GoGovSG

The official Singapore government link shortener.

Content

• GoGovSG
  • Content
  • Introduction
  • Getting Started
    • Running Locally
    • Setting up the infrastructure
      • Server
      • Serverless functions for link migration
      • Batch functions for backups
    • Deploying
  • Operations
    • Transferring links to a new owner or email address
  • Developer Documentation
    • Folder Structure
    • Asset variants
    • Babel
    • ESLint
    • Webpack
    • Webpack dev server
    • ts-node-dev
    • Express
    • Concurrently
    • VSCode + ESLint
    • Redux Devtools
    • Infrastructure

Introduction

GoGovSg is the official Singapore government link shortener, built by the Open Government Products team in GovTech. This repository serves as the codebase to serve three link shortener environments: Go.gov.sg, for.edu.sg, and for.sg.

There are multiple reasons why we built an official government link shortener:

• URLs are too long to fit into tweets or SMSes, and difficult to remember
• Email clients might block other commercial link shorteners if they are listed as spam on their site
• Citizens are afraid of phishing when receiving a shortened link and unsure of where it goes

With GoGovSg, citizens are safe in the knowledge that the links are official and safe. Any authorized user can log in with their government emails and immediately create authenticated and recognisable short links.

Getting Started

Make sure you have node version 14, docker-compose version >= 1.23.1 and Docker version >= 18.09.0 installed.

*For Mac computers with Apple silicon, you will need Rosetta to be installed in order for docker-compose@v1 to work. You can do so using the following command: softwareupdate --install-rosetta.

Start by cloning the repository and installing dependencies.

# Clone the repository
git clone git@github.com:opengovsg/GoGovSG.git gogovsg

# Go inside the directory
cd gogovsg

# Install dependencies
npm install

Running Locally

npm run dev

Docker-compose will spin up a postgresql database and redis container to be connected to the backend server. Once the setup is complete, the local version of GoGovSG can be accessed on your browser via localhost:3000. Note that 3000 is the port number that the webpack dev server listens on; the backend server actually listens on port 8080 instead.

Because redirects are served directly from the backend, shortlinks can be accessed via localhost:3000/shortlink, but that is really being proxied to localhost:8080/shortlink. One-time passwords for all log-in attempts on localhost are obtained using maildev and accessed via http://localhost:1080/.

Setting up the infrastructure

Much of this step will involve setting up key infrastructure components since we do not have docker-compose to do that for us. On top of running the server, GoGovSG minimally requires the following infrastructure to be available:

• A PostgreSQL database (for storing short-long URL mappings)
• A Redis server (transient storage of sessions, one-time passwords, click statistics and frequently used shortlinks)

Other optional infrastructure used in GoGovSG:

• Serverless functions (for migrating user links)
• Batch jobs (for backups of our database to external source)

After these have been set up, set the environment variables according to the table below:

Server

Environment Variable	Required	Description/Value
NODE_ENV	Yes	production
DB_URI	Yes	The postgres connection string, e.g. postgres://postgres:postgres@postgres:5432/postgres
OG_URL	Yes	The origin url, used for both google analytics and circular-redirect prevention. E.g. https://go.gov.sg
AWS_S3_BUCKET	Yes	The bucket name used for storing file uploads.
REDIS_OTP_URI	Yes	Redis connection string, e.g. redis://redis:6379/0
REDIS_SESSION_URI	Yes	Redis connection string, e.g. redis://redis:6379/1
REDIS_REDIRECT_URI	Yes	Redis connection string, e.g. redis://redis:6379/2
REDIS_STAT_URI	Yes	Redis connection string, e.g. redis://redis:6379/3
REDIS_SAFE_BROWSING_URI	Yes	Redis connection string, e.g. redis://redis:6379/4
SESSION_SECRET	Yes	For hashing browser sessions, e.g. change-this
VALID_EMAIL_GLOB_EXPRESSION	Yes	The glob expression used to test if a provided email address is valid. For safety, we have disabled the use of negations, ext-glob, glob stars (**) and braces, e.g. *@youremaildomain.com
GA_TRACKING_ID	No	The Google Analytics tracking ID, e.g. UA-12345678-9
SENTRY_AUTH_TOKEN	No	To get relevant permissions to upload the source maps.
SENTRY_DNS	No	The Sentry DNS used for bug and error tracking. e.g. https://12345@sentry.io/12345
SENTRY_ORG	No	Our Sentry organisation name, e.g. example-org
SENTRY_PROJECT	No	The relevant Sentry project. e.g. project-prod
SENTRY_URL	No	The Sentry url. e.g. https://sentry.io/
LOGIN_MESSAGE	No	A text message that will be displayed on the login page as a snackbar
USER_MESSAGE	No	A text message that will be displayed as a banner, once the user has logged in
ANNOUNCEMENT_MESSAGE	No	The message in the announcement displayed as a modal to users on login
ANNOUNCEMENT_TITLE	No	The title in the announcement displayed as a modal to users on login
ANNOUNCEMENT_SUBTITLE	No	The subtitle in the announcement displayed as a modal to users on login
ROTATED_LINKS	No	List of comma separated path of links to rotate on the landing page
ANNOUNCEMENT_URL	No	The hyperlink for the button in the announcement displayed as a modal to users on login
ANNOUNCEMENT_IMAGE	No	The image in the announcement displayed as a modal to users on login
CSP_REPORT_URI	No	A URI to report CSP violations to.
CSP_ONLY_REPORT_VIOLATIONS	No	Only report CSP violations, do not enforce.
CLOUDMERSIVE_KEY	No	API key for access to Cloudmersive.
SAFE_BROWSING_KEY	No	API key for access to Google Safe Browsing.
SAFE_BROWSING_LOG_ONLY	No	Boolean, whether to log only, or throw error if unsafe link is found by Google SafeBrowsing. Defaults to false
ASSET_VARIANT	Yes	Asset variant specifying environment for deployment, one of gov, edu, health
COOKIE_MAX_AGE	Yes	Session duration of cookie
BULK_UPLOAD_MAX_NUM	No	Maximum number of links that can be bulk uploaded at once. Defaults to 1000
BULK_UPLOAD_RANDOM_STR_LENGTH	No	String length of randomly generated shortUrl in bulk upload. Defaults to 8
REPLICA_URI	Yes	The postgres connection string, e.g. postgres://postgres:postgres@postgres:5432/postgres

