require 'grape/router' require 'grape/api/instance' module Grape # The API class is the primary entry point for creating Grape APIs. Users # should subclass this class in order to build an API. class API # Class methods that we want to call on the API rather than on the API object NON_OVERRIDABLE = (Class.new.methods + %i[call call! configuration compile!]).freeze class << self attr_accessor :base_instance, :instances # Rather than initializing an object of type Grape::API, create an object of type Instance def new(*args, &block) base_instance.new(*args, &block) end # When inherited, will create a list of all instances (times the API was mounted) # It will listen to the setup required to mount that endpoint, and replicate it on any new instance def inherited(api, base_instance_parent = Grape::API::Instance) api.initial_setup(base_instance_parent) api.override_all_methods! make_inheritable(api) end # Initialize the instance variables on the remountable class, and the base_instance # an instance that will be used to create the set up but will not be mounted def initial_setup(base_instance_parent) @instances = [] @setup = [] @base_parent = base_instance_parent @base_instance = mount_instance end # Redefines all methods so that are forwarded to add_setup and be recorded def override_all_methods! (base_instance.methods - NON_OVERRIDABLE).each do |method_override| define_singleton_method(method_override) do |*args, &block| add_setup(method_override, *args, &block) end end end # Configure an API from the outside. If a block is given, it'll pass a # configuration hash to the block which you can use to configure your # API. If no block is given, returns the configuration hash. # The configuration set here is accessible from inside an API with # `configuration` as normal. def configure config = @base_instance.configuration if block_given? yield config self else config end end # This is the interface point between Rack and Grape; it accepts a request # from Rack and ultimately returns an array of three values: the status, # the headers, and the body. See [the rack specification] # (http://www.rubydoc.info/github/rack/rack/master/file/SPEC) for more. # NOTE: This will only be called on an API directly mounted on RACK def call(*args, &block) instance_for_rack.call(*args, &block) end # Allows an API to itself be inheritable: def make_inheritable(api) # When a child API inherits from a parent API. def api.inherited(child_api) # The instances of the child API inherit from the instances of the parent API Grape::API.inherited(child_api, base_instance) end end # Alleviates problems with autoloading by tring to search for the constant def const_missing(*args) if base_instance.const_defined?(*args) base_instance.const_get(*args) else super end end # The remountable class can have a configuration hash to provide some dynamic class-level variables. # For instance, a descripcion could be done using: `desc configuration[:description]` if it may vary # depending on where the endpoint is mounted. Use with care, if you find yourself using configuration # too much, you may actually want to provide a new API rather than remount it. def mount_instance(opts = {}) instance = Class.new(@base_parent) instance.configuration = Grape::Util::EndpointConfiguration.new(opts[:configuration] || {}) instance.base = self replay_setup_on(instance) instance end # Replays the set up to produce an API as defined in this class, can be called # on classes that inherit from Grape::API def replay_setup_on(instance) @setup.each do |setup_step| replay_step_on(instance, setup_step) end end def respond_to?(method, include_private = false) super(method, include_private) || base_instance.respond_to?(method, include_private) end def respond_to_missing?(method, include_private = false) base_instance.respond_to?(method, include_private) end def method_missing(method, *args, &block) # If there's a missing method, it may be defined on the base_instance instead. if respond_to_missing?(method) base_instance.send(method, *args, &block) else super end end def compile! require 'grape/eager_load' instance_for_rack.compile! # See API::Instance.compile! end private def instance_for_rack if never_mounted? base_instance else mounted_instances.first end end # Adds a new stage to the set up require to get a Grape::API up and running def add_setup(method, *args, &block) setup_step = { method: method, args: args, block: block } @setup << setup_step last_response = nil @instances.each do |instance| last_response = replay_step_on(instance, setup_step) end last_response end def replay_step_on(instance, setup_step) return if skip_immediate_run?(instance, setup_step[:args]) args = evaluate_arguments(instance.configuration, *setup_step[:args]) response = instance.send(setup_step[:method], *args, &setup_step[:block]) if skip_immediate_run?(instance, [response]) response else evaluate_arguments(instance.configuration, response).first end end # Skips steps that contain arguments to be lazily executed (on re-mount time) def skip_immediate_run?(instance, args) instance.base_instance? && (any_lazy?(args) || args.any? { |arg| arg.is_a?(Hash) && any_lazy?(arg.values) }) end def any_lazy?(args) args.any? { |argument| argument.respond_to?(:lazy?) && argument.lazy? } end def evaluate_arguments(configuration, *args) args.map do |argument| if argument.respond_to?(:lazy?) && argument.lazy? argument.evaluate_from(configuration) elsif argument.is_a?(Hash) argument.map { |key, value| [key, evaluate_arguments(configuration, value).first] }.to_h elsif argument.is_a?(Array) evaluate_arguments(configuration, *argument) else argument end end end def never_mounted? mounted_instances.empty? end def mounted_instances instances - [base_instance] end end end end