Hi there!

We have been recently adapting our code that worked with Eigen versions 3.2.2 - 3.2.5 to work with Eigen 3.3-alpha1 (I also tried with the mercurial of today without success). And I am running into issues when assigning (= operator) SparseMatrix, the compiler main complaint being "/usr/include/eigen3/Eigen/src/SparseCore/SparseMatrixBase.h:376:34: error: ‘void Eigen::SparseMatrixBase<Derived>::evalTo(Dest&) const [with Dest = Eigen::Matrix<double, -1, -1>; Derived = Eigen::TriangularView<const Eigen::SparseMatrix<double>, 2u>]’ is private".

A bigger snapshot of the offending code is below, but the compilation problem is in these lines using the = operator:

Code: Select all

           SqrJT=Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> >(SJ.topRows(count).transpose());

            if (constrNum >= paramsNum)
                R = SqrJT.matrixR().triangularView<Eigen::Upper>();
            else
                R = SqrJT.matrixR().topRows(constrNum)
                                    .triangularView<Eigen::Upper>();


I think I can get away in the case of SqrJT (first line above), by using the standard constructor and then calling compute(). However, I have not managed to get away with the assignation of the R matrix.

For the sake of completeness, this code does compile perfectly:

Code: Select all

            SqrJT.compute(SJ.topRows(count).transpose());
           
            paramsNum = SqrJT.rows();
            constrNum = SqrJT.cols();
            SqrJT.setPivotThreshold(qrpivotThreshold);
            rank = SqrJT.rank();
           
            if (constrNum >= paramsNum)
                /*R = */SqrJT.matrixR().triangularView<Eigen::Upper>();
            else
                /*R =*/ SqrJT.matrixR().topRows(constrNum)
                .triangularView<Eigen::Upper>(); 


But as you can imagine, my R matrix does not have the result of the QR decomposition, as the assignation is commented out.

This is the larger code snapshot I was referring to:

Code: Select all

    Eigen::MatrixXd J(clist.size(), plist.size());
    int count=0;
    for (std::vector<Constraint *>::iterator constr=clist.begin();
         constr != clist.end(); ++constr) {
        (*constr)->revertParams();
        if ((*constr)->getTag() >= 0) {
            count++;
            for (int j=0; j < int(plist.size()); j++)
                J(count-1,j) = (*constr)->grad(plist[j]);
        }
    }

    Eigen::SparseMatrix<double> SJ;
   
    if(qrAlgorithm==EigenSparseQR){
        // this creation is not optimized (done using triplets)
        // however the time this takes is negligible compared to the
        // time the QR decomposition itself takes
        SJ = J.sparseView();
        SJ.makeCompressed();
    }
   
    Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> > SqrJT;

    Eigen::MatrixXd R;
    int paramsNum = 0;
    int constrNum = 0;
    int rank = 0;
    Eigen::FullPivHouseholderQR<Eigen::MatrixXd> qrJT;

     if (SJ.rows() > 0) {
            SqrJT=Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> >(SJ.topRows(count).transpose());           
            paramsNum = SqrJT.rows();
            constrNum = SqrJT.cols();
            SqrJT.setPivotThreshold(qrpivotThreshold);
            rank = SqrJT.rank();
           
            if (constrNum >= paramsNum)
                R = SqrJT.matrixR().triangularView<Eigen::Upper>();
            else
                R = SqrJT.matrixR().topRows(constrNum)
                                    .triangularView<Eigen::Upper>();   
        }
    }



In case you find it useful, the whole code file is available here:
https://github.com/abdullahtahiriyo/Fre ... cs/GCS.cpp

The relevant member function is "int System::diagnose(Algorithm alg)" and starts in line 1793.

The whole relevant compiler output is:

