Note: This fork should no longer be necessary after version 0.9.1 introduced support for Ruby 3.2.
Apipie-rails is a DSL and Rails engine for documenting your RESTful
API. Instead of traditional use of #comments
, Apipie lets you
describe the code, through the code. This brings advantages like:
- No need to learn yet another syntax, you already know Ruby, right?
- Possibility of reusing the docs for other purposes (such as validation)
- Easier to extend and maintain (no string parsing involved)
- Possibility of reusing other sources for documentation purposes (such as routes etc.)
The documentation is available from within your app (by default under the
/apipie
path.) In development mode, you can see the changes as you
go. It's markup language agnostic, and even provides an API for reusing
the documentation data in JSON.
The easiest way to get Apipie up and running with your app is:
echo "gem 'apipie-rails'" >> Gemfile
bundle install
rails g apipie:install
Now you can start documenting your resources and actions (see DSL Reference for more info):
api :GET, '/users/:id'
param :id, :number, desc: 'id of the requested user'
def show
# ...
end
Run your application and see the result at
http://localhost:3000/apipie
. For further processing, you can
use http://localhost:3000/apipie.json
.
For a more comprehensive getting started guide, see this demo, which includes features such as generating documentation from tests, recording examples etc.
See Contributors page. Special thanks to all of them!
Apipie-rails is released under the MIT License
Table Of Contents
- API Documentation Tool
- Documentation
- DSL Reference
- Configuration Reference
- Rails Routes Integration
- Processing
- Validators
- Versioning
- Markup
- Localization
- Modifying Views
- Static files
- Static Swagger (OpenAPI 2.0) files
- Dynamic Swagger generation
- JSON checksums
- Tests Integration
- Bindings Generator
- Disqus Integration
- External References
You can describe a resource on the controller level. The description is introduced by calling
resource_description do ... end
.
Inheritance is supported, so you can specify common params for group of controllers in their parent class.
The following keywords are available (all are optional):
- resource_id
- How the resource will be referenced in Apipie (paths,
see
command etc.); by default controller_name.downcase is used. - name
- Human readable name of resource. By default
class.name.humanize
is used. - short (also short_description)
- Short description of the resource (it's shown on both the list of resources, and resource details)
- desc (also description and full_description)
- Full description of the resource (shown only in resource details)
- param
- Common params for all methods defined in controller/child controllers.
- returns
- Common responses for all methods defined in controller/child controllers.
- api_base_url
- What URL is the resource available under.
- api_versions (also api_version)
- What versions does the controller define the resource. (See Versioning for details.)
- formats
- Request / response formats.
- error
- Describe every possible error that can happen when calling all methods defined in controller. HTTP response code and description can be provided.
- app_info
- In case of versioning, this sets app info description on a per_version basis.
- meta
- Hash or array with custom metadata.
- deprecated
- Boolean value indicating if the resource is marked as deprecated. (Default false)
resource_description do
short 'Site members'
formats ['json']
param :id, Fixnum, :desc => "User ID", :required => false
param :resource_param, Hash, :desc => 'Param description for all methods' do
param :ausername, String, :desc => "Username for login", :required => true
param :apassword, String, :desc => "Password for login", :required => true
end
api_version "development"
error 404, "Missing"
error 500, "Server crashed for some <%= reason %>", :meta => {:anything => "you can think of"}
error :unprocessable_entity, "Could not save the entity."
returns :code => 403 do
property :reason, String, :desc => "Why this was forbidden"
end
meta :author => {:name => 'John', :surname => 'Doe'}
deprecated false
description <<-EOS
== Long description
Example resource for rest api documentation
These can now be accessed in <tt>shared/header</tt> with:
Headline: <%= headline %>
First name: <%= person.first_name %>
If you need to find out whether a certain local variable has been
assigned a value in a particular render call, you need to use the
following pattern:
<% if local_assigns.has_key? :headline %>
Headline: <%= headline %>
<% end %>
Testing using <tt>defined? headline</tt> will not work. This is an
implementation restriction.
=== Template caching
By default, Rails will compile each template to a method in order
to render it. When you alter a template, Rails will check the
file's modification time and recompile it in development mode.
EOS
end
Then describe methods available to your API.
- api
Describe how this method is exposed, and provide a short description. The first parameter is HTTP method (one of :GET/:POST/:PUT/:DELETE). The second parameter is the relative URL path which is mapped to this method. The last parameter is the methods short description. You can use this +api+ method more than once per method. It could be useful when there are more routes mapped to it.
When providing just one argument (description), or no argument at all, the paths will be loaded from the routes.rb file.
- api!
- Provide a short description and additional option. The last parameter is the methods short description. The paths will be loaded from routes.rb file. See Rails Routes Integration for more details.
- api_versions (also api_version)
- What version(s) does the action belong to. (See Versioning for details.)
- param
- Look at Parameter description section for details.
- returns
- Look at Response description section for details.
- tags
- Adds tags for grouping operations together in Swagger outputs. See swagger for more details. You can also provide tags in the Resource Description block so that they are automatically prepended to all action tags in the controller.
- formats
- Method level request / response formats.
- error
- Describe each possible error that can happen while calling this method. HTTP response code and description can be provided.
- description
- Full method description, which will be converted into HTML by the chosen markup language processor.
- example
- Provide an example of the server response; whole communication or response type. It will be formatted as code.
- see
- Provide reference to another method, this has to be a string with controller_name#method_name.
- meta
- Hash or array with custom metadata.