Serverless functions for link migration

Secrets	Required	Description/Value	Shared across environments
DATABASE_URL	Yes	The postgres connection string, e.g. postgres://postgres:postgres@postgres:5432/postgres	No

Batch functions for backups

Secrets	Required	Description/Value	Shared across environments
DB_URI	Yes	The postgres connection string, e.g. postgres://postgres:postgres@postgres:5432/postgres	No
GCS_CREDENTIALS	Yes	Authorization credentials for writing to backup buckets in GCS	Yes

Environment Variable	Required	Description/Value	Shared across environments
GCS_BUCKET	Yes	Name of bucket in GCS to write to	No
CRONITOR_MONITOR_CODE	No	ID for Cronitor monitor to monitor batch jobs	No

Trigger the typescript compilation and webpack bundling process by calling npm run build.

Finally, start the production server by running npm start.

Deploying

GoGovSG uses Github Actions and Serverless to deploy to AWS Elastic Beanstalk and AWS Lambda. We also use Sentry.io to track client-side errors.

Secrets	Required	Description/Value
AWS_ACCESS_KEY_ID	Yes	AWS credential ID used to deploy to Elastic and Modify files on S3
AWS_SECRET_ACCESS_KEY	Yes	AWS credential secret used to deploy to Elastic Beanstalk and Modify files on S3
SENTRY_AUTH_TOKEN	No	To get relevant permissions to upload the source maps
GITHUB_TOKEN	Yes*	Used by Coveralls to verify test coverage on repo. Does not need to be manually specified as it is specified by Github Actions. More Info
DD_API_KEY	Yes*	Datadog API Key used for integration with Datadog to Trace/Logs collection
DD_SERVICE	No	Datadog service name to be used for the application
DD_ENV	No	Datadog application environment, e.g. staging, production

Environment Variable	Required	Description/Value
EB_ENV_(EDU_/HEALTH_)PRODUCTION, EB_ENV_(EDU_/HEALTH_)STAGING	Yes	Elastic Beanstalk environment name
EB_APP_PRODUCTION, EB_APP_STAGING	Yes	Elastic Beanstalk application name
EB_BUCKET_PRODUCTION, EB_BUCKET_STAGING	Yes	S3 bucket used to store the application bundle
PRODUCTION_BRANCH, STAGING_BRANCH	Yes	Name of Git branches for triggerring deployments to production/staging respectively
ECR_URL	Yes	AWS ECR Docker container registry URI to push built images to
ECR_REPO	Yes	Name of repository in AWS ECR containing images
SENTRY_ORG	No	Sentry.io organisation name
SENTRY_PROJECT_PRODUCTION, SENTRY_PROJECT_STAGING	No	Sentry.io project name
SENTRY_URL	No	Sentry.io URL e.g. https://sentry.io/
SENTRY_DNS_PRODUCTION,SENTRY_DNS_STAGING	No	Sentry.io endpoint to post client-side errors to
API_KEY_SALT	Yes	Salt used for APIKey hashing, guide to salt generation here

Operations

Transferring links to a new owner or email address

Functions to safely transfer links to new owners can be accessed on AWS Lambda console (for authorized users only).

To transfer a single link to a new email address (must be lowercase), please use the relevant Lambda function (gogov-production, edu-production, health-production) to create an event with the following event body:

{
  "shortUrl": "<short url to be transfered>",
  "toUserEmail": "<user email to transfer to>"
}

To transfer all links belonging to an account to another account, please use the relevant Lambda function (gogov-production, edu-production, health-production) to create an event with the following event body:

{
  "fromUserEmail": "<user email to transfer from>",
  "toUserEmail": "<user email to transfer to>"
}

Developer Documentation

Folder Structure

All source code resides in the src directory. Inside src, there is client and server directory. Frontend code (react, css, js and other assets) will be in client directory. Backend Node.js/Express code will be in the server directory.

