Thiago Marinho

June 9, 2022

Intro

I replaced four (sdk, smart-contract, indexer-api, frontend app) GitHub projects with only one using Monorepo / TurboRepo.

I did a post about it. Check it out

The structure of my monorepo:

~/Developer/blog/monorepo (main*) » tree -L 3 --gitignore
.
├── README.md
├── apps
│   ├── frontend
│   │   ├── Procfile
│   │   ├── README.md
│   │   ├── __mocks__
│   │   ├── __tests__
│   │   ├── next-env.d.ts
│   │   ├── next.config.js
│   │   ├── package.json
│   │   ├── public
│   │   ├── src
│   ├── contract
│   │   ├── README.md
│   │   ├── contracts
│   │   ├── hardhat.config.ts
│   │   ├── package.json
│   │   ├── scripts
│   └── backend
│       ├── README.md
│       ├── Procfile
│       ├── build
│       ├── package.json
│       ├── src
│       ├── tsup.config.ts
├── package.json
├── packages
│   ├── contract-types
│   │   ├── README.md
│   │   ├── package.json
│   │   ├── src
│   │   └── tsconfig.json
│   ├── sdk
│   │   ├── README.md
│   │   ├── jest.config.js
│   │   ├── jest.setup.js
│   │   ├── package.json
│   │   ├── src
│   │   └── tsup.config.ts
│   ├── eslint-config-custom
│   │   ├── index.js
│   │   └── package.json
│   └── tsconfig
│       ├── README.md
│       ├── base.json
│       ├── nextjs.json
│       ├── package.json
│       └── react-library.json
├── turbo.json
└── yarn.lock

// Omitting some files and packages unnecessary to this blog post

A little bit of the context

The hard part was hosting, and I'll tell you how to do it on Heroku.

But before, I wish to give you some context on how this project should do in the build process:

The smart-contract should do the build process to generate all types using typechain lib, because instead of using the ABI, I want to use the types (typescript for the win); this build process generates the folder types inside of the smart-contract project; and my script copies this types folder and pastes it to new packages called contract-types (that should be an npm package of types).

The sdk should do the build process and use the contract-types, then the frontend app should build and use the sdk that uses the contract-types.

The indexer-api (backend) should build using the contract-types.

Order of the build:

smart-contract
the types of contract-types should exist
sdk
parallel frontend and backend

TurboRepo does it in a fast and intelligent way without much effort.

There are others things I'm omitting because they are not too important, but we have other packages.

With this in mind, let's see how to set up the Heroku to work on monorepo:

Deployment - Create the Apps

Create two apps on heroku:

frontend - heroku create -a frontend
backend - heroku create -a backend

Add Buildpacks

In both apps, you can connect Heroku apps to the Github. This way, you will save time with CI/CD after committing to the main branch. Also, in both ones, you need to follow the same steps:

Add (GUI: settings -> buildpacks -> Add Buildpack) the buildpacks in this order:

https://github.com/heroku/heroku-buildpack-multi-procfile
heroku/nodejs

Or Heroku CLI:

heroku buildpacks:add -a frontend heroku-community/multi-procfile
heroku buildpacks:add -a frontend heroku/nodejs

heroku buildpacks:add -a backend heroku-community/multi-procfile
heroku buildpacks:add -a backend heroku/nodejs

Create the Procfile

Procfile is a file that receives the commands to run when starting an application; if you have a basic node.js Heroku app, you don't need this once the package.json has the start script instruction.

But in our case, we'll need this for frontend and backend packages:

Frontend:

echo "web: cd apps/frontend && yarn start" > Procfile

Backend:

echo "web: cd apps/backend && yarn start" > Procfile

The command above creates the Procfile file with the content: web: cd apps/backend && yarn start

Setup the new env PROCFILE for file path to Procfile:

Frontend App:

heroku config:set -a frontend PROCFILE=apps/frontend/Procfile

Backend App:

heroku config:set -a backend PROCFILE=apps/backend/Procfile

Setup the root package.json on monorepo

Heroku now knows where to find our Procfiles; however, because we have two separate applications stored within the frontend (client) and backend (server) directories, each has its dependencies.

