puppet 3.5.0.rc1-x86-mingw32

data/lib/puppet/pops/parser/slurp_support.rb

@@ -0,0 +1,95 @@
+# This module is an integral part of the Lexer.
+# It defines the string slurping behavior - finding the string and non string parts in interpolated
+# strings, translating escape sequences in strings to their single character equivalence.
+#
+# PERFORMANCE NOTE: The various kinds of slurping could be made even more generic, but requires
+# additional parameter passing and evaluation of conditional logic.
+# TODO: More detailed performance analysis of excessive character escaping and interpolation.
+#
+module Puppet::Pops::Parser::SlurpSupport
+
+  SLURP_SQ_PATTERN  = /(?:[^\\]|^|[^\\])(?:[\\]{2})*[']/
+  SLURP_DQ_PATTERN  = /(?:[^\\]|^|[^\\])(?:[\\]{2})*(["]|[$]\{?)/
+  SLURP_UQ_PATTERN  = /(?:[^\\]|^|[^\\])(?:[\\]{2})*([$]\{?|\z)/
+  SLURP_ALL_PATTERN = /.*(\z)/m
+  SQ_ESCAPES = %w{ \\ ' }
+  DQ_ESCAPES = %w{ \\  $ ' " r n t s u}+["\r\n", "\n"]
+  UQ_ESCAPES = %w{ \\  $ r n t s u}+["\r\n", "\n"]
+
+  def slurp_sqstring
+    # skip the leading '
+    @scanner.pos += 1
+    str = slurp(@scanner, SLURP_SQ_PATTERN, SQ_ESCAPES, :ignore_invalid_escapes) || lex_error("Unclosed quote after \"'\" followed by '#{followed_by}'")
+    str[0..-2] # strip closing "'" from result
+  end
+
+  def slurp_dqstring
+    scn = @scanner
+    last = scn.matched
+    str = slurp(scn, SLURP_DQ_PATTERN, DQ_ESCAPES, false)
+    unless str
+      lex_error("Unclosed quote after #{format_quote(last)} followed by '#{followed_by}'")
+    end
+
+    # Terminator may be a single char '"', '$', or two characters '${' group match 1 (scn[1]) from the last slurp holds this
+    terminator = scn[1]
+    [str[0..(-1 - terminator.length)], terminator]
+  end
+
+   # Copy from old lexer - can do much better
+   def slurp_uqstring
+     scn = @scanner
+     last = scn.matched
+     ignore = true
+     str = slurp(scn, @lexing_context[:uq_slurp_pattern], @lexing_context[:escapes], :ignore_invalid_escapes)
+
+     # Terminator may be a single char '$', two characters '${', or empty string '' at the end of intput.
+     # Group match 1 holds this.
+     # The exceptional case is found by looking at the subgroup 1 of the most recent match made by the scanner (i.e. @scanner[1]).
+     # This is the last match made by the slurp method (having called scan_until on the scanner).
+     # If there is a terminating character is must be stripped and returned separately.
+     #
+     terminator = scn[1]
+     [str[0..(-1 - terminator.length)], terminator]
+   end
+
+  # Slurps a string from the given scanner until the given pattern and then replaces any escaped
+  # characters given by escapes into their control-character equivalent or in case of line breaks, replaces the
+  # pattern \r?\n with an empty string.
+  # The returned string contains the terminating character. Returns nil if the scanner can not scan until the given
+  # pattern.
+  #
+  def slurp(scanner, pattern, escapes, ignore_invalid_escapes)
+    str = scanner.scan_until(pattern) || return
+
+    # Process unicode escapes first as they require getting 4 hex digits
+    # If later a \u is found it is warned not to be a unicode escape
+    if escapes.include?('u')
+      str.gsub!(/\\u([\da-fA-F]{4})/m) {
+        [$1.hex].pack("U")
+      }
+    end
+
+    str.gsub!(/\\([^\r\n]|(?:\r?\n))/m) {
+      ch = $1
+      if escapes.include? ch
+        case ch
+        when 'r'   ; "\r"
+        when 'n'   ; "\n"
+        when 't'   ; "\t"
+        when 's'   ; " "
+        when 'u'
+          Puppet.warning(positioned_message("Unicode escape '\\u' was not followed by 4 hex digits"))
+          "\\u"
+        when "\n"  ; ''
+        when "\r\n"; ''
+        else      ch
+        end
+      else
+        Puppet.warning(positioned_message("Unrecognized escape sequence '\\#{ch}'")) unless ignore_invalid_escapes
+        "\\#{ch}"
+      end
+    }
+    str
+  end
+end

data/lib/puppet/pops/patterns.rb

@@ -0,0 +1,44 @@
+# The Patterns module contains common regular expression patters for the Puppet DSL language
+module Puppet::Pops::Patterns
+
+  # NUMERIC matches hex, octal, decimal, and floating point and captures three parts
+  # 0 = entire matched number, leading and trailing whitespace included
+  # 1 = hexadecimal number
+  # 2 = non hex integer portion, possibly with leading 0 (octal)
+  # 3 = floating point part, starts with ".", decimals and optional exponent
+  #
+  # Thus, a hex number has group 1 value, an octal value has group 2 (if it starts with 0), and no group 3
+  # and a floating point value has group 2 and group 3.
+  #
+  NUMERIC = %r{^\s*(?:(0[xX][0-9A-Fa-f]+)|(0?\d+)((?:\.\d+)?(?:[eE]-?\d+)?))\s*$}
+
+  # ILLEGAL_P3_1_HOSTNAME matches if a hostname contains illegal characters.
+  # This check does not prevent pathological names like 'a....b', '.....', "---". etc.
+  ILLEGAL_HOSTNAME_CHARS = %r{[^-\w.]}
+
+  # NAME matches a name the same way as the lexer.
+  NAME = %r{\A((::)?[a-z]\w*)(::[a-z]\w*)*\z}
+
+  # CLASSREF_EXT matches a class reference the same way as the lexer - i.e. the external source form
+  # where each part must start with a capital letter A-Z.
+  # This name includes hyphen, which may be illegal in some cases.
+  #
+  CLASSREF_EXT = %r{\A((::){0,1}[A-Z][\w]*)+\z}
+
+  # CLASSREF matches a class reference the way it is represented internally in the
+  # model (i.e. in lower case).
+  # This name includes hyphen, which may be illegal in some cases.
+  #
+  CLASSREF = %r{\A((::){0,1}[a-z][\w]*)+\z}
+
+  # DOLLAR_VAR matches a variable name including the initial $ character
+  DOLLAR_VAR     = %r{\$(::)?(\w+::)*\w+}
+
+  # VAR_NAME matches the name part of a variable (The $ character is not included)
+  # Note, that only the final segment may start with an underscore.
+  VAR_NAME = %r{\A(:?(::)?[a-z]\w*)*(:?(::)?[a-z_]\w*)\z}
+
+  # A Numeric var name must be the decimal number 0, or a decimal number not starting with 0
+  NUMERIC_VAR_NAME = %r{\A(?:0|(?:[1-9][0-9]*))\z}
+
+end

data/lib/puppet/pops/types/class_loader.rb

@@ -0,0 +1,118 @@
+require 'rgen/metamodel_builder'
+
+# The ClassLoader provides a Class instance given a class name or a meta-type.
+# If the class is not already loaded, it is loaded using the Puppet Autoloader.
+# This means it can load a class from a gem, or from puppet modules.
+#
+class Puppet::Pops::Types::ClassLoader
+  @autoloader = Puppet::Util::Autoload.new("ClassLoader", "", :wrap => false)
+
+  # Returns a Class given a fully qualified class name.
+  # Lookup of class is never relative to the calling namespace.
