immutable-ruby 0.0.4 → 0.2.0

data/lib/immutable/_core.rb

@@ -0,0 +1,3067 @@
+require 'immutable/undefined'
+require 'immutable/enumerable'
+require 'immutable/trie'
+require 'immutable/sorted_set'
+require 'set'
+
+module Immutable
+
+  # An `Immutable::Hash` maps a set of unique keys to corresponding values, much
+  # like a dictionary maps from words to definitions. Given a key, it can store
+  # and retrieve an associated value in constant time. If an existing key is
+  # stored again, the new value will replace the old. It behaves much like
+  # Ruby's built-in Hash, which we will call RubyHash for clarity. Like
+  # RubyHash, two keys that are `#eql?` to each other and have the same
+  # `#hash` are considered identical in an `Immutable::Hash`.
+  #
+  # An `Immutable::Hash` can be created in a couple of ways:
+  #
+  #     Immutable::Hash.new(font_size: 10, font_family: 'Arial')
+  #     Immutable::Hash[first_name: 'John', last_name: 'Smith']
+  #
+  # Any `Enumerable` object which yields two-element `[key, value]` arrays
+  # can be used to initialize an `Immutable::Hash`:
+  #
+  #     Immutable::Hash.new([[:first_name, 'John'], [:last_name, 'Smith']])
+  #
+  # Key/value pairs can be added using {#put}. A new hash is returned and the
+  # existing one is left unchanged:
+  #
+  #     hash = Immutable::Hash[a: 100, b: 200]
+  #     hash.put(:c, 500) # => Immutable::Hash[:a => 100, :b => 200, :c => 500]
+  #     hash              # => Immutable::Hash[:a => 100, :b => 200]
+  #
+  # {#put} can also take a block, which is used to calculate the value to be
+  # stored.
+  #
+  #     hash.put(:a) { |current| current + 200 } # => Immutable::Hash[:a => 300, :b => 200]
+  #
+  # Since it is immutable, all methods which you might expect to "modify" a
+  # `Immutable::Hash` actually return a new hash and leave the existing one
+  # unchanged. This means that the `hash[key] = value` syntax from RubyHash
+  # *cannot* be used with `Immutable::Hash`.
+  #
+  # Nested data structures can easily be updated using {#update_in}:
+  #
+  #     hash = Immutable::Hash["a" => Immutable::Vector[Immutable::Hash["c" => 42]]]
+  #     hash.update_in("a", 0, "c") { |value| value + 5 }
+  #     # => Immutable::Hash["a" => Immutable::Hash["b" => Immutable::Hash["c" => 47]]]
+  #
+  # While an `Immutable::Hash` can iterate over its keys or values, it does not
+  # guarantee any specific iteration order (unlike RubyHash). Methods like
+  # {#flatten} do not guarantee the order of returned key/value pairs.
+  #
+  # Like RubyHash, an `Immutable::Hash` can have a default block which is used
+  # when looking up a key that does not exist. Unlike RubyHash, the default
+  # block will only be passed the missing key, without the hash itself:
+  #
+  #     hash = Immutable::Hash.new { |missing_key| missing_key * 10 }
+  #     hash[5] # => 50
+  class Hash
+    include Immutable::Enumerable
+
+    class << self
+      # Create a new `Hash` populated with the given key/value pairs.
+      #
+      # @example
+      #   Immutable::Hash["A" => 1, "B" => 2] # => Immutable::Hash["A" => 1, "B" => 2]
+      #   Immutable::Hash[["A", 1], ["B", 2]] # => Immutable::Hash["A" => 1, "B" => 2]
+      #
+      # @param pairs [::Enumerable] initial content of hash. An empty hash is returned if not provided.
+      # @return [Hash]
+      def [](pairs = nil)
+        (pairs.nil? || pairs.empty?) ? empty : new(pairs)
+      end
+
+      # Return an empty `Hash`. If used on a subclass, returns an empty instance
+      # of that class.
+      #
+      # @return [Hash]
+      def empty
+        @empty ||= new
+      end
+
+      # "Raw" allocation of a new `Hash`. Used internally to create a new
+      # instance quickly after obtaining a modified {Trie}.
+      #
+      # @return [Hash]
+      # @private
+      def alloc(trie = EmptyTrie, block = nil)
+        obj = allocate
+        obj.instance_variable_set(:@trie, trie)
+        obj.instance_variable_set(:@default, block)
+        obj.freeze
+      end
+    end
+
+    # @param pairs [::Enumerable] initial content of hash. An empty hash is returned if not provided.
+    # @yield [key] Optional _default block_ to be stored and used to calculate the default value of a missing key. It will not be yielded during this method. It will not be preserved when marshalling.
+    # @yieldparam key Key that was not present in the hash.
+    def initialize(pairs = nil, &block)
+      @trie    = pairs ? Trie[pairs] : EmptyTrie
+      @default = block
+      freeze
+    end
+
+    # Return the default block if there is one. Otherwise, return `nil`.
+    #
+    # @return [Proc]
+    def default_proc
+      @default
+    end
+
+    # Return the number of key/value pairs in this `Hash`.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3].size  # => 3
+    #
+    # @return [Integer]
+    def size
+      @trie.size
+    end
+    alias length size
+
+    # Return `true` if this `Hash` contains no key/value pairs.
+    #
+    # @return [Boolean]
+    def empty?
+      @trie.empty?
+    end
+
+    # Return `true` if the given key object is present in this `Hash`. More precisely,
+    # return `true` if a key with the same `#hash` code, and which is also `#eql?`
+    # to the given key object is present.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3].key?("B")  # => true
+    #
+    # @param key [Object] The key to check for
+    # @return [Boolean]
+    def key?(key)
+      @trie.key?(key)
+    end
+    alias has_key? key?
+    alias include? key?
+    alias member?  key?
+
+    # Return `true` if this `Hash` has one or more keys which map to the provided value.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3].value?(2)  # => true
+    #
+    # @param value [Object] The value to check for
+    # @return [Boolean]
+    def value?(value)
+      each { |k,v| return true if value == v }
+      false
+    end
+    alias has_value? value?
+
+    # Retrieve the value corresponding to the provided key object. If not found, and
+    # this `Hash` has a default block, the default block is called to provide the
+    # value. Otherwise, return `nil`.
+    #
+    # @example
+    #   h = Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #   h["B"]             # => 2
+    #   h.get("B")         # => 2
+    #   h.get("Elephant")  # => nil
+    #
+    #   # Immutable Hash with a default proc:
+    #   h = Immutable::Hash.new("A" => 1, "B" => 2, "C" => 3) { |key| key.size }
+    #   h.get("B")         # => 2
+    #   h.get("Elephant")  # => 8
+    #
+    # @param key [Object] The key to look up
+    # @return [Object]
+    def get(key)
+      entry = @trie.get(key)
+      if entry
+        entry[1]
+      elsif @default
+        @default.call(key)
+      end
+    end
+    alias [] get
+
+    # Retrieve the value corresponding to the given key object, or use the provided
+    # default value or block, or otherwise raise a `KeyError`.
+    #
+    # @overload fetch(key)
+    #   Retrieve the value corresponding to the given key, or raise a `KeyError`
+    #   if it is not found.
+    #   @param key [Object] The key to look up
+    # @overload fetch(key) { |key| ... }
+    #   Retrieve the value corresponding to the given key, or call the optional
+    #   code block (with the missing key) and get its return value.
+    #   @yield [key] The key which was not found
+    #   @yieldreturn [Object] Object to return since the key was not found
+    #   @param key [Object] The key to look up
+    # @overload fetch(key, default)
+    #   Retrieve the value corresponding to the given key, or else return
+    #   the provided `default` value.
+    #   @param key [Object] The key to look up
+    #   @param default [Object] Object to return if the key is not found
+    #
+    # @example
+    #   h = Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #   h.fetch("B")         # => 2
+    #   h.fetch("Elephant")  # => KeyError: key not found: "Elephant"
+    #
+    #   # with a default value:
+    #   h.fetch("B", 99)         # => 2
+    #   h.fetch("Elephant", 99)  # => 99
+    #
+    #   # with a block:
+    #   h.fetch("B") { |key| key.size }         # => 2
+    #   h.fetch("Elephant") { |key| key.size }  # => 8
+    #
+    # @return [Object]
+    def fetch(key, default = Undefined)
+      entry = @trie.get(key)
+      if entry
+        entry[1]
+      elsif block_given?
+        yield(key)
+      elsif default != Undefined
+        default
+      else
+        raise KeyError, "key not found: #{key.inspect}"
+      end
+    end
+
+    # Return a new `Hash` with the existing key/value associations, plus an association
+    # between the provided key and value. If an equivalent key is already present, its
+    # associated value will be replaced with the provided one.
+    #
+    # If the `value` argument is missing, but an optional code block is provided,
+    # it will be passed the existing value (or `nil` if there is none) and what it
+    # returns will replace the existing value. This is useful for "transforming"
+    # the value associated with a certain key.
+    #
+    # Avoid mutating objects which are used as keys. `String`s are an exception:
+    # unfrozen `String`s which are used as keys are internally duplicated and
+    # frozen. This matches RubyHash's behaviour.
+    #
+    # @example
+    #   h = Immutable::Hash["A" => 1, "B" => 2]
+    #   h.put("C", 3)
+    #   # => Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #   h.put("B") { |value| value * 10 }
+    #   # => Immutable::Hash["A" => 1, "B" => 20]
+    #
+    # @param key [Object] The key to store
+    # @param value [Object] The value to associate it with
+    # @yield [value] The previously stored value, or `nil` if none.
+    # @yieldreturn [Object] The new value to store
+    # @return [Hash]
+    def put(key, value = yield(get(key)))
+      new_trie = @trie.put(key, value)
+      if new_trie.equal?(@trie)
+        self
+      else
+        self.class.alloc(new_trie, @default)
+      end
+    end
+
+    # @private
+    # @raise NoMethodError
+    def []=(*)
+      raise NoMethodError, "Immutable::Hash doesn't support `[]='; use `put' instead"
+    end
+
+    # Return a new `Hash` with a deeply nested value modified to the result of
+    # the given code block.  When traversing the nested `Hash`es and `Vector`s,
+    # non-existing keys are created with empty `Hash` values.
+    #
+    # The code block receives the existing value of the deeply nested key (or
+    # `nil` if it doesn't exist). This is useful for "transforming" the value
+    # associated with a certain key.
+    #
+    # Note that the original `Hash` and sub-`Hash`es and sub-`Vector`s are left
+    # unmodified; new data structure copies are created along the path wherever
+    # needed.
+    #
+    # @example
+    #   hash = Immutable::Hash["a" => Immutable::Hash["b" => Immutable::Hash["c" => 42]]]
+    #   hash.update_in("a", "b", "c") { |value| value + 5 }
+    #   # => Immutable::Hash["a" => Immutable::Hash["b" => Immutable::Hash["c" => 47]]]
+    #
+    # @param key_path [::Array<Object>] List of keys which form the path to the key to be modified
+    # @yield [value] The previously stored value
+    # @yieldreturn [Object] The new value to store
+    # @return [Hash]
+    def update_in(*key_path, &block)
+      if key_path.empty?
+        raise ArgumentError, 'must have at least one key in path'
+      end
+      key = key_path[0]
+      if key_path.size == 1
+        new_value = block.call(get(key))
+      else
+        value = fetch(key, EmptyHash)
+        new_value = value.update_in(*key_path[1..-1], &block)
+      end
+      put(key, new_value)
+    end
+
+    # An alias for {#put} to match RubyHash's API. Does not support {#put}'s
+    # block form.
+    #
+    # @see #put
+    # @param key [Object] The key to store
+    # @param value [Object] The value to associate it with
+    # @return [Hash]
+    def store(key, value)
+      put(key, value)
+    end
+
+    # Return a new `Hash` with `key` removed. If `key` is not present, return
+    # `self`.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3].delete("B")
+    #   # => Immutable::Hash["A" => 1, "C" => 3]
+    #
+    # @param key [Object] The key to remove
+    # @return [Hash]
+    def delete(key)
+      derive_new_hash(@trie.delete(key))
+    end
+
+    # Call the block once for each key/value pair in this `Hash`, passing the key/value
+    # pair as parameters. No specific iteration order is guaranteed, though the order will
+    # be stable for any particular `Hash`.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3].each { |k, v| puts "k=#{k} v=#{v}" }
+    #
+    #   k=A v=1
+    #   k=C v=3
+    #   k=B v=2
+    #   # => Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #
+    # @yield [key, value] Once for each key/value pair.
+    # @return [self]
+    def each(&block)
+      return to_enum if not block_given?
+      @trie.each(&block)
+      self
+    end
+    alias each_pair each
+
+    # Call the block once for each key/value pair in this `Hash`, passing the key/value
+    # pair as parameters. Iteration order will be the opposite of {#each}.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3].reverse_each { |k, v| puts "k=#{k} v=#{v}" }
+    #
+    #   k=B v=2
+    #   k=C v=3
+    #   k=A v=1
+    #   # => Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #
+    # @yield [key, value] Once for each key/value pair.
+    # @return [self]
+    def reverse_each(&block)
+      return enum_for(:reverse_each) if not block_given?
