Awesome
JaSerializer
jsonapi.org formatting of Elixir data structures suitable for serialization by libraries such as Poison.
Usage
See documentation on hexdoc for full serialization and usage details.
Installation
Add JaSerializer to your application
mix.deps
defp deps do
[
# ...
{:ja_serializer, "~> x.x.x"}
# ...
]
end
Serializer Behaviour and DSL
defmodule MyApp.ArticleSerializer do
use JaSerializer
location "/articles/:id"
attributes [:title, :tags, :body, :excerpt]
has_one :author,
serializer: PersonSerializer,
include: true,
field: :authored_by
has_many :comments,
links: [
related: "/articles/:id/comments",
self: "/articles/:id/relationships/comments"
]
def comments(article, _conn) do
Comment.for_article(article)
end
def excerpt(article, _conn) do
[first | _ ] = String.split(article.body, ".")
first
end
end
Attributes
Attributes are defined as a list in the serializer module. The serializer will use the given atom as the key by default. You can also specify a custom method of attribute retrieval by defining a <attribute_name>/2 method. The method will be passed the struct and the connection.
Relationships
Valid relationships are: has_one
, has_many
.
Use has_one
for belongs_to
type of relationships.
For each relationship, you can define the name and a variety of options.
Just like attributes, the serializer will use the given atom
to look up the relationship, unless you specify a custom retrieval method
OR provide a field
option
Relationship options
- serializer - The serializer to use when serializing this resource
- include - boolean - true to always side-load this relationship
- field - custom field to use for relationship retrieval
- links - custom links to use in the
relationships
hash
Direct Usage of Serializer
MyApp.ArticleSerializer
|> JaSerializer.format(struct, conn)
|> Poison.encode!
Formatting options
The format/4
method is able to take in options that can customize the
serialized payload.
Include
By specifying the include
option, the serializer will only side-load
the relationships specified. This option should be a comma separated
list of relationships. Each relationship should be a dot separated path.
Example: include: "author,comments.author"
The format of this string should exactly match the one specified by the JSON-API spec
Note: If specifying the include
option, all "default" includes will
be ignored, and only the specified relationships included, per spec.
Fields
The fields
option satisfies the sparse fieldset portion of the spec. This options should
be a map of resource types whose value is a comma separated list of fields
to include.
Example: fields: %{"articles" => "title,body", "comments" => "body"}
If you're using Plug, you should be able to call fetch_query_params(conn)
and pass the result of conn.query_params["fields"]
as this option.
Phoenix Usage
For an example of starting with Phoenix's JSON generator and updating to work with JaSerializer, see Getting Started with Phoenix.
Simply use JaSerializer.PhoenixView
in your view (or in the Web module) and
define your serializer as above.
The render("index.json-api", data)
and render("show.json-api", data)
are defined
for you. You can just call render as normal from your controller.
By specifying include
s when calling the render function, you can override
the include: false
in the ArticleView.
defmodule PhoenixExample.ArticlesController do
use PhoenixExample.Web, :controller
def index(conn, _params) do
render conn, "index.json-api", data: Repo.all(Article)
end
def show(conn, %{"id" => id}) do
article = Repo.get(Article, id) |> Repo.preload([:comments])
render conn, "show.json-api", data: article,
opts: [include: "comments"]
end
def create(conn, %{"data" => data}) do
attrs = JaSerializer.Params.to_attributes(data)
changeset = Article.changeset(%Article{}, attrs)
case Repo.insert(changeset) do
{:ok, article} ->
conn
|> put_status(201)
|> render("show.json-api", data: article)
{:error, changeset} ->
conn
|> put_status(422)
|> render(:errors, data: changeset)
end
end
end
defmodule PhoenixExample.ArticlesView do
use PhoenixExample.Web, :view
use JaSerializer.PhoenixView # Or use in web/web.ex
attributes [:title]
has_many :comments,
serializer: PhoenixExample.CommentsView,
include: false,
identifiers: :when_included
#has_many, etc.
end
Configuration
To use the Phoenix accepts
plug you must configure Plug to handle the
"application/vnd.api+json" mime type and Phoenix to serialize json-api with
Poison.
Depending on your version of Plug add the following to config.exs
:
Plug ~> "1.2.0"
config :phoenix, :format_encoders,
"json-api": Poison
config :mime, :types, %{
"application/vnd.api+json" => ["json-api"]
}
And then re-compile mime: (per: https://hexdocs.pm/mime/MIME.html)
mix deps.clean mime --build
mix deps.get
Plug < "1.2.0"
config :phoenix, :format_encoders,
"json-api": Poison
config :plug, :mimes, %{
"application/vnd.api+json" => ["json-api"]
}
And then re-compile plug: (per: https://hexdocs.pm/plug/1.1.3/Plug.MIME.html)
mix deps.clean plug --build
mix deps.get
And then add json api to your plug pipeline.
pipeline :api do
plug :accepts, ["json-api"]
end
For strict content-type/accept enforcement and to auto add the proper content-type to responses add the JaSerializer.ContentTypeNegotiation plug.
