HALPresenter is a DSL for creating serializers conforming to JSON HAL. This DSL is highly influenced by ActiveModelSerializers.
Check out this benchmark for a comparison to other serializers.
So, generating some json from an object, whats the big deal? Well if your API is not driven by hypermedia and your payloads most of the time just looks the same, then this might be overkill. But if you do have dynamic payloads (e.g the payload attributes and links depend on the context) then this gem greatly simplifies serialization and puts all the serialization logic in one place.
This documentation might be a bit long and dull, but skim through it and check out the examples. I think you'll get the hang of it.
gem install hal_presenter
With Gemfile:
gem 'hal_presenter'
And then execute:
$ bundle
Lets start with an example. Say you have your typical blog and you want to serialize post resources. Posts have some text, an author and possibly some comments. Only the author of the post may edit or delete it. A serializer could then be written as:
class PostSerializer
extend HALPresenter
model Post
attribute :text
attribute :characters do
resource.text.size
end
link :self do
"/posts/#{resource.id}"
end
link :author do
"/users/#{resource.author.id}"
end
link :edit do
"/posts/#{resource.id}/edit" if resource.author.id == options[:current_user]
end
link :delete do
"/posts/#{resource.id}" if resource.author.id == options[:current_user]
end
embed :recent_comments
end
Instances of Post
can now be serialized with HALPresenter.to_hal(post, current_user: some_user)
which will produce the following (assuming the current user is the author of the post. Else the edit/delete links would not be present):
{
"text": "some very important stuff",
"characters": 25,
"_links": {
"self": {
"href": "/posts/5"
},
"author": {
"href": "/users/8"
},
"edit": {
"href": "/posts/5/edit"
},
"delete": {
"href": "/posts/5"
}
},
"_embedded": {
"recent_comments": {
"count": 2,
"_links": {
"self": {
"href": "/posts/5/recent_comments"
}
},
"_embedded": {
"comments": [
{
"comment": "lorem ipsum",
"_links": {
"self": {
"href": "/posts/5/comment/1"
}
}
},
{
"comment": "dolor sit",
"_links": {
"self": {
"href": "/posts/5/comment/2"
}
}
}
]
}
}
}
}
Note: In the output above, recent_comments
is a collection and collections have their items embedded. Since this collection is
embedded in the post resource that's why we get two levels of embedded resources.
Serializers are defined by extending HALPresenter
in the begining of the class declaration. This will add the following class methods:
::model(clazz)
::attribute(name, value = nil, embed_depth: nil, &block)
::link(rel, value = nil, **options, &block)
::curie(rel, value = nil, embed_depth: nil, &block)
::profile(value = nil, **kwargs, &block)
::policy(clazz)
::namespace(curie, &block)
::embed(name, value = nil, embed_depth: nil, presenter_class: nil, &block)
::collection
::to_hal(resource = nil, **options)
::to_collection(resources = [], **options)
::post_serialize(&block)
::from_hal(payload, resource = nil)
The model
class method is used to register the resource Class that this serializer handles. (There's no Rails magic that automagically maps models to serializers.)
class PostSerializer
extend HALPresenter
model Post
end
This make it possible to serialize Post
instances with HALPresenter.to_hal
. HALPresenter will then lookup the right presenter and delegate the serialization to it
(which in the case above would be PostSerializer
).
post = Post.new(*args)
HALPresenter.to_hal(post)
If a model does not have it's own presenter but one of its superclasses does, then that will be used.
class SubPost < Post; end
sub_post = SubPost.new(*args)
HALPresenter.to_hal(sub_post) # will lookup PostSerializer since there isn't a specific one for SubPost
Using the model
class method is not required for serialization. The serializer can also be called directly.
PostSerializer.to_hal(post)
The serializer class may also be specified as an option, using the :presenter
key.
HALPresenter.to_hal(post, {presenter: PostSerializer})
Even though the model
class method is optional, it is very useful if the serializer should be selected dynamically and when the serializer is used for deserialization.
