Lots of bug fixes and improvements. Had to disable for int32 temporarily

while I sort that out. This also adds the max_k option to radius_search
as discussed in flann-lib#170.

Author: translunar

src/ruby/flann.gemspec

@@ -53,6 +53,7 @@ Gem::Specification.new do |gem|
   gem.add_development_dependency 'rake'
   gem.add_development_dependency 'bundler'
   gem.add_development_dependency 'rspec'
+  gem.add_development_dependency 'rspec-longrun'
   gem.add_development_dependency 'pry'
 end
 

src/ruby/lib/flann.rb

@@ -37,14 +37,17 @@ module Flann
   ffi_lib "libflann"
 
   # Declare enumerators
-  Algorithm   = enum(:linear, :kdtree, :kmeans, :composite, :kdtree_single, :saved, :autotuned)
-  CentersInit = enum(:random, :gonzales, :kmeanspp)
-  LogLevel    = enum(:none, :fatal, :error, :warn, :info)
-  DistanceType = enum(:euclidean, :manhattan, :minkowski, :hist_intersect, :hellinger, :chi_square, :kullback_leibler)
+  Algorithm    = enum(:algorithm, [:linear, :kdtree, :kmeans, :composite, :kdtree_single, :hierarchical, :lsh, :kdtree_cuda, :saved, 254, :autotuned, 255])
+  CentersInit  = enum(:centers_init, [:random, :gonzales, :kmeanspp])
+  LogLevel     = enum(:log_level, [:none, :fatal, :error, :warn, :info, :debug])
+
+  # Note that Hamming and beyond are not supported in the C API. We include them here just in case of future improvements.
+  DistanceType = enum(:distance_type, [:undefined, :euclidean, :l2, :manhattan, :l1, :minkowski, :max, :hist_intersect, :hellinger, :chi_square, :kullback_leibler, :hamming, :hamming_lut, :hamming_popcnt, :l2_simple])
 
   # For NMatrix compatibility
-  typedef :float,  :float32
-  typedef :double, :float64
+  typedef :float,   :float32
+  typedef :double,  :float64
+  typedef :char,    :byte
   typedef :pointer, :index_params_ptr
   typedef :pointer, :index_ptr
 
@@ -123,8 +126,10 @@ def dtype_to_c d #:nodoc:
     # Allocates index space and distance space for storing results from various searches. For a k-nearest neighbors
     # search, for example, you want trows (the number of rows in the testset) times k (the number of nearest neighbors
     # being searched for).
+    #
+    # Note that c_type will produce float for everything except double, which produces double.
     def allocate_results_space result_size, c_type #:nodoc:
-      [FFI::MemoryPointer.new(:int, result_size), FFI::MemoryPointer.new(c_type, result_size)]
+      [FFI::MemoryPointer.new(:int, result_size), FFI::MemoryPointer.new(c_type == :double ? :double : :float, result_size)]
     end
 
 
@@ -175,15 +180,31 @@ def nearest_neighbors dataset, testset, k, parameters = {}
                              indices_int_ptr, distances_t_ptr, k, parameters_ptr
 
       # Return results: two arrays, one of indices and one of distances.
-      [indices_int_ptr.read_array_of_int(result_size), distances_t_ptr.read_array_of_float(result_size)]
+      [indices_int_ptr.read_array_of_int(result_size),
+       c_type == :double ? distances_t_ptr.read_array_of_double(result_size) : distances_t_ptr.read_array_of_float(result_size)]
     end
     alias :nn :nearest_neighbors
 
     # Set the distance function to use when computing distances between data points.
-    def set_distance_type! distance_function, order = 0
-      Flann.send(:flann_set_distance_type, distance_function, order)
+    def set_distance_type! distance_function
+      Flann.send(:flann_set_distance_type, distance_function, get_distance_order)
       self
     end
+    alias :set_distance_type_and_order! :set_distance_type!
+
+    # Get the distance type and order
+    def get_distance_type_and_order
+      [Flann.flann_get_distance_type, Flann.flann_get_distance_order]
+    end
+    def get_distance_type
+      Flann.flann_get_distance_type
+    end
+    def get_distance_order
+      Flann.flann_get_distance_order
+    end
+    alias :distance_type :get_distance_type
+    alias :distance_order :get_distance_order
+
 
     # Perform hierarchical clustering of a set of points.
     #
@@ -249,7 +270,9 @@ def cluster dataset, clusters, parameters = {}
   attach_function :flann_free_index_float,  [:index_ptr, :index_params_ptr], :int
   attach_function :flann_free_index_double, [:index_ptr, :index_params_ptr], :int
 