- show
- Resource is hidden from documentation when set to false (true by default)
# The simplest case: just load the paths from routes.rb
api!
def index
end
# More complex example
api :GET, "/users/:id", "Show user profile"
show false
error :code => 401, :desc => "Unauthorized"
error :code => 404, :desc => "Not Found", :meta => {:anything => "you can think of"}
param :session, String, :desc => "user is logged in", :required => true
param :regexp_param, /^[0-9]* years/, :desc => "regexp param"
param :array_param, [100, "one", "two", 1, 2], :desc => "array validator"
param :boolean_param, [true, false], :desc => "array validator with boolean"
param :proc_param, lambda { |val|
val == "param value" ? true : "The only good value is 'param value'."
}, :desc => "proc validator"
param :param_with_metadata, String, :desc => "", :meta => [:your, :custom, :metadata]
returns :code => 200, :desc => "a successful response" do
property :value1, String, :desc => "A string value"
property :value2, Integer, :desc => "An integer value"
property :value3, Hash, :desc => "An object" do
property :enum1, ['v1', 'v2'], :desc => "One of 2 possible string values"
end
end
tags %w[profiles logins]
tags 'more', 'related', 'resources'
description "method description"
formats ['json', 'jsonp', 'xml']
meta :message => "Some very important info"
example " 'user': {...} "
see "users#showme", "link description"
see :link => "users#update", :desc => "another link description"
def show
#...
end
Use param
to describe every possible parameter. You can use the Hash validator
in conjunction with a block given to the param method to describe nested parameters.
- name
- The first argument is the parameter name as a symbol.
- validator
- Second parameter is the parameter validator, choose one from section Validators
- desc
- Parameter description.
- required
- Set this true/false to make it required/optional. Default is optional
- allow_nil
- Setting this to true means that
nil
can be passed. - allow_blank
- Like
allow_nil
, but for blank values.false
,""
,' '
,nil
,[]
, and{}
are all blank. - as
- Used by the processing functionality to change the name of a key params.
- meta
- Hash or array with custom metadata.
- show
- Parameter is hidden from documentation when set to false (true by default)
- missing_message
- Specify the message to be returned if the parameter is missing as a string or Proc.
Defaults to
Missing parameter #{name}
if not specified. - only_in
- This can be set to
:request
or:response
. Setting to:response
causes the param to be ignored when used as part of a request description. Setting to:request
causes this param to be ignored when used as part of a response description. Ifonly_in
is not specified, the param definition is used for both requests and responses. (Note that the keywordproperty
is similar toparam
, but it has a:only_in => :response
default).
param :user, Hash, :desc => "User info" do
param :username, String, :desc => "Username for login", :required => true
param :password, String, :desc => "Password for login", :required => true
param :membership, ["standard","premium"], :desc => "User membership"
param :admin_override, String, :desc => "Not shown in documentation", :show => false
param :ip_address, String, :desc => "IP address", :required => true, :missing_message => lambda { I18n.t("ip_address.required") }
end
def create
#...
end
Often, params occur together in more actions. Typically, most of the
params for create
and update
actions are shared between them.
These params can be extracted with def_param_group
and
param_group
keywords.
The definition is looked up in the scope of the controller. If the group is defined in a different controller, it might be referenced by specifying the second argument.
# v1/users_controller.rb
def_param_group :address do
param :street, String
param :number, Integer
param :zip, String
end
def_param_group :user do
param :user, Hash do
param :name, String, "Name of the user"
param_group :address
end
end
api :POST, "/users", "Create an user"
param_group :user
def create
# ...
end
api :PUT, "/users/:id", "Update an user"
param_group :user
def update
# ...
end
# v2/users_controller.rb
api :POST, "/users", "Create an user"
param_group :user, V1::UsersController
def create
# ...
end
In CRUD operations, this pattern occurs quite often - params that need to be set are:
- for create action:
required => true
andallow_nil => false
- for update action:
required => false
andallow_nil => false
This makes it hard to share the param definitions across theses
actions. Therefore, you can make the description a bit smarter by
setting :action_aware => true
.
You can specify explicitly how the param group should be evaluated
with :as
option (either :create or :update)
def_param_group :user do
param :user, Hash, :action_aware => true do
param :name, String, :required => true
param :description, String
end
end
api :POST, "/users", "Create an user"
param_group :user
def create
# ...
end
api :PUT, "/users/admin", "Create an admin"
param_group :user, :as => :create
def create_admin
# ...
end
api :PUT, "/users/:id", "Update an user"
param_group :user
def update
# ...
end
In this case, user[name]
will be not be allowed nil for all
actions and required only for create
and create_admin
. Params
with allow_nil
set explicitly don't have this value changed.
Action awareness is inherited from ancestors (in terms of nested params).
The response from an API call can be documented by adding a returns
statement to the method
description. This is especially useful when using Apipie to auto-generate a machine-readable Swagger
definition of your API (see the swagger section for more details).
A returns
statement has several possible formats:
# format #1: reference to a param-group
returns <param-group-name> [, :code => <number>|<http-response-code-symbol>] [, :desc => <human-readable description>]
# format #2: inline response definition
returns :code => <number>|<http-response-code-symbol> [, :desc => <human-readable description>] do
# property ...
# property ...
# param_group ...
end
# format #3: describing an array-of-objects response
returns :array_of => <param-group-name> [, :code => <number>|<http-response-code-symbol>] [, :desc => <human-readable description>]
If the :code
argument is ommitted, 200
is used.
