Creating a Simple REST API Using Elixir and Phoenix

A step-by-step guide on creating a REST API with Phoenix, covering everything from database setup to endpoint implementation.

Headshot of Bryan Anthonio
Bryan Anthonio

When I first dove into the Phoenix framework, I felt intimidated. It quickly became clear that I needed to understand a few key concepts before I could build my first app.

However, reading through the book Programming Phoenix LiveView clarified many of the aspects that confused me. The book gave me enough confidence to build a simple REST API that serves quotes from Stoic philosophers, to apply what I learned.

In this post, I’ll outline the step-by-step process of creating this API using Phoenix. The API provides these endpoints:

  • /api/quotes/ to list all quotes
  • /api/quotes/random to fetch a random quote

Contents

Getting Started

What is Phoenix?

Phoenix is a web development framework built using the Elixir programming language. It’s akin to Ruby on Rails and Django in the Elixir ecosystem as it implements the server-side Model-View-Controller design pattern. This makes Phoenix suitable for crafting backend APIs, full-stack web apps, and other applications.

A key advantage of Phoenix is its robust support for real-time functionalities via websockets, making it ideal for building chat rooms, multiplayer games, and other apps requiring live updates.

Prerequisites

In this tutorial, I’ll assume the following:

Kickstarting Your Phoenix Project

Every Phoenix project begins with the command mix phx.new. Run the commands below in your terminal to get started:

mix phx.new stoic_quotes
cd stoic_quotes
mix ecto.create

This will create a new directory called stoic_quotes/ with all the necessary files and folders and set up a database for the app.

Setting Up a Quotes Database

Next, we’ll create a quotes database. This will involve:

  1. Setting up and running a migration for the quotes table schema.
  2. Implementing validations in the database to prevent duplicate quotes.
  3. Writing a script to insert quotes from a JSON file into the database.

Schema Setup and Migration

First, we’ll create a database table named quotes to store quotes. This table will have these fields:

  • quote: the quote itself
  • author: author of the quote
  • source: the literary work (and chapter) containing the quote

Phoenix provides generators that we can leverage for this. Here, we’ll use the phx.gen.context generator.

This generator executes three tasks:

  1. Creates a schema module, which includes the schema for items in the quotes table and a changeset function used for validating new entries.
  2. Provides a context module that offers functions for creating, reading, updating, and deleting items in the quotes table.
  3. Generates a Migration file, containing instructions for creating the quotes table in our database.

Furthermore, phx.gen.context requires at least 3 arguments: the context module name, the singular and plural names of the generated resources, and any attributes to include in the schema, specifying the data type for each attribute.

Run the following command in your terminal to generate the necessary files for the quotes table:

mix phx.gen.context Quotes Quote quotes quote:text author:string source:string

Preventing Duplicate Quotes in the Database

We need to implement a constraint to prevent users from inserting duplicate quotes into the database. For this, we’ll edit both the migration file and the schema module.

Navigate to the /priv/repo/migrations directory and find a timestamped file labeled create_quotes.

This file contains instructions on creating the quotes table with the fields we specified earlier. The timestamps() macro adds the inserted_at and updated_at fields, which store timestamps at which an entry was inserted and updated, respectively.

Now, edit the migration file to look like this:

defmodule StoicQuotes.Repo.Migrations.CreateQuotes do
	use Ecto.Migration

	def change do
		create table(:quotes) do
			add :quote, :text
			add :author, :string
			add :source, :string
			timestamps()
		end

		# add this line below to enforce unique quotes
		create unique_index(:quotes, [:quote], name: :index_for_duplicate_quotes)
	end
end

The new line we’ve added to the migration file creates a unique index, ensuring each quote is distinct.

After making these edits, execute the command mix ecto.migrate in the terminal to create the quotes table.

The schema module, located in lib/stoic_quotes/quotes/quote.ex, defines the schema and includes a changeset function used for validating new entries before their insertion into the quotes table.

These changeset functions are used by methods in the generated context module StoicQuotes.Quotes, found in lib/stoic_quotes/quotes.ex. For example, the context module provides this function for creating quotes:

def create_quote(attrs \\ %{}) do
	%Quote{}
	|> Quote.changeset(attrs)
	|> Repo.insert()
end

Here, create_quote takes a struct containing the fields of the new quote entry. It applies the Quote.changeset to this input before passing it to Repo.insert() function. Entries that violate any constraints won’t be inserted.

By default, the changeset function in the Quote schema module checks that the necessary fields— quote, author, and source— are provided for each new entry. Let’s add another line to implement the unique constraint:

defmodule StoicQuotes.Quotes.Quote do
	use Ecto.Schema
	import Ecto.Changeset

	schema "quotes" do
		field :author, :string
		field :source, :string
		field :quote, :string
		timestamps()
	end

	@doc false
	def changeset(quote, attrs) do
		quote
		|> cast(attrs, [:quote, :author, :source])
		|> validate_required([:quote, :author, :source])
		|> unique_constraint(:quote, name: :index_for_duplicate_quotes) #Add this new line
	end
end

The new line we’ve added leverages the index we defined in the migration file to prevent the insertion of duplicate quotes and provides an error message when this occurs.

