Pong for RESTful APIs (serverless microservices pattern)

Build scalables RESTful apis under serverless arquitecture for AWS (serverless +v1.x), create a good codebase with scalability while the project grow up could require a lot of efford, time and dedication to know how the framework works, often this process of learning tends to be while we're building the product and this require agility and fast learning, customizing the code could be annoying a take more time than expected, that's the reason of this skeleton.

Why Pong?

Have you play ping-pong? is a sport in which two players (frontend and backend) hit a lightweight ball (HATEOAS) back and forth across a table (Restful) using small bats (Vuejs and Serverless Framework).

Requirements

Installation

First, install Yeoman and generator-pong using npm (we assume you have pre-installed node.js and serverless framework).

npm install -g yo
npm install -g generator-pong

Then generate your new project:

yo pong

Generate new functions inside a project using CLI (recomended for overwrite serverless.yml):

$ yo pong:func
? Path (can include parameters) /users/{uid}
? Which HTTP method? GET
? Your function description Get users by uid
   create functions/users/id.js
✔ GET /users/{uid} ► functions/users/id.handler (get-users-id)

Functions also can be nested resources, just specify the url and pong will create the files and nested folders.

$ pwd
/home/code/myproject

dev @ ~/code/myproject
$ yo pong:func
? Path (can include parameters) /users/{uid}/orgs
? Which HTTP method? GET
? Your function description Get users orgs
   create functions/users/orgs/get.js
✔ GET /users/{uid}/orgs ► functions/users/orgs/get.handler (get-users-orgs)

dev @ ~/code/myproject/functions/users
$ ls
total 16K
total 16K
drwxr-xr-x 3 frang 4,0K feb 14 12:34 .
drwxr-xr-x 5 frang 4,0K feb 14 12:32 ..
-rw-r--r-- 1 frang  268 feb 14 12:32 id.js
drwxr-xr-x 2 frang 4,0K feb 14 12:34 orgs <-------------- Nested folder was created

The func subgenerator will save the path with parameters (if have)

New updates & new nice stuff here?

Thank's to Yeoman 🙌 we have a conflict handler out-of-the-box.

The Conflicter is a module that can be used to detect conflict between files. Each Generator file system helpers pass files through this module to make sure they don't break a user file.

To update a project with the latest features in the boilerplate just run yo pong inside the project, the generator must detect and automatically inform that an update will be made, the generator will only update global files, if you have modifed "core" files be careful while overwriting in updates.

dev @ ~/code/my-api
$ yo pong
Project detected, updating the boilerplate files instead...
? Your project name (my-api)

Resolve conflicts (in case that exists) and continue...

Note: it will ask for some fields in case you want to update basic parameters in serverless.yml, in case nothing change, hit enter to use existing previous values.

Now, let's Rock n' roll

The project structure presented in this boilerplate is Microservices Pattern, where functionality is grouped primarily by function rather than resources. Each job or functionality is isolated within a separate Lambda function.

If you wish to read more about this pattern and comparation with others, please check out this awesome writeup by Eslam Hefnawy.

The basic project contains the following directory structure:

.
├── __tests__
├── .vscode
│   ├── debug.js                    # Debugger for vscode <3
│   ├── event.json                  # Parameters for request when debugging
│   └── launch.json
├── functions
│   └── aws-authorizer-jwt          # AWS authorizer for jwt tokens
│       └── handler.js
│   └── firebase-authorizer-jwt     # Firebase authorizer for jwt tokens
│       └── handler.js
│   └── myresource                  # Basic structure of a resource
│       ├── id.js
│       ├── get.js
│       ├── post.js
│       ├── put.js
│       └── delete.js
├── helpers
│   ├── authpolicy.js
│   ├── index.js
│   └── response.js
├── templates
│   ├── request.vtl
|   └── response.vtl
├── tokens
│   ├── aws.json                    # AWS jwks file
│   └── firebase.json               # Firebase tokens
├── .editorconfig
├── .env.yml.example
├── .gitignore
├── LICENSE
├── package.json
├── README.md
├── serverless.yml

Concepts

What is a service? (Api Gateway)

Due to the current limitations where every service will create an individual API in API Gateway (WIP), we'll be working with a unique service with all the functions (resources) that will be exposed.

How to use multiple environment stages (develop, prod, staging)?

The default stage is "develop", for create a new one, use the package serverless-aws-alias and change the value in serverless.yml or pass it as --option when deployment.

Custom with Apache Velocity Templates

Templates are optionals, used ONLY when the integration is lambda, this method is more complicated and involves a lot more configuration of the http event syntax, more info.

The templates are defined as plain text, however you can also reference an external file with the help of the ${file(templates/response.vtl)} syntax, use Apache Velocity syntax for custom.

Configuration files

serverless.yml

The default provider is aws, see documentation for complete list of options available.

package.js (core required packages)

yortus/asyncawait for avoid callback hell in validation helper.
krachot/options-resolver as port of Symfony component OptionsResolver
HyperBrain/serverless-aws-alias enables use of AWS aliases on Lambda functions.
Brightspace/node-jwk-to-pem used to convert jwks to pem file.
mzabriskie/axios Awesome HTTP client for make request.