The attribute
class method specifies an attribute (property) to be serialized. The first argument, name
, is required and specifies the name of the attribute. When ::attribute
is called with only one argument, the resources being serialized are expected to respond to that argument and the returned value is what ends up in the payload.
class PostSerializer
extend HALPresenter
attribute :title
end
post = OpenStruct.new(title: "hello")
PostSerializer.to_hal(post) # => {"title": "hello"}
If ::attribute
is called with two arguments, then the second arguments is what ends up in the payload.
class PostSerializer
extend HALPresenter
attribute :title, "world"
end
post = OpenStruct.new(title: "ignored")
PostSerializer.to_hal(post) # => {"title": "world"}
When a block is passed to ::attribute
, then the return value of that block is what ends up in the payload.
class PostSerializer
extend HALPresenter
attribute :title do
resource.title.upcase
end
end
post = OpenStruct.new(title: "hello")
PostSerializer.to_hal(post) # => {"title": "HELLO"}
Notice that the object being serialized (post
in the above example) is accessible inside the block by the resource
method.
The keyword argument :embed_depth
may be specified to set a max allowed nesting depth for the corresponding attribute to be serialized. See embed_depth
.
The link
class method specifies a link to be added to the _links property. The first argument, rel
, is required. ::link
must be called with either a second argument (value
) or a block.
class PostSerializer
extend HALPresenter
link :self, '/posts/1'
end
PostSerializer.to_hal # => {"_links": {"self": {"href": "/posts/1"}}}
When a block is passed to ::link
, the block must evaluate to either nil
, a String containing the href or a hash with symbol keys.
class PostSerializer
extend HALPresenter
link :self do
"/posts/#{resource.id}"
end
end
post = OpenStruct.new(id: 5)
PostSerializer.to_hal(post) # => {"_links": {"self": {"href": "/posts/5"}}}
When the block returns a hash the following keys will be used :href
, :type
, :deprecation
, :profile
, :title
and :templated
.
class PostSerializer
extend HALPresenter
link :other do
{
href: "/posts/#{resource.id}",
title: resource.title,
type: resource.type,
}
end
end
post = OpenStruct.new(id: 5, title: 'Foo', type: 'bar')
PostSerializer.to_hal(post) # => {
# "_links": {
# "other": {
# "href": "/posts/5",
# "title": "Foo",
# "type": "bar"
# }
# }
# }
Multiple links can have the same relation. If so, they will be serialized as an array.
class PostSerializer
extend HALPresenter
link :item, "/foo"
link :item do
"/bar"
end
end
PostSerializer.to_hal # =>
# {
# "_links": {
# "item": [
# {
# "href": "/foo"
# },
# {
# "href": "/bar"
# }
# ]
# }
# }
The following options may be given to ::link
:
embed_depth
- sets a max allowed nesting depth for the corresponding link to be serialized. Seeembed_depth
.curie
- prepends a curie to the rel.title
- a string used for labelling the link (e.g. in a user interface).type
- the media type of the resource returned after following this link.deprecation
- a URL providing information about the deprecation of this link.profile
- a URI that hints about the profile of the target resource.
The curie
class method specifies a curie to be added to the curies list. The first argument, rel
, is required. ::curie
must be called with either a second argument (value
) or a block.
class PostSerializer
extend HALPresenter
curie :doc, '/api/docs/{rel}'
link :'doc:user', '/users/5'
end
PostSerializer.to_hal # => {"_links":{"doc:user":{"href":"/users/5"},"curies":[{"name":"doc","href":"/api/docs/{rel}","templated":true}]}}
When a block is passed to ::curie
, the return value of that block is what ends up as the href of the curie.
class PostSerializer
extend HALPresenter
curie :doc do
'/api/docs/{rel}'
end
link :'doc:user', '/users/5'
end
post = OpenStruct.new(id: 5)
PostSerializer.to_hal(post) # => {"_links":{"doc:user":{"href":"/users/5"},"curies":[{"name":"doc","href":"/api/docs/{rel}","templated":true}]}}
The keyword argument :embed_depth
may be specified to set a max allowed nesting depth for the corresponding curie to be serialized. See embed_depth
.
When a resource is embedded in another resource all curies are added to the root resource. This ensures that each curie only appear once in the output.
Note that curies may get renamed if there are conflicts between them. See example in ::embed
for more details.
The profile
class method is used to specify a mediatype profile that the serializer is conforming to. This is optional and does not change anything about how things get serialized.
