jsonapi.org formatting of Elixir data structures suitable for serialization by libraries such as Poison.
Please open an issue or message/mention @alanpeabody in the Elixir Slack.
See documentation on hexdoc for full serialization and usage details.
Add JaSerializer to your application
mix.deps
defp deps do
[
# ...
{:ja_serializer, "~> 0.8.1"}
# ...
]
end
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 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.
Valid relationships are: has_one
, has_many
.
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
- 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
struct
|> MyApp.ArticleSerializer.format(conn)
|> Poison.encode!
The format/3
method is able to take in options that can customize the
serialized payload.
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 exacly 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.
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.
See Relax documentation for building fully compatible jsonapi.org APIs with Plug.
Simply use JaSerializer.PhoenixView
in your view (or in the Web module) and
define your serializer as above.
The render("index.json", data)
and render("show.json", data)
are defined
for you. You can just call render as normal from your controller.
defmodule PhoenixExample.ArticlesController do
use PhoenixExample.Web, :controller
def index(conn, _params) do
render conn, data: Repo.all(Article)
end
def show(conn, params) do
render conn, data: Repo.get(Article, params[:id])
end
def create(conn, %{"data" => %{"attributes" => attrs}}) do
changeset = Article.changeset(%Article{}, attrs)
case Repo.insert(changeset) do
{:ok, article} ->
conn
|> put_status(201)
|> render(:show, 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, etc.
end
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.
Add the following to config.exs
:
config :phoenix, :format_encoders,
"json-api": Poison
config :plug, :mimes, %{
"application/vnd.api+json" => ["json-api"]
}
And then re-compile plug: (per: http://hexdocs.pm/plug/Plug.MIME.html)
touch deps/plug/mix.exs
mix deps.compile plug
# For testing
MIX_ENV=test mix deps.compile plug
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
Set right headers in setup and when passing parameters to put, 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
JaSerializer provides page based pagination integration with Scrivener or custom pagination by passing your owns links in.
JaSerializer allows custom pagination via the page
option. The page
option
expects to receive a Dict
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
MySerializer.format(collection, conn, page: page)
# In Phoenix Controller
render conn, data: collection, opts: [page: page]
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
MySerializer.format(page, conn, [])
# In Phoenix controller
render conn, 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, data: page, opts: [page: [base_url: "http://example.com/foos"]]
Note: The resulting URLs will use the JSON-API recommended page
query
param.
Example URL:
http://example.com/v1/posts?page[page]=2&page[page-size]=50
By default keys are dash-erized
as per the jsonapi.org recommendation, but
keys can be customized via config.
In your config.exs file:
config :ja_serializer,
key_format: :underscored
You may also pass a custom function that accepts 1 binary argument:
defmodule MyStringModule do
def camelize(key), do: key #...
end
config :ja_serializer,
key_format: {:custom, MyStringModule, :camelize}
If you've already compliled your code, be sure to run mix deps.clean ja_serializer && mix deps.get
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
JaSerializer source code is released under Apache 2 License. Check LICENSE file for more information.