require 'yaml' YAML.add_builtin_type("omap") do |type, val| ActiveSupport::OrderedHash[val.map(&:to_a).map(&:first)] end module ActiveSupport # The order of iteration over hashes in Ruby 1.8 is undefined. For example, you do not know the # order in which +keys+ will return keys, or +each+ yield pairs. ActiveSupport::OrderedHash # implements a hash that preserves insertion order, as in Ruby 1.9: # # oh = ActiveSupport::OrderedHash.new # oh[:a] = 1 # oh[:b] = 2 # oh.keys # => [:a, :b], this order is guaranteed # # ActiveSupport::OrderedHash is namespaced to prevent conflicts with other implementations. class OrderedHash < ::Hash #:nodoc: def to_yaml_type "!tag:yaml.org,2002:omap" end def to_yaml(opts = {}) YAML.quick_emit(self, opts) do |out| out.seq(taguri, to_yaml_style) do |seq| each do |k, v| seq.add(k => v) end end end end # Hash is ordered in Ruby 1.9! if RUBY_VERSION < '1.9' # In MRI the Hash class is core and written in C. In particular, methods are # programmed with explicit C function calls and polymorphism is not honored. # # For example, []= is crucial in this implementation to maintain the @keys # array but hash.c invokes rb_hash_aset() originally. This prevents method # reuse through inheritance and forces us to reimplement stuff. # # For instance, we cannot use the inherited #merge! because albeit the algorithm # itself would work, our []= is not being called at all by the C code. def initialize(*args, &block) super @keys = [] end def self.[](*args) ordered_hash = new if (args.length == 1 && args.first.is_a?(Array)) args.first.each do |key_value_pair| next unless (key_value_pair.is_a?(Array)) ordered_hash[key_value_pair[0]] = key_value_pair[1] end return ordered_hash end unless (args.size % 2 == 0) raise ArgumentError.new("odd number of arguments for Hash") end args.each_with_index do |val, ind| next if (ind % 2 != 0) ordered_hash[val] = args[ind + 1] end ordered_hash end def initialize_copy(other) super # make a deep copy of keys @keys = other.keys end def []=(key, value) @keys << key unless has_key?(key) super end def delete(key) if has_key? key index = @keys.index(key) @keys.delete_at index end super end def delete_if super sync_keys! self end def reject! super sync_keys! self end def reject(&block) dup.reject!(&block) end def keys @keys.dup end def values @keys.collect { |key| self[key] } end def to_hash self end def to_a @keys.map { |key| [ key, self[key] ] } end def each_key @keys.each { |key| yield key } end def each_value @keys.each { |key| yield self[key]} end def each @keys.each {|key| yield [key, self[key]]} end alias_method :each_pair, :each def clear super @keys.clear self end def shift k = @keys.first v = delete(k) [k, v] end def merge!(other_hash) if block_given? other_hash.each { |k, v| self[k] = key?(k) ? yield(k, self[k], v) : v } else other_hash.each { |k, v| self[k] = v } end self end alias_method :update, :merge! def merge(other_hash, &block) dup.merge!(other_hash, &block) end # When replacing with another hash, the initial order of our keys must come from the other hash -ordered or not. def replace(other) super @keys = other.keys self end def invert OrderedHash[self.to_a.map!{|key_value_pair| key_value_pair.reverse}] end def inspect "#" end private def sync_keys! @keys.delete_if {|k| !has_key?(k)} end end end end