+  # @param name [String, Array<String>, Array<Symbol>, Puppet::Pops::Types::PObjectType] A fully qualified
+  #   class name String (e.g. '::Foo::Bar', 'Foo::Bar'), a PObjectType, or a fully qualified name in Array form where each part
+  #   is either a String or a Symbol, e.g. `%w{Puppetx Puppetlabs SomeExtension}`.
+  # @return [Class, nil] the looked up class or nil if no such class is loaded
+  # @raise ArgumentError If the given argument has the wrong type
+  # @api public
+  #
+  def self.provide(name)
+    case name
+    when String
+      provide_from_string(name)
+
+    when Array
+      provide_from_name_path(name.join('::'), name)
+
+    when Puppet::Pops::Types::PObjectType, Puppet::Pops::Types::PType
+      provide_from_type(name)
+
+    else
+      raise ArgumentError, "Cannot provide a class from a '#{name.class.name}'"
+    end
+  end
+
+  private
+
+  def self.provide_from_type(type)
+    case type
+    when Puppet::Pops::Types::PRubyType
+      provide_from_string(type.ruby_class)
+
+    when Puppet::Pops::Types::PBooleanType
+      # There is no other thing to load except this Enum meta type
+      RGen::MetamodelBuilder::MMBase::Boolean
+
+    when Puppet::Pops::Types::PType
+      # TODO: PType should have a type argument (a PObjectType)
+      Class
+
+    # Although not expected to be the first choice for getting a concrete class for these
+    # types, these are of value if the calling logic just has a reference to type.
+    #
+    when Puppet::Pops::Types::PArrayType   ; Array
+    when Puppet::Pops::Types::PHashType    ; Hash
+    when Puppet::Pops::Types::PRegexpType ; Regexp
+    when Puppet::Pops::Types::PIntegerType ; Integer
+    when Puppet::Pops::Types::PStringType  ; String
+    when Puppet::Pops::Types::PFloatType   ; Float
+    when Puppet::Pops::Types::PNilType     ; NilClass
+    else
+      nil
+    end
+  end
+
+  def self.provide_from_string(name)
+    name_path = name.split('::')
+    # always from the root, so remove an empty first segment
+    if name_path[0].empty?
+      name_path = name_path[1..-1]
+    end
+    provide_from_name_path(name, name_path)
+  end
+
+  def self.provide_from_name_path(name, name_path)
+    # If class is already loaded, try this first
+    result = find_class(name_path)
+
+    unless result.is_a?(Class)
+      # Attempt to load it using the auto loader
+      loaded_path = nil
+      if paths_for_name(name).find {|path| loaded_path = path; @autoloader.load(path) }
+        result = find_class(name_path)
+        unless result.is_a?(Class)
+          raise RuntimeError, "Loading of #{name} using relative path: '#{loaded_path}' did not create expected class"
+        end
+      end
+    end
+    return nil unless result.is_a?(Class)
+    result
+  end
+
+  def self.find_class(name_path)
+    name_path.reduce(Object) do |ns, name|
+      begin
+        ns.const_get(name)
+      rescue NameError
+        return nil
+      end
+    end
+  end
+
+  def self.paths_for_name(fq_name)
+    [de_camel(fq_name), downcased_path(fq_name)]
+  end
+
+  def self.downcased_path(fq_name)
+    fq_name.to_s.gsub(/::/, '/').downcase
+  end
+
+  def self.de_camel(fq_name)
+    fq_name.to_s.gsub(/::/, '/').
+    gsub(/([A-Z]+)([A-Z][a-z])/,'\1_\2').
+    gsub(/([a-z\d])([A-Z])/,'\1_\2').
+    tr("-", "_").
+    downcase
+  end
+
+end

data/lib/puppet/pops/types/enumeration.rb

@@ -0,0 +1,34 @@
+# The Enumeration class provides default Enumerable::Enumerator creation for Puppet Programming Language
+# runtime objects that supports the concept of enumeration.
+#
+class Puppet::Pops::Types::Enumeration
+  # Produces an Enumerable::Enumerator for Array, Hash, Integer, Integer Range, and String.
+  #
+  def self.enumerator(o)
+    @@singleton ||= new
+    @@singleton.enumerator(o)
+  end
+
+  # Produces an Enumerator for Array, Hash, Integer, Integer Range, and String.
+  #
+  def enumerator(o)
+    case o
+    when String
+      x = o.chars
+      # Ruby 1.8.7 returns Enumerable::Enumerator, Ruby 1.8.9 Enumerator, and 2.0.0 an Array
+      x.is_a?(Array) ? x.each : x
+    when Integer
+      o.times
+    when Array
+      o.each
+    when Hash
+      o.each
+    when Puppet::Pops::Types::PIntegerType
+      # Not enumerable if representing an infinite range
+      return nil if o.to.nil? || o.from.nil?
+      o.each
+    else
+      nil
+    end
+  end
+end

data/lib/puppet/pops/types/type_calculator.rb

@@ -0,0 +1,1482 @@
+# The TypeCalculator can answer questions about puppet types.
+#
+# The Puppet type system is primarily based on sub-classing. When asking the type calculator to infer types from Ruby in general, it
+# may not provide the wanted answer; it does not for instance take module inclusions and extensions into account. In general the type
+# system should be unsurprising for anyone being exposed to the notion of type. The type `Data` may require a bit more explanation; this
+# is an abstract type that includes all scalar types, as well as Array with an element type compatible with Data, and Hash with key
+# compatible with scalar and elements compatible with Data. Expressed differently; Data is what you typically express using JSON (with
+# the exception that the Puppet type system also includes Pattern (regular expression) as a scalar.
+#
+# Inference
+# ---------
+# The `infer(o)` method infers a Puppet type for scalar Ruby objects, and for Arrays and Hashes.
+# The inference result is instance specific for single typed collections
+# and allows answering questions about its embedded type. It does not however preserve multiple types in
+# a collection, and can thus not answer questions like `[1,a].infer() =~ Array[Integer, String]` since the inference
+# computes the common type Scalar when combining Integer and String.
+#
+# The `infer_generic(o)` method infers a generic Puppet type for scalar Ruby object, Arrays and Hashes.
+# This inference result does not contain instance specific information; e.g. Array[Integer] where the integer
+# range is the generic default. Just `infer` it also combines types into a common type.
+#
+# The `infer_set(o)` method works like `infer` but preserves all type information. It does not do any
+# reduction into common types or ranges. This method of inference is best suited for answering questions
+# about an object being an instance of a type. It correctly answers: `[1,a].infer_set() =~ Array[Integer, String]`
+#
+# The `generalize!(t)` method modifies an instance specific inference result to a generic. The method mutates
+# the given argument. Basically, this removes string instances from String, and range from Integer and Float.
+#
+# Assignability
+# -------------
+# The `assignable?(t1, t2)` method answers if t2 conforms to t1. The type t2 may be an instance, in which case
+# its type is inferred, or a type.
+#
+# Instance?
+# ---------
+# The `instance?(t, o)` method answers if the given object (instance) is an instance that is assignable to the given type.
+#
+# String
+# ------
+# Creates a string representation of a type.
+#
+# Creation of Type instances
+# --------------------------
+# Instance of the classes in the {Puppet::Pops::Types type model} are used to denote a specific type. It is most convenient
+# to use the {Puppet::Pops::Types::TypeFactory TypeFactory} when creating instances.
+#
+# @note
+#   In general, new instances of the wanted type should be created as they are assigned to models using containment, and a
+#   contained object can only be in one container at a time. Also, the type system may include more details in each type
+#   instance, such as if it may be nil, be empty, contain a certain count etc. Or put differently, the puppet types are not
+#   singletons.
+#
+# All types support `copy` which should be used when assigning a type where it is unknown if it is bound or not
+# to a parent type. A check can be made with `t.eContainer().nil?`
+#
+# Equality and Hash
+# -----------------
+# Type instances are equal in terms of Ruby eql? and `==` if they describe the same type, but they are not `equal?` if they are not
+# the same type instance. Two types that describe the same type have identical hash - this makes them usable as hash keys.
