d_heap 0.4.0 → 0.5.0

data/ext/d_heap/d_heap.h

@@ -11,11 +11,17 @@
 // comparisons as d gets further from 4.
 #define DHEAP_MAX_D 32
 
-#define DHEAP_DEFAULT_SIZE 16
-#define DHEAP_MAX_SIZE (LONG_MAX / (int)sizeof(long double))
+typedef long double SCORE;
 
-// 10MB
-#define DHEAP_CAPA_INCR_MAX (10 * 1024 * 1024 / (int)sizeof(long double))
+typedef struct dheap_entry {
+    SCORE score;
+    VALUE value;
+} ENTRY;
+
+#define DHEAP_DEFAULT_SIZE 256
+#define DHEAP_MAX_SIZE (LONG_MAX / (int)sizeof(ENTRY))
+
+#define DHEAP_CAPA_INCR_MAX (10 * 1024 * 1024 / (int)sizeof(ENTRY))
 
 VALUE rb_cDHeap;
 

data/ext/d_heap/extconf.rb

@@ -4,6 +4,14 @@ require "mkmf"
 
 # For testing in CI (because I don't otherwise have easy access to Mac OS):
 # $CFLAGS << " -D__D_HEAP_DEBUG" if /darwin/ =~ RUBY_PLATFORM
+# $CFLAGS << " -debug inline-debug-info "
+# $CFLAGS << " -g -ginline-points "
+# $CFLAGS << " -fno-omit-frame-pointer "
+
+# CONFIG["debugflags"] << " -ggdb3 -gstatement-frontiers -ginline-points "
+CONFIG["optflags"]  << " -O3 "
+CONFIG["optflags"]  << " -fno-omit-frame-pointer "
+CONFIG["warnflags"] << " -Werror"
 
 have_func "rb_gc_mark_movable" # since ruby-2.7
 
@@ -12,5 +20,4 @@ check_sizeof("unsigned long long")
 check_sizeof("long double")
 have_macro("LDBL_MANT_DIG", "float.h")
 
-CONFIG["warnflags"] << " -Werror"
 create_makefile("d_heap/d_heap")

data/lib/d_heap.rb

@@ -14,11 +14,35 @@ require "d_heap/version"
 # worst-case time complexity.
 #
 class DHeap
+  alias deq       pop
+  alias enq       push
+  alias first     peek
+  alias pop_below pop_lt
+
+  alias length    size
+  alias count     size
+
   # ruby 3.0+ (2.x can just use inherited initialize_clone)
   if Object.instance_method(:initialize_clone).arity == -1
+    # @!visibility private
     def initialize_clone(other, freeze: nil)
       __init_clone__(other, freeze ? true : freeze)
     end
   end
 
+  # Consumes the heap by popping each minumum value until it is empty.
+  #
+  # If you want to iterate over the heap without consuming it, you will need to
+  # first call +#dup+
+  #
+  # @yieldparam value [Object] each value that would be popped
+  #
+  # @return [Enumerator] if no block is given
+  # @return [nil] if a block is given
+  def each_pop
+    return to_enum(__method__) unless block_given?
+    yield pop until empty?
+    nil
+  end
+
 end

data/lib/d_heap/benchmarks.rb

@@ -93,7 +93,7 @@ module DHeap::Benchmarks
   include Randomness
   include Scenarios
 
-  def initq(klass, count = 0)
+  def initq(klass, count = 0, clear: false)
     queue = klass.new
     while 0 < count
       queue << @dheap_bm_random_vals.fetch(
@@ -101,6 +101,7 @@ module DHeap::Benchmarks
       )
       count -= 1
     end
+    queue.clear if clear
     queue
   end
 

data/lib/d_heap/benchmarks/benchmarker.rb

@@ -87,6 +87,9 @@ module DHeap::Benchmarks
           --runner ips_zero_fail
           benchmarks/#{file}.yml
         ]
+        if file == "push_n"
+          cmd << "--filter" << /dheap|\bstl\b|\bbsearch\b|\brb_heap\b/.to_s
+        end
         env = ENV.to_h.merge(
           "BENCH_N" => size.to_s,
           "RUBYLIB" => File.expand_path("../..", __dir__),

data/lib/d_heap/benchmarks/implementations.rb

@@ -1,13 +1,19 @@
 # frozen_string_literal: true
 
+require "fc"
+
 module DHeap::Benchmarks
 
   # base class for example priority queues
   class ExamplePriorityQueue
     attr_reader :a
 
-    def initialize
+    # quick initialization by simply sorting the array once.
+    def initialize(count = nil, &block)
       @a = []
+      return unless count
+      count.times {|i| @a << block.call(i) }
+      @a.sort!
     end
 
     def clear
@@ -98,70 +104,118 @@ module DHeap::Benchmarks
   # rubocop:disable Metrics/MethodLength, Metrics/AbcSize
   class RbHeap < ExamplePriorityQueue
 