+      @trie.reverse_each(&block)
+      self
+    end
+
+    # Call the block once for each key/value pair in this `Hash`, passing the key as a
+    # parameter. Ordering guarantees are the same as {#each}.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3].each_key { |k| puts "k=#{k}" }
+    #
+    #   k=A
+    #   k=C
+    #   k=B
+    #   # => Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #
+    # @yield [key] Once for each key/value pair.
+    # @return [self]
+    def each_key
+      return enum_for(:each_key) if not block_given?
+      @trie.each { |k,v| yield k }
+      self
+    end
+
+    # Call the block once for each key/value pair in this `Hash`, passing the value as a
+    # parameter. Ordering guarantees are the same as {#each}.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3].each_value { |v| puts "v=#{v}" }
+    #
+    #   v=1
+    #   v=3
+    #   v=2
+    #   # => Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #
+    # @yield [value] Once for each key/value pair.
+    # @return [self]
+    def each_value
+      return enum_for(:each_value) if not block_given?
+      @trie.each { |k,v| yield v }
+      self
+    end
+
+    # Call the block once for each key/value pair in this `Hash`, passing the key/value
+    # pair as parameters. The block should return a `[key, value]` array each time.
+    # All the returned `[key, value]` arrays will be gathered into a new `Hash`.
+    #
+    # @example
+    #   h = Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #   h.map { |k, v| ["new-#{k}", v * v] }
+    #   # => Hash["new-C" => 9, "new-B" => 4, "new-A" => 1]
+    #
+    # @yield [key, value] Once for each key/value pair.
+    # @return [Hash]
+    def map
+      return enum_for(:map) unless block_given?
+      return self if empty?
+      self.class.new(super, &@default)
+    end
+    alias collect map
+
+    # Return a new `Hash` with all the key/value pairs for which the block returns true.
+    #
+    # @example
+    #   h = Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #   h.select { |k, v| v >= 2 }
+    #   # => Immutable::Hash["B" => 2, "C" => 3]
+    #
+    # @yield [key, value] Once for each key/value pair.
+    # @yieldreturn Truthy if this pair should be present in the new `Hash`.
+    # @return [Hash]
+    def select(&block)
+      return enum_for(:select) unless block_given?
+      derive_new_hash(@trie.select(&block))
+    end
+    alias find_all select
+    alias keep_if  select
+
+    # Yield `[key, value]` pairs until one is found for which the block returns true.
+    # Return that `[key, value]` pair. If the block never returns true, return `nil`.
+    #
+    # @example
+    #   h = Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #   h.find { |k, v| v.even? }
+    #   # => ["B", 2]
+    #
+    # @return [Array]
+    # @yield [key, value] At most once for each key/value pair, until the block returns `true`.
+    # @yieldreturn Truthy to halt iteration and return the yielded key/value pair.
+    def find
+      return enum_for(:find) unless block_given?
+      each { |entry| return entry if yield entry }
+      nil
+    end
+    alias detect find
+
+    # Return a new `Hash` containing all the key/value pairs from this `Hash` and
+    # `other`. If no block is provided, the value for entries with colliding keys
+    # will be that from `other`. Otherwise, the value for each duplicate key is
+    # determined by calling the block.
+    #
+    # `other` can be an `Immutable::Hash`, a built-in Ruby `Hash`, or any `Enumerable`
+    # object which yields `[key, value]` pairs.
+    #
+    # @example
+    #   h1 = Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #   h2 = Immutable::Hash["C" => 70, "D" => 80]
+    #   h1.merge(h2)
+    #   # => Immutable::Hash["C" => 70, "A" => 1, "D" => 80, "B" => 2]
+    #   h1.merge(h2) { |key, v1, v2| v1 + v2 }
+    #   # => Immutable::Hash["C" => 73, "A" => 1, "D" => 80, "B" => 2]
+    #
+    # @param other [::Enumerable] The collection to merge with
+    # @yieldparam key [Object] The key which was present in both collections
+    # @yieldparam my_value [Object] The associated value from this `Hash`
+    # @yieldparam other_value [Object] The associated value from the other collection
+    # @yieldreturn [Object] The value to associate this key with in the new `Hash`
+    # @return [Hash]
+    def merge(other)
+      trie = if block_given?
+        other.reduce(@trie) do |trie, (key, value)|
+          if (entry = trie.get(key))
+            trie.put(key, yield(key, entry[1], value))
+          else
+            trie.put(key, value)
+          end
+        end
+      else
+        @trie.bulk_put(other)
+      end
+
+      derive_new_hash(trie)
+    end
+
+    # Retrieve the value corresponding to the given key object, or use the provided
+    # default value or block, or otherwise raise a `KeyError`.
+    #
+    # @overload fetch(key)
+    #   Retrieve the value corresponding to the given key, or raise a `KeyError`
+    #   if it is not found.
+    #   @param key [Object] The key to look up
+    # @overload fetch(key) { |key| ... }
+
+    # Return a sorted {Vector} which contains all the `[key, value]` pairs in
+    # this `Hash` as two-element `Array`s.
+    #
+    # @overload sort
+    #   Uses `#<=>` to determine sorted order.
+    # @overload sort { |(k1, v1), (k2, v2)| ... }
+    #   Uses the block as a comparator to determine sorted order.
+    #
+    #   @example
+    #     h = Immutable::Hash["Dog" => 1, "Elephant" => 2, "Lion" => 3]
+    #     h.sort { |(k1, v1), (k2, v2)| k1.size  <=> k2.size }
+    #     # => Immutable::Vector[["Dog", 1], ["Lion", 3], ["Elephant", 2]]
+    #   @yield [(k1, v1), (k2, v2)] Any number of times with different pairs of key/value associations.
+    #   @yieldreturn [Integer] Negative if the first pair should be sorted
+    #                          lower, positive if the latter pair, or 0 if equal.
+    #
+    # @see ::Enumerable#sort
+    #
+    # @return [Vector]
+    def sort
+      Vector.new(super)
+    end
+
+    # Return a {Vector} which contains all the `[key, value]` pairs in this `Hash`
+    # as two-element Arrays. The order which the pairs will appear in is determined by
+    # passing each pair to the code block to obtain a sort key object, and comparing
+    # the sort keys using `#<=>`.
+    #
+    # @see ::Enumerable#sort_by
+    #
+    # @example
+    #   h = Immutable::Hash["Dog" => 1, "Elephant" => 2, "Lion" => 3]
+    #   h.sort_by { |key, value| key.size }
+    #   # => Immutable::Vector[["Dog", 1], ["Lion", 3], ["Elephant", 2]]
+    #
+    # @yield [key, value] Once for each key/value pair.
+    # @yieldreturn a sort key object for the yielded pair.
+    # @return [Vector]
+    def sort_by
+      Vector.new(super)
+    end
+
+    # Return a new `Hash` with the associations for all of the given `keys` removed.
+    #
+    # @example
+    #   h = Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #   h.except("A", "C")  # => Immutable::Hash["B" => 2]
+    #
+    # @param keys [Array] The keys to remove
+    # @return [Hash]
+    def except(*keys)
+      keys.reduce(self) { |hash, key| hash.delete(key) }
+    end
+
+    # Return a new `Hash` with only the associations for the `wanted` keys retained.
+    #
+    # @example
+    #   h = Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #   h.slice("B", "C")  # => Immutable::Hash["B" => 2, "C" => 3]
+    #
+    # @param wanted [::Enumerable] The keys to retain
+    # @return [Hash]
+    def slice(*wanted)
+      trie = Trie.new(0)
+      wanted.each { |key| trie.put!(key, get(key)) if key?(key) }
+      self.class.alloc(trie, @default)
+    end
+
+    # Return a {Vector} of the values which correspond to the `wanted` keys.
+    # If any of the `wanted` keys are not present in this `Hash`, `nil` will be
+    # placed instead, or the result of the default proc (if one is defined),
+    # similar to the behavior of {#get}.
+    #
+    # @example
+    #   h = Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #   h.values_at("B", "A", "D")  # => Immutable::Vector[2, 1, nil]
+    #
+    # @param wanted [Array] The keys to retrieve
+    # @return [Vector]
+    def values_at(*wanted)
+      Vector.new(wanted.map { |key| get(key) }.freeze)
+    end
+
+    # Return a {Vector} of the values which correspond to the `wanted` keys.
+    # If any of the `wanted` keys are not present in this `Hash`, raise `KeyError`
+    # exception.
+    #
+    # @example
+    #   h = Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #   h.fetch_values("C", "A")  # => Immutable::Vector[3, 1]
+    #   h.fetch_values("C", "Z")  # => KeyError: key not found: "Z"
+    #
+    # @param wanted [Array] The keys to retrieve
+    # @return [Vector]
+    def fetch_values(*wanted)
+      array = wanted.map { |key| fetch(key) }
+      Vector.new(array.freeze)
+    end
+
+    # Return the value of successively indexing into a nested collection.
+    # If any of the keys is not present, return `nil`.
+    #
+    # @example
+    #   h = Immutable::Hash[a: 9, b: Immutable::Hash[c: 'a', d: 4], e: nil]
+    #   h.dig(:b, :c) # => "a"
+    #   h.dig(:b, :f) # => nil
+    #
+    # @return [Object]
+    def dig(key, *rest)
+      value = self[key]
+      if rest.empty? || value.nil?
+        value
+      else
+        value.dig(*rest)
+      end
+    end
+
+    # Return a new {Set} containing the keys from this `Hash`.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3, "D" => 2].keys
+    #   # => Immutable::Set["D", "C", "B", "A"]
+    #
+    # @return [Set]
+    def keys
+      Set.alloc(@trie)
+    end
+
+    # Return a new {Vector} populated with the values from this `Hash`.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3, "D" => 2].values
+    #   # => Immutable::Vector[2, 3, 2, 1]
+    #
+    # @return [Vector]
+    def values
+      Vector.new(each_value.to_a.freeze)
+    end
+
+    # Return a new `Hash` created by using keys as values and values as keys.
+    # If there are multiple values which are equivalent (as determined by `#hash` and
+    # `#eql?`), only one out of each group of equivalent values will be
+    # retained. Which one specifically is undefined.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3, "D" => 2].invert
+    #   # => Immutable::Hash[1 => "A", 3 => "C", 2 => "B"]
+    #
+    # @return [Hash]
+    def invert
+      pairs = []
+      each { |k,v| pairs << [v, k] }
+      self.class.new(pairs, &@default)
+    end
+
+    # Return a new {Vector} which is a one-dimensional flattening of this `Hash`.
+    # If `level` is 1, all the `[key, value]` pairs in the hash will be concatenated
+    # into one {Vector}. If `level` is greater than 1, keys or values which are
+    # themselves `Array`s or {Vector}s will be recursively flattened into the output
+    # {Vector}. The depth to which that flattening will be recursively applied is
+    # determined by `level`.
+    #
+    # As a special case, if `level` is 0, each `[key, value]` pair will be a
+    # separate element in the returned {Vector}.
+    #
+    # @example
+    #   h = Immutable::Hash["A" => 1, "B" => [2, 3, 4]]
+    #   h.flatten
+    #   # => Immutable::Vector["A", 1, "B", [2, 3, 4]]
+    #   h.flatten(2)
+    #   # => Immutable::Vector["A", 1, "B", 2, 3, 4]
+    #
+    # @param level [Integer] The number of times to recursively flatten the `[key, value]` pairs in this `Hash`.
+    # @return [Vector]
+    def flatten(level = 1)
+      return Vector.new(self) if level == 0
+      array = []
+      each { |k,v| array << k; array << v }
+      array.flatten!(level-1) if level > 1
+      Vector.new(array.freeze)
+    end
+
+    # Searches through the `Hash`, comparing `obj` with each key (using `#==`).
+    # When a matching key is found, return the `[key, value]` pair as an array.
+    # Return `nil` if no match is found.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3].assoc("B")  # => ["B", 2]
+    #
+    # @param obj [Object] The key to search for (using #==)
+    # @return [Array]
+    def assoc(obj)
+      each { |entry| return entry if obj == entry[0] }
+      nil
+    end
+
+    # Searches through the `Hash`, comparing `obj` with each value (using `#==`).
+    # When a matching value is found, return the `[key, value]` pair as an array.
+    # Return `nil` if no match is found.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3].rassoc(2)  # => ["B", 2]
+    #
+    # @param obj [Object] The value to search for (using #==)
+    # @return [Array]
+    def rassoc(obj)
+      each { |entry| return entry if obj == entry[1] }
+      nil
+    end
+
+    # Searches through the `Hash`, comparing `value` with each value (using `#==`).
+    # When a matching value is found, return its associated key object.
+    # Return `nil` if no match is found.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3].key(2)  # => "B"
+    #
+    # @param value [Object] The value to search for (using #==)
+    # @return [Object]
+    def key(value)
+      each { |entry| return entry[0] if value == entry[1] }
+      nil
+    end
+
+    # Return a randomly chosen `[key, value]` pair from this `Hash`. If the hash is empty,
+    # return `nil`.
+    #
+    # @example
+    #   Immutable::Hash["A" => 1, "B" => 2, "C" => 3].sample
+    #   # => ["C", 3]
+    #
+    # @return [Array]
+    def sample
+      @trie.at(rand(size))
+    end
+
+    # Return an empty `Hash` instance, of the same class as this one. Useful if you
+    # have multiple subclasses of `Hash` and want to treat them polymorphically.