+#
+# Types and Subclasses
+# --------------------
+# In general, the type calculator should be used to answer questions if a type is a subtype of another (using {#assignable?}, or
+# {#instance?} if the question is if a given object is an instance of a given type (or is a subtype thereof).
+# Many of the types also have a Ruby subtype relationship; e.g. PHashType and PArrayType are both subtypes of PCollectionType, and
+# PIntegerType, PFloatType, PStringType,... are subtypes of PScalarType. Even if it is possible to answer certain questions about
+# type by looking at the Ruby class of the types this is considered an implementation detail, and such checks should in general
+# be performed by the type_calculator which implements the type system semantics.
+#
+# The PRubyType
+# -------------
+# The PRubyType corresponds to a Ruby Class, except for the puppet types that are specialized (i.e. PRubyType should not be
+# used for Integer, String, etc. since there are specialized types for those).
+# When the type calculator deals with PRubyTypes and checks for assignability, it determines the "common ancestor class" of two classes.
+# This check is made based on the superclasses of the two classes being compared. In order to perform this, the classes must be present
+# (i.e. they are resolved from the string form in the PRubyType to a loaded, instantiated Ruby Class). In general this is not a problem,
+# since the question to produce the common super type for two objects means that the classes must be present or there would have been
+# no instances present in the first place. If however the classes are not present, the type calculator will fall back and state that
+# the two types at least have Object in common.
+#
+# @see Puppet::Pops::Types::TypeFactory TypeFactory for how to create instances of types
+# @see Puppet::Pops::Types::TypeParser TypeParser how to construct a type instance from a String
+# @see Puppet::Pops::Types Types for details about the type model
+#
+# Using the Type Calculator
+# -----
+# The type calculator can be directly used via its class methods. If doing time critical work and doing many
+# calls to the type calculator, it is more performant to create an instance and invoke the corresponding
+# instance methods. Note that inference is an expensive operation, rather than infering the same thing
+# several times, it is in general better to infer once and then copy the result if mutation to a more generic form is
+# required.
+#
+# @api public
+#
+class Puppet::Pops::Types::TypeCalculator
+
+  Types = Puppet::Pops::Types
+  TheInfinity = 1.0 / 0.0 # because the Infinity symbol is not defined
+
+  # @api public
+  def self.assignable?(t1, t2)
+    singleton.assignable?(t1,t2)
+  end
+
+  # @api public
+  def self.string(t)
+    singleton.string(t)
+  end
+
+  # @api public
+  def self.infer(o)
+    singleton.infer(o)
+  end
+
+  # @api public
+  def self.generalize!(o)
+    singleton.generalize!(o)
+  end
+
+  # @api public
+  def self.infer_set(o)
+    singleton.infer_set(o)
+  end
+
+  # @api public
+  def self.debug_string(t)
+    singleton.debug_string(t)
+  end
+
+  # @api public
+  def self.enumerable(t)
+    singleton.enumerable(t)
+  end
+
+  # @api private
+  def self.singleton()
+    @tc_instance ||= new
+  end
+
+  # @api public
+  #
+  def initialize
+    @@assignable_visitor ||= Puppet::Pops::Visitor.new(nil,"assignable",1,1)
+    @@infer_visitor ||= Puppet::Pops::Visitor.new(nil,"infer",0,0)
+    @@infer_set_visitor ||= Puppet::Pops::Visitor.new(nil,"infer_set",0,0)
+    @@instance_of_visitor ||= Puppet::Pops::Visitor.new(nil,"instance_of",1,1)
+    @@string_visitor ||= Puppet::Pops::Visitor.new(nil,"string",0,0)
+    @@inspect_visitor ||= Puppet::Pops::Visitor.new(nil,"debug_string",0,0)
+    @@enumerable_visitor ||= Puppet::Pops::Visitor.new(nil,"enumerable",0,0)
+    @@extract_visitor ||= Puppet::Pops::Visitor.new(nil,"extract",0,0)
+    @@generalize_visitor ||= Puppet::Pops::Visitor.new(nil,"generalize",0,0)
+
+    da = Types::PArrayType.new()
+    da.element_type = Types::PDataType.new()
+    @data_array = da
+
+    h = Types::PHashType.new()
+    h.element_type = Types::PDataType.new()
+    h.key_type = Types::PScalarType.new()
+    @data_hash = h
+
+    @data_t = Types::PDataType.new()
+    @scalar_t = Types::PScalarType.new()
+    @numeric_t = Types::PNumericType.new()
+    @t = Types::PObjectType.new()
+
+    # Data accepts a Tuple that has 0-infinity Data compatible entries (e.g. a Tuple equivalent to Array).
+    data_tuple = Types::PTupleType.new()
+    data_tuple.addTypes(Types::PDataType.new())
+    data_tuple.size_type = Types::PIntegerType.new()
+    data_tuple.size_type.from = 0
+    data_tuple.size_type.to = nil # infinity
+    @data_tuple_t = data_tuple
+
+    # Variant type compatible with Data
+    data_variant = Types::PVariantType.new()
+    data_variant.addTypes(@data_hash.copy)
+    data_variant.addTypes(@data_array.copy)
+    data_variant.addTypes(Types::PScalarType.new)
+    data_variant.addTypes(Types::PNilType.new)
+    data_variant.addTypes(@data_tuple_t.copy)
+    @data_variant_t = data_variant
+
+    collection_default_size = Types::PIntegerType.new()
+    collection_default_size.from = 0
+    collection_default_size.to = nil # infinity
+    @collection_default_size_t = collection_default_size
+
+    non_empty_string = Types::PStringType.new
+    non_empty_string.size_type = Types::PIntegerType.new()
+    non_empty_string.size_type.from = 1
+    non_empty_string.size_type.to = nil # infinity
+    @non_empty_string_t = non_empty_string
+  end
+
+  # Convenience method to get a data type for comparisons
+  # @api private the returned value may not be contained in another element
+  #
+  def data
+    @data_t
+  end
+
+  # Convenience method to get a variant compatible with the Data type.
+  # @api private the returned value may not be contained in another element
+  #
+  def data_variant
+    @data_variant_t
+  end
+
+  def self.data_variant
+    singleton.data_variant
+  end
+
+  # Answers the question 'is it possible to inject an instance of the given class'
+  # A class is injectable if it has a special *assisted inject* class method called `inject` taking
+  # an injector and a scope as argument, or if it has a zero args `initialize` method.
