Geek glossary: re-entrant and idempotent

June 23, 2013 Andrew Bruce

Whilst writing some Chef recipes for our project’s Continuous Integration server the other day, my pair and I came across a commit message to some third party code that claimed to make a routine re-entrant. We both realised that we didn’t clearly understand the difference between re-entrancy and idempotency and decided to look the terms up.

Here are my rough re-definitions of the two terms, in the context of Chef, for the wiki-weary:

The ability for a routine to complete successfully after a previous, interrupted call.
The ability for a routine to be called multiple times, producing the same side effects.

Note the difference between mathematical or functional programming idempotency and the kind of idempotency we care about when writing Chef recipes: a system for automating machine configuration, such as Chef, necessarily produces side effects on the machine being configured. We’re not concerned about the result of pure functions, here.

For the uninitiated, a Chef recipe is a declaration of the configuration of some computer software, written in a Domain Specific Language (DSL). The DSL is useful because it hides details. When configuring a new server with an existing recipe, one would ideally like to ignore details specific to, for example, the distribution or version of the operating system. Perhaps more importantly for the system’s maintainer, however, the details of whether certain aspects of the software have already been configured (or failed to be configured) should also be hidden.

Idempotency is commonly used to describe the ideal Chef recipe. One advantage of having a Domain Specific Language (DSL) to create scripts that set up a machine is to reduce the noise involved with becoming idempotent. The built-in Chef methods are idempotent by default.

It’s the difference between having to write:

unless File.exists?('/etc/mysql/my.cnf')


cookbook-file "/etc/mysql/my.cnf" do
  path "my.cnf"

For our Chef recipes, re-entrancy is important for recovering from failure in the middle of our scripts, whereas idempotency (which incorporates re-entrancy) is the property that allows our scripts to work when a system is first configured, and all of the times after that, without causing unwanted side-effects on each subsequent call.

About the Author


More Content by Andrew Bruce
A Rubyist Learning Go – A basic Go program
A Rubyist Learning Go – A basic Go program

Starting July 1st, I am going to have the opportunity to step outside by development comfort zone and begin...

Ember.js and SoundManager2, Part 2 – Tests and StateManager
Ember.js and SoundManager2, Part 2 – Tests and StateManager

If you have read the first part of this series you may have noticed an issue with the implementation propos...

Enter curious. Exit smarter.

Register Now