# ------------------------------------------------
# Example of format #1 (reference to param-group):
# ------------------------------------------------
# the param_group :pet is defined here to describe the output returned by the method below.
def_param_group :pet do
property :pet_name, String, :desc => "Name of pet"
property :animal_type, ['dog','cat','iguana','kangaroo'], :desc => "Type of pet"
end
api :GET, "/pets/:id", "Get a pet record"
returns :pet, :desc => "The pet"
def show_detailed
render JSON({:pet_name => "Skippie", :animal_type => "kangaroo"})
end
# ------------------------------------------------
# Example of format #2 (inline):
# ------------------------------------------------
api :GET, "/pets/:id/with-extra-details", "Get a detailed pet record"
returns :code => 200, :desc => "Detailed info about the pet" do
param_group :pet
property :num_legs, Integer, :desc => "How many legs the pet has"
end
def show
render JSON({:pet_name => "Barkie", :animal_type => "iguana", :legs => 4})
end
# ------------------------------------------------
# Example of format #3 (array response):
# ------------------------------------------------
api :GET, "/pets", "Get all pet records"
returns :array_of => :pet, :code => 200, :desc => "All pets"
def index
render JSON([ {:pet_name => "Skippie", :animal_type => "kangaroo"},
{:pet_name => "Woofie", :animal_type => "cat"} ])
end
Note the use of the property
keyword rather than param
. This is the
preferred mechanism for documenting response-only fields.
property
is very similar to param
with the following differences:
- a
property
is:only_in => :response
by default - a
property
is:required => :true
by default - a
property
can be an:array_of
objects
property :example, :array_of => Hash do
property :number1, Integer
property :number2, Integer
end
To describe multiple possible return codes, the :returns
keyword can be repeated as many times as necessary
(once for each return code). Each one of the :returns
entries can specify a different response format.
api :GET, "/pets/:id/extra_info", "Get extra information about a pet"
returns :desc => "Found a pet" do
param_group :pet
property 'pet_history', Hash do
param_group :pet_history
end
end
returns :code => :unprocessable_entity, :desc => "Fleas were discovered on the pet" do
param_group :pet
property :num_fleas, Integer, :desc => "Number of fleas on this pet"
end
def show_extra_info
# ... implementation here
end
In many cases (such as CRUD implementations), the output from certain API calls is very similar - but not identical - to the inputs of the same or other API calls.
If you already have a :param_group
that defines the input to a create or update routine, it would be quite
frustrating to have to define a completely separate :param_group
to describe the output of the show routine.
To address such situations, it is possible to define a single :param_group
which combines param
and property
statements (as well as :only_in => :request
/ :only_in => :response
) to differentiate between fields that are
only expected in the request, only included in the response, or common to both.
This is somewhat analogous to the way Action Aware params work.
def_param_group :user_record
param :name, String # this is commong to both the request and the response
param :force_update, [true, false], :only_in => :request # this does not show up in responses
property :last_login, String # this shows up only in the response
end
api :POST, "/users", "Create a user"
param_group :user_record # the :last_login field is not expected here, but :force_update is
def create
# ...
end
api :GET, "/users", "Create a user"
returns :array_of => :user_record # the :last_login field will be included in the response, but :force_update will not
def index
# ...
end
If the code creating JSON responses is encapsulated within dedicated classes, it can be more convenient to place the response descriptions outside of the controller and embed them within the response generator.
To support such use cases, Apipie allows any class to provide a describe_own_properties class method which returns a description of the properties such a class would expose. It is then possible to specify that class in the returns statement instead of a param_group.
The describe_own_properties method is expected to return an array of Apipie::prop objects, each one describing a single property.
class Pet
# this method is automatically called by Apipie when Pet is specified as the returned object type
def self.describe_own_properties
[
Apipie::prop(:pet_name, 'string', {:description => 'Name of pet', :required => false}),
Apipie::prop(:animal_type, 'string', {:description => 'Type of pet', :values => ["dog", "cat", "iguana", "kangaroo"]}),
Apipie::additional_properties(false) # this indicates that :pet_name and :animal_type are the only properties in the response
]
end
# this method w
def json
JSON({:pet_name => @name, :animal_type => @type })
end
end
class PetsController
api :GET, "/index", "Get all pets"
returns :array_of => Pet # Pet is a 'self-describing-class'
def index
# ...
end
end
A use case where this is very useful is when JSON generation is done using a reflection mechanism or some other sort of declarative mechanism.
The Apipie::prop function expects the following inputs:
Apipie::prop(<property-name>, <property-type>, <options-hash> [, <array of sub-properties>])