+  #
+  # @param klazz [Class, PRubyType] the class/type to check if it is injectable
+  # @return [Class, nil] the injectable Class, or nil if not injectable
+  # @api public
+  #
+  def injectable_class(klazz)
+    # Handle case when we get a PType instead of a class
+    if klazz.is_a?(Types::PRubyType)
+      klazz = Puppet::Pops::Types::ClassLoader.provide(klazz)
+    end
+
+    # data types can not be injected (check again, it is not safe to assume that given RubyType klazz arg was ok)
+    return false unless type(klazz).is_a?(Types::PRubyType)
+    if (klazz.respond_to?(:inject) && klazz.method(:inject).arity() == -4) || klazz.instance_method(:initialize).arity() == 0
+      klazz
+    else
+      nil
+    end
+  end
+
+  # Answers 'can an instance of type t2 be assigned to a variable of type t'
+  # @api public
+  #
+  def assignable?(t, t2)
+    # nil is assignable to anything except to required types
+    return true if is_pnil?(t2)
+
+    if t.is_a?(Class)
+      t = type(t)
+    end
+
+    if t2.is_a?(Class)
+      t2 = type(t2)
+    end
+
+    @@assignable_visitor.visit_this_1(self, t, t2)
+ end
+
+  # Returns an enumerable if the t represents something that can be iterated
+  def enumerable(t)
+    @@enumerable_visitor.visit_this_0(self, t)
+  end
+
+  # Answers if the two given types describe the same type
+  def equals(left, right)
+    return false unless left.is_a?(Types::PAbstractType) && right.is_a?(Types::PAbstractType)
+    # Types compare per class only - an extra test must be made if the are mutually assignable
+    # to find all types that represent the same type of instance
+    #
+    left  == right || (assignable?(right, left) && assignable?(left, right))
+  end
+
+  # Answers 'what is the Puppet Type corresponding to the given Ruby class'
+  # @param c [Class] the class for which a puppet type is wanted
+  # @api public
+  #
+  def type(c)
+    raise ArgumentError, "Argument must be a Class" unless c.is_a? Class
+
+    # Can't use a visitor here since we don't have an instance of the class
+    case
+    when c <= Integer
+      type = Types::PIntegerType.new()
+    when c == Float
+      type = Types::PFloatType.new()
+    when c == Numeric
+      type = Types::PNumericType.new()
+    when c == String
+      type = Types::PStringType.new()
+    when c == Regexp
+      type = Types::PRegexpType.new()
+    when c == NilClass
+      type = Types::PNilType.new()
+    when c == FalseClass, c == TrueClass
+      type = Types::PBooleanType.new()
+    when c == Class
+      type = Types::PType.new()
+    when c == Array
+      # Assume array of data values
+      type = Types::PArrayType.new()
+      type.element_type = Types::PDataType.new()
+    when c == Hash
+      # Assume hash with scalar keys and data values
+      type = Types::PHashType.new()
+      type.key_type = Types::PScalarType.new()
+      type.element_type = Types::PDataType.new()
+    else
+      type = Types::PRubyType.new()
+      type.ruby_class = c.name
+    end
+    type
+  end
+
+  # Generalizes value specific types. The given type is mutated and returned.
+  # @api public
+  def generalize!(o)
+    @@generalize_visitor.visit_this_0(self, o)
+    o.eAllContents.each { |x| @@generalize_visitor.visit_this_0(self, x) }
+    o
+  end
+
+  def generalize_Object(o)
+    # do nothing, there is nothing to change for most types
+  end
+
+  def generalize_PStringType(o)
+    o.values = []
+    o.size_type = nil
+    []
+  end
+
+  def generalize_PCollectionType(o)
+    # erase the size constraint from Array and Hash (if one exists, it is transformed to -Infinity - + Infinity, which is
+    # not desirable.
+    o.size_type = nil
+  end
+
+  def generalize_PFloatType(o)
+    o.to = nil
+    o.from = nil
+  end
+
+  def generalize_PIntegerType(o)
+    o.to = nil
+    o.from = nil
+  end
+
+  # Answers 'what is the single common Puppet Type describing o', or if o is an Array or Hash, what is the
+  # single common type of the elements (or keys and elements for a Hash).
+  # @api public
+  #
+  def infer(o)
+    @@infer_visitor.visit_this_0(self, o)
+  end
+
+  def infer_generic(o)
+    result = generalize!(infer(o))
+    result
+  end
+
+  # Answers 'what is the set of Puppet Types of o'
+  # @api public
+  #
+  def infer_set(o)
+    @@infer_set_visitor.visit_this_0(self, o)
+  end
+
+  def instance_of(t, o)
+#    return true if o.nil? && !t.is_a?(Types::PRequiredType)
+    @@instance_of_visitor.visit_this_1(self, t, o)
+  end
+
+  def instance_of_Object(t, o)
+    # Undef is Undef and Object, but nothing else when checking instance?
+    return false if (o.nil? || o == :undef) && t.class != Types::PObjectType
+    assignable?(t, infer(o))
+  end
+
+  def instance_of_PArrayType(t, o)
+    return false unless o.is_a?(Array)
+    return false unless o.all? {|element| instance_of(t.element_type, element) }
+    size_t = t.size_type || @collection_default_size_t
+    size_t2 = size_as_type(o)
+    assignable?(size_t, size_t2)
+  end
+
+  def instance_of_PTupleType(t, o)
+    return false unless o.is_a?(Array)
+    # compute the tuple's min/max size, and check if that size matches
+    from, to = size_range(t.size_type)
+    size_t = Types::PIntegerType.new()
+    size_t.from = t.types.size - 1 + from
+    size_t.to = t.types.size - 1 + to
+    # compute the array's size as type
+    size_t2 = size_as_type(o)
+    return false unless assignable?(size_t, size_t2)
+    o.each_with_index do |element, index|
+       return false unless instance_of(t.types[index] || t.types[-1], element)
+    end
+    true
+  end
+
+  def instance_of_PStructType(t, o)
+    return false unless o.is_a?(Hash)
+    h = t.hashed_elements
+    # all keys must be present and have a value (even if nil/undef)
+    (o.keys - h.keys).empty? && h.all? { |k,v| instance_of(v, o[k]) }
+  end
+
+  def instance_of_PHashType(t, o)
+    return false unless o.is_a?(Hash)
+    key_t = t.key_type
+    element_t = t.element_type
+    return false unless o.keys.all? {|key| instance_of(key_t, key) } && o.values.all? {|value| instance_of(element_t, value) }
+    size_t = t.size_type || @collection_default_size_t
+    size_t2 = size_as_type(o)
+    assignable?(size_t, size_t2)
+  end
+
+  def instance_of_PDataType(t, o)
+    instance_of(@data_variant_t, o)
+  end
+
+  def instance_of_PNilType(t, o)
+    return o.nil? || o == :undef
+  end
+
+  def instance_of_POptionalType(t, o)
+    return true if (o.nil? || o == :undef)
+    instance_of(t.optional_type, o)
+  end
+
+  def instance_of_PVariantType(t, o)
+    # instance of variant if o is instance? of any of variant's types
+    t.types.any? { |option_t| instance_of(option_t, o) }
+  end
+
+  # Answers 'is o an instance of type t'
+  # @api public
+  #
+  def self.instance?(t, o)
+    singleton.instance_of(t,o)
+  end
+
+  # Answers 'is o an instance of type t'
+  # @api public
+  #
+  def instance?(t, o)
+    instance_of(t,o)
+  end
+
+  # Answers if t is a puppet type
+  # @api public
+  #
+  def is_ptype?(t)
+    return t.is_a?(Types::PAbstractType)
+  end
+
+  # Answers if t represents the puppet type PNilType
+  # @api public
+  #
+  def is_pnil?(t)
+    return t.nil? || t.is_a?(Types::PNilType)
+  end
+
+  # Answers, 'What is the common type of t1 and t2?'
