# Abstract deserialization service
#
# The main service is `deserialize`.
abstract class Deserializer
	# Deserialize and return an object, storing errors in the attribute `errors`
	#
	# If a `static_type` is given, only subtypes of the `static_type` are accepted.
	#
	# This method behavior varies according to the implementation engines.
	fun deserialize(static_type: nullable String): nullable Object is abstract

	# Deserialize the attribute with `name` from the object open for deserialization
	#
	# The `static_type` restricts what kind of object can be deserialized.
	#
	# Return the deserialized value or null on error, and set
	# `deserialize_attribute_missing` to whether the attribute was missing.
	#
	# Internal method to be implemented by the engines.
	fun deserialize_attribute(name: String, static_type: nullable String): nullable Object is abstract

	# Was the attribute queried by the last call to `deserialize_attribute` missing?
	var deserialize_attribute_missing = false

	# Register a newly allocated object (even if not completely built)
	#
	# Internal method called by objects in creation, to be implemented by the engines.
	fun notify_of_creation(new_object: Object) is abstract

	# Deserialize the next available object as an instance of `class_name`
	#
	# Return the deserialized object on success and
	# record in `errors` if `class_name` is unknown.
	#
	# This method should be redefined for each custom subclass of `Serializable`.
	# All refinement should look for a precise `class_name` and call super
	# on unsupported classes.
	protected fun deserialize_class(class_name: Text): nullable Object do
		if class_name == "Error" then return new Error.from_deserializer(self)
		return deserialize_class_intern(class_name)
	end

	# Generated service to deserialize the next available object as an instance of `class_name`
	#
	# Refinements to this method will be generated by the serialization phase.
	# To avoid conflicts, there should not be any other refinements to this method.
	# You can instead use `deserialize_class`.
	protected fun deserialize_class_intern(class_name: Text): nullable Object do
		errors.add new Error("Deserialization Error: Doesn't know how to deserialize class \"{class_name}\"")
		return null
	end

	# Should `self` keep trying to deserialize an object after an error?
	#
	# This behavior takes effect after each attribute deserialization with
	# errors such as a missing attribute or the value is of the wrong type.
	# If `keep_going`, the attribute will be skipped but the engine will
	# deserialize the next attribute.
	# If `not keep_going`, the engine stops deserializing right away.
	#
	# When at `true`, this may cause the accumulation of a lot of entries in `errors`.
	#
	# Default at `true`.
	var keep_going: nullable Bool = null is writable

	# Errors encountered in the last call to `deserialize`
	var errors = new Array[Error]
end