# property-name should be a symbol
#
# property-type can be any of the following strings:
# "integer": maps to a swagger "integer" with an "int32" format
# "long": maps to a swagger "integer" with an "int64" format
# "number": maps to a swagger "number"(no format specifier)
# "float": maps to a swagger "number" with a "float" format
# "double": maps to a swagger "number" with a "double" format
# "string": maps to a swagger "string" (no format specifier)
# "byte": maps to a swagger "string" with a "byte" format
# "binary": maps to a swagger "string" with a "binary" format
# "boolean": maps to a swagger "boolean" (no format specifier)
# "date": maps to a swagger "string" with a "date" format
# "dateTime": maps to a swagger "string" with a "date-time" format
# "password": maps to a swagger "string" with a "password" format
# "object": the property has sub-properties. include <array of sub-properties> in the call.
# (see https://github.com/OAI/OpenAPI-Specification/blob/master/versions/2.0.md#data-types for more information
# about the mapped swagger types)
#
# options-hash can include any of the options fields allowed in a :returns statement.
# additionally, it can include the ':is_array => true', in which case the property is understood to be
# an array of the described type.
To describe an embedded object:
#
# PetWithMeasurements is a self-describing class with an embedded object
#
class PetWithMeasurements
def self.describe_own_properties
[
Apipie::prop(:pet_name, 'string', {:description => 'Name of pet', :required => false}),
Apipie::prop('animal_type', 'string', {:description => 'Type of pet', :values => ["dog", "cat", "iguana", "kangaroo"]}),
Apipie::prop(:pet_measurements, 'object', {}, [
Apipie::prop(:weight, 'number', {:description => "Weight in pounds" }),
Apipie::prop(:height, 'number', {:description => "Height in inches" }),
Apipie::prop(:num_legs, 'number', {:description => "Number of legs", :required => false }),
Apipie::additional_properties(false)
])
]
end
end
#
# PetWithManyMeasurements is a self-describing class with an embedded array of objects
#
class PetWithManyMeasurements
def self.describe_own_properties
[
Apipie::prop(:pet_name, 'string', {:description => 'Name of pet', :required => false}),
Apipie::prop(:many_pet_measurements, 'object', {is_array: true}, [
Apipie::prop(:weight, 'number', {:description => "Weight in pounds" }),
Apipie::prop(:height, 'number', {:description => "Height in inches" }),
])
]
end
end
Sometimes, the actions are not defined in the controller class
directly but included from a module instead. You can load the Apipie
DSL into the module by extending it with Apipie::DSL::Concern
.
The module can be used in more controllers. Therefore there is a way to
substitute parts of the documentation in the module with controller
specific values. These substitutions can be stated explicitly with
apipie_concern_subst(:key => "value")
(needs to be called before
the module is included to take effect). The substitutions are
performed in the paths and descriptions of APIs and names and descriptions
of params.
There are some default substitutions available:
- :controller_path
- value of
controller.controller_path
, e.g.api/users
forApi::UsersController
. Only if not using theapi!
keyword. - :resource_id
- Apipie identifier of the resource, e.g.
users
forApi::UsersController
or set byresource_id
# users_module.rb
module UsersModule
extend Apipie::DSL::Concern
api :GET, '/:controller_path', 'List :resource_id'
def index
# ...
end
api! 'Show a :resource'
def show
# ...
end
api :POST, '/:resource_id', "Create a :resource"
param :concern, Hash, :required => true
param :name, String, 'Name of a :resource'
param :resource_type, ['standard','vip']
end
def create
# ...
end
api :GET, '/:resource_id/:custom_subst'
def custom
# ...
end
end
# users_controller.rb
class UsersController < ApplicationController
resource_description { resource_id 'customers' }
apipie_concern_subst(:custom_subst => 'custom', :resource => 'customer')
include UsersModule
# the following paths are documented
# api :GET, '/users'
# api :GET, '/customers/:id', 'Show a customer'
# api :POST, '/customers', 'Create a customer'
# param :customer, :required => true do
# param :name, String, 'Name of a customer'
# param :customer_type, ['standard', 'vip']
# end
# api :GET, '/customers/:custom'
end
Sometimes, it's needed to extend an existing controller method with additional parameters (usually when extending exiting API from plugins/rails engines). The concern can be also used for this purposed, using update_api method. The params defined in this block are merged with the params of the original method in the controller this concern is included to.
module Concerns
module OauthConcern
extend Apipie::DSL::Concern
update_api(:create, :update) do
param :user, Hash do
param :oauth, String, :desc => 'oauth param'
end
end
end
end
The concern needs to be included to the controller after the methods are defined
(either at the end of the class, or by using
Controller.send(:include, Concerns::OauthConcern)
.
The swagger definitions created by Apipie can be used to auto-generate clients that access the described APIs. Those clients will break if the responses returned from the API do not match the declarations. As such, it is very important to include unit tests that validate the actual responses against the swagger definitions.
The implemented mechanism provides two ways to include such validations in RSpec unit tests: manual (using an RSpec matcher) and automated (by injecting a test into the http operations 'get', 'post', raising an error if there is no match).
require 'apipie/rspec/response_validation_helper'
RSpec.describe MyController, :type => :controller, :show_in_doc => true do
describe "GET stuff with response validation" do
render_views # this makes sure the 'get' operation will actually
# return the rendered view even though this is a Controller spec
it "does something" do
response = get :index, {format: :json}
# the following expectation will fail if the returned object
# does not match the 'returns' declaration in the Controller,
# or if there is no 'returns' declaration for the returned
# HTTP status code
expect(response).to match_declared_responses
end
end
end
require 'apipie/rspec/response_validation_helper'
RSpec.describe MyController, :type => :controller, :show_in_doc => true do
describe "GET stuff with response validation" do
render_views
auto_validate_rendered_views
it "does something" do
get :index, {format: :json}
end
it "does something else" do
get :another_index, {format: :json}
end
end
describe "GET stuff without response validation" do
it "does something" do
get :index, {format: :json}
end
it "does something else" do
get :another_index, {format: :json}
end
end
end
Create a configuration file in e.g. /config/initializers/apipie.rb
.
You can set the application name, footer text, API and documentation base URL
and turn off validations. You can also choose your favorite markup language
for full descriptions.
- app_name
- Name of your application; used in breadcrumbs navigation.
- copyright
- Copyright information (shown in page footer).
- compress_examples
- If
true
recorded examples are compressed usingZlib
. Useful for big test-suits. - doc_base_url
- Documentation frontend base url.