+  #
+  # TODO: The current implementation should be optimized for performance
+  #
+  # @api public
+  #
+  def common_type(t1, t2)
+    raise ArgumentError, 'two types expected' unless (is_ptype?(t1) || is_pnil?(t1)) && (is_ptype?(t2) || is_pnil?(t2))
+
+    # if either is nil, the common type is the other
+    if is_pnil?(t1)
+      return t2
+    elsif is_pnil?(t2)
+      return t1
+    end
+
+    # Simple case, one is assignable to the other
+    if assignable?(t1, t2)
+      return t1
+    elsif assignable?(t2, t1)
+      return t2
+    end
+
+    # when both are arrays, return an array with common element type
+    if t1.is_a?(Types::PArrayType) && t2.is_a?(Types::PArrayType)
+      type = Types::PArrayType.new()
+      type.element_type = common_type(t1.element_type, t2.element_type)
+      return type
+    end
+
+    # when both are hashes, return a hash with common key- and element type
+    if t1.is_a?(Types::PHashType) && t2.is_a?(Types::PHashType)
+      type = Types::PHashType.new()
+      type.key_type = common_type(t1.key_type, t2.key_type)
+      type.element_type = common_type(t1.element_type, t2.element_type)
+      return type
+    end
+
+    # when both are host-classes, reduce to PHostClass[] (since one was not assignable to the other)
+    if t1.is_a?(Types::PHostClassType) && t2.is_a?(Types::PHostClassType)
+      return Types::PHostClassType.new()
+    end
+
+    # when both are resources, reduce to Resource[T] or Resource[] (since one was not assignable to the other)
+    if t1.is_a?(Types::PResourceType) && t2.is_a?(Types::PResourceType)
+      result = Types::PResourceType.new()
+      # only Resource[] unless the type name is the same
+      if t1.type_name == t2.type_name then result.type_name = t1.type_name end
+      # the cross assignability test above has already determined that they do not have the same type and title
+      return result
+    end
+
+    # Integers have range, expand the range to the common range
+    if t1.is_a?(Types::PIntegerType) && t2.is_a?(Types::PIntegerType)
+      t1range = from_to_ordered(t1.from, t1.to)
+      t2range = from_to_ordered(t2.from, t2.to)
+      t = Types::PIntegerType.new()
+      from = [t1range[0], t2range[0]].min
+      to = [t1range[1], t2range[1]].max
+      t.from = from unless from == TheInfinity
+      t.to = to unless to == TheInfinity
+      return t
+    end
+
+    # Floats have range, expand the range to the common range
+    if t1.is_a?(Types::PFloatType) && t2.is_a?(Types::PFloatType)
+      t1range = from_to_ordered(t1.from, t1.to)
+      t2range = from_to_ordered(t2.from, t2.to)
+      t = Types::PFloatType.new()
+      from = [t1range[0], t2range[0]].min
+      to = [t1range[1], t2range[1]].max
+      t.from = from unless from == TheInfinity
+      t.to = to unless to == TheInfinity
+      return t
+    end
+
+    if t1.is_a?(Types::PStringType) && t2.is_a?(Types::PStringType)
+      t = Types::PStringType.new()
+      t.values = t1.values | t2.values
+      return t
+    end
+
+    if t1.is_a?(Types::PPatternType) && t2.is_a?(Types::PPatternType)
+      t = Types::PPatternType.new()
+      # must make copies since patterns are contained types, not data-types
+      t.patterns = (t1.patterns | t2.patterns).map {|p| p.copy }
+      return t
+    end
+
+    if t1.is_a?(Types::PEnumType) && t2.is_a?(Types::PEnumType)
+      # The common type is one that complies with either set
+      t = Types::PEnumType.new
+      t.values = t1.values | t2.values
+      return t
+    end
+
+    if t1.is_a?(Types::PVariantType) && t2.is_a?(Types::PVariantType)
+      # The common type is one that complies with either set
+      t = Types::PVariantType.new
+      t.types = (t1.types | t2.types).map {|opt_t| opt_t.copy }
+      return t
+    end
+
+    if t1.is_a?(Types::PRegexpType) && t2.is_a?(Types::PRegexpType)
+      # if they were identical, the general rule would return a parameterized regexp
+      # since they were not, the result is a generic regexp type
+      return Types::PPatternType.new()
+    end
+
+    # Common abstract types, from most specific to most general
+    if common_numeric?(t1, t2)
+      return Types::PNumericType.new()
+    end
+
+    if common_scalar?(t1, t2)
+      return Types::PScalarType.new()
+    end
+
+    if common_data?(t1,t2)
+      return Types::PDataType.new()
+    end
+
+    # Meta types Type[Integer] + Type[String] => Type[Data]
+    if t1.is_a?(Types::PType) && t2.is_a?(Types::PType)
+      type = Types::PType.new()
+      type.type = common_type(t1.type, t2.type)
+      return type
+    end
+
+    if t1.is_a?(Types::PRubyType) && t2.is_a?(Types::PRubyType)
+      if t1.ruby_class == t2.ruby_class
+        return t1
+      end
+      # finding the common super class requires that names are resolved to class
+      c1 = Types::ClassLoader.provide_from_type(t1)
+      c2 = Types::ClassLoader.provide_from_type(t2)
+      if c1 && c2
+        c2_superclasses = superclasses(c2)
+        superclasses(c1).each do|c1_super|
+          c2_superclasses.each do |c2_super|
+            if c1_super == c2_super
+              result = Types::PRubyType.new()
+              result.ruby_class = c1_super.name
+              return result
+            end
+          end
+        end
+      end
+    end
+    # If both are RubyObjects
+
+    if common_pobject?(t1, t2)
+      return Types::PObjectType.new()
+    end
+  end
+
+  # Produces the superclasses of the given class, including the class
+  def superclasses(c)
+    result = [c]
+    while s = c.superclass
+      result << s
+      c = s
+    end
+    result
+  end
+
+  # Produces a string representing the type
+  # @api public
+  #
+  def string(t)
+    @@string_visitor.visit_this_0(self, t)
+  end
+
+  # Produces a debug string representing the type (possibly with more information that the regular string format)
+  # @api public
+  #
+  def debug_string(t)
+    @@inspect_visitor.visit_this_0(self, t)
+  end
+
+
+  # Reduces an enumerable of types to a single common type.
+  # @api public
+  #
+  def reduce_type(enumerable)
+    enumerable.reduce(nil) {|memo, t| common_type(memo, t) }
+  end
+
+  # Reduce an enumerable of objects to a single common type
+  # @api public
+  #
+  def infer_and_reduce_type(enumerable)
+    reduce_type(enumerable.collect() {|o| infer(o) })
+  end
+
+  # The type of all classes is PType
+  # @api private
+  #
+  def infer_Class(o)
+    Types::PType.new()
+  end
+
+  # @api private
+  def infer_Object(o)
+    type = Types::PRubyType.new()
+    type.ruby_class = o.class.name
+    type
+  end
+
+  # The type of all types is PType
+  # @api private
+  #
+  def infer_PObjectType(o)
+    type = Types::PType.new()
+    type.type = o.copy
+    type
+  end
+
+  # The type of all types is PType
+  # This is the metatype short circuit.
+  # @api private
+  #
+  def infer_PType(o)
+    type = Types::PType.new()
+    type.type = o.copy
+    type
+  end
+
+  # @api private
+  def infer_String(o)
+    t = Types::PStringType.new()
+    t.addValues(o)
+    t.size_type = size_as_type(o)
+    t
+  end
+
+  # @api private
+  def infer_Float(o)
+    t = Types::PFloatType.new()
+    t.from = o
+    t.to = o
+    t
+  end
+
+  # @api private
+  def infer_Integer(o)
+    t = Types::PIntegerType.new()
+    t.from = o
+    t.to = o
+    t
+  end
+
+  # @api private
+  def infer_Regexp(o)
+    t = Types::PRegexpType.new()
+    t.pattern = o.source
+    t
+  end
+
+  # @api private
+  def infer_NilClass(o)
+    Types::PNilType.new()
+  end
+
+  # Inference of :undef as PNilType, all other are Ruby[Symbol]
+  # @api private
+  def infer_Symbol(o)
+    o == :undef ? infer_NilClass(o) : infer_Object(o)
+  end
+
+  # @api private
+  def infer_TrueClass(o)
+    Types::PBooleanType.new()
+  end
+
+  # @api private
+  def infer_FalseClass(o)
+    Types::PBooleanType.new()
+  end
+
+  # @api private
+  # A Puppet::Parser::Resource, or Puppet::Resource
+  #
+  def infer_Resource(o)
+    t = Types::PResourceType.new()
+    t.type_name = o.type.to_s
+    t.title = o.title
+    t
+  end
+
+  # @api private
+  def infer_Array(o)
+    type = Types::PArrayType.new()
+    type.element_type =
+    if o.empty?
+      Types::PNilType.new()
+    else
+      infer_and_reduce_type(o)
+    end
+    type.size_type = size_as_type(o)
+    type
+  end
+
+  # @api private
+  def infer_Hash(o)
+    type = Types::PHashType.new()
+    if o.empty?
