Switch List implementation to use Trie-based lookup

lib/public_suffix.rb

@@ -9,6 +9,7 @@
 require_relative "public_suffix/errors"
 require_relative "public_suffix/rule"
 require_relative "public_suffix/list"
+require_relative "public_suffix/trie"
 
 # PublicSuffix is a Ruby domain name parser based on the Public Suffix List.
 #

lib/public_suffix/list.rb

@@ -104,6 +104,7 @@ def self.parse(input, private_domains: true)
     # @yieldparam [PublicSuffix::List] self The newly created instance.
     def initialize
       @rules = {}
+      @trie = PublicSuffix::Trie.new
       yield(self) if block_given?
     end
 
@@ -137,7 +138,9 @@ def each(&block)
     # @param  rule [PublicSuffix::Rule::*] the rule to add to the list
     # @return [self]
     def add(rule)
-      @rules[rule.value] = rule_to_entry(rule)
+      entry = rule_to_entry(rule)
+      @rules[rule.value] = entry
+      @trie.insert(rule.value, type: entry.type, private: entry.private)
       self
     end
     alias << add
@@ -170,14 +173,26 @@ def clear
     # @param  default [PublicSuffix::Rule::*] the default rule to return in case no rule matches
     # @return [PublicSuffix::Rule::*]
     def find(name, default: default_rule, **options)
+      if ENV["WHAT"] == "hash"
+        find_hash(name, default: default, **options)
+      else
+        find_trie(name, default: default, **options)
+      end
+    end
+
+    def find_hash(name, default: default_rule, **options)
       rule = select(name, **options).inject do |l, r|
         return r if r.class == Rule::Exception
         l.length > r.length ? l : r
       end
       rule || default
     end
 
-    # Selects all the rules matching given hostame.
+    def find_trie(name, default: default_rule, ignore_private: false)
+      @trie.longest_prefix(name, ignore_private: ignore_private) || default
+    end
+
+    # Selects all the rules matching given hostname.
     #
     # If `ignore_private` is set to true, the algorithm will skip the rules that are flagged as
     # private domain. Note that the rules will still be part of the loop.

lib/public_suffix/trie.rb

@@ -0,0 +1,97 @@
+module PublicSuffix
+
+  # Implements a Trie data structure used to store the List.
+  class Trie
+
+    # @return [PublicSuffix::Node]
+    attr_reader :root
+
+    def initialize
+      @root = self.class::Node.new
+    end
+
+    def insert(word, type:, private:)
+      node = @root
+      word.split(DOT).reverse.each do |token|
+        node = node.put(token)
+      end
+      node.end!(type: type, private: private) && node
+    end
+
+    def longest_prefix(word, ignore_private: false)
+      node = @root
+
+      results = []
+      leaf = nil
+      excp = nil
+
+      word.split(DOT).reverse.each_with_index do |token, index|
+        break unless (child = node.get(token))
+        results << [child, token]
+        node = child
+
+        if node.end? && (ignore_private == false || node.private == false)
+          leaf = index + 1
+          excp = index + 1 if node.type == Rule::Exception
+        end
+      end
+
+      return nil if leaf.nil?
+
+      path = excp ? results[0, excp] : results[0, leaf]
+      node = path.last.first
+
+      tokens = []
+      (path.size - 1).downto(0).each do |index|
+        tokens << path[index].last
+      end
+      node.type.new(value: tokens.join(DOT), private: node.private)
+    end
+
+
+    # Node is a node of the Trie and contains references to all the children nodes.
+    #
+    # A node marked as "end" represents the final part of a rule. It contains the rule information
+    # such as the rule type and whether it belongs to PRIVATE.
+    class Node
+      attr_accessor :children
+      attr_accessor :type
+      attr_accessor :private
+
+      def initialize
+        @children = nil
+      end
+
+      def contains?(key)
+        return false if @children.nil?
+        !@children[index(key)].nil?
+      end
+
+      def put(key)
+        @children ||= {}
+        @children[index(key)] ||= self.class.new
+      end
+
+      def get(key)
+        return nil if @children.nil?
+        @children[index(key)]
+      end
+
+      def end?
+        [email protected]?
+      end
+
+      def end!(type:, private:)
+        @type = type
+        @private = private
+      end
+
+      private
+
+      def index(key)
+        key
+      end
+    end
+
+  end
+end

test/profilers/list_profsize.rb

@@ -8,4 +8,5 @@
 
 prof = ObjectBinsize.new
 prof.report(PublicSuffix::List.default, label: "PublicSuffix::List size")
-prof.report(PublicSuffix::List.default.instance_variable_get(:@rules), label: "Size of rules")
+prof.report(PublicSuffix::List.default.instance_variable_get(:@rules), label: "Size of @rules")
+prof.report(PublicSuffix::List.default.instance_variable_get(:@trie),  label: "Size of @trie")

test/profilers/tries_profiler.rb

@@ -0,0 +1,15 @@
+$LOAD_PATH.unshift File.expand_path("../../lib", __dir__)
+
+require "memory_profiler"
+require "public_suffix"
+require "public_suffix/trie"
+
+list = PublicSuffix::List.default
+puts "#{list.size} rules:"
+
+report = MemoryProfiler.report do
+  @trie = PublicSuffix::Trie.new
+  list.instance_variable_get(:@rules).keys { |word| @trie.insert(word.split(".").reverse.join(".")) }
+end
+
+report.pretty_print

test/profilers/tries_prosize.rb

@@ -0,0 +1,15 @@
+$LOAD_PATH.unshift File.expand_path("../../lib", __dir__)
+
+require_relative "object_binsize"
+require "public_suffix"
+require "public_suffix/trie"
+
+list = PublicSuffix::List.default
+rules = list.instance_variable_get(:@rules)
+
+@trie = PublicSuffix::Trie.new
+rules.keys.each { |word| @trie.insert(word.split(".").reverse.join(".")) }
+
+prof = ObjectBinsize.new
+prof.report(rules, label: "@rules")
+prof.report(@trie, label: "@trie")