Squashed commit of the following:

commit 79baa1c Author: Manos Papadakis <[email protected]> Date: Tue Dec 5 13:29:46 2023 +0200 Update nn.Rd commit 6c1531c Author: Manos Papadakis <[email protected]> Date: Tue Dec 5 13:29:43 2023 +0200 Update system.Rd commit ad2c5a7 Author: Manos Papadakis <[email protected]> Date: Tue Dec 5 13:29:41 2023 +0200 Update .Rbuildignore commit fc9f68f Author: Manos Papadakis <[email protected]> Date: Mon Dec 4 17:20:24 2023 +0200 Update nn.Rd commit b686cf8 Author: Manos Papadakis <[email protected]> Date: Mon Dec 4 17:16:55 2023 +0200 Update NEWS.md commit 6eccaf4 Author: Manos Papadakis <[email protected]> Date: Mon Dec 4 17:16:52 2023 +0200 Update DESCRIPTION commit a6c50ed Author: Manos Papadakis <[email protected]> Date: Mon Dec 4 17:07:11 2023 +0200 Update NEWS.md commit 0273aba Author: Manos Papadakis <[email protected]> Date: Mon Dec 4 17:06:58 2023 +0200 Update KDTreeArmadilloAdaptor.hpp commit b46a7c6 Author: Manos Papadakis <[email protected]> Date: Mon Dec 4 17:04:58 2023 +0200 Update helpers.h commit 8af14f4 Author: Manos Papadakis <[email protected]> Date: Thu Nov 23 19:25:23 2023 +0200 optimize log calculation commit 226d257 Author: Manos Papadakis <[email protected]> Date: Thu Nov 23 19:06:20 2023 +0200 Update .gitignore commit f76c078 Author: Manos Papadakis <[email protected]> Date: Thu Nov 23 19:05:13 2023 +0200 optimize jensen_shannon commit aa1759e Author: Manos Papadakis <[email protected]> Date: Thu Nov 16 13:28:36 2023 +0200 Update nn.R commit 470bfd5 Author: Manos Papadakis <[email protected]> Date: Wed Nov 15 16:41:38 2023 +0200 Update RcppExports.cpp commit 055399c Author: Manos Papadakis <[email protected]> Date: Wed Nov 15 16:41:34 2023 +0200 Update nn.R commit 9c9542c Author: Manos Papadakis <[email protected]> Date: Wed Nov 15 16:41:29 2023 +0200 Update .gitignore commit 65c7a2f Author: Manos Papadakis <[email protected]> Date: Wed Nov 15 16:41:17 2023 +0200 upload proj commit 2049da8 Author: Manos Papadakis <[email protected]> Date: Wed Nov 15 15:01:24 2023 +0200 add equations commit b8b19fd Author: Manos Papadakis <[email protected]> Date: Wed Nov 15 13:50:39 2023 +0200 Update .gitignore commit e352ba3 Author: Manos Papadakis <[email protected]> Date: Wed Nov 15 13:46:08 2023 +0200 fix bugs This reverts commit 99a5c98. commit d976a3e Merge: f20c937 99a5c98 Author: Manos Papadakis <[email protected]> Date: Tue Nov 14 22:04:01 2023 +0200 Merge branch 'main' into dev commit f20c937 Author: Manos Papadakis <[email protected]> Date: Wed Nov 1 17:07:37 2023 +0200 Create r.yml (#2)
ManosPapadakis95 · Dec 6, 2023 · 070c322 · 070c322
1 parent 3691013
commit 070c322
Show file tree

Hide file tree

Showing 12 changed files with 1,384 additions and 131 deletions.
diff --git a/.Rbuildignore b/.Rbuildignore
@@ -1,4 +1,8 @@
 ^.*\.Rproj$
 ^\.Rproj\.user$
-^.RData
-^.Rhistory
+^docs
+^.todo
+^.github
+^.html
+^.gitignore
+^.vscode
diff --git a/.gitignore b/.gitignore
@@ -1,7 +1,10 @@
-
-*src/.dll
-*src/.o
-*.dll
+*.swp
 *.o
+*.o.tmp
+*\src-i386
+*\src-x64
+*.dll
+*.html
 *.RData
 *.Rhistory
+.Rproj.user
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -1,8 +1,8 @@
 Package: Rnanoflann
 Type: Package
 Title: Extremely Fast Nearest Neighbor Search 
-Version: 0.0.1
-Date: 2023-10-30
+Version: 0.0.2
+Date: 2023-12-04
 Authors@R: c(
                 person("Manos Papadakis", email = "[email protected]", role = c("aut", "cre", "cph")), 
                 person("Jose Luis Blanco", email = "[email protected]" , role = c("aut","cph")),