+      ktype = Types::PNilType.new()
+      etype = Types::PNilType.new()
+    else
+      ktype = infer_and_reduce_type(o.keys())
+      etype = infer_and_reduce_type(o.values())
+    end
+    type.key_type = ktype
+    type.element_type = etype
+    type.size_type = size_as_type(o)
+    type
+  end
+
+  def size_as_type(collection)
+    size = collection.size
+    t = Types::PIntegerType.new()
+    t.from = size
+    t.to = size
+    t
+  end
+
+  # Common case for everything that intrinsically only has a single type
+  def infer_set_Object(o)
+    infer(o)
+  end
+
+  def infer_set_Array(o)
+    type = Types::PArrayType.new()
+    type.element_type = if o.empty?
+      Types::PNilType.new()
+    else
+      t = Types::PVariantType.new()
+      t.types = o.map() {|x| infer_set(x) }
+      t.types.size == 1 ? t.types[0] : t
+    end
+    type.size_type = size_as_type(o)
+    type
+  end
+
+  def infer_set_Hash(o)
+    type = Types::PHashType.new()
+    if o.empty?
+      ktype = Types::PNilType.new()
+      etype = Types::PNilType.new()
+    else
+      ktype = Types::PVariantType.new()
+      ktype.types = o.keys.map() {|k| infer_set(k) }
+      etype = Types::PVariantType.new()
+      etype.types = o.values.map() {|e| infer_set(e) }
+    end
+    type.key_type = unwrap_single_variant(ktype)
+    type.element_type = unwrap_single_variant(vtype)
+    type.size_type = size_as_type(o)
+    type
+  end
+
+  def unwrap_single_variant(possible_variant)
+    if possible_variant.is_a?(Types::PVariantType) && possible_variant.types.size == 1
+      possible_variant.types[0]
+    else
+      possible_variant
+    end
+  end
+  # False in general type calculator
+  # @api private
+  def assignable_Object(t, t2)
+    false
+  end
+
+  # @api private
+  def assignable_PObjectType(t, t2)
+    t2.is_a?(Types::PObjectType)
+  end
+
+  # @api private
+  def assignable_PNilType(t, t2)
+    # Only undef/nil is assignable to nil type
+    t2.is_a?(Types::PNilType)
+  end
+
+  # @api private
+  def assignable_PScalarType(t, t2)
+    t2.is_a?(Types::PScalarType)
+  end
+
+  # @api private
+  def assignable_PNumericType(t, t2)
+    t2.is_a?(Types::PNumericType)
+  end
+
+  # @api private
+  def assignable_PIntegerType(t, t2)
+    return false unless t2.is_a?(Types::PIntegerType)
+    trange =  from_to_ordered(t.from, t.to)
+    t2range = from_to_ordered(t2.from, t2.to)
+    # If t2 min and max are within the range of t
+    trange[0] <= t2range[0] && trange[1] >= t2range[1]
+  end
+
+  # Transform int range to a size constraint
+  # if range == nil the constraint is 1,1
+  # if range.from == nil min size = 1
+  # if range.to == nil max size == Infinity
+  #
+  def size_range(range)
+    return [1,1] if range.nil?
+    from = range.from
+    to = range.to
+    x = from.nil? ? 1 : from
+    y = to.nil? ? TheInfinity : to
+    if x < y
+      [x, y]
+    else
+      [y, x]
+    end
+  end
+
+  # @api private
+  def from_to_ordered(from, to)
+    x = (from.nil? || from == :default) ? -TheInfinity : from
+    y = (to.nil? || to == :default) ? TheInfinity : to
+    if x < y
+      [x, y]
+    else
+      [y, x]
+    end
+  end
+
+  # @api private
+  def assignable_PVariantType(t, t2)
+    # Data is a specific variant
+    t2 = @data_variant_t if t2.is_a?(Types::PDataType)
+    if t2.is_a?(Types::PVariantType)
+      # A variant is assignable if all of its options are assignable to one of this type's options
+      return true if t == t2
+      t2.types.all? do |other|
+        # if the other is a Variant, all if its options, but be assignable to one of this type's options
+        other = other.is_a?(Types::PDataType) ? @data_variant_t : other
+        if other.is_a?(Types::PVariantType)
+          assignable?(t, other)
+        else
+          t.types.any? {|option_t| assignable?(option_t, other) }
+        end
+      end
+    else
+      # A variant is assignable if t2 is assignable to any of its types
+      t.types.any? { |option_t| assignable?(option_t, t2) }
+    end
+  end
+
+  def assignable_PTupleType(t, t2)
+    return true if t == t2 || t.types.empty? && (t2.is_a?(Types::PArrayType))
+    t_regular = t.types[0..-2]
+    t_ranged = t.types[-1]
+    t_from, t_to = size_range(t.size_type)
+    t_required = t_regular.size + t_from
+
+    if t2.is_a?(Types::PTupleType)
+      t2_regular = t2.types[0..-2]
+      t2_ranged = t2.types[-1]
+      t2_from, t2_to = size_range(t2.size_type)
+      t2_required = t2_regular.size + t2_from
+
+      # tuples with fewer required entries can not be assigned
+      return false if t_required > t2_required
+      # tuples with more optionally available entries can not be assigned
+      return false if t_regular.size + t_to < t2_regular.size + t2_to
+
+      t_required.times do |index|
+        t_entry = tuple_entry_at(t, t_from, t_to, index)
+        t2_entry = tuple_entry_at(t2, t2_from, t2_to, index)
+        return false if t2_entry.nil? || !assignable?(t_entry, t2_entry)
+      end
+      # Handle remainder in t2's required
+      (t2_required - t_required).times do |index|
+        t_entry = tuple_entry_at(t, t_from, t_to, t_required + index)
+        t2_entry = tuple_entry_at(t2, t2_from, t2_to, t_required + index)
+        return false if t2_entry.nil? || !assignable?(t_entry, t2_entry)
+      end
+      # Now only a trailing optional type remains - the last type must always be compatible
+      # irrespective of optionality and count
+      #
+      return assignable?(t_ranged, t2_ranged)
+
+    elsif t2.is_a?(Types::PArrayType)
+      t2_entry = t2.element_type
+
+      # Array of anything can not be assigned (unless tuple is tuple of anything) - this case
+      # was handled at the top of this method.
+      #
+      return false if t2_entry.nil?
+
+      # array type may be size constrained
+      size_t = t2.size_type || @collection_default_size_t
+      min, max = size_t.range
+      # Array with fewer min entries can not be assigned
+      return false if t_required > min
+      # Array with more optionally available entries can not be assigned
+      return false if t_regular.size + t_to < max
+      # each tuple type must be assignable to the element type
+      t_required.times do |index|
+        t_entry = tuple_entry_at(t, t_from, t_to, index)
+        return false unless assignable?(t_entry, t2_entry)
+      end
+      # ... and so must the last, possibly optional (ranged) type
+      return assignable?(t_ranged, t2_entry)
+    end
+  end
+
+  # Produces the tuple entry at the given index given a tuple type, its from/to constraints on the last
+  # type, and an index.
