fix for cityslam

linear/SubgraphPreconditioner.h

@@ -8,7 +8,7 @@
 
 #include <gtsam/linear/GaussianFactorGraph.h>
 #include <gtsam/linear/GaussianBayesNet.h>
-#include <gtsam/inference/Ordering.h>
+#include <gtsam/nonlinear/Ordering.h>
 
 namespace gtsam {
 
@@ -56,7 +56,11 @@ namespace gtsam {
 		VectorValues x(const VectorValues& y) const;
 
 		/* A zero VectorValues with the structure of xbar */
-		VectorValues zero() const { return VectorValues::zero(*xbar_);}
+		VectorValues zero() const {
+			VectorValues V(*xbar_) ;
+			V.makeZero();
+			return V ;
+		}
 
 		/* error, given y */
 		double error(const VectorValues& y) const;

linear/VectorValues.h

@@ -46,6 +46,7 @@ public:
    * slow reallocation of space at runtime.
    */
   VectorValues() : varStarts_(1,0) {}
+  VectorValues(const VectorValues &V) : values_(V.values_), varStarts_(V.varStarts_) {}
 
   /** Construct from a container of variable dimensions (in variable order). */
   template<class Container>
@@ -76,6 +77,10 @@ public:
    */
   size_t dim() const { return varStarts_.back(); }
 
+  /* dot product */
+  double dot(const VectorValues& V) const { return gtsam::dot(this->values_, V.values_) ; }
+
+
   /** Total dimensions capacity allocated */
   size_t dimCapacity() const { return values_.size(); }
 
@@ -163,6 +168,13 @@ public:
 
 protected:
   void checkVariable(Index variable) const { assert(variable < varStarts_.size()-1); }
+
+
+public:
+  friend double dot(const VectorValues& V1, const VectorValues& V2) { return gtsam::dot(V1.values_, V2.values_) ; }
+  friend void scal(double alpha, VectorValues& x) {	gtsam::scal(alpha, x.values_) ; }
+  friend void axpy(double alpha, const VectorValues& x, VectorValues& y) { gtsam::axpy(alpha, x.values_, y.values_) ; }
+
 };
 
 

linear/iterative-inl.h

@@ -32,8 +32,8 @@ namespace gtsam {
 			k = 0;
 			verbose = verb;
 			steepest = steep;
-			maxIterations = (maxIt > 0) ? maxIt : dim(x) * (steepest ? 10 : 1);
-			reset = (size_t) (sqrt(dim(x)) + 0.5); // when to reset
+			maxIterations = (maxIt > 0) ? maxIt : x.dim() * (steepest ? 10 : 1);
+			reset = (size_t) (sqrt(x.dim()) + 0.5); // when to reset
 
 			// Start with g0 = A'*(A*x0-b), d0 = - g0
 			// i.e., first step is in direction of negative gradient