By the end of this lesson, students should be able to:
- Explain why an API is necessary.
- List some of the responsibilities of a typical API, and identify which components within MVC map to those responsibilities.
- Diagram the process flow from Client through Database and back
- Map MVC roles to specific components of Rails.
- Indicate where different types of files can be found within a Rails application.
- Recall how to use
rails generate
for startingmodels
,controllers
, andserializers
.
No preparation necessary for developers.
By now, you've built a client application: it had a user interface (UI) and behavioral logic, allowing users to play Tic-Tac-Toe in the browser. Isn't that enough? Why might we want an API?
One possible reason is that when two different people are served up a web page, they each have a separate copy of everything on the page. As a result, if we have two users interacting with our app, they can never interact with each other, or even be aware of each other.
Another possible reason is that we might want want to store the state of the game over time. What if we wanted to be able to 'save' our game, close the browser, walk away, and pick it up again the next day? Having an API where data is stored means that we can do just that. This is known as persistence (things being saved over time).
For your client, we provided an API that could do all that. But how did it work?
In your squads, take ten minutes (and a patch of whiteboard), and write out all the different things that you think that the API for your Tic-Tac-Toe clients needed to do in order to work correctly.
Then, once you've done this, take a second look at what you've written down. Assuming you were to try to build a Ruby program to handle all of these different tasks, what abstractions would you need? How would the different responsibilities be broken up? Make a list. It doesn't matter how "correct" it is. For this part, recall our strategies in modeling. What are the relevant nouns and verbs? Big ideas and details?
Once all groups are finished, we'll discuss our results.
Suppose we wanted to build an API app that records a user's height and weight. It might work as follows:
- A user interacts with a front-end application, triggering a POST, and this POST request contains data - specifically, the latest measurements of height and weight.
- When the POST request is received by the server, the API app parses the request and extracts the relevant information.
- The data from the POST request gets added to our records.
- As confirmation, a JSON with a unique identifier (so we can refer to this specific measurement in the future) gets sent to the front end.
Our app can also retrieve records, like so:
- Something on the front-end (probably driven by user interaction) triggers a GET request that asks for a specific record.
- When the GET request is received by the server, the API app parses the request and identifies which measurement record is being requested.
- The data for the desired measurement gets retrieved from our set of records.
- Finally, the desired data gets put into a JSON and sent back to the front end.
If we were to try to generalize and abstract away the differences between these two steps, we might say that our web application needs to:
- receive incoming requests from a front-end
- execute specific behaviors in response to those requests
- create, read, update, or destroy data records through some kind of data storage system
- share information back to the browser
The quartet of 'create', 'read', 'update', and 'destroy' is commonly known as 'CRUD'; each refers to a specific type of action that can be performed on our data storage system. Each of these types might have more than one specific action associated with it ('read one' vs 'read all', for instance).
One common way of dividing up these four responsibilities is the Model-View-Controller (MVC) architecture pattern. This pattern involves making three core types of components, each responsible for a different part of the API's functionality.
A Model directly manages the data in our application, and provides a representation of that data for the rest of the application to use.
A View is like it sounds - it's data that gets sent back to the client for the user to consume.
A Controller responds to user requests as they come in, and utilizes both the model and the view components to perform the desired behavior and produce a response.
In addition to these three types of components, however, there is a fourth piece that it's important to consider with web development particularly: routing. Routes indicate to the server which controllers should be triggered (and how) by which kinds of requests. It's a critical piece of the puzzle, and one we'll be looking at later today in more detail.
What which part(s) of an HTTP request does the router use to determine which code to run?
MVC architecture is very common in web applications, and Rails gives us the tools to spin up applications that are roughly in line with the idea of MVC.
We're going to act our the various parts of an MVC application. Link up with another squad so that you're in a group of eight, and assign each member of your 'super-group' one of the roles below. These roles are:
-
Client
-
Server
-
Controllers
-
People Controller
-
Places Controller
-
Things Controller
-
Models
-
People Model
-
Places Model
-
Things Model
Once you've divvied up the roles, take two minutes to read through the
directions for your role, found ./roles
. These directions
explain what your responsibilities are and how you should carry them out.
However, you may only communicate with the teammates listed in your role
card.
We'll work through one example request together, and then each group will work independently to answer all remaining requests. Once all groups have finished, we'll have a retro-style discussion about how everything went.
Rails is a web framework - a tool that helps us quickly and easily build web applications - written in Ruby, and designed and created by David Heinemeier Hannson (also known as 'DHH'). Although Rails applications don't match up perfectly with the abstract idea of MVC, their architecture is fairly similar.
Let's take a look at an actual Rails app and see how it lines up. Fork and clone this repo. Recognize it? It's the Tic-Tac-Toe API!
Open it up in Atom. Have a look at the file structure. Let's take a step back and just look at directories at the top level of the app.