diff --git a/NEWS.md b/NEWS.md
@@ -4,6 +4,25 @@ Rnanoflann
 
 </h1>
 
+### **Version 0.0.2**
+
+------------------------------------------------------------------------
+
+> **New**
+>
+> | Function | What's new!                              |
+> |----------|------------------------------------------|
+> | nn       | add new metrics with parallel version, "hellinger","canberra","kullback_leibler","jensen_shannon","itakura_saito","bhattacharyya","jeffries_matusita","minimum","maximum","total_variation","sorensen,"cosine","gower","minkowski","soergel","kulczynski","wave_hedges","motyka","harmonic_mean". |
+> | nn       | fix bug in metric "euclidean". |
+>
+> **LinkingTo**
+>
+> | Function | What's new!                              |
+> |----------|------------------------------------------|
+> | nn       | add new metrics with parallel version, "hellinger","canberra","kullback_leibler","jensen_shannon","itakura_saito","bhattacharyya","jeffries_matusita","minimum","maximum","total_variation","sorensen,"cosine","gower","minkowski","soergel","kulczynski","wave_hedges","motyka","harmonic_mean". |
+> | nn       | fix bug in metric "euclidean". |
+
+
 ### **Version 0.0.1**
 
 ------------------------------------------------------------------------

diff --git a/R/nn.R b/R/nn.R
@@ -1,8 +1,11 @@
 #[export]
-nn <- function(data, points = data, k = nrow(data), method = "euclidean", search = "standard", eps = 0.0, square = FALSE,
-                sorted = FALSE, radius = 0.0, trans = TRUE, leafs = 10, parallel = FALSE, cores = 0){
-    res <- .Call(`_Rnanoflann_nn`, t(data), t(points), k, method, search, eps, square,
-                 sorted, radius, leafs, parallel, cores)
+nn <- function(data, points, k = nrow(data), method = "euclidean", search = "standard", 
+eps = 0.0, square = FALSE, sorted = FALSE, radius = 0.0, trans = TRUE, leafs = 10L, p = 0.0, parallel = FALSE, cores = 0L) {
+    if(method == "hellinger"){
+		data <- sqrt(data)
+		points <- sqrt(points)
+	}
+    res <- .Call(`_Rnanoflann_nn`, t(data), t(points), k, method, search, eps, square, sorted, radius, leafs, p, parallel, cores)
     if(trans){
         res$indices <- t(res$indices)
         res$distances <- t(res$distances)

diff --git a/inst/include/internal/KDTreeArmadilloAdaptor.hpp b/inst/include/internal/KDTreeArmadilloAdaptor.hpp
@@ -6,25 +6,25 @@
 using namespace arma;
 using namespace nanoflann;
 
-
-namespace Rnanoflann {
+namespace Rnanoflann
+{
     // Define an Armadillo KDTreeAdaptor class
     template <class MatrixType, class Distance, int DIM = -1>
     struct KDTreeArmadilloAdaptor
     {
         using self_t = KDTreeArmadilloAdaptor<MatrixType, Distance, DIM>;
         using num_t = typename MatrixType::elem_type;
-        using IndexType =  uword;
-        
-        using metric_t = typename Distance::template traits<num_t, self_t>::distance_t;// You can change the distance metric as needed.
-        
+        using IndexType = uword;
+
+        using metric_t = typename Distance::template traits<num_t, self_t>::distance_t; // You can change the distance metric as needed.
+
         using index_t = nanoflann::KDTreeSingleIndexAdaptor<metric_t, self_t, DIM, IndexType>;
-        
-        index_t* index_;
-        
+
+        index_t *index_;
+
         const std::reference_wrapper<const MatrixType> m_data_matrix;
-        
-        explicit KDTreeArmadilloAdaptor(const uword dimensionality, const std::reference_wrapper<const MatrixType>& mat,const uword leafs = 10)
+
+        explicit KDTreeArmadilloAdaptor(const uword dimensionality, const std::reference_wrapper<const MatrixType> &mat, const uword leafs = 10)
             : m_data_matrix(mat)
         {
             const auto dims = mat.get().n_rows; // Assumes column-major Armadillo matrix
@@ -35,47 +35,51 @@ namespace Rnanoflann {
             index_ = new index_t(dims, *this /* adaptor */, nanoflann::KDTreeSingleIndexAdaptorParams(leafs));
             index_->buildIndex();
         }
-        
+
         ~KDTreeArmadilloAdaptor() { delete index_; }
-        
-        const self_t& derived() const { return *this; }
-        self_t& derived() { return *this; }
-        
+
+        const self_t &derived() const { return *this; }
+        self_t &derived() { return *this; }
+
         uword kdtree_get_point_count() const
         {
             return m_data_matrix.get().n_cols;
         }
-        
+
         num_t kdtree_get_pt(uword idx, size_t dim) const
         {
-            return m_data_matrix.get()(dim,idx);
+            return m_data_matrix.get()(dim, idx);
+        }
+
+        colvec col(uword idx) const
+        {
+            return m_data_matrix.get().col(idx);
         }
-        
+
         template <class BBOX>
-        bool kdtree_get_bbox(BBOX& bb) const
+        bool kdtree_get_bbox(BBOX &bb) const
         {
             return false; // Optional bounding-box computation (not used in this example)
         }
     };
 
