag/cpp/Lift.h
>> >  demo/undocumented/lift-drag/cpp/Pressure.h
>> >  demo/undocumented/neumann-poisson/cpp/Poisson.h
>> >  demo/undocumented/nonlinear-poisson/cpp/NonlinearPoisson.h
>> >  demo/undocumented/nonmatching-interpolation/cpp/P1.h
>> >  demo/undocumented/nonmatching-interpolation/cpp/P3.h
>> >  demo/undocumented/nonmatching-projection/cpp/P1_projection.h
>> >  demo/undocumented/nonmatching-projection/cpp/P3.h
>> >  demo/undocumented/periodic/cpp/Poisson.h
>> >  demo/undocumented/poisson1D/cpp/Poisson.h
>> >  demo/undocumented/reaction-diffusion/cpp/ReactionDiffusion.h
>> >  demo/undocumented/simple/cpp/MassMatrix3D.h
>> >  demo/undocumented/simple/cpp/StiffnessMatrix3D.h
>> >  demo/undocumented/spatial-coordinates/cpp/SpatialCoordinates.h
>> >  demo/undocumented/stokes-stabilized/cpp/Stokes.h
>> >  demo/undocumented/stokes-taylor-hood/cpp/Stokes.h
>> >  demo/undocumented/sym-dirichlet-bc/cpp/Poisson.h
>> >  demo/undocumented/waveguide/cpp/Forms.h
>> >  dolfin/adaptivity/ErrorControl.cpp
>> >  dolfin/adaptivity/Extrapolation.cpp
>> >  dolfin/adaptivity/GenericAdaptiveVariationalSolver.cpp
>> >  dolfin/adaptivity/adapt.cpp
>> >  dolfin/ale/Poisson1D.h
>> >  dolfin/ale/Poisson2D.h
>> >  dolfin/ale/Poisson3D.h
>> >  dolfin/fem/BoundaryCondition.cpp
>> >  dolfin/fem/DirichletBC.cpp
>> >  dolfin/fem/Form.cpp
>> >  dolfin/fem/PeriodicBC.cpp
>> >  dolfin/fem/PointSource.cpp
>> >  dolfin/function/Function.cpp
>> >  dolfin/function/FunctionSpace.cpp
>> >  dolfin/function/FunctionSpace.h
>> >  dolfin/function/SubSpace.cpp
>> >  dolfin/io/RAWFile.cpp
>> >  dolfin/io/VTKFile.cpp
>> >  dolfin/io/VTKWriter.cpp
>> >  dolfin/io/XYZFile.cpp
>> >  dolfin/plot/plot.cpp
>> >  site-packages/dolfin_utils/wrappers/goalfunctional.py
>> >  test/unit/function/cpp/Projection.h
>> > The size of the diff (167257 lines) is larger than your specified limit of 1000 lines
>> >
>> >
>> > Your team DOLFIN Core Team is subscribed to branch lp:dolfin.
>> > To unsubscribe from this branch go to https://code.launchpad.net/~dolfin-core/dolfin/main/+edit-subscription
>> >
>> > _______________________________________________
>> > Mailing list: https://launchpad.net/~dolfin
>> > Post to     : dolfin@xxxxxxxxxxxxxxxxxxx
>> > Unsubscribe : https://launchpad.net/~dolfin
>> > More help   : https://help.launchpad.net/ListHelp
>> >
>> >
>


Follow ups

References