- api_base_url
- Base url for default version of your API. To set it for specific version use
config.api_base_url[version] = url
. - default_version
- Default API version to be used (1.0 by default)
- validate
- Parameters validation is turned off when set to false. When set to
:explicitly
, you must invoke parameter validation yourself by calling controller methodapipie_validations
(typically in a before_action). When set to:implicitly
(or just true), your controller's action methods are wrapped with generated methods which callapipie_validations
, and then call the action method. (:implicitly
by default) - validate_value
- Check the value of params against specified validators (true by default)
- validate_presence
- Check the params presence against the documentation.
- validate_key
- Check the received params to ensure they are defined in the API. (false by default)
- process_params
- Process and extract the parameter defined from the params of the request to the api_params variable
- app_info
- Application long description.
- reload_controllers
- Set to enable/disable reloading controllers (and the documentation with it). Enabled by default in development.
- api_controllers_matcher
- For reloading to work properly you need to specify where your API controllers are. Can be an array if multiple paths are needed
- api_routes
- Set if your application uses a custom API router, different from the Rails default
- routes_formatter
- An object providing the translation from the Rails routes to the
format usable in the documentation when using the api! keyword. By
default, the
Apipie::RoutesFormatter
is used. - markup
- You can choose markup language for descriptions of your application, resources and methods. RDoc is the default but you can choose from Apipie::Markup::Markdown.new or Apipie::Markup::Textile.new. In order to use Markdown you need Maruku gem and for Textile you need RedCloth. Add those to your gemfile and run bundle if you want to use them. You can also add any other markup language processor.
- layout
- Name of a layout template to use instead of Apipie's layout. You can use Apipie.include_stylesheets and Apipie.include_javascripts helpers to include Apipie's stylesheets and javascripts.
- ignored
- An array of controller names (strings) (might include actions as well)
to be ignored when generationg the documentation
e.g.
%w[Api::CommentsController Api::PostsController#post]
- namespaced_resources
- Use controller paths instead of controller names as resource id. This prevents same named controllers overwriting each other.
- authenticate
- Pass a proc in order to authenticate user. Pass nil for no authentication (by default).
- authorize
- Pass a proc in order to authorize controllers and methods. The Proc is evaluated in the controller context.
- show_all_examples
- Set this to true to set show_in_doc=1 in all recorded examples
- link_extension
- The extension to use for API pages ('.html' by default). Link extensions in static API docs cannot be changed from '.html'.
- languages
- List of languages the API documentation should be translated into. Empty by default.
- default_locale
- Locale used for generating documentation when no specific locale is set. Set to 'en' by default.
- locale
- Pass locale setter/getter
config.locale = lambda { |loc| loc ? FastGettext.set_locale(loc) : FastGettext.locale }
- translate
- Pass proc to translate strings using the localization library your project uses. For example see Localization
Example:
Apipie.configure do |config|
config.app_name = "Test app"
config.copyright = "© 2012 Pavel Pokorny"
config.doc_base_url = "/apidoc"
config.api_base_url = "/api"
config.validate = false
config.markup = Apipie::Markup::Markdown.new
config.reload_controllers = Rails.env.development?
config.api_controllers_matcher = File.join(Rails.root, "app", "controllers", "**","*.rb")
config.api_routes = Rails.application.routes
config.app_info["1.0"] = "
This is where you can inform user about your application and API
in general.
"
config.authenticate = Proc.new do
authenticate_or_request_with_http_basic do |username, password|
username == "test" && password == "supersecretpassword"
end
end
config.authorize = Proc.new do |controller, method, doc|
!method # show all controller doc, but no method docs.
end
end
- checksum_path
- Used in ChecksumInHeaders middleware (see JSON checksums for more info). It contains path prefix(es) where the header with checksum is added. If set to nil, checksum is added in headers in every response. e.g.
%w[/api /apipie]
- update_checksum
- If set to true, the checksum is recalculated with every documentation_reload call
Apipie is able to load the information about the paths based on the routes defined in the Rails application, by using the api! keyword in the DSL.
It should be usable out of box, however, one might want
to do some customization (such as omitting some implicit parameters in
the path etc.). For this kind of customizations one can create a new
formatter and pass as the Apipie.configuration.routes_formatter
option, like this:
class MyFormatter < Apipie::RoutesFormatter
def format_path(route)
super.gsub(/\(.*?\)/, '').gsub('//','') # hide all implicit parameters
end
end
Apipie.configure do |config|
...
config.routes_formatter = MyFormatter.new
...
end
A similar way can be used to influence things like order, or a description of the loaded APIs, even omitting some paths if needed.
The goal is to extract and pre-process parameters of the request.
For example Rails, by default, transforms an empty array to nil value. Perhaps you want to transform it again into an empty array. Or you want to support an enumeration type (comma separated values) and you want to automatically transform this string into an array.
To use it, set the process_params
configuration variable to true.
Also by using as
you can separate your API parameter
names from the names you are using inside your code.
To implement it, you just have to write a process_value function in your validator:
For an enumeration type:
def process_value(value)
value ? value.split(',') : []
end
Every parameter needs to have an associated validator. For now there are some basic validators. You can always provide your own to achieve complex results.
If validations are enabled (default state) the parameters of every request are validated. If the value is wrong an +ArgumentError+ exception is raised and can be rescued and processed. It contains a description of the parameter value expectations. Validations can be turned off in the configuration file.
Parameter validation normally happens after before_actions, just before
your controller method is invoked. If you prefer to control when parameter
validation occurs, set the configuration parameter validate
to :explicitly
.