+  # Produces nil if the index is out of bounds
+  # from must be less than to, and from may not be less than 0
+  #
+  # @api private
+  #
+  def tuple_entry_at(tuple_t, from, to, index)
+    regular = (tuple_t.types.size - 1)
+    if index < regular
+      tuple_t.types[index]
+    elsif index < regular + to
+      # in the varargs part
+      tuple_t.types[-1]
+    else
+      nil
+    end
+  end
+
+  # @api private
+  #
+  def assignable_PStructType(t, t2)
+    return true if t == t2 || t.elements.empty? && (t2.is_a?(Types::PHashType))
+    h = t.hashed_elements
+    if t2.is_a?(Types::PStructType)
+      h2 = t2.hashed_elements
+      h.size == h2.size && h.all? {|k, v| assignable?(v, h2[k]) }
+    elsif t2.is_a?(Types::PHashType)
+      size_t2 = t2.size_type || @collection_default_size_t
+      size_t = Types::PIntegerType.new
+      size_t.from = size_t.to = h.size
+      # compatible size
+      # hash key type must be string of min 1 size
+      # hash value t must be assignable to each key
+      element_type = t2.element_type
+      assignable?(size_t, size_t2) &&
+        assignable?(@non_empty_string_t, t2.key_type) &&
+        h.all? {|k,v| assignable?(v, element_type) }
+    else
+      false
+    end
+  end
+
+  # @api private
+  def assignable_POptionalType(t, t2)
+    return true if t2.is_a?(Types::PNilType)
+    if t2.is_a?(Types::POptionalType)
+      assignable?(t.optional_type, t2.optional_type)
+    else
+      assignable?(t.optional_type, t2)
+    end
+  end
+
+  # @api private
+  def assignable_PEnumType(t, t2)
+    return true if t == t2 || (t.values.empty? && (t2.is_a?(Types::PStringType) || t2.is_a?(Types::PEnumType)))
+    if t2.is_a?(Types::PStringType)
+      # if the set of strings are all found in the set of enums
+      t2.values.all? { |s| t.values.any? { |e| e == s }}
+    else
+      false
+    end
+  end
+
+  # @api private
+  def assignable_PStringType(t, t2)
+    if t.values.empty?
+      # A general string is assignable by any other string or pattern restricted string
+      # if the string has a size constraint it does not match since there is no reasonable way
+      # to compute the min/max length a pattern will match. For enum, it is possible to test that
+      # each enumerator value is within range
+      size_t = t.size_type || @collection_default_size_t
+      case t2
+      when Types::PStringType
+        # true if size compliant
+        size_t2 = t2.size_type || @collection_default_size_t
+        assignable?(size_t, size_t2)
+
+      when Types::PPatternType
+        # true if size constraint is at least 0 to +Infinity (which is the same as the default)
+        assignable?(size_t, @collection_default_size_t)
+
+      when Types::PEnumType
+        if t2.values
+          # true if all enum values are within range
+          min, max = t2.values.map(&:size).minmax
+          trange =  from_to_ordered(size_t.from, size_t.to)
+          t2range = [min, max]
+          # If t2 min and max are within the range of t
+          trange[0] <= t2range[0] && trange[1] >= t2range[1]
+        else
+          # no string can match this enum anyway since it does not accept anything
+          false
+        end
+      end
+    elsif t2.is_a?(Types::PStringType)
+      # A specific string acts as a set of strings - must have exactly the same strings
+      # In this case, size does not matter since the definition is very precise anyway
+      Set.new(t.values) == Set.new(t2.values)
+    else
+      # All others are false, since no other type describes the same set of specific strings
+      false
+    end
+  end
+
+  # @api private
+  def assignable_PPatternType(t, t2)
+    return true if t == t2
+    return false unless t2.is_a?(Types::PStringType) || t2.is_a?(Types::PEnumType)
+
+    if t2.values.empty?
+      # Strings / Enums (unknown which ones) cannot all match a pattern, but if there is no pattern it is ok
+      # (There should really always be a pattern, but better safe than sorry).
+      return t.patterns.empty? ? true : false
+    end
+    # all strings in String/Enum type must match one of the patterns in Pattern type
+    regexps = t.patterns.map {|p| p.regexp }
+    t2.values.all? { |v| regexps.any? {|re| re.match(v) } }
+  end
+
+  # @api private
+  def assignable_PFloatType(t, t2)
+    return false unless t2.is_a?(Types::PFloatType)
+    trange =  from_to_ordered(t.from, t.to)
+    t2range = from_to_ordered(t2.from, t2.to)
+    # If t2 min and max are within the range of t
+    trange[0] <= t2range[0] && trange[1] >= t2range[1]
+  end
+
+  # @api private
+  def assignable_PBooleanType(t, t2)
+    t2.is_a?(Types::PBooleanType)
+  end
+
+  # @api private
+  def assignable_PRegexpType(t, t2)
+    t2.is_a?(Types::PRegexpType) && (t.pattern.nil? || t.pattern == t2.pattern)
+  end
+
+  # @api private
+  def assignable_PCollectionType(t, t2)
+    size_t = t.size_type || @collection_default_size_t
+    case t2
+    when Types::PCollectionType
+      size_t2 = t2.size_type || @collection_default_size_t
+      assignable?(size_t, size_t2)
+    when Types::PTupleType
+      # compute the tuple's min/max size, and check if that size matches
+      from, to = size_range(t2.size_type)
+      t2s = Types::PIntegerType.new()
+      t2s.from = t2.types.size - 1 + from
+      t2s.to = t2.types.size - 1 + to
+      assignable?(size_t, t2s)
+    when Types::PStructType
+      from = to = t2.elements.size
+      t2s = Types::PIntegerType.new()
+      t2s.from = from
+      t2s.to = to
+      assignable?(size_t, t2s)
+    else
+      false
+    end
+  end
+
+  # @api private
+  def assignable_PType(t, t2)
+    return false unless t2.is_a?(Types::PType)
+    return true if t.type.nil? # wide enough to handle all types
+    return false if t2.type.nil? # wider than t
+    assignable?(t.type, t2.type)
+  end
+
+  # Array is assignable if t2 is an Array and t2's element type is assignable, or if t2 is a Tuple
+  # where 
+  # @api private
+  def assignable_PArrayType(t, t2)
+    if t2.is_a?(Types::PArrayType)
+      return false unless assignable?(t.element_type, t2.element_type)
+      assignable_PCollectionType(t, t2)
+
+    elsif t2.is_a?(Types::PTupleType)
+      return false unless t2.types.all? {|t2_element| assignable?(t.element_type, t2_element) }
+      t2_regular = t2.types[0..-2]
+      t2_ranged = t2.types[-1]
+      t2_from, t2_to = size_range(t2.size_type)
+      t2_required = t2_regular.size + t2_from
+
+      t_entry = t.element_type
+
+      # Tuple of anything can not be assigned (unless array is tuple of anything) - this case
+      # was handled at the top of this method.
+      #
+      return false if t_entry.nil?
+
+      # array type may be size constrained
+      size_t = t.size_type || @collection_default_size_t
+      min, max = size_t.range
+      # Tuple with fewer min entries can not be assigned
+      return false if t2_required < min
+      # Tuple with more optionally available entries can not be assigned
+      return false if t2_regular.size + t2_to > max
+      # each tuple type must be assignable to the element type
+      t2_required.times do |index|
+        t2_entry = tuple_entry_at(t2, t2_from, t2_to, index)
+        return false unless assignable?(t_entry, t2_entry)
+      end
+      # ... and so must the last, possibly optional (ranged) type
+      return assignable?(t_entry, t2_ranged)
+    else
+      false
+    end
+  end
+
+  # Hash is assignable if t2 is a Hash and t2's key and element types are assignable
+  # @api private
+  def assignable_PHashType(t, t2)
+    case t2
+    when Types::PHashType
+      return false unless assignable?(t.key_type, t2.key_type) && assignable?(t.element_type, t2.element_type)
+      assignable_PCollectionType(t, t2)
+    when Types::PStructType
+      # hash must accept String as key type
+      # hash must accept all value types
+      # hash must accept the size of the struct
+      size_t = t.size_type || @collection_default_size_t
+      min, max = size_t.range
+      struct_size = t2.elements.size
+      element_type = t.element_type
+      ( struct_size >= min && struct_size <= max &&
+        assignable?(t.key_type, @non_emptry_string_t)  &&
+        t2.hashed_elements.all? {|k,v| assignable?(element_type, v) })
+    else
+      false
+    end
+  end
+
+  # @api private
+  def assignable_PCatalogEntryType(t1, t2)
+    t2.is_a?(Types::PCatalogEntryType)
+  end
+
+  # @api private
+  def assignable_PHostClassType(t1, t2)
+    return false unless t2.is_a?(Types::PHostClassType)
+    # Class = Class[name}, Class[name] != Class
+    return true if t1.class_name.nil?
