Add operator[] for MarginalDistribution (#487)

This change is in support of the PIC implementation.

include/albatross/src/core/distribution.hpp

@@ -20,6 +20,8 @@ inline bool operator==(const albatross::DiagonalMatrixXd &x,
 
 namespace albatross {
 
+struct MarginalDistribution;
+
 constexpr double cDefaultApproximatelyEqualEpsilon = 1e-3;
 
 template <typename Derived> struct DistributionBase {
@@ -48,6 +50,8 @@ template <typename Derived> struct DistributionBase {
     return derived().get_diagonal(i);
   }
 
+  MarginalDistribution operator[](std::size_t index) const;
+
   Eigen::VectorXd mean;
   std::map<std::string, std::string> metadata;
 
@@ -245,6 +249,13 @@ inline void set_subset(const DistributionBase<DistributionType> &from,
   to->derived().set_subset(from, indices);
 }
 
+template <typename Derived>
+MarginalDistribution
+DistributionBase<Derived>::operator[](std::size_t index) const {
+  return MarginalDistribution(mean[cast::to_index(index)],
+                              get_diagonal(cast::to_index(index)));
+}
+
 inline MarginalDistribution
 concatenate_marginals(const MarginalDistribution &x,
                       const MarginalDistribution &y) {

tests/test_core_distribution.cc

@@ -236,14 +236,27 @@ TYPED_TEST_P(DistributionTest, test_subtract) {
   EXPECT_EQ(actual, expected);
 };
 
+TYPED_TEST_P(DistributionTest, test_operator_indexing) {
+
+  TypeParam test_case;
+  const auto dist = test_case.create();
+
+  for (std::size_t idx = 0; idx < dist.size(); ++idx) {
+    MarginalDistribution m = dist[idx];
+
+    EXPECT_EQ(m.mean[0], dist.mean[cast::to_index(idx)]);
+    EXPECT_EQ(m.get_diagonal(0), dist.get_diagonal(cast::to_index(idx)));
+  }
+};
+
 REGISTER_TYPED_TEST_SUITE_P(DistributionTest, test_subset,
                             test_multiply_with_matrix_joint,
                             test_multiply_with_matrix_marginal,
                             test_multiply_with_sparse_matrix_joint,
                             test_multiply_with_sparse_matrix_marginal,
                             test_multiply_with_vector, test_multiply_by_scalar,
                             test_equal, test_approximately_equal, test_add,
-                            test_subtract);
+                            test_subtract, test_operator_indexing);
 
 Eigen::VectorXd arange(Eigen::Index k = 5) {
   Eigen::VectorXd mean(k);