You must then call the apipie_validations
method yourself, e.g.:
before_action :apipie_validations
This is useful if you have before_actions which use parameter values: just add them
after the apipie_validations
before_action.
Check the parameter type. Only String, Hash and Array are supported for the sake of simplicity. Read more to find out how to add your own validator.
param :session, String, :desc => "user is logged in", :required => true
param :facts, Hash, :desc => "Additional optional facts about the user"
Check parameter value against given regular expression.
param :regexp_param, /^[0-9]* years/, :desc => "regexp param"
Check if parameter value is included in the given array.
param :enum_param, [100, "one", "two", 1, 2], :desc => "enum validator"
If you need more complex validation and you know you won't reuse it, you can use the Proc/lambda validator. Provide your own Proc, taking the value of the parameter as the only argument. Return true if value passes validation or return some text about what is wrong otherwise. _Don't use the keyword return if you provide an instance of Proc (with lambda it is ok), just use the last statement return property of ruby.
param :proc_param, lambda { |val|
val == "param value" ? true : "The only good value is 'param value'."
}, :desc => "proc validator"
You can describe hash parameters in depth if you provide a block with a description of nested values.
param :user, Hash, :desc => "User info" do
param :username, String, :desc => "Username for login", :required => true
param :password, String, :desc => "Password for login", :required => true
param :membership, ["standard","premium"], :desc => "User membership"
end
In fact there isn't any NilValidator, but setting it to nil can be used to override parameters described on the resource level.
param :user, nil
def destroy
#...
end
Check if the parameter is a positive integer number or zero
param :product_id, :number, :desc => "Identifier of the product", :required => true
param :quantity, :number, :desc => "Number of products to order", :required => true
Check if the parameter is a decimal number
param :latitude, :decimal, :desc => "Geographic latitude", :required => true
param :longitude, :decimal, :desc => "Geographic longitude", :required => true
Check if the parameter is an array
- of
- Specify the type of items. If not given it accepts an array of any item type
- in
- Specify an array of valid item values.
Assert things is an array of any items
param :things, Array
Assert hits must be an array of integer values
param :hits, Array, of: Integer
Assert colors must be an array of valid string values
param :colors, Array, in: ["red", "green", "blue"]
The retrieving of valid items can be deferred until needed using a lambda. It is evaluated only once
param :colors, Array, in: -> { Color.all.pluck(:name) }
You can describe nested parameters in depth if you provide a block with a description of nested values.
param :comments, Array, :desc => "User comments" do
param :name, String, :desc => "Name of the comment", :required => true
param :comment, String, :desc => "Full comment", :required => true
end
Only basic validators are included but it is really easy to add your own.
Create a new initializer with a subclass of Apipie::Validator::BaseValidator.
Two methods are required to implement this - instance method
validate(value)
and class method
build(param_description, argument, options, block)
.
When searching for the validator +build+ method, every subclass of Apipie::Validator::BaseValidator is called. The first one that returns the constructed validator object is used.
Example: Adding IntegerValidator
We want to check if the parameter value is an integer like this:
param :id, Integer, :desc => "Company ID"
So we create apipie_validators.rb initializer with this content:
class IntegerValidator < Apipie::Validator::BaseValidator
def initialize(param_description, argument)
super(param_description)
@type = argument
end
def validate(value)
return false if value.nil?
!!(value.to_s =~ /^[-+]?[0-9]+$/)
end
def self.build(param_description, argument, options, block)
if argument == Integer || argument == Fixnum
self.new(param_description, argument)
end
end
def description
"Must be #{@type}."
end
end
Parameters of the build method:
- param_description
- Instance of Apipie::ParamDescription contains all given information about the validated parameter.
- argument
- Specified validator; in our example it is +Integer+
- options
- Hash with specified options, for us just
{:desc => "Company ID"}
- block
- Block converted into Proc, use it as you desire. In this example nil.
Every resource/method can belong to one or more versions. The version is specified with the api_version DSL keyword. When not specified, the resource belongs to config.default_version ("1.0" by default)
resource_description do
api_versions "1", "2"
end
api :GET, "/api/users/", "List: users"
api_version "1"
def index
# ...
end
api :GET, "/api/users/", "List: users", :deprecated => true
In the example above we say the whole controller/resource is defined for versions "1" and "2", but we override this by explicitly saying index belongs only to version "1". Also, inheritance works (therefore we can specify the api_version for the parent controller, and all children will know about that). Routes can be flagged as deprecated, and an annotation will be added to them when viewing in the API documentation.
From the Apipie API perspective, the resources belong to the version. With versioning, there are paths like this provided by apipie:
/apipie/1/users/index /apipie/2/users/index
When not specifying the version explicitly in the path (or in DSL), default version (Apipie.configuration.default_version) is used instead ("1.0" by default). Therefore, an application that doesn't need versioning should work as before.
The static page generator takes a version parameter (or uses default).
You can specify the versions for the examples, with the versions keyword. It specifies the versions the example is used for. When not specified, it's shown in all versions with the given method.
When referencing or quering the resource/method descripion, this format should be used: "version#resource#method". When not specified, the default version is used instead.
The default markup language is RDoc. It can be changed in
the config file (config.markup=
) to one of these:
- Markdown
- Use Apipie::Markup::Markdown.new. You need Maruku gem.
- Textile
- Use Apipie::Markup::Textile.new. You need RedCloth gem.
Or provide you own object with a to_html(text)
method.