+    # Maintains the default block, if there is one.
+    #
+    # @return [Hash]
+    def clear
+      if @default
+        self.class.alloc(EmptyTrie, @default)
+      else
+        self.class.empty
+      end
+    end
+
+    # Return true if `other` has the same type and contents as this `Hash`.
+    #
+    # @param other [Object] The collection to compare with
+    # @return [Boolean]
+    def eql?(other)
+      return true if other.equal?(self)
+      instance_of?(other.class) && @trie.eql?(other.instance_variable_get(:@trie))
+    end
+
+    # Return true if `other` has the same contents as this `Hash`. Will convert
+    # `other` to a Ruby `Hash` using `#to_hash` if necessary.
+    #
+    # @param other [Object] The object to compare with
+    # @return [Boolean]
+    def ==(other)
+      eql?(other) || (other.respond_to?(:to_hash) && to_hash == other.to_hash)
+    end
+
+    # Return true if this `Hash` is a proper superset of `other`, which means
+    # all `other`'s keys are contained in this `Hash` with identical
+    # values, and the two hashes are not identical.
+    #
+    # @param other [Immutable::Hash] The object to compare with
+    # @return [Boolean]
+    def >(other)
+      self != other && self >= other
+    end
+
+    # Return true if this `Hash` is a superset of `other`, which means all
+    # `other`'s keys are contained in this `Hash` with identical values.
+    #
+    # @param other [Immutable::Hash] The object to compare with
+    # @return [Boolean]
+    def >=(other)
+      other.each do |key, value|
+        if self[key] != value
+          return false
+        end
+      end
+      true
+    end
+
+    # Return true if this `Hash` is a proper subset of `other`, which means all
+    # its keys are contained in `other` with the identical values, and the two
+    # hashes are not identical.
+    #
+    # @param other [Immutable::Hash] The object to compare with
+    # @return [Boolean]
+    def <(other)
+      other > self
+    end
+
+    # Return true if this `Hash` is a subset of `other`, which means all its
+    # keys are contained in `other` with the identical values, and the two
+    # hashes are not identical.
+    #
+    # @param other [Immutable::Hash] The object to compare with
+    # @return [Boolean]
+    def <=(other)
+      other >= self
+    end
+
+    # See `Object#hash`.
+    # @return [Integer]
+    def hash
+      keys.to_a.sort.reduce(0) do |hash, key|
+        (hash << 32) - hash + key.hash + get(key).hash
+      end
+    end
+
+    # Return the contents of this `Hash` as a programmer-readable `String`. If all the
+    # keys and values are serializable as Ruby literal strings, the returned string can
+    # be passed to `eval` to reconstitute an equivalent `Hash`. The default
+    # block (if there is one) will be lost when doing this, however.
+    #
+    # @return [String]
+    def inspect
+      result = "#{self.class}["
+      i = 0
+      each do |key, val|
+        result << ', ' if i > 0
+        result << key.inspect << ' => ' << val.inspect
+        i += 1
+      end
+      result << ']'
+    end
+
+    # Return `self`. Since this is an immutable object duplicates are
+    # equivalent.
+    # @return [Hash]
+    def dup
+      self
+    end
+    alias clone dup
+
+    # Allows this `Hash` to be printed at the `pry` console, or using `pp` (from the
+    # Ruby standard library), in a way which takes the amount of horizontal space on
+    # the screen into account, and which indents nested structures to make them easier
+    # to read.
+    #
+    # @private
+    def pretty_print(pp)
+      pp.group(1, "#{self.class}[", ']') do
+        pp.breakable ''
+        pp.seplist(self, nil) do |key, val|
+          pp.group do
+            key.pretty_print(pp)
+            pp.text ' => '
+            pp.group(1) do
+              pp.breakable ''
+              val.pretty_print(pp)
+            end
+          end
+        end
+      end
+    end
+
+    # Convert this `Immutable::Hash` to an instance of Ruby's built-in `Hash`.
+    #
+    # @return [::Hash]
+    def to_hash
+      output = {}
+      each do |key, value|
+        output[key] = value
+      end
+      output
+    end
+    alias to_h to_hash
+
+    # Return a `Proc` which accepts a key as an argument and returns the value.
+    # The `Proc` behaves like {#get} (when the key is missing, it returns nil or
+    # the result of the default proc).
+    #
+    # @example
+    #   h = Immutable::Hash["A" => 1, "B" => 2, "C" => 3]
+    #   h.to_proc.call("B")
+    #   # => 2
+    #   ["A", "C", "X"].map(&h)   # The & is short for .to_proc in Ruby
+    #   # => [1, 3, nil]
+    #
+    # @return [Proc]
+    def to_proc
+      lambda { |key| get(key) }
+    end
+
+    # @return [::Hash]
+    # @private
+    def marshal_dump
+      to_hash
+    end
+
+    # @private
+    def marshal_load(dictionary)
+      @trie = Trie[dictionary]
+    end
+
+    private
+
+    # Return a new `Hash` which is derived from this one, using a modified {Trie}.
+    # The new `Hash` will retain the existing default block, if there is one.
+    #
+    def derive_new_hash(trie)
+      if trie.equal?(@trie)
+        self
+      elsif trie.empty?
+        if @default
+          self.class.alloc(EmptyTrie, @default)
+        else
+          self.class.empty
+        end
+      else
+        self.class.alloc(trie, @default)
+      end
+    end
+  end
+
+  # The canonical empty `Hash`. Returned by `Hash[]` when
+  # invoked with no arguments; also returned by `Hash.empty`. Prefer using this
+  # one rather than creating many empty hashes using `Hash.new`.
+  #
+  # @private
+  EmptyHash = Immutable::Hash.empty
+
+
+  # A `Vector` is an ordered, integer-indexed collection of objects. Like
+  # Ruby's `Array`, `Vector` indexing starts at zero and negative indexes count
+  # back from the end.
+  #
+  # `Vector` has a similar interface to `Array`. The main difference is methods
+  # that would destructively update an `Array` (such as {#insert} or
+  # {#delete_at}) instead return new `Vectors` and leave the existing one
+  # unchanged.
+  #
+  # ### Creating New Vectors
+  #
+  #     Immutable::Vector.new([:first, :second, :third])
+  #     Immutable::Vector[1, 2, 3, 4, 5]
+  #
+  # ### Retrieving Items from Vectors
+  #
+  #     vector = Immutable::Vector[1, 2, 3, 4, 5]
+  #
+  #     vector[0]      # => 1
+  #     vector[-1]     # => 5
+  #     vector[0,3]    # => Immutable::Vector[1, 2, 3]
+  #     vector[1..-1]  # => Immutable::Vector[2, 3, 4, 5]
+  #     vector.first   # => 1
+  #     vector.last    # => 5
+  #
+  # ### Creating Modified Vectors
+  #
+  #     vector.add(6)            # => Immutable::Vector[1, 2, 3, 4, 5, 6]
+  #     vector.insert(1, :a, :b) # => Immutable::Vector[1, :a, :b, 2, 3, 4, 5]
+  #     vector.delete_at(2)      # => Immutable::Vector[1, 2, 4, 5]
+  #     vector + [6, 7]          # => Immutable::Vector[1, 2, 3, 4, 5, 6, 7]
+  #
+  class Vector
+    include Immutable::Enumerable
+
+    # @private
+    BLOCK_SIZE = 32
+    # @private
+    INDEX_MASK = BLOCK_SIZE - 1
+    # @private
+    BITS_PER_LEVEL = 5
+
+    # Return the number of items in this `Vector`
+    # @return [Integer]
+    attr_reader :size
+    alias length size
+
+    class << self
+      # Create a new `Vector` populated with the given items.
+      # @return [Vector]
+      def [](*items)
+        new(items.freeze)
+      end
+
+      # Return an empty `Vector`. If used on a subclass, returns an empty instance
+      # of that class.
+      #
+      # @return [Vector]
+      def empty
+        @empty ||= new
+      end
+
+      # "Raw" allocation of a new `Vector`. Used internally to create a new
+      # instance quickly after building a modified trie.
+      #
+      # @return [Vector]
+      # @private
+      def alloc(root, size, levels)
+        obj = allocate
+        obj.instance_variable_set(:@root, root)
+        obj.instance_variable_set(:@size, size)
+        obj.instance_variable_set(:@levels, levels)
+        obj.freeze
+      end
+    end
+
+    def initialize(items=[].freeze)
+      items = items.to_a
+      if items.size <= 32
+        items = items.dup.freeze if !items.frozen?
+        @root, @size, @levels = items, items.size, 0
+      else
+        root, size, levels = items, items.size, 0
+        while root.size > 32
+          root = root.each_slice(32).to_a
+          levels += 1
+        end
+        @root, @size, @levels = root.freeze, size, levels
+      end
+      freeze
+    end
+
+    # Return `true` if this `Vector` contains no items.
+    #
+    # @return [Boolean]
+    def empty?
+      @size == 0
+    end
+
+    # Return the first item in the `Vector`. If the vector is empty, return `nil`.
+    #
+    # @example
+    #   Immutable::Vector["A", "B", "C"].first  # => "A"
+    #
+    # @return [Object]
+    def first
+      get(0)
+    end
+
+    # Return the last item in the `Vector`. If the vector is empty, return `nil`.
+    #
+    # @example
+    #   Immutable::Vector["A", "B", "C"].last  # => "C"
+    #
+    # @return [Object]
+    def last
+      get(-1)
+    end
+
+    # Return a new `Vector` with `item` added after the last occupied position.
+    #
+    # @example
+    #   Immutable::Vector[1, 2].add(99)  # => Immutable::Vector[1, 2, 99]
+    #
+    # @param item [Object] The object to insert at the end of the vector
+    # @return [Vector]
+    def add(item)
+      update_root(@size, item)
+    end
+    alias << add
+    alias push add
+
+    # Return a new `Vector` with a new value at the given `index`. If `index`
+    # is greater than the length of the vector, the returned vector will be
+    # padded with `nil`s to the correct size.
+    #
+    # @overload set(index, item)
+    #   Return a new `Vector` with the item at `index` replaced by `item`.
+    #
+    #   @param item [Object] The object to insert into that position
+    #   @example
+    #     Immutable::Vector[1, 2, 3, 4].set(2, 99)
+    #     # => Immutable::Vector[1, 2, 99, 4]
+    #     Immutable::Vector[1, 2, 3, 4].set(-1, 99)
+    #     # => Immutable::Vector[1, 2, 3, 99]
+    #     Immutable::Vector[].set(2, 99)
+    #     # => Immutable::Vector[nil, nil, 99]
+    #
+    # @overload set(index)
+    #   Return a new `Vector` with the item at `index` replaced by the return
+    #   value of the block.
+    #
+    #   @yield (existing) Once with the existing value at the given `index`.
+    #   @example
+    #     Immutable::Vector[1, 2, 3, 4].set(2) { |v| v * 10 }
+    #     # => Immutable::Vector[1, 2, 30, 4]
+    #
+    # @param index [Integer] The index to update. May be negative.
+    # @return [Vector]
+    def set(index, item = yield(get(index)))
+      raise IndexError, "index #{index} outside of vector bounds" if index < -@size
+      index += @size if index < 0
+      if index > @size
+        suffix = Array.new(index - @size, nil)
+        suffix << item
+        replace_suffix(@size, suffix)
+      else
+        update_root(index, item)
+      end
+    end
+
+    # Return a new `Vector` with a deeply nested value modified to the result
+    # of the given code block.  When traversing the nested `Vector`s and
+    # `Hash`es, non-existing keys are created with empty `Hash` values.
+    #
+    # The code block receives the existing value of the deeply nested key (or
+    # `nil` if it doesn't exist). This is useful for "transforming" the value
+    # associated with a certain key.
+    #
+    # Note that the original `Vector` and sub-`Vector`s and sub-`Hash`es are
+    # left unmodified; new data structure copies are created along the path
+    # wherever needed.
+    #
+    # @example
+    #   v = Immutable::Vector[123, 456, 789, Immutable::Hash["a" => Immutable::Vector[5, 6, 7]]]
+    #   v.update_in(3, "a", 1) { |value| value + 9 }
+    #   # => Immutable::Vector[123, 456, 789, Immutable::Hash["a" => Immutable::Vector[5, 15, 7]]]
+    #
+    # @param key_path [Object(s)] List of keys which form the path to the key to be modified
+    # @yield [value] The previously stored value
+    # @yieldreturn [Object] The new value to store
+    # @return [Vector]
+    def update_in(*key_path, &block)
+      if key_path.empty?
+        raise ArgumentError, 'must have at least one key in path'
+      end
+      key = key_path[0]
+      if key_path.size == 1
+        new_value = block.call(get(key))
+      else
+        value = fetch(key, Immutable::EmptyHash)
+        new_value = value.update_in(*key_path[1..-1], &block)
+      end
+      set(key, new_value)
+    end
+
+    # Retrieve the item at `index`. If there is none (either the provided index
+    # is too high or too low), return `nil`.
