/// \brief A map for filtering the edge-set to those edges 

   /*! 

   /// which are tight w.r.t. some node_potential map and 

     \brief A map for filtering the edge-set to those edges 

   /// edge_distance map.

     which are tight w.r.t. a node-potential and 

   ///

     edge-distance.

   /// A node-map node_potential is said to be a potential w.r.t. 

   /// an edge-map edge_distance 

     Let \f$G=(V,A)\f$ be a directed graph (graph for short) and 

   /// if and only if for each edge e, node_potential[g.target(e)] 

     let \f$\mathbb{F}\f$ be a number type. 

   /// <= edge_distance[e]+node_potential[g.source(e)] 

     Given a distance function 

   /// (or the reverse inequality holds for each edge).

     \f$d:E\to\mathbb{F}\f$, 

   /// An edge is said to be tight if this inequality holds with equality, 

     \f$\pi:V\to\mathbb{F}\f$ is said to be a potetial 

   /// and the map returns true exactly for those edges.

     w.r.t. \f$d\f$ 

   /// To avoid rounding errors, it is recommended to use this class with exact 

     if and only if 

   /// types, e.g. with int.

     \f$\pi(v)\le d(uv)+\pi(u)\f$ holds for each edge \f$uv\in E\f$ 

     (or the reverse inequality holds for each edge). 

     An edge is said to be tight if this inequality holds with equality, 

     and the map returns \c true exactly for those edges. 

     To avoid rounding errors, it is recommended to use this class with exact 

     number types, e.g. with \c int.

   */