Installing PostgreSQL for Rails development on OS X

Edit: Some of the commands below have changed from their initial versions to reflect things I’ve learned since getting it all up and running. It has also been broken out into sections.

I’m playing around with writing a Rails application, just to learn new things. At the same time, I wanted to understand why everyone in the Rails community seems to be using Heroku. And since Heroku defaults to using PostgreSQL for the database backend when you use Rails, I have to now set up a PostgreSQL environment on my Mac for development. This turned out to be harder than I first expected. This is how I got it to work.

First, I had to realize that PostgreSQL doesn’t just magically get installed when you create a Rails application that is configured to use it :wink: This was pretty easy because the first time I attempted to launch the “Hello, world!” type thing I had written I got the pretty red Rails error message page saying it couldn’t connect to the database. I quickly googled1 for instructions on how to properly install PostgreSQL for Rails development on the Mac and was quite happy to find a StackOverflow answer for this very issue. I thought I was done, but it seems that as the answer was posted over three years ago … let’s just say that the contents may have settled during transit :laughing: But I was able to use it as a template for how I got things to work and each time I got stuck, it helped me find my way.

Installation and Configuration of PGSQL

Instead of downloading the installer and using that, I used Homebrew to install PostgreSQL:

$ brew install postgresql

Then I needed to initialize a database environment2 since PostgreSQL needs you to tell it where to put the databases you create. The Homebrew recipe was kind enough to give me the command:

$ initdb /usr/local/var/postgres -E utf8

Next, I started the server using the command that initdb conveniently handed me:

$ postgres -D /usr/local/var/postgres

In the future, I’ll probably use the pg_ctl command to start and stop the server process in the background. But for today, this was good enough.

Configuration of the Project to Work with PGSQL

At this point, I could actually start messing around with the server and getting things to communicate between Rails and PostgreSQL. Or so I thought. It turns out that Rails, by default, attempts to communicate with PostgreSQL via domain sockets. This is a good thing because they’re system local and are much more performant than TCP/IP connections. On the other hand, Rails expects the domain sockets to be created in one directory by default and my PostgreSQL installation was using a slightly different one. After taking a stab at reconfiguring the directory used, I just told Rails to connect via TCP/IP by editing the config/database.yml file and adding the following line to the development section:

host: localhost

Which is when I started getting database errors! Progress! :sunglasses:

The database errors told me first that the user Rails was trying to connect with didn’t exist. So I created it by entering:

$ createuser --createdb username

The user needs to be created with the createdb permission so that when executing unit tests, the test database can be dropped and created. I’ve since created a db:create_user Rake task in my project to perform this for me.

In Homebrew installations it seems that all of the PostgreSQL administration commands are also symlinked to the /usr/local/bin directory. Very convenient!

Once that was complete, I could execute the standard Rails commands:

$ rake db:create
$ rake db:migrate
$ rails s

to start the server and everything was up and running!

To tell the truth, I was really impressed with how easy everything was once I used Homebrew to install PostgreSQL. One thing led me to another and in many cases if the next step wasn’t obvious, it was provided in the output of the previous step. Hats off to the Homebrew and PostgreSQL folks! :beer:

  1. Am I supposed to capitalize it when it is being used as a verb or not? What is the convention on that? Hmmmmm … 

  2. The documentation calls it a “database cluster”, but I find that to be a poor choice of phrase given that most people who aren’t database wonks use the term “database” and “server” interchangeably. 


Copyright © 2010-2018 by Lee Dohm