+    #
+    # @example
+    #   v = Immutable::Vector["A", "B", "C", "D"]
+    #   v.get(2)   # => "C"
+    #   v.get(-1)  # => "D"
+    #   v.get(4)   # => nil
+    #
+    # @param index [Integer] The index to retrieve
+    # @return [Object]
+    def get(index)
+      return nil if @size == 0
+      index += @size if index < 0
+      return nil if index >= @size || index < 0
+      leaf_node_for(@root, @levels * BITS_PER_LEVEL, index)[index & INDEX_MASK]
+    end
+    alias at get
+
+    # Retrieve the value at `index` with optional default.
+    #
+    # @overload fetch(index)
+    #   Retrieve the value at the given index, or raise an `IndexError` if not
+    #   found.
+    #
+    #   @param index [Integer] The index to look up
+    #   @raise [IndexError] if index does not exist
+    #   @example
+    #     v = Immutable::Vector["A", "B", "C", "D"]
+    #     v.fetch(2)       # => "C"
+    #     v.fetch(-1)      # => "D"
+    #     v.fetch(4)       # => IndexError: index 4 outside of vector bounds
+    #
+    # @overload fetch(index) { |index| ... }
+    #   Retrieve the value at the given index, or return the result of yielding
+    #   the block if not found.
+    #
+    #   @yield Once if the index is not found.
+    #   @yieldparam [Integer] index The index which does not exist
+    #   @yieldreturn [Object] Default value to return
+    #   @param index [Integer] The index to look up
+    #   @example
+    #     v = Immutable::Vector["A", "B", "C", "D"]
+    #     v.fetch(2) { |i| i * i }   # => "C"
+    #     v.fetch(4) { |i| i * i }   # => 16
+    #
+    # @overload fetch(index, default)
+    #   Retrieve the value at the given index, or return the provided `default`
+    #   value if not found.
+    #
+    #   @param index [Integer] The index to look up
+    #   @param default [Object] Object to return if the key is not found
+    #   @example
+    #     v = Immutable::Vector["A", "B", "C", "D"]
+    #     v.fetch(2, "Z")  # => "C"
+    #     v.fetch(4, "Z")  # => "Z"
+    #
+    # @return [Object]
+    def fetch(index, default = (missing_default = true))
+      if index >= -@size && index < @size
+        get(index)
+      elsif block_given?
+        yield(index)
+      elsif !missing_default
+        default
+      else
+        raise IndexError, "index #{index} outside of vector bounds"
+      end
+    end
+
+    # Return the value of successively indexing into a nested collection.
+    # If any of the keys is not present, return `nil`.
+    #
+    # @example
+    #   v = Immutable::Vector[9, Immutable::Hash[c: 'a', d: 4]]
+    #   v.dig(1, :c) # => "a"
+    #   v.dig(1, :f) # => nil
+    #
+    # @return [Object]
+    def dig(key, *rest)
+      value = self[key]
+      if rest.empty? || value.nil?
+        value
+      else
+        value.dig(*rest)
+      end
+    end
+
+    # Return specific objects from the `Vector`. All overloads return `nil` if
+    # the starting index is out of range.
+    #
+    # @overload vector.slice(index)
+    #   Returns a single object at the given `index`. If `index` is negative,
+    #   count backwards from the end.
+    #
+    #   @param index [Integer] The index to retrieve. May be negative.
+    #   @return [Object]
+    #   @example
+    #     v = Immutable::Vector["A", "B", "C", "D", "E", "F"]
+    #     v[2]  # => "C"
+    #     v[-1] # => "F"
+    #     v[6]  # => nil
+    #
+    # @overload vector.slice(index, length)
+    #   Return a subvector starting at `index` and continuing for `length`
+    #   elements or until the end of the `Vector`, whichever occurs first.
+    #
+    #   @param start [Integer] The index to start retrieving items from. May be
+    #                          negative.
+    #   @param length [Integer] The number of items to retrieve.
+    #   @return [Vector]
+    #   @example
+    #     v = Immutable::Vector["A", "B", "C", "D", "E", "F"]
+    #     v[2, 3]  # => Immutable::Vector["C", "D", "E"]
+    #     v[-2, 3] # => Immutable::Vector["E", "F"]
+    #     v[20, 1] # => nil
+    #
+    # @overload vector.slice(index..end)
+    #   Return a subvector starting at `index` and continuing to index
+    #   `end` or the end of the `Vector`, whichever occurs first.
+    #
+    #   @param range [Range] The range of indices to retrieve.
+    #   @return [Vector]
+    #   @example
+    #     v = Immutable::Vector["A", "B", "C", "D", "E", "F"]
+    #     v[2..3]    # => Immutable::Vector["C", "D"]
+    #     v[-2..100] # => Immutable::Vector["E", "F"]
+    #     v[20..21]  # => nil
+    def slice(arg, length = (missing_length = true))
+      if missing_length
+        if arg.is_a?(Range)
+          from, to = arg.begin, arg.end
+          from += @size if from < 0
+          to   += @size if to < 0
+          to   += 1     if !arg.exclude_end?
+          length = to - from
+          length = 0 if length < 0
+          subsequence(from, length)
+        else
+          get(arg)
+        end
+      else
+        arg += @size if arg < 0
+        subsequence(arg, length)
+      end
+    end
+    alias [] slice
+
+    # Return a new `Vector` with the given values inserted before the element
+    # at `index`. If `index` is greater than the current length, `nil` values
+    # are added to pad the `Vector` to the required size.
+    #
+    # @example
+    #   Immutable::Vector["A", "B", "C", "D"].insert(2, "X", "Y", "Z")
+    #   # => Immutable::Vector["A", "B", "X", "Y", "Z", "C", "D"]
+    #   Immutable::Vector[].insert(2, "X", "Y", "Z")
+    #   # => Immutable::Vector[nil, nil, "X", "Y", "Z"]
+    #
+    # @param index [Integer] The index where the new items should go
+    # @param items [Array] The items to add
+    # @return [Vector]
+    # @raise [IndexError] if index exceeds negative range.
+    def insert(index, *items)
+      raise IndexError if index < -@size
+      index += @size if index < 0
+
+      if index < @size
+        suffix = flatten_suffix(@root, @levels * BITS_PER_LEVEL, index, [])
+        suffix.unshift(*items)
+      elsif index == @size
+        suffix = items
+      else
+        suffix = Array.new(index - @size, nil).concat(items)
+        index = @size
+      end
+
+      replace_suffix(index, suffix)
+    end
+
+    # Return a new `Vector` with the element at `index` removed. If the given `index`
+    # does not exist, return `self`.
+    #
+    # @example
+    #   Immutable::Vector["A", "B", "C", "D"].delete_at(2)
+    #   # => Immutable::Vector["A", "B", "D"]
+    #
+    # @param index [Integer] The index to remove
+    # @return [Vector]
+    def delete_at(index)
+      return self if index >= @size || index < -@size
+      index += @size if index < 0
+
+      suffix = flatten_suffix(@root, @levels * BITS_PER_LEVEL, index, [])
+      replace_suffix(index, suffix.tap(&:shift))
+    end
+
+    # Return a new `Vector` with the last element removed. Return `self` if
+    # empty.
+    #
+    # @example
+    #   Immutable::Vector["A", "B", "C"].pop  # => Immutable::Vector["A", "B"]
+    #
+    # @return [Vector]
+    def pop
+      return self if @size == 0
+      replace_suffix(@size-1, [])
+    end
+
+    # Return a new `Vector` with `object` inserted before the first element,
+    # moving the other elements upwards.
+    #
+    # @example
+    #   Immutable::Vector["A", "B"].unshift("Z")
+    #   # => Immutable::Vector["Z", "A", "B"]
+    #
+    # @param object [Object] The value to prepend
+    # @return [Vector]
+    def unshift(object)
+      insert(0, object)
+    end
+
+    # Return a new `Vector` with the first element removed. If empty, return
+    # `self`.
+    #
+    # @example
+    #   Immutable::Vector["A", "B", "C"].shift  # => Immutable::Vector["B", "C"]
+    #
+    # @return [Vector]
+    def shift
+      delete_at(0)
+    end
+
+    # Call the given block once for each item in the vector, passing each
+    # item from first to last successively to the block. If no block is given,
+    # an `Enumerator` is returned instead.
+    #
+    # @example
+    #   Immutable::Vector["A", "B", "C"].each { |e| puts "Element: #{e}" }
+    #
+    #   Element: A
+    #   Element: B
+    #   Element: C
+    #   # => Immutable::Vector["A", "B", "C"]
+    #
+    # @return [self, Enumerator]
+    def each(&block)
+      return to_enum unless block_given?
+      traverse_depth_first(@root, @levels, &block)
+      self
+    end
+
+    # Call the given block once for each item in the vector, from last to
+    # first.
+    #
+    # @example
+    #   Immutable::Vector["A", "B", "C"].reverse_each { |e| puts "Element: #{e}" }
+    #
+    #   Element: C
+    #   Element: B
+    #   Element: A
+    #
+    # @return [self]
+    def reverse_each(&block)
+      return enum_for(:reverse_each) unless block_given?
+      reverse_traverse_depth_first(@root, @levels, &block)
+      self
+    end
+
+    # Return a new `Vector` containing all elements for which the given block returns
+    # true.
+    #
+    # @example
+    #   Immutable::Vector["Bird", "Cow", "Elephant"].select { |e| e.size >= 4 }
+    #   # => Immutable::Vector["Bird", "Elephant"]
+    #
+    # @return [Vector]
+    # @yield [element] Once for each element.
+    def select
+      return enum_for(:select) unless block_given?
+      reduce(self.class.empty) { |vector, item| yield(item) ? vector.add(item) : vector }
+    end
+    alias find_all select
+    alias keep_if  select
+
+    # Return a new `Vector` with all items which are equal to `obj` removed.
+    # `#==` is used for checking equality.
+    #
+    # @example
+    #   Immutable::Vector["C", "B", "A", "B"].delete("B")  # => Immutable::Vector["C", "A"]
+    #
+    # @param obj [Object] The object to remove (every occurrence)
+    # @return [Vector]
+    def delete(obj)
+      select { |item| item != obj }
+    end
+
+    # Invoke the given block once for each item in the vector, and return a new
+    # `Vector` containing the values returned by the block. If no block is
+    # provided, return an enumerator.
+    #
+    # @example
+    #   Immutable::Vector[3, 2, 1].map { |e| e * e }  # => Immutable::Vector[9, 4, 1]
+    #
+    # @return [Vector, Enumerator]
+    def map
+      return enum_for(:map) if not block_given?
+      return self if empty?
+      self.class.new(super)
+    end
+    alias collect map
+
+    # Return a new `Vector` with the concatenated results of running the block once
+    # for every element in this `Vector`.
+    #
+    # @example
+    #   Immutable::Vector[1, 2, 3].flat_map { |x| [x, -x] }
+    #   # => Immutable::Vector[1, -1, 2, -2, 3, -3]
+    #
+    # @return [Vector]
+    def flat_map
+      return enum_for(:flat_map) if not block_given?
+      return self if empty?
+      self.class.new(super)
+    end
+
+    # Return a new `Vector` with the same elements as this one, but randomly permuted.
+    #
+    # @example
+    #   Immutable::Vector[1, 2, 3, 4].shuffle  # => Immutable::Vector[4, 1, 3, 2]
+    #
+    # @return [Vector]
+    def shuffle
+      self.class.new(((array = to_a).frozen? ? array.shuffle : array.shuffle!).freeze)
+    end
+
+    # Return a new `Vector` with no duplicate elements, as determined by `#hash` and
+    # `#eql?`. For each group of equivalent elements, only the first will be retained.
+    #
+    # @example
+    #   Immutable::Vector["A", "B", "C", "B"].uniq      # => Immutable::Vector["A", "B", "C"]
+    #   Immutable::Vector["a", "A", "b"].uniq(&:upcase) # => Immutable::Vector["a", "b"]
+    #
+    # @return [Vector]
+    def uniq(&block)
+      array = to_a
+      if array.frozen?
+        self.class.new(array.uniq(&block).freeze)
+      elsif array.uniq!(&block) # returns nil if no changes were made
+        self.class.new(array.freeze)
+      else
+        self
+      end
+    end
+
+    # Return a new `Vector` with the same elements as this one, but in reverse order.
+    #
+    # @example
+    #   Immutable::Vector["A", "B", "C"].reverse  # => Immutable::Vector["C", "B", "A"]
+    #
+    # @return [Vector]
+    def reverse
+      self.class.new(((array = to_a).frozen? ? array.reverse : array.reverse!).freeze)
+    end
+
+    # Return a new `Vector` with the same elements, but rotated so that the one at
+    # index `count` is the first element of the new vector. If `count` is positive,
+    # the elements will be shifted left, and those shifted past the lowest position
+    # will be moved to the end. If `count` is negative, the elements will be shifted
+    # right, and those shifted past the last position will be moved to the beginning.