-  attach_function :flann_set_distance_type, [DistanceType, :int], :void
+  attach_function :flann_set_distance_type, [:distance_type, :int], :void
+  attach_function :flann_get_distance_type, [], :distance_type
+  attach_function :flann_get_distance_order, [], :int
 
   attach_function :flann_compute_cluster_centers_byte,    [:pointer, :int, :int, :int, :pointer, :index_params_ptr], :int
   attach_function :flann_compute_cluster_centers_int,     [:pointer, :int, :int, :int, :pointer, :index_params_ptr], :int

src/ruby/lib/flann/index.rb

@@ -29,7 +29,9 @@ class FFI::Pointer
   class << self
     def new_from_nmatrix nm
       raise(StorageError, "dense storage expected") unless nm.dense?
-      ::FFI::Pointer.new(nm.data_pointer).tap { |p| p.autorelease = false }
+      c_type = Flann::dtype_to_c(nm.dtype)
+      c_type = :uchar if c_type == :byte
+      ::FFI::Pointer.new(c_type, nm.data_pointer).tap { |p| p.autorelease = false }
     end
   end
 end
@@ -43,9 +45,11 @@ class Index
     #
     # * https://github.com/mariusmuja/flann/tree/master/src/cpp/flann/algorithms
     #
-    def initialize dataset = nil, dtype: :float64, parameters: Flann::Parameters::DEFAULT
-      @dataset        = dataset
-      @dtype          = (!dataset.nil? && dataset.is_a?(NMatrix)) ? dataset.dtype : dtype
+    def initialize index_dataset = nil, dtype: :float64, parameters: Flann::Parameters::DEFAULT
+      @dataset        = index_dataset
+      #require 'pry'
+      #binding.pry if @dataset.nil?
+      @dtype          = (!index_dataset.nil? && index_dataset.is_a?(NMatrix)) ? index_dataset.dtype : dtype
       @index_ptr      = nil
 
       @parameters_ptr, @parameters = Flann::handle_parameters(parameters)
@@ -55,17 +59,24 @@ def initialize dataset = nil, dtype: :float64, parameters: Flann::Parameters::DE
     attr_reader :dtype, :dataset, :parameters, :parameters_ptr, :index_ptr
 
     # Assign a new dataset. Requires that the old index be freed.
-    def dataset= new_dataset
+    def dataset= index_dataset
       free!
+      @dataset = index_dataset
     end
 
     # Build an index
     def build!
       raise("no dataset specified") if dataset.nil?
-
-      c_method = "flann_build_index_#{Flann::dtype_to_c(dtype)}".to_sym
+      c_type   = Flann::dtype_to_c(dtype)
+      c_method = "flann_build_index_#{c_type}".to_sym
       speedup_float_ptr = FFI::MemoryPointer.new(:float)
       @index_ptr = Flann.send(c_method, FFI::Pointer.new_from_nmatrix(dataset), dataset.shape[0], dataset.shape[1], speedup_float_ptr, parameters_ptr)
+      if index_ptr.address == 0
+        require 'pry'
+        binding.pry
+        raise("failed to allocate index_ptr")
+      end
+
 
       # Return the speedup
       speedup_float_ptr.read_float
@@ -91,12 +102,14 @@ def nearest_neighbors testset, k, parameters = {}
                            k,
                            parameters_ptr
 
-      [indices_int_ptr.read_array_of_int(result_size), distances_t_ptr.read_array_of_float(result_size)]
+
+      [indices_int_ptr.read_array_of_int(result_size),
+       c_type == :double ? distances_t_ptr.read_array_of_double(result_size) : distances_t_ptr.read_array_of_float(result_size)]
     end
 
     # Perform a radius search on a single query point
-    def radius_search query, radius, parameters = {}
-      max_k      = parameters[:max_neighbors] || dataset.shape[1]
+    def radius_search query, radius, max_k=nil, parameters = {}
+      max_k    ||= dataset.shape[1]
       parameters = Parameters.new(Flann::Parameters::DEFAULT.merge(parameters))
 
       self.build! if index_ptr.nil?
@@ -109,7 +122,8 @@ def radius_search query, radius, parameters = {}
       Flann.send(c_method, index_ptr, FFI::Pointer.new_from_nmatrix(query), indices_int_ptr, distances_t_ptr, max_k, radius, parameters_ptr)
 
       # Return results: two arrays, one of indices and one of distances.
-      [indices_int_ptr.read_array_of_int(max_k), distances_t_ptr.read_array_of_float(max_k)]
+      [indices_int_ptr.read_array_of_int(max_k),
+       c_type == :double ? distances_t_ptr.read_array_of_double(max_k) : distances_t_ptr.read_array_of_float(max_k)]
     end
 