+    # Class[name] = Class[name]
+    return t1.class_name == t2.class_name
+  end
+
+  # @api private
+  def assignable_PResourceType(t1, t2)
+    return false unless t2.is_a?(Types::PResourceType)
+    return true if t1.type_name.nil?
+    return false if t1.type_name != t2.type_name
+    return true if t1.title.nil?
+    return t1.title == t2.title
+  end
+
+  # Data is assignable by other Data and by Array[Data] and Hash[Scalar, Data]
+  # @api private
+  def assignable_PDataType(t, t2)
+    t2.is_a?(Types::PDataType) || assignable?(@data_variant_t, t2)
+  end
+
+  # Assignable if t2's ruby class is same or subclass of t1's ruby class
+  # @api private
+  def assignable_PRubyType(t1, t2)
+    return false unless t2.is_a?(Types::PRubyType)
+    return true if t1.ruby_class.nil?   # t1 is wider
+    return false if t2.ruby_class.nil?  # t1 not nil, so t2 can not be wider
+    c1 = class_from_string(t1.ruby_class)
+    c2 = class_from_string(t2.ruby_class)
+    return false unless c1.is_a?(Class) && c2.is_a?(Class)
+    !!(c2 <= c1)
+  end
+
+  # @api private
+  def debug_string_Object(t)
+    string(t)
+  end
+
+  # @api private
+  def string_PType(t)
+    if t.type.nil?
+      "Type"
+    else
+      "Type[#{string(t.type)}]"
+    end
+  end
+
+  # @api private
+  def string_NilClass(t)     ; '?'       ; end
+
+  # @api private
+  def string_String(t)       ; t         ; end
+
+  # @api private
+  def string_PObjectType(t)  ; "Object"  ; end
+
+  # @api private
+  def string_PNilType(t)     ; 'Undef'   ; end
+
+  # @api private
+  def string_PBooleanType(t) ; "Boolean" ; end
+
+  # @api private
+  def string_PScalarType(t) ; "Scalar" ; end
+
+  # @api private
+  def string_PDataType(t)    ; "Data"    ; end
+
+  # @api private
+  def string_PNumericType(t) ; "Numeric" ; end
+
+  # @api private
+  def string_PIntegerType(t)
+    range = range_array_part(t)
+    unless range.empty?
+      "Integer[#{range.join(', ')}]"
+    else
+      "Integer"
+    end
+  end
+
+  # Produces a string from an Integer range type that is used inside other type strings
+  # @api private
+  def range_array_part(t)
+    return [] if t.nil? || (t.from.nil? && t.to.nil?)
+    [t.from.nil? ? 'default' : t.from , t.to.nil? ? 'default' : t.to ]
+  end
+
+  # @api private
+  def string_PFloatType(t)
+    range = range_array_part(t)
+    unless range.empty?
+      "Float[#{range.join(', ')}]"
+    else
+      "Float"
+    end
+  end
+
+  # @api private
+  def string_PRegexpType(t)
+    t.pattern.nil? ? "Regexp" : "Regexp[#{t.regexp.inspect}]"
+  end
+
+  # @api private
+  def string_PStringType(t)
+    # skip values in regular output - see debug_string
+    range = range_array_part(t.size_type)
+    unless range.empty?
+      "String[#{range.join(', ')}]"
+    else
+      "String"
+    end
+  end
+
+  # @api private
+  def debug_string_PStringType(t)
+    range = range_array_part(t.size_type)
+    range_part = range.empty? ? '' : '[' << range.join(' ,') << '], '
+    "String[" << range_part << (t.values.map {|s| "'#{s}'" }).join(', ') << ']'
+  end
+
+  # @api private
+  def string_PEnumType(t)
+    return "Enum" if t.values.empty?
+    "Enum[" << t.values.map {|s| "'#{s}'" }.join(', ') << ']'
+  end
+
+  # @api private
+  def string_PVariantType(t)
+    return "Variant" if t.types.empty?
+    "Variant[" << t.types.map {|t2| string(t2) }.join(', ') << ']'
+  end
+
+  # @api private
+  def string_PTupleType(t)
+    range = range_array_part(t.size_type)
+    return "Tuple" if t.types.empty?
+    s = "Tuple[" << t.types.map {|t2| string(t2) }.join(', ')
+    unless range.empty?
+      s << ", " << range.join(', ')
+    end
+    s << "]"
+    s
+  end
+
+  # @api private
+  def string_PStructType(t)
+    return "Struct" if t.elements.empty?
+    "Struct[{" << t.elements.map {|element| string(element) }.join(', ') << "}]"
+  end
+
+  def string_PStructElement(t)
+    "'#{t.name}'=>#{string(t.type)}"
+  end
+
+  # @api private
+  def string_PPatternType(t)
+    return "Pattern" if t.patterns.empty?
+    "Pattern[" << t.patterns.map {|s| "#{s.regexp.inspect}" }.join(', ') << ']'
+  end
+
+  # @api private
+  def string_PCollectionType(t)
+    range = range_array_part(t.size_type)
+    unless range.empty?
+      "Collection[#{range.join(', ')}]"
+    else
+      "Collection"
+    end
+  end
+
+  # @api private
+  def string_PRubyType(t)   ; "Ruby[#{string(t.ruby_class)}]"  ; end
+
+  # @api private
+  def string_PArrayType(t)
+    parts = [string(t.element_type)] + range_array_part(t.size_type)
+    "Array[#{parts.join(', ')}]"
+  end
+
+  # @api private
+  def string_PHashType(t)
+    parts = [string(t.key_type), string(t.element_type)] + range_array_part(t.size_type)
+    "Hash[#{parts.join(', ')}]"
+  end
+
+  # @api private
+  def string_PCatalogEntryType(t)
+    "CatalogEntry"
+  end
+
+  # @api private
+  def string_PHostClassType(t)
+    if t.class_name
+      "Class[#{t.class_name}]"
+    else
+      "Class"
+    end
+  end
+
+  # @api private
+  def string_PResourceType(t)
+    if t.type_name
+      if t.title
+        "#{t.type_name.capitalize}['#{t.title}']"
+      else
+        "#{t.type_name.capitalize}"
+      end
+    else
+      "Resource"
+    end
+  end
+
+  def string_POptionalType(t)
+    if t.optional_type.nil?
+      "Optional"
+    else
+      "Optional[#{string(t.optional_type)}]"
+    end
+  end
+
+  # Catches all non enumerable types
+  # @api private
+  def enumerable_Object(o)
+    nil
+  end
+
+  # @api private
+  def enumerable_PIntegerType(t)
+    # Not enumerable if representing an infinite range
+    return nil if t.size == TheInfinity
+    t
+  end
+
+  private
+
+  def class_from_string(str)
+    begin
+      str.split('::').inject(Object) do |memo, name_segment|
+        memo.const_get(name_segment)
+      end
+    rescue NameError
+      return nil
+    end
+  end
+
+  def common_data?(t1, t2)
+    assignable?(@data_t, t1) && assignable?(@data_t, t2)
+  end
+
+  def common_scalar?(t1, t2)
+    assignable?(@scalar_t, t1) && assignable?(@scalar_t, t2)
+  end
+
+  def common_numeric?(t1, t2)
+    assignable?(@numeric_t, t1) && assignable?(@numeric_t, t2)
+  end
+
+  def common_pobject?(t1, t2)
+    assignable?(@t, t1) && assignable?(@t, t2)
+  end
+end