-    def <<(score)
-      raise ArgumentError unless score
-      @a.push(score)
+    def <<(value)
+      raise ArgumentError unless value
+      @a.push(value)
       # shift up
       index = @a.size - 1
       while 0 < index # rubocop:disable Style/NumericPredicate
         parent_index = (index - 1) / 2
-        break if @a[parent_index] <= @a[index]
-        @a[index] = @a[parent_index]
+        parent_value = @a[parent_index]
+        break if parent_value <= value
+        @a[index] = parent_value
         index = parent_index
-        @a[index] = score
-        # check_heap!(index)
       end
-      self
+      @a[index] = value
+      # dbg "__push__(%p)" % [value]
+      # check_heap!(index)
     end
 
     def pop
       return if @a.empty?
       popped = @a.first
-      @a[0] = shifting = @a.last
-      @a.pop
-      # shift down
-      index = 0
+      value = @a.pop
       last_index = @a.size - 1
-      while (child_index = index * 2 + 1) <= last_index
+      last_parent = (last_index - 1) / 2
+      return popped unless 0 <= last_index
+
+      # sift down from 0
+      index = 0
+      child_index = 1
+      while index <= last_parent
+        child_value = @a[child_index]
         # select min child
-        if child_index < last_index && @a[child_index + 1] < @a[child_index]
-          child_index += 1
+        if child_index < last_index
+          other_child_index = child_index + 1
+          other_child_value = @a[other_child_index]
+          if other_child_value < child_value
+            child_value = other_child_value
+            child_index = other_child_index
+          end
         end
-        break if @a[index] <= @a[child_index]
-        @a[index] = @a[child_index]
+        break if value <= child_value
+        @a[index] = child_value
         index = child_index
-        @a[index] = shifting
+        child_index = index * 2 + 1
       end
+      @a[index] = value
+
       popped
     end
 
     private
 
-    def check_heap!(idx, last = @a.size - 1)
-      pscore = @a[idx]
-      child = idx * 2 + 1
-      if child <= last
-        cscore = check_heap!(child)
-        raise "#{pscore} > #{cscore}" if pscore > cscore
-      end
-      child += 1
-      if child <= last
-        check_heap!(child)
-        cscore = check_heap!(child)
-        raise "#{pscore} > #{cscore}" if pscore > cscore
+    def check_heap!(idx)
+      check_heap_up!(idx)
+      check_heap_dn!(idx)
+    end
+
+    # compares index to its parent
+    def check_heap_at!(idx)
+      value = @a[idx]
+      unless idx <= 0
+        pidx = (idx - 1) / 2
+        pval = @a[pidx]
+        raise "@a[#{idx}] == #{value}, #{pval} > #{value}" if pval > value
       end
-      pscore
+      value
+    end
+
+    def check_heap_up!(idx)
+      return if idx <= 0
+      pidx = (idx - 1) / 2
+      check_heap_at!(pidx)
+      check_heap_up!(pidx)
+    end
+
+    def check_heap_dn!(idx)
+      return unless @a.size <= idx
+      check_heap_at!(idx)
+      check_heap_down!(idx * 2 + 1)
+      check_heap_down!(idx * 2 + 2)
     end
 
   end
   # rubocop:enable Metrics/MethodLength, Metrics/AbcSize
 
+  # minor adjustments to the "priority_queue_cxx" gem, to match the API
+  class CppSTL
+
+    def initialize
+      clear
+    end
+
+    def <<(value); @q.push(value, value) end
+
+    def clear
+      @q = FastContainers::PriorityQueue.new(:min)
+    end
+
+    def pop
+      @q.pop
+    rescue RuntimeError
+      nil
+    end
+
+  end
+
   # Different duck-typed priority queue implemenations
   IMPLEMENTATIONS = [
     OpenStruct.new(name: " push and resort", klass: Sorting).freeze,
     OpenStruct.new(name: "  find min + del", klass: FindMin).freeze,
     OpenStruct.new(name: "bsearch + insert", klass: BSearch).freeze,
     OpenStruct.new(name: "ruby binary heap", klass: RbHeap).freeze,
+    OpenStruct.new(name: "C++STL PriorityQ", klass: CppSTL).freeze,
     OpenStruct.new(name: "quaternary DHeap", klass: DHeap).freeze,
   ].freeze
 

data/lib/d_heap/version.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
 class DHeap
-  VERSION = "0.4.0"
+  VERSION = "0.5.0"
 
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: d_heap
 version: !ruby/object:Gem::Version
-  version: 0.4.0
+  version: 0.5.0
 platform: ruby
 authors:
 - nicholas a. evans
 autorequire: 
 bindir: exe
 cert_chain: []
-date: 2021-01-12 00:00:00.000000000 Z
+date: 2021-01-17 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: benchmark_driver
@@ -60,9 +60,11 @@ files:
 - LICENSE.txt
 - README.md
 - Rakefile
+- benchmarks/perf.rb
 - benchmarks/push_n.yml
 - benchmarks/push_n_pop_n.yml
 - benchmarks/push_pop.yml
+- benchmarks/stackprof.rb
 - bin/bench_n
 - bin/benchmark-driver
 - bin/benchmarks
@@ -74,6 +76,7 @@ files:
 - bin/setup
 - d_heap.gemspec
 - docs/benchmarks-2.txt
+- docs/benchmarks-mem.txt
 - docs/benchmarks.txt
 - docs/profile.txt
 - ext/d_heap/d_heap.c