This method exist only to specify that a given serializer will produce content with semantic meaning described in the specified mediatype profile.
The profile is retrieved with the class method semantic_profile
.
class PostSerializer
extend HALPresenter
profile "foobar"
end
PostSerializer.semantic_profile # => "foobar"
# Or with a block
class PostSerializer
extend HALPresenter
profile do
"foo#{options[:foo]}"
end
end
PostSerializer.semantic_profile(foo: 'bar') # => "foobar"
The policy
class method is used to register a policy class that should be used during serialization. The purpose of using a policy class is to specify rules about which properties should be serialized (depending on the context). E.g hide some attributes and/or links if current_user
is not an admin. If a policy is specified, then by default no properties will be serialized unless the policy explicitly allows them to be serialized.
Using polices is not required, but its a nice way to structure rules about what should be shown and what actions (links) are possible to perform on a resource. The latter is usually tightly coupled with authorization in controllers. This means we can create polices with a bunch of rules and use the same policy in both serialization and in controllers. This plays nicely with gems like Pundit.
Instances of the class registered with this method needs to respond to the following methods:
initialize(current_user, resource, options = {})
attribute?(name)
link?(rel)
embed?(name)
Additional methods will be needed for authorization in controller. Such as create?
, update?
etc when using Pundit.
A policy instance will be instantiated with the resource being serialized and the option :current_user
passed to ::to_hal
. For each attribute being serialized a call to policy_instance.attribute?(name)
will be made. If that call returns true
then the attribute will be serialized. Else it will not end up in the serialized payload. Same goes for links and embedded resources. Curies are ignored by policies and always serialized.
Using the following Policy would discard everything except the title attribute, the self link and only embedded resources if current_user
is an admin user.
class SomePolicy
attr_reader :current_user, :resource, :options
def initialize(current_user, resource, options = {})
@current_user = current_user
@resource = resource
@options = options
end
def attribute?(name)
name.to_s == 'title'
end
def link?(rel)
rel == :self
end
def embed?(_name)
return false unless current_user
current_user.admin?
end
end
This gem includes a DSL that simplifies creating policies. See HALPresenter::Policy::DSL
.
Multiple links and embedded resources may be grouped together inside a curie namespace. This is done by wrapping them inside a block passed to ::namespace
.
class DogSerializer
extend HALPresenter
attribute :name
link :self do
"/dogs/#{resource.name}"
end
end
class PetSerializer
extend HALPresenter
namespace :foo do
link :owner do
"/users/#{resource.owner_id}"
end
embed :animal, presenter_class: DogSerializer
end
curie :foo do
'/some_link/{rel}'
end
end
dog = OpenStruct.new(id: 3, name: 'Milo')
pet = OpenStruct.new(owner_id: 5, animal: dog)
PetSerializer.to_hal(pet) # => {"_links":{"foo:owner":{"href":"/users/5"},"curies":[{"name":"foo","href":"/some_link/{rel}","templated":true}]},"_embedded":{"foo:animal":{"name":"Milo","_links":{"self":{"href":"/dogs/Milo"}}}}}
The embed
class method specifies a nested resource to be embedded. The first argument, name
, is required. When ::embed
is called with only one argument, the resource being serialized is expected to respond to the value of that argument and the returned value is what ends up in the payload. The keyword argument presenter_class
specifies the serializer to be used for serializing the embedded resource.
class UserSerializer
extend HALPresenter
attribute :name
end
class PostSerializer
extend HALPresenter
embed :author, presenter_class: UserSerializer
end
user = OpenStruct.new(name: "bengt")
post = OpenStruct.new(title: "hello", author: user)
PostSerializer.to_hal(post) # => {"_embedded":{"author":{"name":"bengt"}}}
If ::embed
is called with two arguments, then the second arguments is embedded.
class UserSerializer
extend HALPresenter
attribute :name
end
class PostSerializer
extend HALPresenter
embed :author, OpenStruct.new(name: "bengt"), presenter_class: UserSerializer
end
post = OpenStruct.new(title: "hello")
PostSerializer.to_hal(post) # => {"_embedded":{"author":{"name":"bengt"}}}
When a block is passed to ::embed
, then the return value of that block is embedded.
class UserSerializer
extend HALPresenter
attribute :name
end
class PostSerializer
extend HALPresenter
embed :author, presenter_class: UserSerializer do
OpenStruct.new(name: "bengt")
end
end
post = OpenStruct.new(title: "hello")
PostSerializer.to_hal(post) # => {"_embedded":{"author":{"name":"bengt"}}}
The keyword argument :embed_depth
may be specified to set a max allowed nesting depth for the corresponding resource to be embedded. See embed_depth
.