+    #
+    # @example
+    #   v = Immutable::Vector["A", "B", "C", "D", "E", "F"]
+    #   v.rotate(2)   # => Immutable::Vector["C", "D", "E", "F", "A", "B"]
+    #   v.rotate(-1)  # => Immutable::Vector["F", "A", "B", "C", "D", "E"]
+    #
+    # @param count [Integer] The number of positions to shift items by
+    # @return [Vector]
+    def rotate(count = 1)
+      return self if (count % @size) == 0
+      self.class.new(((array = to_a).frozen? ? array.rotate(count) : array.rotate!(count)).freeze)
+    end
+
+    # Return a new `Vector` with all nested vectors and arrays recursively "flattened
+    # out". That is, their elements inserted into the new `Vector` in the place where
+    # the nested array/vector originally was. If an optional `level` argument is
+    # provided, the flattening will only be done recursively that number of times.
+    # A `level` of 0 means not to flatten at all, 1 means to only flatten nested
+    # arrays/vectors which are directly contained within this `Vector`.
+    #
+    # @example
+    #   v = Immutable::Vector["A", Immutable::Vector["B", "C", Immutable::Vector["D"]]]
+    #   v.flatten(1)
+    #   # => Immutable::Vector["A", "B", "C", Immutable::Vector["D"]]
+    #   v.flatten
+    #   # => Immutable::Vector["A", "B", "C", "D"]
+    #
+    # @param level [Integer] The depth to which flattening should be applied
+    # @return [Vector]
+    def flatten(level = -1)
+      return self if level == 0
+      array = to_a
+      if array.frozen?
+        self.class.new(array.flatten(level).freeze)
+      elsif array.flatten!(level) # returns nil if no changes were made
+        self.class.new(array.freeze)
+      else
+        self
+      end
+    end
+
+    # Return a new `Vector` built by concatenating this one with `other`. `other`
+    # can be any object which is convertible to an `Array` using `#to_a`.
+    #
+    # @example
+    #   Immutable::Vector["A", "B", "C"] + ["D", "E"]
+    #   # => Immutable::Vector["A", "B", "C", "D", "E"]
+    #
+    # @param other [Enumerable] The collection to concatenate onto this vector
+    # @return [Vector]
+    def +(other)
+      other = other.to_a
+      other = other.dup if other.frozen?
+      replace_suffix(@size, other)
+    end
+    alias concat +
+
+    # Combine two vectors by "zipping" them together. `others` should be arrays
+    # and/or vectors. The corresponding elements from this `Vector` and each of
+    # `others` (that is, the elements with the same indices) will be gathered
+    # into arrays.
+    #
+    # If `others` contains fewer elements than this vector, `nil` will be used
+    # for padding.
+    #
+    # @overload zip(*others)
+    #   Return a new vector containing the new arrays.
+    #
+    #   @return [Vector]
+    #
+    # @overload zip(*others)
+    #   @yield [pair] once for each array
+    #   @return [nil]
+    #
+    # @example
+    #   v1 = Immutable::Vector["A", "B", "C"]
+    #   v2 = Immutable::Vector[1, 2]
+    #   v1.zip(v2)
+    #   # => Immutable::Vector[["A", 1], ["B", 2], ["C", nil]]
+    #
+    # @param others [Array] The arrays/vectors to zip together with this one
+    # @return [Vector]
+    def zip(*others)
+      if block_given?
+        super
+      else
+        self.class.new(super)
+      end
+    end
+
+    # Return a new `Vector` with the same items, but sorted.
+    #
+    # @overload sort
+    #   Compare elements with their natural sort key (`#<=>`).
+    #
+    #   @example
+    #     Immutable::Vector["Elephant", "Dog", "Lion"].sort
+    #     # => Immutable::Vector["Dog", "Elephant", "Lion"]
+    #
+    # @overload sort
+    #   Uses the block as a comparator to determine sorted order.
+    #
+    #   @yield [a, b] Any number of times with different pairs of elements.
+    #   @yieldreturn [Integer] Negative if the first element should be sorted
+    #                          lower, positive if the latter element, or 0 if
+    #                          equal.
+    #   @example
+    #     Immutable::Vector["Elephant", "Dog", "Lion"].sort { |a,b| a.size <=> b.size }
+    #     # => Immutable::Vector["Dog", "Lion", "Elephant"]
+    #
+    # @return [Vector]
+    def sort
+      self.class.new(super)
+    end
+
+    # Return a new `Vector` with the same items, but sorted. The sort order is
+    # determined by mapping the items through the given block to obtain sort
+    # keys, and then sorting the keys according to their natural sort order
+    # (`#<=>`).
+    #
+    # @yield [element] Once for each element.
+    # @yieldreturn a sort key object for the yielded element.
+    # @example
+    #   Immutable::Vector["Elephant", "Dog", "Lion"].sort_by { |e| e.size }
+    #   # => Immutable::Vector["Dog", "Lion", "Elephant"]
+    #
+    # @return [Vector]
+    def sort_by
+      self.class.new(super)
+    end
+
+    # Drop the first `n` elements and return the rest in a new `Vector`.
+    #
+    # @example
+    #   Immutable::Vector["A", "B", "C", "D", "E", "F"].drop(2)
+    #   # => Immutable::Vector["C", "D", "E", "F"]
+    #
+    # @param n [Integer] The number of elements to remove
+    # @return [Vector]
+    # @raise ArgumentError if `n` is negative.
+    def drop(n)
+      return self if n == 0
+      return self.class.empty if n >= @size
+      raise ArgumentError, 'attempt to drop negative size' if n < 0
+      self.class.new(flatten_suffix(@root, @levels * BITS_PER_LEVEL, n, []))
+    end
+
+    # Return only the first `n` elements in a new `Vector`.
+    #
+    # @example
+    #   Immutable::Vector["A", "B", "C", "D", "E", "F"].take(4)
+    #   # => Immutable::Vector["A", "B", "C", "D"]
+    #
+    # @param n [Integer] The number of elements to retain
+    # @return [Vector]
+    def take(n)
+      return self if n >= @size
+      self.class.new(super)
+    end
+
+    # Drop elements up to, but not including, the first element for which the
+    # block returns `nil` or `false`. Gather the remaining elements into a new
+    # `Vector`. If no block is given, an `Enumerator` is returned instead.
+    #
+    # @example
+    #   Immutable::Vector[1, 3, 5, 7, 6, 4, 2].drop_while { |e| e < 5 }
+    #   # => Immutable::Vector[5, 7, 6, 4, 2]
+    #
+    # @return [Vector, Enumerator]
+    def drop_while
+      return enum_for(:drop_while) if not block_given?
+      self.class.new(super)
+    end
+
+    # Gather elements up to, but not including, the first element for which the
+    # block returns `nil` or `false`, and return them in a new `Vector`. If no block
+    # is given, an `Enumerator` is returned instead.
+    #
+    # @example
+    #   Immutable::Vector[1, 3, 5, 7, 6, 4, 2].take_while { |e| e < 5 }
+    #   # => Immutable::Vector[1, 3]
+    #
+    # @return [Vector, Enumerator]
+    def take_while
+      return enum_for(:take_while) if not block_given?
+      self.class.new(super)
+    end
+
+    # Repetition. Return a new `Vector` built by concatenating `times` copies
+    # of this one together.
+    #
+    # @example
+    #   Immutable::Vector["A", "B"] * 3
+    #   # => Immutable::Vector["A", "B", "A", "B", "A", "B"]
+    #
+    # @param times [Integer] The number of times to repeat the elements in this vector
+    # @return [Vector]
+    def *(times)
+      return self.class.empty if times == 0
+      return self if times == 1
+      result = (to_a * times)
+      result.is_a?(Array) ? self.class.new(result) : result
+    end
+
+    # Replace a range of indexes with the given object.
+    #
+    # @overload fill(object)
+    #   Return a new `Vector` of the same size, with every index set to
+    #   `object`.
+    #
+    #   @param [Object] object Fill value.
+    #   @example
+    #     Immutable::Vector["A", "B", "C", "D", "E", "F"].fill("Z")
+    #     # => Immutable::Vector["Z", "Z", "Z", "Z", "Z", "Z"]
+    #
+    # @overload fill(object, index)
+    #   Return a new `Vector` with all indexes from `index` to the end of the
+    #   vector set to `object`.
+    #
+    #   @param [Object] object Fill value.
+    #   @param [Integer] index Starting index. May be negative.
+    #   @example
+    #     Immutable::Vector["A", "B", "C", "D", "E", "F"].fill("Z", 3)
+    #     # => Immutable::Vector["A", "B", "C", "Z", "Z", "Z"]
+    #
+    # @overload fill(object, index, length)
+    #   Return a new `Vector` with `length` indexes, beginning from `index`,
+    #   set to `object`. Expands the `Vector` if `length` would extend beyond
+    #   the current length.
+    #
+    #   @param [Object] object Fill value.
+    #   @param [Integer] index Starting index. May be negative.
+    #   @param [Integer] length
+    #   @example
+    #     Immutable::Vector["A", "B", "C", "D", "E", "F"].fill("Z", 3, 2)
+    #     # => Immutable::Vector["A", "B", "C", "Z", "Z", "F"]
+    #     Immutable::Vector["A", "B"].fill("Z", 1, 5)
+    #     # => Immutable::Vector["A", "Z", "Z", "Z", "Z", "Z"]
+    #
+    # @return [Vector]
+    # @raise [IndexError] if index is out of negative range.
+    def fill(object, index = 0, length = nil)
+      raise IndexError if index < -@size
+      index += @size if index < 0
+      length ||= @size - index # to the end of the array, if no length given
+
+      if index < @size
+        suffix = flatten_suffix(@root, @levels * BITS_PER_LEVEL, index, [])
+        suffix.fill(object, 0, length)
+      elsif index == @size
+        suffix = Array.new(length, object)
+      else
+        suffix = Array.new(index - @size, nil).concat(Array.new(length, object))
+        index = @size
+      end
+
+      replace_suffix(index, suffix)
+    end
+
+    # When invoked with a block, yields all combinations of length `n` of items
+    # from the `Vector`, and then returns `self`. There is no guarantee about
+    # which order the combinations will be yielded.
+    #
+    # If no block is given, an `Enumerator` is returned instead.
+    #
+    # @example
+    #   v = Immutable::Vector[5, 6, 7, 8]
+    #   v.combination(3) { |c| puts "Combination: #{c}" }
+    #
+    #   Combination: [5, 6, 7]
+    #   Combination: [5, 6, 8]
+    #   Combination: [5, 7, 8]
+    #   Combination: [6, 7, 8]
+    #   #=> Immutable::Vector[5, 6, 7, 8]
+    #
+    # @return [self, Enumerator]
+    def combination(n)
+      return enum_for(:combination, n) if not block_given?
+      return self if n < 0 || @size < n
+      if n == 0
+        yield []
+      elsif n == 1
+        each { |item| yield [item] }
+      elsif n == @size
+        yield to_a
+      else
+        combos = lambda do |result,index,remaining|
+          while @size - index > remaining
+            if remaining == 1
+              yield result.dup << get(index)
+            else
+              combos[result.dup << get(index), index+1, remaining-1]
+            end
+            index += 1
+          end
+          index.upto(@size-1) { |i| result << get(i) }
+          yield result
+        end
+        combos[[], 0, n]
+      end
+      self
+    end
+
+    # When invoked with a block, yields all repeated combinations of length `n` of
+    # items from the `Vector`, and then returns `self`. A "repeated combination" is
+    # one in which any item from the `Vector` can appear consecutively any number of
+    # times.
+    #
+    # There is no guarantee about which order the combinations will be yielded in.
+    #
+    # If no block is given, an `Enumerator` is returned instead.
+    #
+    # @example
+    #   v = Immutable::Vector[5, 6, 7, 8]
+    #   v.repeated_combination(2) { |c| puts "Combination: #{c}" }
+    #
+    #   Combination: [5, 5]
+    #   Combination: [5, 6]
+    #   Combination: [5, 7]
+    #   Combination: [5, 8]
+    #   Combination: [6, 6]
+    #   Combination: [6, 7]
+    #   Combination: [6, 8]
+    #   Combination: [7, 7]
+    #   Combination: [7, 8]
+    #   Combination: [8, 8]
+    #   # => Immutable::Vector[5, 6, 7, 8]
+    #
+    # @return [self, Enumerator]
+    def repeated_combination(n)
+      return enum_for(:repeated_combination, n) if not block_given?
+      if n < 0
+        # yield nothing
+      elsif n == 0
+        yield []
+      elsif n == 1
+        each { |item| yield [item] }
+      elsif @size == 0
+        # yield nothing
+      else
+        combos = lambda do |result,index,remaining|
+          while index < @size-1
+            if remaining == 1
+              yield result.dup << get(index)
+            else
+              combos[result.dup << get(index), index, remaining-1]
+            end
+            index += 1
+          end
+          item = get(index)
+          remaining.times { result << item }
+          yield result
+        end
+        combos[[], 0, n]
+      end
+      self
+    end
+
+    # Yields all permutations of length `n` of items from the `Vector`, and then
+    # returns `self`. If no length `n` is specified, permutations of all elements
+    # will be yielded.
+    #
+    # There is no guarantee about which order the permutations will be yielded in.
+    #
+    # If no block is given, an `Enumerator` is returned instead.
+    #
+    # @example
+    #   v = Immutable::Vector[5, 6, 7]
+    #   v.permutation(2) { |p| puts "Permutation: #{p}" }
+    #
+    #   Permutation: [5, 6]
+    #   Permutation: [5, 7]
+    #   Permutation: [6, 5]
+    #   Permutation: [6, 7]
+    #   Permutation: [7, 5]
+    #   Permutation: [7, 6]
+    #   # => Immutable::Vector[5, 6, 7]
+    #
+    # @return [self, Enumerator]
+    def permutation(n = @size)
+      return enum_for(:permutation, n) if not block_given?