For inspiration, this is how Textile markup usage is implemented:
class Textile
def initialize
require 'RedCloth'
end
def to_html(text)
RedCloth.new(text).to_html
end
end
Apipie has support for localized API documentation in both formats (JSON and HTML).
Apipie uses the library I18n for localization of itself.
Check config/locales
directory for available translations.
A major part of strings in the documentation comes from the API. As preferences regarding localization libraries differ amongst project, Apipie needs to know how to set the locale for your project, and how to translate a string using the library your project uses. That can be done using lambdas in configuration.
Sample configuration when your project uses FastGettext
Apipie.configure do |config|
...
config.languages = ['en', 'cs']
config.default_locale = 'en'
config.locale = lambda { |loc| loc ? FastGettext.set_locale(loc) : FastGettext.locale }
config.translate = lambda do |str, loc|
old_loc = FastGettext.locale
FastGettext.set_locale(loc)
trans = _(str)
FastGettext.set_locale(old_loc)
trans
end
end
And the strings in the API documentation need to be marked with the N_()
function
api :GET, "/users/:id", N_("Show user profile")
param :session, String, :desc => N_("user is logged in"), :required => true
When your project use I18n, localization related configuration could appear as follows
Apipie.configure do |config|
...
config.languages = ['en', 'cs']
config.default_locale = 'en'
config.locale = lambda { |loc| loc ? I18n.locale = loc : I18n.locale }
config.translate = lambda do |str, loc|
return '' if str.blank?
I18n.t str, locale: loc, scope: 'doc'
end
end
And the strings in the API documentation needs to be in the form of translation keys
api :GET, "/users/:id", "show_user_profile"
param :session, String, :desc => "user_is_logged_in", :required => true
The localized versions of the documentation are distinguished by language in the filename.
E.g. doc/apidoc/apidoc.cs.html
is static documentation in the Czech language.
If the language is missing, e.g. doc/apidoc/apidoc.html
,
the documentation is localized with the default_locale
.
The dynamic documentation follows the same schema. The http://localhost:3000/apidoc/v1.cs.html
is documentation for version '1' of the API in the Czech language. For JSON descriptions, the API applies the same format: http://localhost:3000/apidoc/v1.cs.json
To modify the views of your documentation, run rails g apipie:views
.
This will copy the Apipie views to app/views/apipie/apipies
and
app/views/layouts/apipie
.
To generate a static version of documentation (perhaps to put it on
your project site or something), run the rake apipie:static
task. It will
create a set of HTML files (multi-pages, single-page, plain) in your doc
directory. If you prefer a JSON version run rake apipie:static_json
.
By default the documentation for the default API version is
used. You can specify the version with rake apipie:static[2.0]
When you want to avoid any unnecessary computation in production mode,
you can generate a cache with rake apipie:cache
and configure the
app to use it in production with config.use_cache = Rails.env.production?
Default cache dir is File.join(Rails.root, "public", "apipie-cache")
,
you can change it to where you want, example: config.cache_dir = File.join(Rails.root, "doc", "apidoc")
.
If, for some complex cases, you need to generate/re-generate just part of the cache
use rake apipie:cache cache_part=index
resp. rake apipie:cache cache_part=resources
To generate it for different locations for further processing use rake apipie:cache OUT=/tmp/apipie_cache
.
To generate a static Swagger definition file from the api, run rake apipie:static_swagger_json
.
By default the documentation for the default API version is
used. You can specify the version with rake apipie:static_swagger_json[2.0]
. A swagger file will be
generated for each locale. The files will be generated in the same location as the static_json files, but
instead of being named schema_apipie[.locale].json
, they will be called schema_swagger[.locale].json
.
Swagger allows method definitions to include an indication of the the default value for each parameter. To include such
indications, use :default_value => <some value>
in the parameter definition DSL. For example:
param :do_something, Boolean, :desc => "take an action", :required => false, :default_value => false
The help identify potential improvements to your documentation, the swagger generation process issues warnings if it identifies various shortcomings of the DSL documentation. Each warning has a code to allow selective suppression (see swagger-specific configuration below)
100: | missing short description for method |
---|---|
101: | added missing / at beginning of path |
102: | no return codes specified for method |
103: | a parameter is a generic Hash without an internal type specification |
104: | a parameter is an 'in-path' parameter, but specified as 'not required' in the DSL |
105: | a parameter is optional but does not have a default value specified |
106: | a parameter was ommitted from the swagger output because it is a Hash without fields in a formData specification |
107: | a path parameter is not described |
108: | inferring that a parameter type is boolean because described as an enum with [false,true] values |
There are several configuration parameters that determine the structure of the generated swagger file:
config.swagger_content_type_input
If the value is
:form_data
- the swagger file will indicate that the server consumes the content typesapplication/x-www-form-urlencoded
andmultipart/form-data
. Non-path parameters will have the value"in": "formData"
. Note that parameters of type Hash that do not have any fields in them will be ommitted from the resulting files, as there is no way to describe them in swagger.If the value is
:json
- the swagger file will indicate that the server consumes the content typeapplication/json
. All non-path parameters will be included in the schema of a single"in": "body"
parameter of typeobject
.You can specify the value of this configuration parameter as an additional input to the rake command (e.g.,
rake apipie:static_swagger_json[2.0,form_data]
).config.swagger_json_input_uses_refs
This parameter is only relevant if
swagger_content_type_input
is:json
.If
true
: the schema of the"in": "body"
parameter of each method is given its own entry in thedefinitions
section, and is referenced using$ref
from the method definition.If
false
: the body parameter definitions are inlined within the method definitions.config.swagger_include_warning_tags
- If
true
: in addition to tagging methods with the name of the resource they belong to, methods for which warnings have been issued will be tagged with. config.swagger_suppress_warnings
If
false
: no warnings will be suppressedIf
true
: all warnings will be suppressedIf an array of values (e.g.,
[100,102,107]
), only the warnings identified by the numbers in the array will be suppressed.config.swagger_api_host
The value to place in the swagger host field.