If the resource to be embedded has a registered Serializer then presenter_class
is not needed.
class User
def name; "bengt"; end
end
class UserSerializer
extend HALPresenter
model User
attribute :name
end
class PostSerializer
extend HALPresenter
embed :author
end
post = OpenStruct.new(title: "hello", author: User.new)
PostSerializer.to_hal(post) # => {"_embedded":{"author":{"name":"bengt"}}}
If the embedded resource has a curie then it will be added to the root resource rather than the embedded resource. If there is a curie name conflict between the root resource and any embedded resources, one or many of them will get renamed.
class FooSerializer
extend HALPresenter
attribute :info
link :foo, curie: :doc do
'/foo'
end
curie :doc do
'/doc/{rel}'
end
end
class BarSerializer
extend HALPresenter
attribute :info
link :bar, curie: :doc do
'/bar'
end
curie :doc do
'/conflicting/{rel}'
end
end
class PostSerializer
extend HALPresenter
curie :doc do
'/doc/{rel}'
end
embed :foo, presenter_class: FooSerializer
embed :bar, presenter_class: BarSerializer
end
post = OpenStruct.new(
foo: OpenStruct.new(info: "doc means the same thing for foo"),
bar: OpenStruct.new(info: "doc conflicts with Foo and Post. It must be renamed!")
)
PostSerializer.to_hal(post) # =>
# {
# "_links": {
# "curies": [
# {
# "name": "doc",
# "href": "/doc/{rel}",
# "templated": true
# },
# {
# "name": "doc0",
# "href": "/conflicting/{rel}",
# "templated": true
# }
# ]
# },
# "_embedded": {
# "foo": {
# "info": "doc means the same thing for foo",
# "_links": {
# "doc:foo": {
# "href": "/foo"
# }
# }
# },
# "bar": {
# "info": "doc conflicts with Foo and Post. It must be renamed!",
# "_links": {
# "doc0:bar": {
# "href": "/bar"
# }
# }
# }
# }
# }
Note that the curie in FooSerializer
has the same href as the curie in PostSerializer
, thus it remains unaffected. The curie in BarSerializer
however has another href and needs to be renamed. The new name is doc0
. Any rels (links and embedded) refering to the renamed curie will be updated with the new curie name (e.g. doc0:bar
).
The collection
class method is used to make a serializer capable of serializing an array of resources. Serializing collections may of course be done with separate serializer, but should we want to use the same serializer class for both then ::collection
will make that work. The method takes a required keyword paramter named :of
, which will be used as the key in the corresponding _embedded property. Each entry in the array given to ::to_collection
will then be serialized with this serializer.
class PostSerializer
extend HALPresenter
attribute :id
attribute :title
collection of: 'posts'
end
list = (1..2).map do |i|
OpenStruct.new(id: i, title: "hello#{i}")
end
PostSerializer.to_collection(list) # => {"_embedded":{"posts":[{"id":1,"title":"hello1"},{"id":2,"title":"hello2"}]}}
The collection
class method takes an optional block. The purpose of this block is to be able to set attributes, links and embedded resources on the serialized collection.
class PostSerializer
extend HALPresenter
attribute :id
attribute :title
collection of: 'posts' do
attribute(:number_of_posts) { resources.count }
link :self do
format "/posts%s", (options[:page] && "?page=#{options[:page]}")
end
link :next do
"/posts?page=#{options[:next]}" if options[:next]
end
end
end
list = (1..2).map do |i|
OpenStruct.new(id: i, title: "hello#{i}")
end
PostSerializer.to_collection(list, page: 1, next: 2) # => {"number_of_posts":2,"_links":{"self":{"href":"/posts?page=1"},"next":{"href":"/posts?page=2"}},"_embedded":{"posts":[{"id":1,"title":"hello1"},{"id":2,"title":"hello2"}]}}"
The response above with some newlines.