+      if n < 0 || @size < n
+        # yield nothing
+      elsif n == 0
+        yield []
+      elsif n == 1
+        each { |item| yield [item] }
+      else
+        used, result = [], []
+        perms = lambda do |index|
+          0.upto(@size-1) do |i|
+            next if used[i]
+            result[index] = get(i)
+            if index < n-1
+              used[i] = true
+              perms[index+1]
+              used[i] = false
+            else
+              yield result.dup
+            end
+          end
+        end
+        perms[0]
+      end
+      self
+    end
+
+    # When invoked with a block, yields all repeated permutations of length `n` of
+    # items from the `Vector`, and then returns `self`. A "repeated permutation" is
+    # one where any item from the `Vector` can appear any number of times, and in
+    # any position (not just consecutively)
+    #
+    # If no length `n` is specified, permutations of all elements will be yielded.
+    # There is no guarantee about which order the permutations will be yielded in.
+    #
+    # If no block is given, an `Enumerator` is returned instead.
+    #
+    # @example
+    #   v = Immutable::Vector[5, 6, 7]
+    #   v.repeated_permutation(2) { |p| puts "Permutation: #{p}" }
+    #
+    #   Permutation: [5, 5]
+    #   Permutation: [5, 6]
+    #   Permutation: [5, 7]
+    #   Permutation: [6, 5]
+    #   Permutation: [6, 6]
+    #   Permutation: [6, 7]
+    #   Permutation: [7, 5]
+    #   Permutation: [7, 6]
+    #   Permutation: [7, 7]
+    #   # => Immutable::Vector[5, 6, 7]
+    #
+    # @return [self, Enumerator]
+    def repeated_permutation(n = @size)
+      return enum_for(:repeated_permutation, n) if not block_given?
+      if n < 0
+        # yield nothing
+      elsif n == 0
+        yield []
+      elsif n == 1
+        each { |item| yield [item] }
+      else
+        result = []
+        perms = lambda do |index|
+          0.upto(@size-1) do |i|
+            result[index] = get(i)
+            if index < n-1
+              perms[index+1]
+            else
+              yield result.dup
+            end
+          end
+        end
+        perms[0]
+      end
+      self
+    end
+
+    # Cartesian product or multiplication.
+    #
+    # @overload product(*vectors)
+    #   Return a `Vector` of all combinations of elements from this `Vector` and each
+    #   of the given vectors or arrays. The length of the returned `Vector` is the product
+    #   of `self.size` and the size of each argument vector or array.
+    #   @example
+    #     v1 = Immutable::Vector[1, 2, 3]
+    #     v2 = Immutable::Vector["A", "B"]
+    #     v1.product(v2)
+    #     # => [[1, "A"], [1, "B"], [2, "A"], [2, "B"], [3, "A"], [3, "B"]]
+    # @overload product
+    #   Return the result of multiplying all the items in this `Vector` together.
+    #
+    #   @example
+    #     Immutable::Vector[1, 2, 3, 4, 5].product  # => 120
+    #
+    # @return [Vector]
+    def product(*vectors)
+      # if no vectors passed, return "product" as in result of multiplying all items
+      return super if vectors.empty?
+
+      vectors.unshift(self)
+
+      if vectors.any?(&:empty?)
+        return block_given? ? self : []
+      end
+
+      counters = Array.new(vectors.size, 0)
+
+      bump_counters = lambda do
+        i = vectors.size-1
+        counters[i] += 1
+        while counters[i] == vectors[i].size
+          counters[i] = 0
+          i -= 1
+          return true if i == -1 # we are done
+          counters[i] += 1
+        end
+        false # not done yet
+      end
+      build_array = lambda do
+        array = []
+        counters.each_with_index { |index,i| array << vectors[i][index] }
+        array
+      end
+
+      if block_given?
+        loop do
+          yield build_array[]
+          return self if bump_counters[]
+        end
+      else
+        result = []
+        loop do
+          result << build_array[]
+          return result if bump_counters[]
+        end
+      end
+    end
+
+    # Assume all elements are vectors or arrays and transpose the rows and columns.
+    # In other words, take the first element of each nested vector/array and gather
+    # them together into a new `Vector`. Do likewise for the second, third, and so on
+    # down to the end of each nested vector/array. Gather all the resulting `Vectors`
+    # into a new `Vector` and return it.
+    #
+    # This operation is closely related to {#zip}. The result is almost the same as
+    # calling {#zip} on the first nested vector/array with the others supplied as
+    # arguments.
+    #
+    # @example
+    #   Immutable::Vector[["A", 10], ["B", 20], ["C", 30]].transpose
+    #   # => Immutable::Vector[Immutable::Vector["A", "B", "C"], Immutable::Vector[10, 20, 30]]
+    #
+    # @return [Vector]
+    # @raise [IndexError] if elements are not of the same size.
+    # @raise [TypeError] if an element does not respond to #size and #[]
+    def transpose
+      return self.class.empty if empty?
+      result = Array.new(first.size) { [] }
+
+      0.upto(@size-1) do |i|
+        source = get(i)
+        if source.size != result.size
+          raise IndexError, "element size differs (#{source.size} should be #{result.size})"
+        end
+
+        0.upto(result.size-1) do |j|
+          result[j].push(source[j])
+        end
+      end
+
+      result.map! { |a| self.class.new(a) }
+      self.class.new(result)
+    rescue NoMethodError
+      if any? { |x| !x.respond_to?(:size) || !x.respond_to?(:[]) }
+        bad = find { |x| !x.respond_to?(:size) || !x.respond_to?(:[]) }
+        raise TypeError, "'#{bad.inspect}' must respond to #size and #[] to be transposed"
+      else
+        raise
+      end
+    end
+
+    # Finds a value from this `Vector` which meets the condition defined by the
+    # provided block, using a binary search. The vector must already be sorted
+    # with respect to the block.  See Ruby's `Array#bsearch` for details,
+    # behaviour is equivalent.
+    #
+    # @example
+    #   v = Immutable::Vector[1, 3, 5, 7, 9, 11, 13]
+    #   # Block returns true/false for exact element match:
+    #   v.bsearch { |e| e > 4 }      # => 5
+    #   # Block returns number to match an element in 4 <= e <= 7:
+    #   v.bsearch { |e| 1 - e / 4 }  # => 7
+    #
+    # @yield Once for at most `log n` elements, where `n` is the size of the
+    #        vector. The exact elements and ordering are undefined.
+    # @yieldreturn [Boolean] `true` if this element matches the criteria, `false` otherwise.
+    # @yieldreturn [Integer] See `Array#bsearch` for details.
+    # @yieldparam [Object] element element to be evaluated
+    # @return [Object] The matched element, or `nil` if none found.
+    # @raise TypeError if the block returns a non-numeric, non-boolean, non-nil
+    #                  value.
+    def bsearch
+      return enum_for(:bsearch) if not block_given?
+      low, high, result = 0, @size, nil
+      while low < high
+        mid = (low + ((high - low) >> 1))
+        val = get(mid)
+        v   = yield val
+        if v.is_a? Numeric
+          if v == 0
+            return val
+          elsif v > 0
+            high = mid
+          else
+            low = mid + 1
+          end
+        elsif v == true
+          result = val
+          high = mid
+        elsif !v
+          low = mid + 1
+        else
+          raise TypeError, "wrong argument type #{v.class} (must be numeric, true, false, or nil)"
+        end
+      end
+      result
+    end
+
+    # Return an empty `Vector` instance, of the same class as this one. Useful if you
+    # have multiple subclasses of `Vector` and want to treat them polymorphically.
+    #
+    # @return [Vector]
+    def clear
+      self.class.empty
+    end
+
+    # Return a randomly chosen item from this `Vector`. If the vector is empty, return `nil`.
+    #
+    # @example
+    #   Immutable::Vector[1, 2, 3, 4, 5].sample  # => 2
+    #
+    # @return [Object]
+    def sample
+      get(rand(@size))
+    end
+
+    # Return a new `Vector` with only the elements at the given `indices`, in the
+    # order specified by `indices`. If any of the `indices` do not exist, `nil`s will
+    # appear in their places.
+    #
+    # @example
+    #   v = Immutable::Vector["A", "B", "C", "D", "E", "F"]
+    #   v.values_at(2, 4, 5)   # => Immutable::Vector["C", "E", "F"]
+    #
+    # @param indices [Array] The indices to retrieve and gather into a new `Vector`
+    # @return [Vector]
+    def values_at(*indices)
+      self.class.new(indices.map { |i| get(i) }.freeze)
+    end
+
+    # Find the index of an element, starting from the end of the vector.
+    # Returns `nil` if no element is found.
+    #
+    # @overload rindex(obj)
+    #   Return the index of the last element which is `#==` to `obj`.
+    #
+    #   @example
+    #     v = Immutable::Vector[7, 8, 9, 7, 8, 9]
+    #     v.rindex(8) # => 4
+    #
+    # @overload rindex
+    #   Return the index of the last element for which the block returns true.
+    #
+    #   @yield [element] Once for each element, last to first, until the block
+    #                    returns true.
+    #   @example
+    #     v = Immutable::Vector[7, 8, 9, 7, 8, 9]
+    #     v.rindex { |e| e.even? }  # => 4
+    #
+    # @return [Integer]
+    def rindex(obj = (missing_arg = true))
+      i = @size - 1
+      if missing_arg
+        if block_given?
+          reverse_each { |item| return i if yield item; i -= 1 }
+          nil
+        else
+          enum_for(:rindex)
+        end
+      else
+        reverse_each { |item| return i if item == obj; i -= 1 }
+        nil
+      end
+    end
+
+    # Assumes all elements are nested, indexable collections, and searches through them,
+    # comparing `obj` with the first element of each nested collection. Return the
+    # first nested collection which matches, or `nil` if none is found.
+    # Behaviour is undefined when elements do not meet assumptions (i.e. are
+    # not indexable collections).
+    #
+    # @example
+    #   v = Immutable::Vector[["A", 10], ["B", 20], ["C", 30]]
+    #   v.assoc("B")  # => ["B", 20]
+    #
+    # @param obj [Object] The object to search for
+    # @return [Object]
+    def assoc(obj)
+      each do |array|
+        next if !array.respond_to?(:[])
+        return array if obj == array[0]
+      end
+      nil
+    end
+
+    # Assumes all elements are nested, indexable collections, and searches through them,
+    # comparing `obj` with the second element of each nested collection. Return
+    # the first nested collection which matches, or `nil` if none is found.
+    # Behaviour is undefined when elements do not meet assumptions (i.e. are
+    # not indexable collections).
+    #
+    # @example
+    #   v = Immutable::Vector[["A", 10], ["B", 20], ["C", 30]]
+    #   v.rassoc(20)  # => ["B", 20]
+    #
+    # @param obj [Object] The object to search for
+    # @return [Object]
+    def rassoc(obj)
+      each do |array|
+        next if !array.respond_to?(:[])
+        return array if obj == array[1]
+      end
+      nil
+    end
+
+    # Return an `Array` with the same elements, in the same order. The returned
+    # `Array` may or may not be frozen.
+    #
+    # @return [Array]
+    def to_a
+      if @levels == 0
+        # When initializing a Vector with 32 or less items, we always make
+        # sure @root is frozen, so we can return it directly here
+        @root
+      else
+        flatten_node(@root, @levels * BITS_PER_LEVEL, [])
+      end
+    end
+    alias to_ary to_a
+
+    # Return true if `other` has the same type and contents as this `Vector`.
+    #
+    # @param other [Object] The collection to compare with
+    # @return [Boolean]
+    def eql?(other)
+      return true if other.equal?(self)
+      return false unless instance_of?(other.class) && @size == other.size
+      @root.eql?(other.instance_variable_get(:@root))
+    end
+
+    # See `Object#hash`.
+    # @return [Integer]
+    def hash
+      reduce(0) { |hash, item| (hash << 5) - hash + item.hash }
+    end
+
+    # Return `self`. Since this is an immutable object duplicates are
+    # equivalent.
