Charlie Massry

Metaprogramming with Method Missing

November 12, 2014

In Ruby, you can extend the language significantly to suit your preferences, and even overide methods already used by the language itself. This is typically derogatorily called Monkey Patching as I had covered previously. One significant feature that I think all Ruby developers should know about is how the error NoMethodError gets generated. Because Ruby uses classical inheritance, all classes have ancestors. And as you may already know, everything in Ruby is an object, even classes are objects. Because of this, classes you create come with methods already defined on them. One method is called #method_missing. This very important method is defined on BasicObject, which is the first Ruby object, and all of its children get access to this method. The way Ruby uses this method is if a method is called on a Ruby object that doesn’t have that method defined, it calls #method_missing which walks up the chain of inheritance to BasicObject to see where #method_missing is defined. The way it is defined on BasicObject would look something like this if Ruby was written in Ruby.

class BasicObject
  def method_missing
  raise NoMethodError
  end
end

The reason why the language works like this is so you can override it and even dynamically create an API so you can override #method_missing Let’s say you wanted to write an API where the user can type in a camelcased number to access its index in an array.

class Array
  def method_missing(name)
    index = EnglishNumber.word_to_number(name.to_s) - 1
    self[index] || raise
    rescue
    super
  end
end

array = [1] array.one # => 1 array.two # => NoMethodError undefined method ‘two’ for [1]:Array

Let’s suppose that EnglishNumber.word_to_number takes a camelcased string, and converts it to a number. If you call a number on the array it will return the index or give a NoMethodError.

When #one is called on an Array instance, the interpreter checks for #one and if it doesn’t find it, it then checks if #method_missing is defined, and if it is, it is called. This method, like all Ruby methods, can be overridden. In case there is no index on the array, you might want to throw an error, self[index] will return nil if the array doesn’t have that index. You can call raise to throw a generic error, but rescue will handle that error by passing it off up the chain of inheritance. This will make sure that you will get what you are expecting, as opposed to just getting nil, as that is very nondescript, and NoMethodError is very descriptive. While this can seem like a very contrived example, there are some very powerful libraries that take control of this feature such as ActiveRecord, specifically how .find_by worked in Rails 3.

In Rails 3, there was no .find_by method but a .find_by_#{attribute} where attribute was defined by #method_missing. This was a very cool feature, but it can lead to a slowdown in performance as every time the Ruby interpreter has to call #method_missing and walk up the chain of inheritance. This code was removed for Rails 4 in favor of .find_by(attribute: attribute) for what I assume is slight performance gains.

You can use #method_missing for some really cool programs, just be sure to stay out of trouble by using super to get its default behavior if the method doesn’t get what it wants.