Default is
localhost:3000
If
nil
then then host field will not be included.config.swagger_allow_additional_properties_in_response
If
false
(default): response descriptions in the generated swagger will include anadditional-properties: false
fieldIf
true
: theadditional-properties: false
field will not be included in response object descriptions
- There is currently no way to document the structure and content-type of the data returned from a method
- Recorded examples are currently not included in the generated swagger file
- The apipie
formats
value is ignored. - It is not possible to specify the "consumed" content type on a per-method basis
- It is not possible to leverage all of the parameter type/format capabilities of swagger
- Only OpenAPI 2.0 is supported
- Responses are defined inline and not as a $ref
To generate swagger dynamically, use http://localhost:3000/apipie.json?type=swagger
.
Note that authorization is not supported for dynamic swagger generation, so if config.authorize
is defined,
dynamic swagger generation will be disabled.
Dynamically generated swagger is not cached, and is always generated on the fly.
If the API client needs to be sure that the JSON didn't changed, add
the ApipieChecksumInHeaders
middleware in your rails app.
It can add a checksum of the entire JSON document in the response headers.
"Apipie-Checksum"=>"fb81460e7f4e78d059f826624bdf9504"
Apipie bindings uses this feature to refresh its JSON cache.
To set it up add the following to your application.rb
require 'apipie/middleware/checksum_in_headers' # Add JSON checksum in headers for smarter caching config.middleware.use "Apipie::Middleware::ChecksumInHeaders"
And in your apipie initializer allow checksum calculation
Apipie.configuration.update_checksum = true
By default the header is added to responses for config.doc_base_url
and /api
.
It can be changed in configuration (see Configuration Reference for details).
The checksum calculation is lazy, and done with the first request. If you run with use_cache = true
,
do not forget to run the rake task apipie:cache
.
Apipie integrates with automated testing in two ways. Documentation bootstrapping and examples recording.
Let's say you have an application without REST API documentation. However you have a set of tests that are run against this API. A lot of information is already included in these tests, it just needs to be extracted somehow. Luckily, Apipie provides such a feature.
When running the tests, set the APIPIE_RECORD=params
environment
variable or call Apipie.record('params')
from specs starter. You can either use it with functional tests:
APIPIE_RECORD=params rake test:functionals
or you can run your server with this param, in case you run the tests against running server:
APIPIE_RECORD=params rails server
When the process quits, the data from requests/responses are used to determine the documentation. It's quite raw, but it makes the initial phase much easier.
You can also use the tests to generate up-to-date examples for your
code. Similar to the bootstrapping process, you can use it with functional
tests or a running server, setting APIPIE_RECORD=examples
or calling Apipie.record('examples')
in your specs starter.
APIPIE_RECORD=examples rake test:functionals APIPIE_RECORD=examples rails server
The data is written into doc/apipie_examples.yml
. By default,
only the first example is shown for each action. You can customize
this by setting the show_in_doc
attribute at each example.
You can add a title to the examples (useful when showing more than one example per method) by adding a 'title' attribute.
--- !omap - announcements#index: - !omap - title: This is a custom title for this example - verb: :GET - path: /api/blabla/1 - versions: - '1.0' - query: - request_data: - response_data: ... - code: 200 - show_in_doc: 1 # If 1, show. If 0, do not show. - recorded: true
In RSpec you can add metadata to examples. We can use that feature to mark selected examples - the ones that perform the requests that we want to show as examples in the documentation.
For example, we can add show_in_doc
to examples, like this:
describe "This is the correct path" do
it "some test", :show_in_doc do
....
end
end
context "These are edge cases" do
it "Can't authenticate" do
....
end
it "record not found" do
....
end
end
And then configure RSpec in this way:
RSpec.configure do |config|
config.treat_symbols_as_metadata_keys_with_true_values = true
config.filter_run :show_in_doc => true if ENV['APIPIE_RECORD']
end
This way, when running in recording mode, only the tests that have been marked with the
:show_in_doc
metadata will be run, and hence only those will be used as examples.
Make sure to enable config.render_views
in your config/rails_helper.rb
or
config/spec_helper.rb
if you're using jbuilder, or you will get back empty results
In earlier versions (<= 0.0.13), there was a simple client generator as a part of Apipie gem. As more features and users came to Apipie, there was a greater need for changes on a per project basis. It's hard (or even impossible) to provide a generic solution for the client code. We also don't want to tell you what's the right way to do it (what gems to use, how the API should look like etc.).
Therefore you can't generate client code directly by a rake task in further versions.
There is, however, an even better and more flexible way to reuse your API documentation for this purpose: using the API the Apipie provides in the generator code. Check out our sister project apipie-bindings, as they use exactly this approach. You also don't need to run the service, provided it uses Apipie as a backend.
And if you write one on your own, don't hesitate to share it with us!
You can setup Disqus discussion within your documentation. Just set the credentials in the Apipie configuration:
config.disqus_shortname = "MyProjectDoc"
- Getting started tutorial - including examples of using the tests integration and versioning.
- Real-world application usage
- Read-world application usage with versioning
- Using Apipie API to generate bindings