{
"number_of_posts": 2,
"_links": {
"self": {
"href": "/posts?page=1"
},
"next": {
"href": "/posts?page=2"
}
},
"_embedded": {
"posts": [
{
"id": 1,
"title": "hello1"
},
{
"id": 2,
"title": "hello2"
}
]
}
}
Note: the block given to the :number_of_posts
attribute is using the method resources
. This is just and alias for resource
which looks better inside collections.
The :embed_depth
keyword arguments specifies for which levels of embedding the correponding property should be serialized.
nil
: The property it is always serialized.0
: The property is only serialized when the resource is not embedded.1
: The property is serialized when it's embedded at most 1 level deep.- etc..
Consider the following payload representing a post resource:
{
"id": 1,
"message": "lorem ipsum..",
"_links": {
"self": {
"href": "/posts/1"
},
},
"_embedded": {
"comments": [
{
"id": 1,
"comment": "hello1"
"_embedded": {
"user": {
"id": 2,
"name": "foo",
"_links": {
"self": {
"href": "/users/2"
}
}
}
}
},
{
"id": 2,
"comment": "hello2"
"_embedded": {
"user": {
"id": 3,
"name": "foo",
"_links": {
"self": {
"href": "/users/3"
}
}
}
}
}
]
}
}
Here the post attributes id
, message
as well as the self link and the embedded comments all have a depth of 0. The properties of each embedded comment (attributes id
, comment
and embedded user) have a depth of 1. The properties of the user resources (embedded in the comments, which in turn are embedded in the post resource) have a depth of 2.
When a collection is serialized (using ::to_collection
) the embedded resources will have the same depth as attributes and links on collection (e.g. depth = 0 unless the collection itself is embedded).
The purpose of specifying embed_depth
is to be able skip serializing properties when embeddeed.
For example, when you serialize a collection of resources, perhaps you would like for each resource in that collection to only serialize a few properties, making it kind of like a "preview" of each resource.
Blocks passes to ::attribute
, ::link
, ::curie
and ::embed
have access to the resource being serialized throught the resource
method. These blocks also have access to an optional options
hash that can be passed to ::to_hal
.
class PostSerializer
extend HALPresenter
attribute :title do
"#{resource.id} -- #{resource.title} -- #{options[:extra]}"
end
end
post = OpenStruct.new(id: 5, title: "hello")
PostSerializer.to_hal(post, extra: 'world') # => {"title": "5 -- hello -- world"}
These blocks also have access to the scope where the block was created (e.g. the Serializer class)
class PostSerializer
extend HALPresenter
def self.prefix; "-->"; end
attribute :title do
"#{prefix} #{resource.title}"
end
end
post = OpenStruct.new(id: 5, title: "hello")
PostSerializer.to_hal(post) # => {"title": "--> hello"}
Note: this does not mean that self
inside the block is the serializer class. The access to the serializer class methods is done by delegation.
If the block passed to ::attribute
evaluates to nil
then the serialized value will be null
. If the block passed to ::link
, ::curie
or ::embed
evaluates to nil
,
then the corresponding property will not be serialized.
class PostSerializer
extend HALPresenter
attribute :title
attribute :foo { nil }
link :self { "/posts/#{resource.id}" }
link :edit do
"/posts/#{resource.id}" if resource.author_id == options[:current_user].id
end
end
user = OpenStruct.new(id: 5)
post = OpenStruct.new(id: 1, title: "hello", author_id: 2)
PostSerializer.to_hal(post, current_user: user) # => {"title":"hello","foo":null,"_links":{"self":{"href":"/posts/1"}}}
user = OpenStruct.new(id: 2)
PostSerializer.to_hal(post, current_user: user) # => "{"title":"hello","foo":null,"_links":{"self":{"href":"/posts/1"},"edit":{"href":"/posts/1"}}}"
See examples in ::attribute
, ::link
, ::curie
and ::embed
.