-
     // Define an Armadillo KDTreeAdaptor class
-    template <class MatrixType, class Distance, bool Square = false, int DIM = -1>
+    template <class MatrixType, class Distance, bool Square, int DIM = -1>
     struct KDTreeArmadilloAdaptor2
     {
         using self_t = KDTreeArmadilloAdaptor2<MatrixType, Distance, Square, DIM>;
         using num_t = typename MatrixType::elem_type;
-        using IndexType =  uword;
-        
-        using metric_t = typename Distance::template traits<num_t, self_t, Square>::distance_t;// You can change the distance metric as needed.
-        
+        using IndexType = uword;
+
+        using metric_t = typename Distance::template traits<num_t, self_t, Square>::distance_t; // You can change the distance metric as needed.
+
         using index_t = nanoflann::KDTreeSingleIndexAdaptor<metric_t, self_t, DIM, IndexType>;
-        
-        index_t* index_;
-        
+
+        index_t *index_;
+
         const std::reference_wrapper<const MatrixType> m_data_matrix;
-        
-        explicit KDTreeArmadilloAdaptor2(const uword dimensionality, const std::reference_wrapper<const MatrixType>& mat,const uword leafs = 10)
+
+        explicit KDTreeArmadilloAdaptor2(const uword dimensionality, const std::reference_wrapper<const MatrixType> &mat, const uword leafs = 10)
             : m_data_matrix(mat)
         {
             const auto dims = mat.get().n_rows; // Assumes column-major Armadillo matrix
@@ -86,24 +90,169 @@ namespace Rnanoflann {
             index_ = new index_t(dims, *this /* adaptor */, nanoflann::KDTreeSingleIndexAdaptorParams(leafs));
             index_->buildIndex();
         }
-        
+
         ~KDTreeArmadilloAdaptor2() { delete index_; }
-        
-        const self_t& derived() const { return *this; }
-        self_t& derived() { return *this; }
-        
+
+        const self_t &derived() const { return *this; }
+        self_t &derived() { return *this; }
+
         uword kdtree_get_point_count() const
         {
             return m_data_matrix.get().n_cols;
         }
-
+
+        colvec col(uword idx) const
+        {
+            return m_data_matrix.get().col(idx);
+        }
+
         num_t kdtree_get_pt(uword idx, size_t dim) const
         {
-            return m_data_matrix.get()(dim,idx);
+            return m_data_matrix.get()(dim, idx);
+        }
+
+        template <class BBOX>
+        bool kdtree_get_bbox(BBOX &bb) const
+        {
+            return false; // Optional bounding-box computation (not used in this example)
         }
-
+    };
+
+    template <class MatrixType, class Distance, int DIM = -1>
+    struct KDTreeArmadilloAdaptor3
+    {
+        using self_t = KDTreeArmadilloAdaptor3<MatrixType, Distance, DIM>;
+        using num_t = typename MatrixType::elem_type;
+        using IndexType = uword;
+
+        using metric_t = typename Distance::template traits<num_t, self_t>::distance_t; // You can change the distance metric as needed.
+
+        using index_t = nanoflann::KDTreeSingleIndexAdaptor<metric_t, self_t, DIM, IndexType>;
+
+        index_t *index_;
+
+        const double p;
+        const double p_1;
+
+        const std::reference_wrapper<const MatrixType> m_data_matrix;
+
+        explicit KDTreeArmadilloAdaptor3(const uword dimensionality, const std::reference_wrapper<const MatrixType> &mat, const double p, const uword leafs = 10)
+            : p(p), p_1(1.0 / p), m_data_matrix(mat)
+        {
+            const auto dims = mat.get().n_rows; // Assumes column-major Armadillo matrix
+            if (static_cast<uword>(dims) != dimensionality)
+                throw std::runtime_error("Error: 'dimensionality' must match the column count in the data matrix");
+            if (DIM > 0 && static_cast<int>(dims) != DIM)
+                throw std::runtime_error("Data set dimensionality does not match the 'DIM' template argument");
