Ruby Annotations

26 October 2006

One of Ruby's most popular features is its support for metaprogramming, that is features that act like they change the language itself - introducing things like new keywords.

Although the mainstream curly-brace languages have less support for metaprogramming in general, one useful feature they do have is Annotations. Annotations are an important capability for the language enhancement InternalDslStyle. At first glance Ruby doesn't support annotations, but if tilt your head you'll find it does.

With annotations you tag a language element (class, method, field etc) with a marker - essentially metadata about that element. You can then use that metadata at runtime or compile time.

A good example of this is marking of tests that was pioneered by NUnit2.

[Test] public void SomeTestMethod() {...

At run time the NUnit framework looks at classes, finds the methods annotated with [Test] and runs them.

Annotations can take parameters. So if you have a hankering for the subranges of Pascal, which allow you to indicate a valid range for a variable, you define them with something like this

@ValidRange(lower = 1, upper = 1000)
  private int weight; // in lb
  @ValidRange(lower = 1, upper = 120)
  private int height; // in inches

This isn't quite the same as the Pascal subrange idea, this only defines the valid range, you still have to build the mechanism to check the value, such as calling an isValid method on your object (I agree that's not much of a ContextualValidation). But the key here is that you have defined your own declarative statement about that variable.

So how do you do this in Ruby? Something like this:

	validate_range :@height, :with => 1..120
	validate_range :@weight, :with => 1..1000

In terms of syntax, the big difference is that with Java (or .NET) annotations you place the annotation immediately before the element you're annotating. In ruby you name the element you're annotating in the annotation. Although this adds some typing (a rare statement for Ruby) it does give you the freedom to place your annotations anywhere in your class. It also makes it easier to construct annotations that reference more than one language element.

A deeper difference between the two styles is how they get implemented. Curly Brace annotations are a special language construct that attach special objects as metadata to language elements. These annotation objects can be queried and processed either at compile time (with Java's apt) or at runtime.

Ruby's annotations, on the other hand, are class methods usually defined in a superclass or included module. By writing them directly in the class definition they are executed when the class is loaded. As a result they aren't a particular language feature, more of a way of using class methods. You don't have to create metadata objects (although you could if you wanted to), more likely you'll just build the objects to carry out the task you want them to.

Due to Ruby's dynamic nature, you can do a lot more interesting things with annotations. In particular you can do code generation. The most obvious example of this is things like attr_accessor :height which will generate get and set methods for the field, effectively modifying the class itself. With apt Java can do some similar things, but you can't modify the class itself. With build scripts and partial classes you probably could do something like this in C#, but this kind of run-time code generation is certainly more natural in Ruby, it's one of its Lispish elements.

Rubyists don't call these things annotations. One of the things I like doing is to find common techniques that cross languages, for me this is a common technique and 'annotation' seems like a good generic word for it. I don't know if Rubyists will agree.