Asset variants

This repository serves as the codebase to serve three link shortener environments: Go.gov.sg, for.edu.sg, and for.sg. These environments are run on separate infrastructure, and the deployment pipeline is set up to deploy any code changes in this codebase across all infrastructure environments. The environments are identical apart from the assets, copy and list of authorized users.

Babel

Babel helps us to write code in the latest version of JavaScript. If an environment does not support certain features natively, Babel will help us to compile those features down to a supported version. It also helps us to convert JSX to Javascript. babel.config.json file is used to describe the configurations required for Babel.

Babel requires plugins to do the transformation. Presets are the set of plugins defined by Babel. Preset env allows to use babel-preset-es2015, babel-preset-es2016, and babel-preset-es2017 and it will transform them to ES5. Preset react allows us to use JSX syntax and it will transform JSX to Javascript.

ESLint

ESLint is a pluggable and configurable linter tool for identifying and reporting on patterns in JavaScript.

.eslintrc.json file (alternatively configurations can we written in Javascript or YAML as well) is used describe the configurations required for ESLint.

I am using Airbnb's Javascript Style Guide which is used by many JavaScript developers worldwide. Since we are going to write both client (browser) and server side (Node.js) code, I am setting the env to browser and node. Optionally, we can override the Airbnb's configurations to suit our needs. I have turned off no-console, comma-dangle and react/jsx-filename-extension rules.

Webpack

Webpack is a module bundler. Its main purpose is to bundle JavaScript files for usage in a browser.

webpack.config.js file is used to describe the configurations required for webpack.

1. entry: entry: ./src/client/app/index.tsx is where the application starts executing and webpack starts bundling. Note: babel-polyfill is added to support async/await. Read more here.
2. output path and filename: the target directory and the filename for the bundled output
3. resolve: We use aliasing at bundle time to inject and resolve the right asset variant path, which allows us to easily switch between asset folders for the different environments.
4. module loaders: Module loaders are transformations that are applied on the source code of a module. We pass all the js file through babel-loader to transform JSX to Javascript. Fonts and images are loaded through file-loader.
5. Dev Server: Configurations for the webpack-dev-server which will be described incoming section.
6. plugins: clean-webpack-plugin is a webpack plugin to remove the build folder(s) before building. html-webpack-plugin simplifies creation of HTML files to serve your webpack bundles. It loads the template (public/index.html) and injects the output bundle.

Webpack dev server

Webpack dev server is used along with webpack. It provides a development server that provides live reloading for the client side code. This should be used for development only.

The devServer section of webpack.config.js contains the configuration required to run webpack-dev-server which is given below.

devServer: {
    port: 3000,
    open: true,
    proxy: {
        "/api": "http://localhost:8080"
    }
}

Port specifies the Webpack dev server to listen on this particular port (3000 in this case). When open is set to true, it will automatically open the home page on startup. Proxying URLs can be useful when we have a separate API backend development server and we want to send API requests on the same domain. In our case, we have a Node.js/Express backend where we want to send the API requests to.

ts-node-dev

ts-node-dev will monitor for any changes in the server source code and automatically restart the server. This is used in development only. It watches changes to all files required from the entry point onwards, and will restart the node server whenever such files are modified.

ts-node-dev also allows debugging via Inspector. The quickest way to debug the application will be to use Chrome DevTools via chrome://inspect. VSCode users may want to add the following to their launch.json to quickly attach VSCode to the application for debugging:

    {
      "type": "node",
      "request": "attach",
      "name": "Inspect",
      "protocol": "inspector",
      "port": 9229,
      "restart": true,
      "cwd": "${workspaceFolder}"
    }

Express

Express is a web application framework for Node.js. It is used to build our backend API's.

src/server/index.ts is the entry point to the server application. This starts a server and listens on port 8080 for connections. It is also configured to serve the static files from dist directory.

Concurrently

Concurrently is used to run multiple commands concurrently. I am using it to run the webpack dev server and the backend node server concurrently in the development environment. Below are the npm/yarn script commands used.

"client": "webpack-dev-server --mode development --devtool inline-source-map --hot",
"server": "nodemon src/server/index.js",
"dev": "concurrently \"npm run server\" \"npm run client\""

VSCode + ESLint

VSCode is a lightweight but powerful source code editor. ESLint takes care of the code-quality.

Redux Devtools

Developer Tools to power-up Redux development workflow.

It can be used as a browser extension (for Chrome, Edge and Firefox), as a standalone app or as a React component integrated in the client app.

Infrastructure

Diagrams for our infrastructure setup can be found here.

Mics

Salt generation

let salt = bcrypt.genSaltSync(10)