See examples in ::collection
The ::post_serialize
class method can used to run a hook after each serialization. This method must be called with a block taking one parameter (which will be a Hash of the serialized result). This can be convenient when dynamic properties needs to be added to the serialized payload. As an example say that you have a Form class and a FormSerializer that should be used to serialize different kinds of forms.
class Field
attr_reader :name, :type, :value
def initialize(name, params = {})
@name = name
@type = params[:type]
@has_value = params.key? :value
@value = params[:value]
end
def has_value?
@has_value
end
end
class Form
attr_accessor :resource, :name, :title, :href, :method, :type, :self_link, :fields
def initialize(params = {})
@name = params[:name]
@title = params[:title]
@method = params[:method] || :post
@type = params[:type] || 'application/json'
@fields = (params[:fields] || {}).map { |name, args| Field.new(name, args) }
end
end
class FormSerializer
extend HALPresenter
model Form
profile 'shaf-form'
attribute :method do
(resource&.method || 'POST').to_s.upcase
end
attribute :name do
resource&.name
end
attribute :title do
resource&.title
end
attribute :href do
resource&.href
end
attribute :type do
resource&.type
end
link :self do
resource&.self_link
end
link :profile,
"https://gist.githubusercontent.com/sammyhenningsson/39c8aafeaf60192b082762cbf3e08d57/raw/shaf-form.md"
post_serialize do |hash|
fields = resource&.fields
break if fields.nil? || fields.empty?
hash[:fields] = fields.map do |field|
{ name: field.name, type: field.type }.tap do |f|
f[:value] = field.value if field.has_value?
end
end
end
end
Now this setup can be used to serialize different kinds of forms with a single serializer.
common_fields = {
email: { type: "string"},
password: { type: "string"}
}
create_form = Form.new(name: 'create-user', title: 'Create User', fields: common_fields.merge({username: { type: "string"}}))
create_form.href = '/users'
edit_form = Form.new(name: 'edit-user', title: 'Update User', method: :put, fields: common_fields)
edit_form.href = '/users/5/edit'
FormSerializer.to_hal(create_form)
FormSerializer.to_hal(edit_form)
This would give the following:
{
"href": "/users",
"method": "POST",
"name": "create-user",
"title": "Create User",
"type": "application/json",
"fields": [
{
"name": "email",
"type": "string"
},
{
"name": "password",
"type": "string"
},
{
"name": "username",
"type": "string"
}
]
}
{
"href": "/users/5/edit",
"method": "PUT",
"name": "edit-user",
"title": "Update User",
"type": "application/json",
"fields": [
{
"name": "email",
"type": "string"
},
{
"name": "password",
"type": "string"
}
]
}
The class method from_hal
is used to deserialize a payload into a model instance. If there are links in the payload they will be discarded. Fields in the payload that
does not have a corresponding attribute or embed in the serializer will be ignored.
class User
attr_accessor :name
end
class Post
attr_accessor :title, :author
end
class UserSerializer
extend HALPresenter
model User
attribute :name
link :self, '/user'
end
class PostSerializer
extend HALPresenter
model Post
attribute :title
link :self, '/post'
embed :author, presenter_class: UserSerializer
end
user = User.new
user.name = "bengt"
post = Post.new
post.title= "hello"
post.author = user
payload = PostSerializer.to_hal(post) # => {"title":"hello","_links":{"self":{"href":"/post"}},"_embedded":{"author":{"name":"bengt","_links":{"self":{"href":"/user"}}}}}"
post = PostSerializer.from_hal(payload)
post.title # => "hello"
post.author.name # => "bengt"
Instances are created by calling ::new
on the class registered by ::model
without any arguments. Then each attribute is set with #attribute_name=
(e.g.
post.title = 'hello'
)
Thus, all models used for deserialization must respond to attribute_name=
for all attributes used in the serializer.