Code: Select all

[ 93%] Built target ImportGui
In file included from /usr/include/eigen3/Eigen/SparseCore:30:0,
                 from /usr/include/eigen3/Eigen/Sparse:19,
                 from /home/abdullah/github/free-cad-code/src/Mod/Sketcher/App/planegcs/GCS.cpp:57:
/usr/include/eigen3/Eigen/src/Core/AssignEvaluator.h: In instantiation of ‘static void Eigen::internal::Assignment<DstXprType, SrcXprType, Functor, Eigen::internal::EigenBase2EigenBase, Scalar>::run(DstXprType&, const SrcXprType&, const Eigen::internal::assign_op<typename DstXprType::Scalar>&) [with DstXprType = Eigen::Matrix<double, -1, -1>; SrcXprType = Eigen::TriangularView<const Eigen::SparseMatrix<double>, 2u>; Functor = Eigen::internal::assign_op<double>; Scalar = double; typename DstXprType::Scalar = double]’:
/usr/include/eigen3/Eigen/src/Core/AssignEvaluator.h:765:70:   required from ‘void Eigen::internal::call_assignment_no_alias(Dst&, const Src&, const Func&) [with Dst = Eigen::Matrix<double, -1, -1>; Src = Eigen::TriangularView<const Eigen::SparseMatrix<double>, 2u>; Func = Eigen::internal::assign_op<double>]’
/usr/include/eigen3/Eigen/src/Core/AssignEvaluator.h:718:42:   required from ‘void Eigen::internal::call_assignment(Dst&, const Src&, const Func&, typename Eigen::internal::enable_if<(Eigen::internal::evaluator_traits<Rhs>::AssumeAliasing == 0), void*>::type) [with Dst = Eigen::Matrix<double, -1, -1>; Src = Eigen::TriangularView<const Eigen::SparseMatrix<double>, 2u>; Func = Eigen::internal::assign_op<double>; typename Eigen::internal::enable_if<(Eigen::internal::evaluator_traits<Rhs>::AssumeAliasing == 0), void*>::type = void*]’
/usr/include/eigen3/Eigen/src/Core/AssignEvaluator.h:699:72:   required from ‘void Eigen::internal::call_assignment(Dst&, const Src&) [with Dst = Eigen::Matrix<double, -1, -1>; Src = Eigen::TriangularView<const Eigen::SparseMatrix<double>, 2u>]’
/usr/include/eigen3/Eigen/src/Core/Assign.h:75:55:   required from ‘Derived& Eigen::MatrixBase<Derived>::operator=(const Eigen::EigenBase<OtherDerived>&) [with OtherDerived = Eigen::TriangularView<const Eigen::SparseMatrix<double>, 2u>; Derived = Eigen::Matrix<double, -1, -1>]’
/usr/include/eigen3/Eigen/src/Core/PlainObjectBase.h:506:38:   required from ‘Derived& Eigen::PlainObjectBase<Derived>::operator=(const Eigen::EigenBase<OtherDerived>&) [with OtherDerived = Eigen::TriangularView<const Eigen::SparseMatrix<double>, 2u>; Derived = Eigen::Matrix<double, -1, -1>]’
/usr/include/eigen3/Eigen/src/Core/Matrix.h:238:35:   required from ‘Eigen::Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>& Eigen::Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>::operator=(const Eigen::EigenBase<OtherDerived>&) [with OtherDerived = Eigen::TriangularView<const Eigen::SparseMatrix<double>, 2u>; _Scalar = double; int _Rows = -1; int _Cols = -1; int _Options = 0; int _MaxRows = -1; int _MaxCols = -1]’
/home/abdullah/github/free-cad-code/src/Mod/Sketcher/App/planegcs/GCS.cpp:1922:19:   required from here
/usr/include/eigen3/Eigen/src/SparseCore/SparseMatrixBase.h:376:34: error: ‘void Eigen::SparseMatrixBase<Derived>::evalTo(Dest&) const [with Dest = Eigen::Matrix<double, -1, -1>; Derived = Eigen::TriangularView<const Eigen::SparseMatrix<double>, 2u>]’ is private
     template<typename Dest> void evalTo(Dest &) const;
                                  ^
In file included from /usr/include/eigen3/Eigen/Core:349:0,
                 from /home/abdullah/github/free-cad-code/src/Mod/Sketcher/App/planegcs/SubSystem.h:29,
                 from /home/abdullah/github/free-cad-code/src/Mod/Sketcher/App/planegcs/GCS.h:26,
                 from /home/abdullah/github/free-cad-code/src/Mod/Sketcher/App/planegcs/GCS.cpp:32:
/usr/include/eigen3/Eigen/src/Core/AssignEvaluator.h:820:5: error: within this context
     src.evalTo(dst);
     ^
make[2]: *** [src/Mod/Sketcher/App/CMakeFiles/Sketcher.dir/planegcs/GCS.cpp.o] Error 1
make[1]: *** [src/Mod/Sketcher/App/CMakeFiles/Sketcher.dir/all] Error 2
make: *** [all] Error 2


I wonder if this is a bug in Eigen 3.3 as this used to work fine in Eigen 3.2, or it is intended and I should change my code somehow.