     # Save an index to a file (without the dataset).

src/ruby/spec/flann_spec.rb

@@ -68,15 +68,18 @@
 
   context "#set_distance_type!" do
     it "sets the distance functor without error" do
-      pending "distance type unsupported in the C bindings, use the C++ bindings instead"
+      #pending "distance type unsupported in the C bindings, use the C++ bindings instead"
       Flann.set_distance_type! :euclidean
+      d = Flann.get_distance_type
+      expect(d).to eq(:euclidean)
     end
   end
 
   [:byte, :int32, :float32, :float64].each do |dtype|
     before :each do
-      @dataset = NMatrix.random([1000,128], dtype: dtype)
-      @testset = NMatrix.random([100,128],  dtype: dtype)
+      scale = [:byte, :int32, :int64].include?(dtype) ? 255 : 1.0
+      @dataset = NMatrix.random([1000,128], dtype: dtype, scale: scale)
+      @testset = NMatrix.random([100,128],  dtype: dtype, scale: scale)
     end
 
     context "#nearest_neighbors" do

src/ruby/spec/index_spec.rb

@@ -29,61 +29,70 @@
 
 
 describe Flann::Index do
-
-  before :each do
-    @dataset = NMatrix.random([1000,128])
-    @testset = NMatrix.random([100,128])
-    @index   = Flann::Index.new(@dataset) do |t|
-      t[:algorithm] = :kdtree
-      t[:trees]     = 4
-    end
-    @index.build!
-  end
-
-
-  context "#build!" do
-    it "builds a kdtree index with block parameters" do
-      # Empty: handled in :each, above
-    end
-  end
-
-
-  context "#nearest_neighbors" do
-    it "runs without error" do
-      @index.nearest_neighbors @testset, 5
+  [:byte, :float32, :float64].each do |dtype|
+    context dtype.inspect do
+      before :each do
+        scale = [:byte, :int32, :int64].include?(dtype) ? 255 : 1.0
+
+        @dataset = NMatrix.random([1000,128], dtype: dtype, scale: scale)
+        @testset = NMatrix.random([100,128],  dtype: dtype, scale: scale)
+        @index   = Flann::Index.new(@dataset) do |t|
+          t[:algorithm] = :kdtree
+          t[:trees]     = 4
+        end
+        @index.build!
+      end
+
+
+      context "#build!" do
+        it "builds a kdtree index with block parameters" do
+          scale = [:byte, :int].include?(dtype) ? 255 : 1.0
+
+          @dataset = NMatrix.random([1000,128], dtype: dtype, scale: scale)
+          @testset = NMatrix.random([100,128],  dtype: dtype, scale: scale)
+          @index   = Flann::Index.new(@dataset) do |t|
+            t[:algorithm] = :kdtree
+            t[:trees]     = 4
+          end
+          @index.build!
+        end
+      end
+
+
+      context "#nearest_neighbors" do
+        it "runs without error" do
+          @index.nearest_neighbors @testset, 5
+        end
+      end
+
+
+      context "#radius_search" do
+        it "runs without error" do
+          query = NMatrix.random([1,128])
+          @index.radius_search query, 0.4
+        end
+      end
+
+
+      context "#save" do
+        it "saves an index to a file which can be loaded again" do
+          FileUtils.rm("temp_index.save_file", :force => true)
+          @index.save("temp_index.save_file")
+
+          raise(IOError, "save failed") unless File.exists?("temp_index.save_file")
+
+          post_index = Flann::Index.new(@dataset)
+          post_index.load!("temp_index.save_file")
+          FileUtils.rm("temp_index.save_file", :force => true)
+        end
+      end
+
+
+      context "#free!" do
+        it "frees an index" do
+          @index.free!
+        end
+      end
     end
   end
-
-
-  context "#radius_search" do
-    it "runs without error" do
-      query = NMatrix.random([1,128])
-      @index.radius_search query, 0.4
-    end
-  end
-
-
-  context "#save" do
-    it "saves an index to a file which can be loaded again" do
-      FileUtils.rm("temp_index.save_file", :force => true)
-      @index.save("temp_index.save_file")
-
-      raise(IOError, "save failed") unless File.exists?("temp_index.save_file")
-
-      post_index = Flann::Index.new(@dataset)
-      post_index.load!("temp_index.save_file")
-      FileUtils.rm("temp_index.save_file", :force => true)
-    end
-  end
-
-
-  context "#free!" do
-    it "frees an index" do
-      @index.free!
-    end
-  end
-
-
-
-
 end

src/ruby/spec/spec_helper.rb

@@ -26,5 +26,6 @@
 # THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 require 'rspec'
+require 'rspec/longrun'
 
 require "./lib/flann"