.env.yml.example

Environment variables used by your function, variables are grouped by stage, so this meas variables will only be available depending of the stage where you defined them, variables are loaded automatically, there is not need to "require a file early as possible", so copy the file IN CASE IF NOT EXISTS (CREATED AUTOMATICALLY BY BOILERPLATE) .env.yml.example to .env.yml and write the real values, depending the value for stage in serverless.yml file, values will be loaded, eg:

Create your final env vars file

$ cp .env.yml.example .env.yml

Now add the values per stage

develop:
  AWS_SECRET_KEY: dontsavethiscredentialsstringsincode

prod:
  AWS_SECRET_KEY: dontsavethiscredentialsstringsincode

And access them natively in you code from process.env:

module.exports.handler = (event, context, callback) => {
    console.log(process.env.AWS_SECRET_KEY)
}

.env.yml.example is added to VCS for keep reference of the variables, not values (good practice). .env.yml is not uploaded either aws when create the package or vcs.

Some helpers...

Helpers are just custom reusable functions for facilitate some repetitive tasks like validations, custom response, etc.

Here the current availables:

jwks-to-pem.js if a helper file to convert AWS jwks json to pem file used in aws-authorizer-jwt function, eg: in root inside a project:

$ node ./helpers/jwks-to-pem.js <url to jwks.json>

jwks.json is usually located in https://cognito-idp.{region}.amazonaws.com/{userPoolId}/.well-known/jwks.json.

This will generate a json file with the pem keys in it, aws-authorizer-jwt use this file to authenticate using JSON Web Tokens with cognito integration for secure your API resources, more info.

The authorizer needs to khow the iss of the token, so add the value to .env.yml replacing the values of region, userPoolId, like this:

develop:
  AWS_ISS: https://cognito-idp.{region}.amazonaws.com/{userPoolId}

If a function needs to be secured using aws jwt authorizer, remember to add it inside the function template in serverless.yml file, eg like this:

  get-users-orgs:
    name: test-api-get-users-orgs
    description: Get users orgs
    handler: functions/users/orgs/get.handler
    events:
      - http:
          path: '/users/{sub}/orgs'
          method: GET
          cors: true
          authorizer: aws-authorizer-jwt

And that's it, API Gateway will run the authorizer before the lambda execution automatically 💃

validate() this method return a Promise and throw an Error if the validation fails.
response() is a shorcut for the callback received in the lambda handler, but this add the json body for integration response in API Gateway at the same time implementing JSON API standard, eg:

Samples of the response using lambda-proxy integration, more info of integrations.

const { response } = require('path/to/helpers')

response(201)
// {"statusCode": 201, "body": "{\"data\":null}","headers": {}}

response(403, new Error('my custom error message'))
// {"statusCode": 403, "body": "{\"error\":{\"title\":\"my custom error message\",\"meta\":{}}}", "headers":{}}

response(501, {key: 'value'})
//{"statusCode": 501, "body": "{\"data\":{\"key\":\"value\"}}", "headers": {}}
 
response(403)
// {"statusCode": 403, "body": "{\"data\":null}", "headers": {}}

response()
// Error - Invalid arguments supplied for response

[WIP] Customization of header using the new response is not supported for now...

resolver just an object to interact with krachot/options-resolver For use response helpers it's extremely required add this code at the very begining of the handler, the reason is that response helper use the lambda callback function for finish the execution of the lambda and is not cool always send it by parameter...

module.exports.handler = (event, context, callback) => {
    // needed for response scope
    global.cb = callback

The other issue is related to request body, from Serverless docs and AWS Developer Guide:

Note: When the body is a JSON-Document, you must parse it yourself

{
    "resource": "Resource path",
    "path": "Path parameter",
    "httpMethod": "Incoming request's method name",
    "headers": {},
    "queryStringParameters": {},
    "pathParameters":  {},
    "stageVariables": {},
    "requestContext": {},
    "body": "---------------A JSON STRING OF THE REQUEST PAYLOAD.-------------------",
    "isBase64Encoded": "A boolean flag to indicate if the applicable request payload is Base64-encode"
}

That means we must to parse the body received, in every functions, is not an object, is an string, so

module.exports.handler = (event, context, callback) => {

    // needed for response scope
    global.cb = callback

    // parse the body string to object
    let body = JSON.parse(event.body)

    ...

After adding the code bellow, just import the helper lib built-in and that's it... ^_^

const { validate, resolver, response } = require('../../helpers')

module.exports.handler = (event, context, callback) => {
    // needed for response scope
    global.cb = callback

    // parse the body string to object
    let body = JSON.parse(event.body)

    // marking as required some parameters
    resolver.setRequired([
        'email',
        'password'
    ])

    // { email: '[email protected]' }
    validate(event)
        // all good!
        .then((body) => console.log('passed!'))
        /*
            400 Bad Request
            {"message": "The required options \"password\" are missing"}
        */
        .catch((err) => response(err)) 
}

Development

Install dependencies, Run test:

npm install
npm run test

Cheap and reliable Node.js hosting starts at $3/month, and $1/month static HTML hosting

frangeris / pong

Programming Languages

Labels

Projects that are alternatives of or similar to pong