Feature/dynamicpolymorphism

This is the first MR in a series to implement #26 (closed).

The solver interfaces use dynamic polymorphism, but currently it is not possible to fully benefit from this feature. We want to allow the user to take full benefit from a dynamic interface for solver/precoditioner/scalarproduct/linearoperator.

The main change is that the category isn't an enum anymore, but virtual method catergory().

This MR contains the following changes:

• new method category() to check for the same SolverCategory of different components at run time. This is a slight drawback compared to the current compile time checks, but it adds a lot of flexibility as we will be able to write factory classes for solvers and compose/exchange solvers at runtime.
• add SolverCategory::category(object) to be be backwards compatible
• add a cmake switch to enable/disable the backwards compatibility
• update solvers to the new interface
• update preconditioners to the new interface
• update scalar products to the new interface
• update operators to the new interface
• update AMG implementation to the new interface
  • remove static dependency of PropertiesGraphCreator on the category

This MR partially incorporates commits from !5 (closed) and !1 (closed).

dune/istl/overlappingschwarz.hh

@@ -795,11 +795,6 @@ namespace Dune
     /** @brief The vector type containing subdomain solvers. */
     typedef std::vector<slu, typename TA::template rebind<slu>::other> slu_vector;
 
-    enum {
-      //! \brief The category the precondtioner is part of.
-      category = SolverCategory::sequential
-    };
-
     /**
      * @brief Construct the overlapping Schwarz method.
      * @param mat The matrix to precondition.
@@ -861,6 +856,12 @@ namespace Dune
     template<bool forward>
     void apply(X& v, const X& d);
 
+    //! Category of the preconditioner (see SolverCategory::Category)
+    virtual SolverCategory::Category category() const
+    {
+      return SolverCategory::sequential;
+    }
+
   private:
     const M& mat;
     slu_vector solvers;