.
├── app
├── bin
├── config
├── db
├── lib
├── log
├── public
└── vendor
For now, you only need to think about app
, config
, and db
- you probably
won't be touching any code outside of those three directories. How is that
possible? Because Rails actually builds out most of these files and folders for
you, every time you use it to create a new application. That's why it's called
a 'framework' - it gives you the skeleton for a brand new app, which you can
then customize.
- The
app
directory holds the code for our application itself. We'll be writing a lot of code here. bin
holds the binary code for Rails itself, as well as some of the programs that it depends on - better not touch this directory.config
holds configuration settings for our app and for the things that plug into it. This includes things like environmental variable settings and secret keys, but also things like the routing configuration for our server (which is not strictly part of our Rails app, but which our app uses); theroutes.rb
file, in particular, defines all of the routing for our app.db
holds files related to the structure of the application's database. The database, like the server, is separate from Rails and is not strictly part of the app, so it makes sense to keep this outside of theapp
directory.log
is a place where Rails keeps its log files - records of things that have happened when the app was running. Looking through here can sometimes be helpful if you've hit a bug.public
holds some static HTML pages that anyone can see, regardless of whether or not they're logged into our app. These are typically pages that represent different errors our app might have hit, such as404
.vendor
holds non-gem software that our app will use. Don't worry too much about this one for now.
Let's dive into the app
directory. This app in particular has more going on
than yours probably will, but it still has all the basic components.
./app
├── channels
├── controllers
├── jobs
├── mailers
├── models
├── serializers
└── views
Three of these directories should jump out at you: controllers
, models
, and
views
. Each holds the different Ruby files that Rails uses to handle the
respective responsibilities of MVC.
Don't worry about assets
, serializers
, mailers
, or helpers
for now. In
API-style applications, serializers
replace all the views that we care about
for this program, so we can safely skip the views
folder as well.
Now that we've conceptually wrapped our heads around what goes into a backend lets make one ourselves.
WARNING: Do not ask "why?". None of this should "make sense", yet. Instead,
focus on the commands I'm running and watch how Rails automatically organizes
our code. The rails generate
commands are particularly interesting.
-
Use the quite-excellent Rails API Template, already included in this repository for our convenience.
-
Scaffold our Posts and Comments (Users come with the template!)
bin/rails generate scaffold post title:string body:text user:references
bin/rails generate scaffold comment body:text user:references post:references
-
Now lets create and migrate our database by typing in:
bin/rake db:create
bin/rake db:migrate
-
Start the server!
bin/rails server
-
Now navigate to
localhost:4741/users
(empty brackets is a good sign) -
Let's actually see some data, by adding it to our
db/examples.rb
file.%w(alice bob charlie dana).each do |name| email = "#{name}@#{name}.com" next if User.exists? email: email User.create!(email: email, password: 'abc123', password_confirmation: nil) end %w(alice bob).each_with_index do |name, i| title = "Dear Diary Number #{i}" email = "#{name}@#{name}.com" next if Post.exists? title: title Post.create!(title: title, body: "Another beautiful day!" user: User.find_by(email: email)) end %w(charlie dana).each do |name| body = "Great post!" email = "#{name}@#{name}.com" next if Comment.exists? body: body, email: email Comment.create!(body: body user: User.find_by(email: email)) end
-
Now we have to update the models with relationships. Put the following in the user model (
model/user.rb
):class User < ActiveRecord::Base has_many :posts has_many :comments end
-
And add this to the post model (
model/post.rb
):class Post < ActiveRecord::Base has_many :comments end
-
Now fill the database with examples by running
bin/rake db:examples
-
If we need to reset entirely,
bin/rake db:drop db:create db:migrate db:seed db:examples
-
Try navigating to
localhost:4741/users
or/posts
or/comments
. You should see the JSON you seeded. Try making a curl request to send JSON to your API:curl http://localhost:4741/posts \ --include \ --request POST \ --header "Content-Type: application/json" \ --data '{ "post": { "title": "a sample title", "body": "a sample body" } }'
-
Navigate to
localhost:4741/users
and see what we have. In the serializer files try adding moreattributes
as keys, and see how this changes. (check thedb/schema
file for some ideas.) -
After adding attribute fields to the user, post and comment serializers add a
has_many
relationship to user serializer to match the relationship added to the user model.
Developers should run these often!
bin/rake routes
lists the endpoints available in your API.bin/rake test
runs automated tests.bin/rails console
opens a REPL that pre-loads the API.bin/rails db
opens your database client and loads the correct database.bin/rails server
starts the API.scripts/*.sh
run variouscurl
commands to test the API. See below.
- All content is licensed under a CCBYNCSA 4.0 license.
- All software code is licensed under GNU GPLv3. For commercial use or alternative licensing, please contact legal@ga.co.