How Git Subtree Helped Me Deploy to Heroku

Posted by hcarnes on August 14, 2018

I'm currently looking to join an awesome team of developers to continue my career development. If that's your team, please book a meeting with me at your convenience.

One thing to consider when you start developing an application is the organization of your git repos. Monorepo is an organization structure where all of your code is one repo, and multi-repo involves organizing code across multiple repos. When I built Greenthumb Gardens, I used something similar to the monorepo structure because I nested the client in a folder inside of the Rails API. Typically, monorepos are a bunch of different high level folders in the same repo.

If you’ve ever deployed to Heroku, you know that you need a repo for your application. Well, I had 2 applications that I needed to deploy (frontend and backend), but I only had 1 git repo. This was a problem. With that said, this post will explain how I used git subtree to deploy my app to Heroku, despite the fact that the client and server were in the same repo.

Feel free to check out Greenthumb Gardens. Unfortunately, due to changes in the Google Maps API, the map now has some restrictions. Womp womp. On the bright side, the garden search and plant adding feature still works.

Here is quick tip on changing from Sqlite3 to Postgres. In order to deploy the backend to Heroku, I had to change my database from Sqlite3 to Postgres. In the gemfile, I changed gem sqlite3 to gem pg. Next, I updated theconfig/database.yml file. Instead of manually changing anything that referenced Sqlite3 to Postgres, I generated a new Rails app and installed Postgres as its database (I ran rails new foobar --database=postgresql), so that I could simply copy the fresh config/database.yml over to my current project. I ran rake db:create and then rake db:migrate to create a new db and ensure my changes were updated.

Using git subtree

As I mentioned before, my client app was in a folder nested inside of the backend app, so I had to figure out a way to create a new repo within my main repo. That’s where git subtree comes in. git subtree allowed me to nest one repository inside another as a sub-directory.

First, I created a new Heroku repo for the client and pushed it the remote.

heroku create greenthumb-gardens --remote heroku-frontend

Next, I used git subtree push to push that repo to the remote instance of the Heroku application. I included --prefix to specify the path in the repo to the subtree I wanted to manipulate.

git subtree push --prefix greenthumb-client/ heroku-frontend master

Next, I created a new Heroku repo for the backend and pushed it the remote.

heroku create greenthumb-gardens-backend --remote heroku-backend

Next, I pushed the repo to the instance of the Heroku application.

git push heroku-backend heroku:master

Getting ready for production

Now that the backend app is running using a separate repo, I set the new environment variable in the greenthumb-client/.env.production file.

REACT_APP_BACKEND_HOST="https://greenthumb-gardens-backend.herokuapp.com"

I also updated the Redux action file that made an async request to the backend application.

export function addPlant(plantToAdd) {
  return async (dispatch) => {
    dispatch({ type: 'START_ADDING_PLANT' });
    const plantJSON = JSON.stringify({ plant: plantToAdd })
    const plantResponse = await fetch(`${process.env.REACT_APP_BACKEND_HOST}/plants`, {
      headers: {
        'Accept': 'application/json',
        'Content-Type': 'application/json'
      },
      body: plantJSON,
      method: "post"
    })
    const plant = await plantResponse.json()
    dispatch({ type: 'ADD_PLANT', plant })
  };
}

Next time

Next time I am creating a relatively simple full stack applicaiton, I would start out by creating different repositories for the frontend and backend. Also, I would deploy them to Heroku at the beginning of the development process instead of waiting until the end.