+    # @return [Vector]
+    def dup
+      self
+    end
+    alias clone dup
+
+    # @return [::Array]
+    # @private
+    def marshal_dump
+      to_a
+    end
+
+    # @private
+    def marshal_load(array)
+      initialize(array.freeze)
+    end
+
+    private
+
+    def traverse_depth_first(node, level, &block)
+      return node.each(&block) if level == 0
+      node.each { |child| traverse_depth_first(child, level - 1, &block) }
+    end
+
+    def reverse_traverse_depth_first(node, level, &block)
+      return node.reverse_each(&block) if level == 0
+      node.reverse_each { |child| reverse_traverse_depth_first(child, level - 1, &block) }
+    end
+
+    def leaf_node_for(node, bitshift, index)
+      while bitshift > 0
+        node = node[(index >> bitshift) & INDEX_MASK]
+        bitshift -= BITS_PER_LEVEL
+      end
+      node
+    end
+
+    def update_root(index, item)
+      root, levels = @root, @levels
+      while index >= (1 << (BITS_PER_LEVEL * (levels + 1)))
+        root = [root].freeze
+        levels += 1
+      end
+      new_root = update_leaf_node(root, levels * BITS_PER_LEVEL, index, item)
+      if new_root.equal?(root)
+        self
+      else
+        self.class.alloc(new_root, @size > index ? @size : index + 1, levels)
+      end
+    end
+
+    def update_leaf_node(node, bitshift, index, item)
+      slot_index = (index >> bitshift) & INDEX_MASK
+      if bitshift > 0
+        old_child = node[slot_index] || []
+        item = update_leaf_node(old_child, bitshift - BITS_PER_LEVEL, index, item)
+      end
+      existing_item = node[slot_index]
+      if existing_item.equal?(item)
+        node
+      else
+        node.dup.tap { |n| n[slot_index] = item }.freeze
+      end
+    end
+
+    def flatten_range(node, bitshift, from, to)
+      from_slot = (from >> bitshift) & INDEX_MASK
+      to_slot   = (to   >> bitshift) & INDEX_MASK
+
+      if bitshift == 0 # are we at the bottom?
+        node.slice(from_slot, to_slot-from_slot+1)
+      elsif from_slot == to_slot
+        flatten_range(node[from_slot], bitshift - BITS_PER_LEVEL, from, to)
+      else
+        # the following bitmask can be used to pick out the part of the from/to indices
+        #   which will be used to direct path BELOW this node
+        mask   = ((1 << bitshift) - 1)
+        result = []
+
+        if from & mask == 0
+          flatten_node(node[from_slot], bitshift - BITS_PER_LEVEL, result)
+        else
+          result.concat(flatten_range(node[from_slot], bitshift - BITS_PER_LEVEL, from, from | mask))
+        end
+
+        (from_slot+1).upto(to_slot-1) do |slot_index|
+          flatten_node(node[slot_index], bitshift - BITS_PER_LEVEL, result)
+        end
+
+        if to & mask == mask
+          flatten_node(node[to_slot], bitshift - BITS_PER_LEVEL, result)
+        else
+          result.concat(flatten_range(node[to_slot], bitshift - BITS_PER_LEVEL, to & ~mask, to))
+        end
+
+        result
+      end
+    end
+
+    def flatten_node(node, bitshift, result)
+      if bitshift == 0
+        result.concat(node)
+      elsif bitshift == BITS_PER_LEVEL
+        node.each { |a| result.concat(a) }
+      else
+        bitshift -= BITS_PER_LEVEL
+        node.each { |a| flatten_node(a, bitshift, result) }
+      end
+      result
+    end
+
+    def subsequence(from, length)
+      return nil if from > @size || from < 0 || length < 0
+      length = @size - from if @size < from + length
+      return self.class.empty if length == 0
+      self.class.new(flatten_range(@root, @levels * BITS_PER_LEVEL, from, from + length - 1))
+    end
+
+    def flatten_suffix(node, bitshift, from, result)
+      from_slot = (from >> bitshift) & INDEX_MASK
+
+      if bitshift == 0
+        if from_slot == 0
+          result.concat(node)
+        else
+          result.concat(node.slice(from_slot, 32)) # entire suffix of node. excess length is ignored by #slice
+        end
+      else
+        mask = ((1 << bitshift) - 1)
+        if from & mask == 0
+          from_slot.upto(node.size-1) do |i|
+            flatten_node(node[i], bitshift - BITS_PER_LEVEL, result)
+          end
+        elsif (child = node[from_slot])
+          flatten_suffix(child, bitshift - BITS_PER_LEVEL, from, result)
+          (from_slot+1).upto(node.size-1) do |i|
+            flatten_node(node[i], bitshift - BITS_PER_LEVEL, result)
+          end
+        end
+        result
+      end
+    end
+
+    def replace_suffix(from, suffix)
+      # new suffix can go directly after existing elements
+      raise IndexError if from > @size
+      root, levels = @root, @levels
+
+      if (from >> (BITS_PER_LEVEL * (@levels + 1))) != 0
+        # index where new suffix goes doesn't fall within current tree
+        # we will need to deepen tree
+        root = [root].freeze
+        levels += 1
+      end
+
+      new_size = from + suffix.size
+      root = replace_node_suffix(root, levels * BITS_PER_LEVEL, from, suffix)
+
+      if !suffix.empty?
+        levels.times { suffix = suffix.each_slice(32).to_a }
+        root.concat(suffix)
+        while root.size > 32
+          root = root.each_slice(32).to_a
+          levels += 1
+        end
+      else
+        while root.size == 1 && levels > 0
+          root = root[0]
+          levels -= 1
+        end
+      end
+
+      self.class.alloc(root.freeze, new_size, levels)
+    end
+
+    def replace_node_suffix(node, bitshift, from, suffix)
+      from_slot = (from >> bitshift) & INDEX_MASK
+
+      if bitshift == 0
+        if from_slot == 0
+          suffix.shift(32)
+        else
+          node.take(from_slot).concat(suffix.shift(32 - from_slot))
+        end
+      else
+        mask = ((1 << bitshift) - 1)
+        if from & mask == 0
+          if from_slot == 0
+            new_node = suffix.shift(32 * (1 << bitshift))
+            while bitshift != 0
+              new_node = new_node.each_slice(32).to_a
+              bitshift -= BITS_PER_LEVEL
+            end
+            new_node
+          else
+            result = node.take(from_slot)
+            remainder = suffix.shift((32 - from_slot) * (1 << bitshift))
+            while bitshift != 0
+              remainder = remainder.each_slice(32).to_a
+              bitshift -= BITS_PER_LEVEL
+            end
+            result.concat(remainder)
+          end
+        elsif (child = node[from_slot])
+          result = node.take(from_slot)
+          result.push(replace_node_suffix(child, bitshift - BITS_PER_LEVEL, from, suffix))
+          remainder = suffix.shift((31 - from_slot) * (1 << bitshift))
+          while bitshift != 0
+            remainder = remainder.each_slice(32).to_a
+            bitshift -= BITS_PER_LEVEL
+          end
+          result.concat(remainder)
+        else
+          raise "Shouldn't happen"
+        end
+      end
+    end
+  end
+
+  # The canonical empty `Vector`. Returned by `Vector[]` when
+  # invoked with no arguments; also returned by `Vector.empty`. Prefer using this
+  # one rather than creating many empty vectors using `Vector.new`.
+  #
+  # @private
+  EmptyVector = Immutable::Vector.empty
+
+
+  # `Immutable::Set` is a collection of unordered values with no duplicates. Testing whether
+  # an object is present in the `Set` can be done in constant time. `Set` is also `Enumerable`, so you can
+  # iterate over the members of the set with {#each}, transform them with {#map}, filter
+  # them with {#select}, and so on. Some of the `Enumerable` methods are overridden to
+  # return `immutable-ruby` collections.
+  #
+  # Like the `Set` class in Ruby's standard library, which we will call RubySet,
+  # `Immutable::Set` defines equivalency of objects using `#hash` and `#eql?`. No two
+  # objects with the same `#hash` code, and which are also `#eql?`, can coexist in the
+  # same `Set`. If one is already in the `Set`, attempts to add another one will have
+  # no effect.
+  #
+  # `Set`s have no natural ordering and cannot be compared using `#<=>`. However, they
+  # define {#<}, {#>}, {#<=}, and {#>=} as shorthand for {#proper_subset?},
+  # {#proper_superset?}, {#subset?}, and {#superset?} respectively.
+  #
+  # The basic set-theoretic operations {#union}, {#intersection}, {#difference}, and
+  # {#exclusion} work with any `Enumerable` object.
+  #
+  # A `Set` can be created in either of the following ways:
+  #
+  #     Immutable::Set.new([1, 2, 3]) # any Enumerable can be used to initialize
+  #     Immutable::Set['A', 'B', 'C', 'D']
+  #
+  # The latter 2 forms of initialization can be used with your own, custom subclasses
+  # of `Immutable::Set`.
+  #
+  # Unlike RubySet, all methods which you might expect to "modify" an `Immutable::Set`
+  # actually return a new set and leave the existing one unchanged.
+  #
+  # @example
+  #   set1 = Immutable::Set[1, 2] # => Immutable::Set[1, 2]
+  #   set2 = Immutable::Set[1, 2] # => Immutable::Set[1, 2]
+  #   set1 == set2              # => true
+  #   set3 = set1.add("foo")    # => Immutable::Set[1, 2, "foo"]
+  #   set3 - set2               # => Immutable::Set["foo"]
+  #   set3.subset?(set1)        # => false
+  #   set1.subset?(set3)        # => true
+  #
+  class Set
+    include Immutable::Enumerable
+
+    class << self
+      # Create a new `Set` populated with the given items.
+      # @return [Set]
+      def [](*items)
+        items.empty? ? empty : new(items)
+      end
+
+      # Return an empty `Set`. If used on a subclass, returns an empty instance
+      # of that class.
+      #
+      # @return [Set]
+      def empty
+        @empty ||= new
+      end
+
+      # "Raw" allocation of a new `Set`. Used internally to create a new
+      # instance quickly after obtaining a modified {Trie}.
+      #
+      # @return [Set]
+      # @private
+      def alloc(trie = EmptyTrie)
+        allocate.tap { |s| s.instance_variable_set(:@trie, trie) }.freeze
+      end
+    end
+
+    def initialize(items=[])
+      @trie = Trie.new(0)
+      items.each { |item| @trie.put!(item, nil) }
+      freeze
+    end
+
+    # Return `true` if this `Set` contains no items.
+    # @return [Boolean]
+    def empty?
+      @trie.empty?
+    end
+
+    # Return the number of items in this `Set`.
+    # @return [Integer]
+    def size
+      @trie.size
+    end
+    alias length size
+
+    # Return a new `Set` with `item` added. If `item` is already in the set,
+    # return `self`.
+    #
+    # @example
+    #   Immutable::Set[1, 2, 3].add(4) # => Immutable::Set[1, 2, 4, 3]
+    #   Immutable::Set[1, 2, 3].add(2) # => Immutable::Set[1, 2, 3]
+    #
+    # @param item [Object] The object to add
+    # @return [Set]
+    def add(item)
+      include?(item) ? self : self.class.alloc(@trie.put(item, nil))
+    end
+    alias << add
+
+    # If `item` is not a member of this `Set`, return a new `Set` with `item` added.
+    # Otherwise, return `false`.
+    #
+    # @example
+    #   Immutable::Set[1, 2, 3].add?(4) # => Immutable::Set[1, 2, 4, 3]
+    #   Immutable::Set[1, 2, 3].add?(2) # => false
+    #
+    # @param item [Object] The object to add
+    # @return [Set, false]
+    def add?(item)
+      !include?(item) && add(item)
+    end
+
+    # Return a new `Set` with `item` removed. If `item` is not a member of the set,
+    # return `self`.
+    #
+    # @example
+    #   Immutable::Set[1, 2, 3].delete(1)  # => Immutable::Set[2, 3]
+    #   Immutable::Set[1, 2, 3].delete(99) # => Immutable::Set[1, 2, 3]
+    #
+    # @param item [Object] The object to remove
+    # @return [Set]
+    def delete(item)
+      trie = @trie.delete(item)
+      new_trie(trie)
+    end
+
+    # If `item` is a member of this `Set`, return a new `Set` with `item` removed.
+    # Otherwise, return `false`.
+    #
+    # @example
+    #   Immutable::Set[1, 2, 3].delete?(1)  # => Immutable::Set[2, 3]
+    #   Immutable::Set[1, 2, 3].delete?(99) # => false
+    #
+    # @param item [Object] The object to remove
+    # @return [Set, false]
+    def delete?(item)
+      include?(item) && delete(item)
+    end
+
+    # Call the block once for each item in this `Set`. No specific iteration order
+    # is guaranteed, but the order will be stable for any particular `Set`. If
+    # no block is given, an `Enumerator` is returned instead.
+    #
+    # @example
+    #   Immutable::Set["Dog", "Elephant", "Lion"].each { |e| puts e }
+    #   Elephant
+    #   Dog
+    #   Lion
+    #   # => Immutable::Set["Dog", "Elephant", "Lion"]
+    #
+    # @yield [item] Once for each item.
+    # @return [self, Enumerator]
+    def each
+      return to_enum if not block_given?
+      @trie.each { |key, _| yield(key) }
+      self
+    end
+
+    # Call the block once for each item in this `Set`. Iteration order will be
+    # the opposite of {#each}. If no block is given, an `Enumerator` is
+    # returned instead.
+    #
+    # @example
+    #   Immutable::Set["Dog", "Elephant", "Lion"].reverse_each { |e| puts e }
+    #   Lion
+    #   Dog
+    #   Elephant
+    #   # => Immutable::Set["Dog", "Elephant", "Lion"]
+    #
+    # @yield [item] Once for each item.
+    # @return [self]
+    def reverse_each
+      return enum_for(:reverse_each) if not block_given?
+      @trie.reverse_each { |key, _| yield(key) }
+      self
+    end
+
+    # Return a new `Set` with all the items for which the block returns true.
+    #
+    # @example
+    #   Immutable::Set["Elephant", "Dog", "Lion"].select { |e| e.size >= 4 }
+    #   # => Immutable::Set["Elephant", "Lion"]
+    # @yield [item] Once for each item.