Thanks in advance,
Abdullah

Fixed, thank you for the report.

https://bitbucket.org/eigen/eigen/commits/c70219b49063/

You are welcome!!

I have just tested it. Thank you very much for fixing it so fast

Regards,
Abdullah

Just one thing you might want to fix. This was working in Eigen 3.2.2-3.2.5:

Code: Select all

SqrJT=Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> >(SJ.topRows(count).transpose());           


Now it does not, with this compiler error:

Code: Select all

[ 93%] Built target ImportGui
In file included from /usr/include/eigen3/Eigen/Core:295:0,
                 from /home/abdullah/github/free-cad-code/src/Mod/Sketcher/App/planegcs/SubSystem.h:29,
                 from /home/abdullah/github/free-cad-code/src/Mod/Sketcher/App/planegcs/GCS.h:26,
                 from /home/abdullah/github/free-cad-code/src/Mod/Sketcher/App/planegcs/GCS.cpp:32:
/usr/include/eigen3/Eigen/src/Core/util/Meta.h: In member function ‘Eigen::SparseSolverBase<Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> > >& Eigen::SparseSolverBase<Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> > >::operator=(const Eigen::SparseSolverBase<Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> > >&)’:
/usr/include/eigen3/Eigen/src/Core/util/Meta.h:195:40: error: ‘const Eigen::internal::noncopyable& Eigen::internal::noncopyable::operator=(const Eigen::internal::noncopyable&)’ is private
   EIGEN_DEVICE_FUNC const noncopyable& operator=(const noncopyable&);
                                        ^
In file included from /usr/include/eigen3/Eigen/SparseCore:64:0,
                 from /usr/include/eigen3/Eigen/Sparse:26,
                 from /home/abdullah/github/free-cad-code/src/Mod/Sketcher/App/planegcs/GCS.cpp:57:
/usr/include/eigen3/Eigen/src/SparseCore/SparseSolverBase.h:53:7: error: within this context
 class SparseSolverBase : internal::noncopyable
       ^
In file included from /usr/include/eigen3/Eigen/SparseQR:33:0,
                 from /usr/include/eigen3/Eigen/Sparse:30,
                 from /home/abdullah/github/free-cad-code/src/Mod/Sketcher/App/planegcs/GCS.cpp:57:
/usr/include/eigen3/Eigen/src/SparseQR/SparseQR.h: In member function ‘Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> >& Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> >::operator=(const Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> >&)’:
/usr/include/eigen3/Eigen/src/SparseQR/SparseQR.h:71:7: note: synthesized method ‘Eigen::SparseSolverBase<Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> > >& Eigen::SparseSolverBase<Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> > >::operator=(const Eigen::SparseSolverBase<Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> > >&)’ first required here
 class SparseQR : public SparseSolverBase<SparseQR<_MatrixType,_OrderingType> >
       ^
/home/abdullah/github/free-cad-code/src/Mod/Sketcher/App/planegcs/GCS.cpp: In member function ‘int GCS::System::diagnose(GCS::Algorithm)’:
/home/abdullah/github/free-cad-code/src/Mod/Sketcher/App/planegcs/GCS.cpp:1893:18: note: synthesized method ‘Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> >& Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> >::operator=(const Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> >&)’ first required here
             SqrJT=Eigen::SparseQR<Eigen::SparseMatrix<double>, Eigen::COLAMDOrdering<int> >(SJ.topRows(count).transpose());
                  ^
make[2]: *** [src/Mod/Sketcher/App/CMakeFiles/Sketcher.dir/planegcs/GCS.cpp.o] Error 1
make[1]: *** [src/Mod/Sketcher/App/CMakeFiles/Sketcher.dir/all] Error 2
make: *** [all] Error 2


For our project I can get it work-around by using:

Code: Select all

SqrJT.compute(SJ.topRows(count).transpose());


instead, but I think you may consider it a bug

Thanks again,
abdullah

This is on purpose. Factorisations are not copyable, the fact it worked with 3.2 is just luck (and that we forgot to make it non-copyable). This syntax is also more expensive in c++98 (memory copy and consumption).

Ok. Thank you very much for the explanation.