Heroku typically tries to install dependencies as specified in the package.json at the project's root and will try to run the build script set here. To ensure we install the correct dependencies and run the proper build scripts for our application, we need to specify a heroku-postbuild script at the root of our project.

The secret ingredient of the recipe: In the package.json file in the project root, add the following scripts:

"build:frontend": "turbo run build --filter=frontend",
"build:backend": "turbo run build --filter=backend",
"heroku-postbuild": "if [ $CLIENT_ENV ]; then yarn run prod-frontend; elif [ $SERVER_ENV ]; then yarn run prod-backend; else echo no environment detected, please set CLIENT_ENV or SERVER_ENV; fi",
"prod-frontend": "yarn run build:frontend",
"prod-backend": "yarn run build:backend"

We’ve added three scripts: heroku-postbuild, prod-frontend, and prod-backend.

Heroku will automatically run the heroku-postbuild script for us upon deployment.

Our heroku-postbuild script looks for environment variables $CLIENT_ENV or $SERVER_ENV to determine which script to run prod-frontend or prod-backend.

Setting environment variables on Heroku

Now add new CLIENT_ENV and SERVER_ENV on heroku apps:

Frontend App:

heroku config:set -a frontend CLIENT_ENV=true

Backend App:

heroku config:set -a backend SERVER_ENV=true

Now our heroku-postbuild script will be able to run the correct install scripts for each of our applications on deployment.

See the package.json complete:

{
  "name": "my-monorepo",
  "version": "0.0.0",
  "private": true,
  "workspaces": [
    "apps/*",
    "packages/*"
  ],
  "scripts": {
    "build": "turbo run build",
    "dev": "turbo run dev --parallel",
    "dev:app": "turbo run dev --filter=frontend",
    "lint": "turbo run lint",
    "format": "prettier --write \"**/*.{ts,tsx,md}\"",
    "build:app": "turbo run build --filter=frontend",
    "build:api": "turbo run build --filter=backend",
    "start:app": "turbo run start --filter=frontend",
    "start:api": "turbo run start --filter=backend",
    "heroku-postbuild": "if [ $CLIENT_ENV ]; then yarn run prod-frontend; elif [ $SERVER_ENV ]; then yarn run prod-backend; else echo no environment detected, please set CLIENT_ENV or SERVER_ENV; fi",
    "prod-frontend": "yarn run build:app",
    "prod-backend": "yarn run build:api"
  },
  "devDependencies": {
    "eslint-config-custom": "latest",
    "prettier": "latest",
    "turbo": "latest",
    "tsup": "^5.12.6"
  },
  "engines": {
    "npm": ">=7.0.0",
    "node": ">=8.0.0 <=16.14.2"
  },
  "dependencies": {},
  "packageManager": "yarn@1.22.18",
}

🚨 I recommend not using the caches, but it's not a best practice; it's nice to study a better solution; I was facing issues keeping it true:

heroku config:set USE_YARN_CACHE=false -a frontend
heroku config:set NODE_MODULES_CACHE=false -a frontend
heroku config:set YARN_PRODUCTION=false -a frontend

heroku config:set USE_YARN_CACHE=false -a backend
heroku config:set NODE_MODULES_CACHE=false -a backend
heroku config:set YARN_PRODUCTION=false -a backend

My turbo.json:

{
  "pipeline": {
    "build": {
      "dependsOn": ["^build"],
      "outputs": ["dist/**", ".next/**"]
    },
    "start": {
      "dependsOn": [
        "^build"
      ]
    },
    "start:app": {

    },
    "lint": {
      "outputs": []
    },
    "dev": {
      "cache": false
    }
  }
}