+            index_ = new index_t(dims, *this /* adaptor */, nanoflann::KDTreeSingleIndexAdaptorParams(leafs));
+            index_->buildIndex();
+        }
+
+        ~KDTreeArmadilloAdaptor3() { delete index_; }
+
+        const self_t &derived() const { return *this; }
+        self_t &derived() { return *this; }
+
+        uword kdtree_get_point_count() const
+        {
+            return m_data_matrix.get().n_cols;
+        }
+
+        num_t kdtree_get_pt(uword idx, size_t dim) const
+        {
+            return m_data_matrix.get()(dim, idx);
+        }
+
+        colvec col(uword idx) const
+        {
+            return m_data_matrix.get().col(idx);
+        }
+
+        const double getP() const
+        {
+            return p;
+        }
+
+        const double getP_1() const
+        {
+            return p_1;
+        }
+
+        template <class BBOX>
+        bool kdtree_get_bbox(BBOX &bb) const
+        {
+            return false; // Optional bounding-box computation (not used in this example)
+        }
+    };
+
+    // Define an Armadillo KDTreeAdaptor class
+    template <class MatrixType, class Distance, int DIM = -1>
+    struct KDTreeArmadilloAdaptor4
+    {
+        using self_t = KDTreeArmadilloAdaptor4<MatrixType, Distance, DIM>;
+        using num_t = typename MatrixType::elem_type;
+        using IndexType = uword;
+
+        using metric_t = typename Distance::template traits<num_t, self_t>::distance_t; // You can change the distance metric as needed.
+
+        using index_t = nanoflann::KDTreeSingleIndexAdaptor<metric_t, self_t, DIM, IndexType>;
+
+        index_t *index_;
+
+        const std::reference_wrapper<const MatrixType> m_data_matrix;
+        const MatrixType xlogx, ylogy;
+        const num_t *begin_points;
+
+        explicit KDTreeArmadilloAdaptor4(const uword dimensionality, const std::reference_wrapper<const MatrixType> &mat, const MatrixType &points, const uword leafs = 10)
+            : m_data_matrix(mat), xlogx(mat.get() % arma::log(mat.get())), ylogy(points % arma::log(points)), begin_points(points.memptr())
+        {
+            const auto dims = mat.get().n_rows; // Assumes column-major Armadillo matrix
+            if (static_cast<uword>(dims) != dimensionality)
+                throw std::runtime_error("Error: 'dimensionality' must match the column count in the data matrix");
+            if (DIM > 0 && static_cast<int>(dims) != DIM)
+                throw std::runtime_error("Data set dimensionality does not match the 'DIM' template argument");
+            index_ = new index_t(dims, *this /* adaptor */, nanoflann::KDTreeSingleIndexAdaptorParams(leafs));
+            index_->buildIndex();
+        }
+
+        ~KDTreeArmadilloAdaptor4() { delete index_; }
+
+        const self_t &derived() const { return *this; }
+        self_t &derived() { return *this; }
+
+        uword kdtree_get_point_count() const
+        {
+            return m_data_matrix.get().n_cols;
+        }
+
+        num_t kdtree_get_pt(uword idx, size_t dim) const
+        {
+            return m_data_matrix.get()(dim, idx);
+        }
+
+        colvec col(uword idx) const
+        {
+            return m_data_matrix.get().col(idx);
+        }
+
+        colvec col_xlogx(uword idx) const
+        {
+            return xlogx.col(idx);
+        }
+
+        colvec col_ylogy(const num_t *a) const
+        {
+            auto index = std::floor((a - begin_points) / ylogy.n_rows);
+            return ylogy.col(index);
+        }
+
+        double log0_5() const
+        {
+            return std::log(0.5);
+        }
+
         template <class BBOX>
-        bool kdtree_get_bbox(BBOX& bb) const
+        bool kdtree_get_bbox(BBOX &bb) const
         {
             return false; // Optional bounding-box computation (not used in this example)
         }