Skip to content


Folders and files

Last commit message
Last commit date

Latest commit



42 Commits

Repository files navigation


Dead simple enumerated types for Ruby.


This gem implements the familiar notion of enumerated types in Ruby.

"But this is Ruby," you say, "where we haven't any use for such things." Yep. In Ruby, you can get a long way without any formalized concept of enumerated types by using symbols. Let's take the fairly typical example of a "status" field on the Job class:

class Job
  attr_reader :status

  def initialize
    @status = :pending

  def process
      # Do work here...
      @status = :success
      @status = :failure

At first pass this seems fine. Any code that needs to act based on a job's status has to have magic symbols (i.e. job.status == :success), but maybe that's OK for a little while. Later, though, we might want to add a little logic around the Job's status, something like:

# In something like a JobNotifier class
if job.status == :failure or job.status == :success
  # Email user to let them know about their job

At this point, its starting to feel like a little bit too much knowledge about the Job has slipped into other classes; any changes to the in the way status is handled in Job will require change in other classes because they've been exposed to the details of its implementation. What, for example, if we want to add another "finished" state that a user should be notified of (say, :partial_success)? To deal with this we might create a predicate method that lets you interrogate a Job more abstractly about it's status:

class Job
  # ...
  def done?
    [:failure, :success, :partial_success].include?(@status)

Now, say we need another kind of job: the AdminJob. It needs to have the same set of statuses (with the same behavior as Job's status). We could certainly move the status related code into a StatusHaving module and mix it in to both Job and AdminJob, but there are some drawbacks here, chief among them that we'd have to add a good bit of coupling between the Job, AdminJob and the StatusHaving mix-in module. For example both classes and the mix-in would need to agree on the @status instance variable. I would argue at this point the idea of a JobStatus should be promoted to its own class, maybe with a little bit of error checking:

class JobStatus
  NAMES = [:pending, :success, :failure]

  def initialize(name)
    unless NAMES.include?(name)
      raise"Invalid status #{name.inspect}")

    @name = name

and then in Job:

class Job
  def initialize
    @status =

There are some advantages to this approach:

  1. The list of all the possible statuses lives in only one place. I think it's way easier to look at the JobStatus class and see what the possible statuses are than it is to hunt through the Job class Looking for symbols assigned to @status (or, even worse, to look outside the Job class looking for assignments to Job#status).
  2. It's now possible to interact with the list of all legal statuses programmatically (perhaps in an admin console that has a drop down for of all statuses so that they may be manually updated).
  3. We can separate behavior (methods) that apply to the enumerated types from the classes that include one of these types (Job in our example). Although the status handling code in our version of Job was fairly simple, it often becomes clear that handling the status of a job is a very separate concern from the actual processing of a Job, and thus implementing both in a single class becomes a Single Responsibility Principle violation.
  4. By separating the JobStatus behavior from the Job, we're able to more easily respond to change in requirements around the JobStatus in the future. Perhaps at some point we'll want to transform JobStatus into a state machine where only certain transitions are allowed, or include code that audits the change of state, etc.

"But all did was explain enumerated types, the kind that are present all over the place in other languages." Yep. That is correct. The only downside of the approach shown is that you have to re-create very similar, one-off implementations of an enumerated type. The EnumeratedType gem is just a clean and simple Ruby implementation of a well known concept.


Define an enumerated type:

class JobStatus
  include EnumeratedType

  declare :pending
  declare :success
  declare :failure

Get an instance of an enumerated type:

# Via constant...
@status = JobStatus::PENDING

# Or via symbol...
@status = JobStatus[:pending]
@status = JobStatus[:wrong] #=> raises an ArgumentError

All instances have predicate methods defined automatically:

@status = JobStatus::PENDING
@status.pending? # => true
@status.failure? # => false

Get the original symbol used to define the type: # => :pending

A class that includes EnumeratedType is just a regular Ruby class, so feel free to (and definitely do) add your own methods:

class JobStatus
  def done?
    success? or failure?

The only exception exception to the "just a regular ruby class" rule is that the constructor ( is privatized.

Classes that include enumerated types themselves become enumerable, so you can do things like: { |n| puts "JobStatus: #{n}" }

Bonus Features

Create enumerated types with ultra-low ceremony:

JobStatus =, :success, :failure)

Add arbitrary attributes:

class JobStatus
  include EnumeratedType

  declare :pending, :code => 1, :message => "Your Job is waiting to be processed"
  declare :success, :code => 2, :message => "Your Job has completed"
  declare :failure, :code => 3, :message => "Oops, it looks like there was a problem"

JobStatus::SUCCESS.message # => "Your job has completed"

Coerce from other types:

JobStatus.coerce(:pending)           # => #<JobStatus:pending>
JobStatus.coerce("pending")          # => #<JobStatus:pending>
JobStatus.coerce(JobStatus::PENDING) # => #<JobStatus:pending>
JobStatus.coerce(nil)                # => raises a TypeError
JobStatus.coerce(1)                  # => raises a TypeError
JobStatus.coerce(:wrong)             # => raises an ArgumentError

.coerce is particularly useful for scrubbing parameters, allowing you to succinctly assert that arguments are valid for your EnumeratedType, while also broadening the range of types that can be used as input. Using .coerce at the boundaries of your code allows clients the freedom to pass in full fledged EnumeratedType objects, symbols or even strings, and allows you to use the .coerced input with confidence (i.e without any type or validity checking beyond the call to .coerce).

Look up an enum by a property:, 2) # => #<JobStatus:success>, 4) # => raises a TypeError

The first parameter specifies the method that will be called on instances of your EnumeratedType during lookup. This works for attributes specified with declare (such as :code above) but also for any instance method on your EnumeratedType.

class JobStatus
  def display
end, "pending-1") # => #<JobStatus:pending>

If more than one instance of your EnumeratedType matches, the first match will be returned, in the order the types were declared.

If you pass a bock to .by, it will be executed and its value returned (just like Hash#fetch). This is useful for defining a default value or a custom error message., "does-not-exist") { JobStatus::FAILURE }  # => #<JobStatus:failure>, "does-not-exist") { |v| raise("I have no idea what to do with #{v.inspect}") } # => raises an exception


To run the tests (assuming you have already run gem install bundler):

bundle install && bundle exec rake