If the model can't be created without arguments (or if the instance already exit), then the instance can be passed to ::from_hal
.
class User
attr_accessor :name
def initialize(name)
@name = name
end
end
class Post
attr_accessor :title, :author
def initialize(title, author)
@title = title
@author = author
end
end
class UserSerializer
extend HALPresenter
model User
attribute :name
link :self, '/user'
end
class PostSerializer
extend HALPresenter
model Post
attribute :title
link :self, '/post'
embed :author, presenter_class: UserSerializer
end
payload = JSON.generate(
{
"title": "hello",
"_embedded": {
"author": {
"name": "bengt"
}
}
}
)
user = User.new('will_be_overwritten')
post = Post.new('will_be_overwritten', user)
post = PostSerializer.from_hal(payload, post)
post.title # => "hello"
post.author.name # => "bengt"
Collections can be deserialized into an array as long as the serializer has a collection. In this case the model instance cannot be passed as an argument so it must be possbile to create new instances with ModelName.new (whithout any arguments).
class User
attr_accessor :name
end
class UserSerializer
extend HALPresenter
model User
attribute :name
link :self, '/user'
collection of: 'users'
end
users = (1..5).map do |i|
{
name: "user#{i}",
foo: "ignored"
}
end
payload = JSON.generate(
{
_embedded: {
users: users
}
}
)
users = UserSerializer.from_hal(payload)
users.class # => Array
users.first.class # => User
users.map(&:name) # => ["user1", "user2", "user3", "user4", "user5"]
Serializers may use inheritance and should work just as expected
class BaseSerializer
extend HALPresenter
attribute :base, "will_be_overwritten"
attribute :title
link :self do
resource_path(resource.id)
end
end
class PostSerializer < BaseSerializer
attribute :base, "child_attribute"
def self.resource_path(id)
"/posts/#{id}"
end
end
post = OpenStruct.new(id: 5, title: 'hello')
PostSerializer.to_hal(post) # => {"title": "hello", "base": "child_attribute", "_links": {"self": {"href": "/posts/5"}}}
This module method can be used to specify a base url that will get prepended to links hrefs.
HALPresenter.base_href = 'https://localhost:3000/'
class PostSerializer
extend HALPresenter
link :self, '/posts/1'
end
PostSerializer.to_hal # => {"_links": {"self": {"href": "https://localhost:3000/posts/1"}}}
Setting HALPresenter.paginate = true
will add next/prev links for collections when possible. Requirements for this is:
- The resource being serialized is a paginated collection (Kaminari, will_paginate and Sequel are supported)
- The serializer being used has a collection block which declares a self link
HALPresenter includes a DSL for creating polices. By including HALPresenter::Policy::DSL
into your policy class you get the following class methods:
::attribute(*names, &block)
::link(*rels, &block)
::embed(*names, &block)
These methods all work the same way and creates one or more rules for each name
argument (rel
for links). If no block is given then the corresponding attribute/link/embedded resource will always be serialized. If the block evaluates to true
then the attribute/link/embedded resource will be serialized. Otherwise it will not be serialized. The block has access to the current user, the resource that is being serialized, as well as any options passed to ::to_hal
from the methods current_user
, resource
resp. options
.
class UserPolicy
include HALPresenter::Policy::DSL
attribute :first_name, :last_name
attribute :email do
# show name and email attributes if user is logged in
!current_user.nil?
end
attribute :ssn do
# Only show ssn if the resource belongs to current_user
current_user && resource.user.id == current_user.id
end
link :self
link :edit do
edit?
end
embed :posts do
current_user && !current_user.posts.empty?
end
def edit?
current_user && resource.user.id == current_user.id
end
Notice the instance method #edit?
which is typically used by Pundit. That method is called from the block belonging to the rule for the edit link. This means that we can use the same policy class both for serialization and for authorization (and have all the rules in one place). This is great since we should only provide links to actions that are possible (authorized) and we don't want to sync this between controller code and serialization code.
Policies can be inherited, so probably it makes sense to have a base policy that other policies can inherit from. Like:
class BasePolicy
include HALPresenter::Policy::DSL
def authenticated?
!!current_user
end
end
class CommentPolicy < BasePolicy
link :post
link :comment do
authenticated?
end
end
Sometimes policies may depend on other policies. For thoses cases we can delegate to another policy. Like:
class CommentPolicy < BasePolicy
def read?
delegate_to PostPolicy, :read?, resource: resource.post
end
end
Besides #delegate_to(policy_class, method, resource: nil, args: nil, **opts)
, there is also:
#delegate_attribute(policy_class, attr, **opts)
#delegate_link(policy_class, rel, **opts)
#delegate_embed(policy_class, rel, **opts)