Add Data to the Database

Now, let’s add quotes to the database. Create a JSON file named quotes.json in the priv/repo/ directory. We’ll populate the JSON file with 3 quotes to start:

[
  {
    "quote": "Seldom are any found unhappy from not observing what is in the minds of others. But such as observe not well the stirrings of their own souls must of necessity be unhappy.",
    "author": "Marcus Aurelius",
    "source": "Book II, Meditations"
  },

  {
    "quote": "Consider whence each thing came, of what it was compounded, into what it will be changed, how it will be with it when changed, and that it will suffer no evil.",
    "author": "Marcus Aurelius",
    "source": "Book XI, Meditations"
  },

  {
    "quote": "Accustom yourself as much as possible, when any one takes any action, to consider only: To what end is he working? But begin at home; and examine yourself first of all.",
    "author": "Marcus Aurelius",
    "source": "Book X, Meditations"
  }
]

Next, we’ll write a script to insert these quotes into the database. Create a script called seeds.exs in priv/repo/ with the following contents:

alias StoicQuotes.Quotes

# Read quotes from the JSON file
quotes_path = "priv/repo/quotes.json"
quotes_path
|> File.read!()
|> Jason.decode!()
|> Enum.each(fn attrs ->
	# Construct a quote struct and attempt to insert it
	quote = %{quote: attrs["quote"], author: attrs["author"], source: attrs["source"]}
	case Quotes.create_quote(quote) do
		{:ok, _quote} -> :ok
		{:error, _changeset} -> :duplicate
	end
end)

This script uses the Quotes context module to insert each quote into the database.

Run the following command to execute the script:

mix run priv/repo/seeds.exs

To verify the quotes in the database:

  1. Execute iex -S mix in the terminal to open an Elixir shell.
  2. Enter alias StoicQuotes.Quotes to alias the Quotes module.
  3. Run Quotes.list_quotes() to list all quotes in the database.

Building API Endpoints

Now that the database is set up, we need to establish endpoints for our REST API. These include the following:

  • /api/quotes/ to list all quotes
  • /api/quotes/random to fetch a random quote

Define Endpoints

In this project, routes are defined in lib/stoic_quotes_web/router.ex. We can implement the REST API routes by adding the following declaration into the body of this file:

scope "/api", StoicQuotesWeb do
	pipe_through :api
	get "/quotes", QuotesController, :index
	get "/quotes/random", QuotesController, :show
end

This snippet sets up the /api/ routes to use the pipeline defined by pipe_through :api. Requests to these routes will be processed by the :api pipeline, which is defined in the file as

pipeline :api do
	plug :accepts, ["json"]
end

Routes using this pipeline will accept JSON requests.

Furthermore, the defined routes specify that GET requests to the /api/quotes/ and /api/quotes/random will be handled by the :index and :show actions in the QuotesController module, respectively.

In the following section, we’ll learn how to create these controller actions.

Create Controllers

Phoenix will look for controller definitions in the directory lib/stoic_quotes_web/controllers/. Let’s create a file named quotes_controller.ex here to define our API’s behavior:

defmodule StoicQuotesWeb.QuotesController do
	use Phoenix.Controller, formats: [:json]
	alias StoicQuotes.Quotes

	def index(conn, _params) do
		quotes = %{quotes: Quotes.list_quotes()}
		render(conn, :index, quotes)
	end

	def show(conn, _params) do
		quote = %{quote: Quotes.get_random_quote()}
		render(conn, :show, quote)
	end
end

The index and show functions implement the logic for the :index and :show actions defined in the router, respectively. These require two arguments: the connection struct conn and a set of parameters params, which we won’t use (for more info on actions read here).

Outputs of these functions are dictated by the render/3 function, which generates the views for these actions. Here, render takes a connection struct, an atom corresponding to the view template and data to pass to the template.

In addition, the macro use Phoenix.Controller, formats: [:json] specifies that this controller renders JSON responses. Phoenix will look for the view templates in the file quotes_json.ex, which we’ll define in the next section.

For the show action, we’ll need to define the get_random_quote in the StoicQuotes.Quotes module located in lib/stoic_quotes/quotes.ex. Here, we can define this function as shown below:

def get_random_quote() do
	query =
		from q in Quote,
		order_by: fragment("RANDOM()"),
		limit: 1

	Repo.one(query)
end

Make View Templates

In the lib/stoic_quotes_web/controllers/ directory, let’s create the quotes_json.ex file to format data for the API responses:

defmodule StoicQuotesWeb.QuotesJSON do
	alias StoicQuotes.Quotes.Quote

	def index(%{quotes: quotes}) do
		%{data: for(quote <- quotes, do: data(quote))}
	end

	def show(%{quote: quote}) do
		%{data: data(quote)}
	end

	defp data(%Quote{} = datum) do
		%{
			quote: datum.quote,
			author: datum.author,
			source: datum.source
		}
	end
end

These functions define the view templates for the index and the show actions. Both functions return a struct with a data field, which Phoenix automatically converts into appropriate JSON response formats.

Testing the API

Now, let’s test this API. An easy way to do this is to use the curl command, available in terminals, that transfers data from or to a server.

First, run mix phx.server to start the server. After that, test the endpoint routes using these commands:

curl -i localhost:4000/api/quotes/ #get all quotes
curl -i localhost:4000/api/quotes/random #get random quote

Conclusion

In this blog post, we covered the necessary steps to build a simple REST API using the Phoenix Framework. First, we set up a database, populated it with quote entries, and implemented the endpoints for our REST API. Finally, we created controllers and corresponding views to generate JSON responses.