To normalize attributes to underscores include the JaSerializer.Deserializer plug.
pipeline :api do
plug :accepts, ["json-api"]
plug JaSerializer.ContentTypeNegotiation
plug JaSerializer.Deserializer
end
If you're rendering JSON API errors, like 404.json-api
, then you must add json-api
to the accepts
of your render_errors
within your existing configuration in config.exs
, like so:
config :phoenix, PhoenixExample.Endpoint,
render_errors: [view: PhoenixExample.ErrorView, accepts: ~w(html json json-api)]
If you're rendering both JSON-API and HTML, you need to include the html
option in the config:
config :phoenix, :format_encoders,
html: Phoenix.Template.HTML,
"json-api": Poison
Testing controllers
Set the right headers in setup
and when passing parameters to put and post requests,
you should pass them as a binary. That is because for map and list parameters,
the content-type will be automatically changed to multipart.
defmodule Sample.SomeControllerTest do
use Sample.ConnCase
setup %{conn: conn} do
conn =
conn
|> put_req_header("accept", "application/vnd.api+json")
|> put_req_header("content-type", "application/vnd.api+json")
{:ok, conn: conn}
end
test "create action", %{conn: conn} do
params = Poison.encode!(%{data: %{attributes: @valid_attrs}})
conn = post conn, "/some_resource", params
...
end
...
end
Pagination
JaSerializer provides page based pagination integration with Scrivener or custom pagination by passing your owns links in.
Custom
JaSerializer allows custom pagination via the page
option. The page
option
expects to receive a Map
with URL values for first
, next
, prev
,
and last
.
For example:
page = %{
first: "http://example.com/api/v1/posts?page[cursor]=1&page[per]=20",
prev: nil
next: "http://example.com/api/v1/posts?page[cursor]=20&page[per]=20",
last: "http://example.com/api/v1/posts?page[cursor]=60&page[per]=20"
}
# Direct call
JaSerializer.format(MySerializer, collection, conn, page: page)
# In Phoenix Controller
render conn, "index.json-api", data: collection, opts: [page: page]
Builder
You can build the pagination links with
JaSerializer.Builder.PaginationLinks.build/2
Simply pass in the following:
links =
JaSerializer.Builder.PaginationLinks.build(
%{
number: 2,
size: 10,
total: 20
},
conn
)
See JaSerializer.Builder.PaginationLinks
for how to customize.
Scrivener Integration
If you are using Scrivener for pagination, all you need to do is pass the
results of paginate/2
to your serializer.
page = MyRepo.paginate(MyModel, params.page)
# Direct call
JaSerializer.format(MySerializer, page, conn, [])
# In Phoenix controller
render conn, "index.json-api", data: page
When integrating with Scrivener, the URLs generated will be based on the
Plug.Conn
's path. This can be overridden by passing in the page[:base_url]
option.
render conn, "index.json-api", data: page, opts: [base_url: "http://example.com/foos"]
You can also configure ja_serializer
to use a global default URL
base for all links.
config :ja_serializer,
page_base_url: "http://example.com:4000/v1/"
Note: The resulting URLs will use the JSON-API recommended page
query
param.
Example URL:
http://example.com:4000/v1/posts?page[page]=2&page[page-size]=50
Meta Data
JaSerializer allows adding top level meta information via the meta
option. The meta
option
expects to receive a Map
containing the data which will be rendered under the top level meta key.
meta_data = %{
"key" => "value"
}
# Direct call
JaSerializer.format(MySerializer, data, conn, meta: meta_data)
# In Phoenix controller
render conn, "index.json-api", data: data, opts: [meta: meta_data]
Customization
Key Format (for Attribute, Relationship and Query Param)
By default keys are dash-erized
as per the JSON:API 1.0 recommendation, but keys can be customized via config.
In your config.exs
file you can use camel_cased
recommended by upcoming JSON:API 1.1:
config :ja_serializer,
key_format: :camel_cased
Or underscored
:
config :ja_serializer,
key_format: :underscored
You may also pass custom function for serialization and a second optional one for deserialization. Both accept a single binary argument:
defmodule MyStringModule do
def camelize(key), do: key #...
def underscore(key), do: key #...
end
config :ja_serializer,
key_format: {:custom, MyStringModule, :camelize, :underscore}
Custom Attribute Value Formatters
When serializing attribute values more complex than string, numbers, atoms or list of those things it is recommended to implement a custom formatter.
To implement a custom formatter:
defimpl JaSerializer.Formatter, for: [MyStruct] do
def format(struct), do: struct
end
Pluralizing All Types By Default
You can opt-in to pluralizing all types for default:
config :ja_serializer,
pluralize_types: true
Complimentary Libraries
- JaResource - WIP behaviour for creating JSON-API controllers in Phoenix.
- voorhees - Testing tool for JSON API responses
- inquisitor - Composable query builder for Ecto
- scrivener - Ecto pagination
License
JaSerializer source code is released under Apache 2 License. Check LICENSE file for more information.