+    # @return [Set]
+    def select
+      return enum_for(:select) unless block_given?
+      trie = @trie.select { |key, _| yield(key) }
+      new_trie(trie)
+    end
+    alias find_all select
+    alias keep_if  select
+
+    # Call the block once for each item in this `Set`. All the values returned
+    # from the block will be gathered into a new `Set`. If no block is given,
+    # an `Enumerator` is returned instead.
+    #
+    # @example
+    #   Immutable::Set["Cat", "Elephant", "Dog", "Lion"].map { |e| e.size }
+    #   # => Immutable::Set[8, 4, 3]
+    #
+    # @yield [item] Once for each item.
+    # @return [Set]
+    def map
+      return enum_for(:map) if not block_given?
+      return self if empty?
+      self.class.new(super)
+    end
+    alias collect map
+
+    # Return `true` if the given item is present in this `Set`. More precisely,
+    # return `true` if an object with the same `#hash` code, and which is also `#eql?`
+    # to the given object is present.
+    #
+    # @example
+    #   Immutable::Set["A", "B", "C"].include?("B") # => true
+    #   Immutable::Set["A", "B", "C"].include?("Z") # => false
+    #
+    # @param object [Object] The object to check for
+    # @return [Boolean]
+    def include?(object)
+      @trie.key?(object)
+    end
+    alias member? include?
+
+    # Return a member of this `Set`. The member chosen will be the first one which
+    # would be yielded by {#each}. If the set is empty, return `nil`.
+    #
+    # @example
+    #   Immutable::Set["A", "B", "C"].first # => "C"
+    #
+    # @return [Object]
+    def first
+      (entry = @trie.at(0)) && entry[0]
+    end
+
+    # Return a {SortedSet} which contains the same items as this `Set`, ordered by
+    # the given comparator block.
+    #
+    # @example
+    #   Immutable::Set["Elephant", "Dog", "Lion"].sort
+    #   # => Immutable::SortedSet["Dog", "Elephant", "Lion"]
+    #   Immutable::Set["Elephant", "Dog", "Lion"].sort { |a,b| a.size <=> b.size }
+    #   # => Immutable::SortedSet["Dog", "Lion", "Elephant"]
+    #
+    # @yield [a, b] Any number of times with different pairs of elements.
+    # @yieldreturn [Integer] Negative if the first element should be sorted
+    #                        lower, positive if the latter element, or 0 if
+    #                        equal.
+    # @return [SortedSet]
+    def sort(&comparator)
+      SortedSet.new(to_a, &comparator)
+    end
+
+    # Return a {SortedSet} which contains the same items as this `Set`, ordered
+    # by mapping each item through the provided block to obtain sort keys, and
+    # then sorting the keys.
+    #
+    # @example
+    #   Immutable::Set["Elephant", "Dog", "Lion"].sort_by { |e| e.size }
+    #   # => Immutable::SortedSet["Dog", "Lion", "Elephant"]
+    #
+    # @yield [item] Once for each item to create the set, and then potentially
+    #               again depending on what operations are performed on the
+    #               returned {SortedSet}. As such, it is recommended that the
+    #               block be a pure function.
+    # @yieldreturn [Object] sort key for the item
+    # @return [SortedSet]
+    def sort_by(&mapper)
+      SortedSet.new(to_a, &mapper)
+    end
+
+    # Return a new `Set` which contains all the members of both this `Set` and `other`.
+    # `other` can be any `Enumerable` object.
+    #
+    # @example
+    #   Immutable::Set[1, 2] | Immutable::Set[2, 3] # => Immutable::Set[1, 2, 3]
+    #
+    # @param other [Enumerable] The collection to merge with
+    # @return [Set]
+    def union(other)
+      if other.is_a?(Immutable::Set)
+        if other.size > size
+          small_set_pairs = @trie
+          large_set_trie = other.instance_variable_get(:@trie)
+        else
+          small_set_pairs = other.instance_variable_get(:@trie)
+          large_set_trie = @trie
+        end
+      else
+        if other.respond_to?(:lazy)
+          small_set_pairs = other.lazy.map { |e| [e, nil] }
+        else
+          small_set_pairs = other.map { |e| [e, nil] }
+        end
+        large_set_trie = @trie
+      end
+
+      trie = large_set_trie.bulk_put(small_set_pairs)
+      new_trie(trie)
+    end
+    alias | union
+    alias + union
+    alias merge union
+
+    # Return a new `Set` which contains all the items which are members of both
+    # this `Set` and `other`. `other` can be any `Enumerable` object.
+    #
+    # @example
+    #   Immutable::Set[1, 2] & Immutable::Set[2, 3] # => Immutable::Set[2]
+    #
+    # @param other [Enumerable] The collection to intersect with
+    # @return [Set]
+    def intersection(other)
+      if other.size < @trie.size
+        if other.is_a?(Immutable::Set)
+          trie = other.instance_variable_get(:@trie).select { |key, _| include?(key) }
+        else
+          trie = Trie.new(0)
+          other.each { |obj| trie.put!(obj, nil) if include?(obj) }
+        end
+      else
+        trie = @trie.select { |key, _| other.include?(key) }
+      end
+      new_trie(trie)
+    end
+    alias & intersection
+
+    # Return a new `Set` with all the items in `other` removed. `other` can be
+    # any `Enumerable` object.
+    #
+    # @example
+    #   Immutable::Set[1, 2] - Immutable::Set[2, 3] # => Immutable::Set[1]
+    #
+    # @param other [Enumerable] The collection to subtract from this set
+    # @return [Set]
+    def difference(other)
+      trie = if (@trie.size <= other.size) && (other.is_a?(Immutable::Set) || (defined?(::Set) && other.is_a?(::Set)))
+        @trie.select { |key, _| !other.include?(key) }
+      else
+        @trie.bulk_delete(other)
+      end
+      new_trie(trie)
+    end
+    alias subtract difference
+    alias - difference
+
+    # Return a new `Set` which contains all the items which are members of this
+    # `Set` or of `other`, but not both. `other` can be any `Enumerable` object.
+    #
+    # @example
+    #   Immutable::Set[1, 2] ^ Immutable::Set[2, 3] # => Immutable::Set[1, 3]
+    #
+    # @param other [Enumerable] The collection to take the exclusive disjunction of
+    # @return [Set]
+    def exclusion(other)
+      ((self | other) - (self & other))
+    end
+    alias ^ exclusion
+
+    # Return `true` if all items in this `Set` are also in `other`.
+    #
+    # @example
+    #   Immutable::Set[2, 3].subset?(Immutable::Set[1, 2, 3]) # => true
+    #
+    # @param other [Set]
+    # @return [Boolean]
+    def subset?(other)
+      return false if other.size < size
+
+      # This method has the potential to be very slow if 'other' is a large Array, so to avoid that,
+      #   we convert those Arrays to Sets before checking presence of items
+      # Time to convert Array -> Set is linear in array.size
+      # Time to check for presence of all items in an Array is proportional to set.size * array.size
+      # Note that both sides of that equation have array.size -- hence those terms cancel out,
+      #   and the break-even point is solely dependent on the size of this collection
+      # After doing some benchmarking to estimate the constants, it appears break-even is at ~190 items
+      # We also check other.size, to avoid the more expensive #is_a? checks in cases where it doesn't matter
+      #
+      if other.size >= 150 && @trie.size >= 190 && !(other.is_a?(Immutable::Set) || other.is_a?(::Set))
+        other = ::Set.new(other)
+      end
+      all? { |item| other.include?(item) }
+    end
+    alias <= subset?
+
+    # Return `true` if all items in `other` are also in this `Set`.
+    #
+    # @example
+    #   Immutable::Set[1, 2, 3].superset?(Immutable::Set[2, 3]) # => true
+    #
+    # @param other [Set]
+    # @return [Boolean]
+    def superset?(other)
+      other.subset?(self)
+    end
+    alias >= superset?
+
+    # Returns `true` if `other` contains all the items in this `Set`, plus at least
+    # one item which is not in this set.
+    #
+    # @example
+    #   Immutable::Set[2, 3].proper_subset?(Immutable::Set[1, 2, 3])    # => true
+    #   Immutable::Set[1, 2, 3].proper_subset?(Immutable::Set[1, 2, 3]) # => false
+    #
+    # @param other [Set]
+    # @return [Boolean]
+    def proper_subset?(other)
+      return false if other.size <= size
+      # See comments above
+      if other.size >= 150 && @trie.size >= 190 && !(other.is_a?(Immutable::Set) || other.is_a?(::Set))
+        other = ::Set.new(other)
+      end
+      all? { |item| other.include?(item) }
+    end
+    alias < proper_subset?
+
+    # Returns `true` if this `Set` contains all the items in `other`, plus at least
+    # one item which is not in `other`.
+    #
+    # @example
+    #   Immutable::Set[1, 2, 3].proper_superset?(Immutable::Set[2, 3])    # => true
+    #   Immutable::Set[1, 2, 3].proper_superset?(Immutable::Set[1, 2, 3]) # => false
+    #
+    # @param other [Set]
+    # @return [Boolean]
+    def proper_superset?(other)
+      other.proper_subset?(self)
+    end
+    alias > proper_superset?
+
+    # Return `true` if this `Set` and `other` do not share any items.
+    #
+    # @example
+    #   Immutable::Set[1, 2].disjoint?(Immutable::Set[8, 9]) # => true
+    #
+    # @param other [Set]
+    # @return [Boolean]
+    def disjoint?(other)
+      if other.size <= size
+        other.each { |item| return false if include?(item) }
+      else
+        # See comment on #subset?
+        if other.size >= 150 && @trie.size >= 190 && !(other.is_a?(Immutable::Set) || other.is_a?(::Set))
+          other = ::Set.new(other)
+        end
+        each { |item| return false if other.include?(item) }
+      end
+      true
+    end
+
+    # Return `true` if this `Set` and `other` have at least one item in common.
+    #
+    # @example
+    #   Immutable::Set[1, 2].intersect?(Immutable::Set[2, 3]) # => true
+    #
+    # @param other [Set]
+    # @return [Boolean]
+    def intersect?(other)
+      !disjoint?(other)
+    end
+
+    # Recursively insert the contents of any nested `Set`s into this `Set`, and
+    # remove them.
+    #
+    # @example
+    #   Immutable::Set[Immutable::Set[1, 2], Immutable::Set[3, 4]].flatten
+    #   # => Immutable::Set[1, 2, 3, 4]
+    #
+    # @return [Set]
+    def flatten
+      reduce(self.class.empty) do |set, item|
+        next set.union(item.flatten) if item.is_a?(Set)
+        set.add(item)
+      end
+    end
+
+    alias group group_by
+    alias classify group_by
+
+    # Return a randomly chosen item from this `Set`. If the set is empty, return `nil`.
+    #
+    # @example
+    #   Immutable::Set[1, 2, 3, 4, 5].sample # => 3
+    #
+    # @return [Object]
+    def sample
+      empty? ? nil : @trie.at(rand(size))[0]
+    end
+
+    # Return an empty `Set` instance, of the same class as this one. Useful if you
+    # have multiple subclasses of `Set` and want to treat them polymorphically.
+    #
+    # @return [Set]
+    def clear
+      self.class.empty
+    end
+
+    # Return true if `other` has the same type and contents as this `Set`.
+    #
+    # @param other [Object] The object to compare with
+    # @return [Boolean]
+    def eql?(other)
+      return true if other.equal?(self)
+      return false if not instance_of?(other.class)
+      other_trie = other.instance_variable_get(:@trie)
+      return false if @trie.size != other_trie.size
+      @trie.each do |key, _|
+        return false if !other_trie.key?(key)
+      end
+      true
+    end
+    alias == eql?
+
+    # See `Object#hash`.
+    # @return [Integer]
+    def hash
+      reduce(0) { |hash, item| (hash << 5) - hash + item.hash }
+    end
+
+    # Return `self`. Since this is an immutable object duplicates are
+    # equivalent.
+    # @return [Set]
+    def dup
+      self
+    end
+    alias clone dup
+
+    undef :"<=>" # Sets are not ordered, so Enumerable#<=> will give a meaningless result
+    undef :each_index # Set members cannot be accessed by 'index', so #each_index is not meaningful
+
+    # Return `self`.
+    #
+    # @return [self]
+    def to_set
+      self
+    end
+
+    # @private
+    def marshal_dump
+      output = {}
+      each do |key|
+        output[key] = nil
+      end
+      output
+    end
+
+    # @private
+    def marshal_load(dictionary)
+      @trie = dictionary.reduce(EmptyTrie) do |trie, key_value|
+        trie.put(key_value.first, nil)
+      end
+    end
+
+    private
+
+    def new_trie(trie)
+      if trie.empty?
+        self.class.empty
+      elsif trie.equal?(@trie)
+        self
+      else
+        self.class.alloc(trie)
+      end
+    end
+  end
+
+  # The canonical empty `Set`. Returned by `Set[]` when
+  # invoked with no arguments; also returned by `Set.empty`. Prefer using this
+  # one rather than creating many empty sets using `Set.new`.
+  #
+  # @private
+  EmptySet = Immutable::Set.empty
+end