Style Guide

Generally, RSpec specs describe the expected behavior of code. While RSpec is fairly young, there are some conventions for writing specs. The RubySpecs cover a wide variety of components, so we have developed some pragmatic conventions to handle the various situations. As noted below, some conventions are more rigid than others.

These conventions apply to all specs. Existing specs that deviate from these conventions need to be fixed. Consistency is the principle that will almost always trump other conventions. Consistency aids understanding and readability. There are many thousands of lines of code in the spec files, so the value of consistency cannot be overstated.

The specs uniformly use describe not context. The use of it is preferred over specify except in situations when the first word of the string is not a verb. The word "should" is unnecessary noise in the spec description strings and is not used. (The rationale is this: the spec string describes the expected behavior unconditionally. The code examples, on the other hand, set up an expectation that is tested with the call to the should method. The code examples can violate the expectation, but the spec string does not. The value of the spec string is as clearly as possible describing the behavior. Including "should" in that description adds no value.)

Whenever possible, the spec strings should be written to conform to very basic English sentence structure: subject + predicate. The spec strings also uniformly use double-quotes, not single-quotes. The minimum number of words should be used to describe the behavior. Only make distinctions when they add significant value to understanding the behavior. This is explained further below. The general rule across all the specs is to use the least amount of detail to unambiguously describe behavior. Add to the detail conservatively. This is conceptually consistent with doing the simplest thing that could work.

Ruby is a beautifully expressive language with optional parentheses. There is a distinct preference for omitting parentheses in the specs whenever they are not needed. In other words, parentheses should not be used unless necessary to make an expression syntactically or semantically correct.

1. Core and Standard Library

The specs for the Ruby core and standard libraries use one describe block per method. For particularly complex methods, such as Array#[], more than one describe block may be used according to the nature of arguments the method takes.

The describe string should be "Constant.method" for class methods and "Constant#method" for instance methods. "Constant" is either a class or module name. For subclasses or submodules, the "Constant" name should be "Super::Sub". The describe string should not include arguments to the methods unless absolutely necessary to describe the behavior of the method. Keep in mind that in Ruby duck-typing is a deeply embedded concept. Many methods will take any object that responds to a particular method or acts like an instance of a particular class.

Nested describe blocks should not be used. Various automated process scripts depend on the describe string having the format explained above. Also, nested describe blocks complicate the structure of the specs. If a particular situation appears to greatly benefit from nested blocks, open a discussion about it on the mailing list.

  # This is correct
  describe "String#eql?" do
    it "returns true if other has the same length and content" do
      ...
    end
  end

  describe "Array#[]= with [index, count]" do
    it "returns non-array value if non-array value assigned" do
      ...
    end
  end

Contrast the good example above with the one below. The following example deviates from the conventions for describe strings and uses "should" and single-quotes for the descriptions.

  # This is NOT correct
  describe "String#eql?(string)" do
    it 'should return true if other has the same length and content' do
      ...
    end
  end

  describe 'Array#[]=(index, count)' do
    it 'returns non-array value if non-array value assigned' do
      ...
    end
  end

The vast majority of the spec files for the core library have already been created. To create template files for the standard library classes, refer to the mkspec documentation.

1.1 Utility Classes

Many spec code examples refer to a particular class. To prevent name clashes with these different class definitions across all the specs, the classes should be scoped to a module. The convention is as follows:

module ObjectSpecs
  class SomeClass
  end
end

The module is named after the class for which the specs are being written. So, for the specs for Object, the module name is ObjectSpecs.

These utility classes are also referred to as fixtures. In the directory for each class, there is also a fixtures directory. Refer to the existing files for examples.

1.2 Aliased or Identical Methods

Ruby has a significant number of aliased methods. True aliases are identical methods, so the specs should be exactly the same for each aliased method. The following illustrates the convention for specs for aliased methods (or just otherwise identical interfaces.)

In rubyspec/1.8/core/array/shared/collect.rb

shared :array_collect do |cmd|
  describe "Array##{cmd}" do
    it "returns a copy of array with each element replaced by the value returned by block" do
      a = ['a', 'b', 'c', 'd']
      b = a.send(cmd) { |i| i + '!' }
      b.should == ["a!", "b!", "c!", "d!"]
    end

    it "does not return subclass instances" do
      MyArray[1, 2, 3].send(cmd) { |x| x + 1 }.class.should == Array
    end
  end
end

In rubyspec/1.8/core/array/collect_spec.rb

require File.dirname(__FILE__) + '/../../spec_helper'
require File.dirname(__FILE__) + '/shared/collect.rb'

describe "Array#collect" do
  it_behaves_like :array_collect, :collect
end

In rubyspec/1.8/core/array/map_spec.rb

require File.dirname(__FILE__) + '/../../spec_helper'
require File.dirname(__FILE__) + '/shared/collect.rb'

describe "Array#map" do
  it_behaves_like :array_collect, :map
end

1.3 Floating Point Values

Writing specs that use floating point values poses a problem because two values that look the same when rendered to a string may not actually be bitwise equal. Also, floating point operations can result in a value that differs based on the way the FPU carried out the operations.

Specs that compare floating point values should use #should_be_close with the TOLERANCE constant. For floating point values that are exact, but larger than the precision formatted with #to_s (e.g. 1093840198347109283720.00), use the expanded float literal not the truncated precision format that #to_s provides (e.g. don't use 1.09384019834711e+21).

1.4 Private methods

Generally, no specs are written for private methods. A notable exeception are the specs for #initialize on some classes. These specs are primarily written to illustrate the behavior of #initialize for subclasses, where the subclass #initialize behavior is contrasted with the superclass's. Another exeception is #initialize_copy.

1.5 Ruby Ducktyping Interface

Ruby method dispatch behavior calls #method_missing if an instance has no method corresponding to a particular selector. Ruby also defines a number of methods, for example, #to_ary, #to_int, #to_str, that form an interface to Ruby's ducktyping behavior. String methods, for instance, may call #to_str when passed an argument that is not a String.

The point of the RubySpecs is to describe behavior in such a way that if two different implementations pass a spec, Ruby code that relies on behavior described by the spec will execute with the same result on either implementation.

If a spec asserts that a method calls #to_int on an object, it is immaterial to the final outcome whether an implementation calls #to_int and handles the possibility that the method is missing in some way, or first calls #respond_to?(:to_int) and then calls #to_int. There are only two significant aspects to this from the perspective of user code (i.e. code using the interface, not the Ruby implementation code): 1) #to_int is called and performs some action; or 2) #to_int is not called.

It is conceivable that user code like the following exists:

  class Silly
    def method_missing(sym, *args)
      return 1 if sym == :to_int
    end
  end

In such case, the behavior of the following code would be different:

  # The implementation calls #to_int without checking #respond_to?
  [1, 2].at(silly) # => 2

  # The implementation calls #respond_to? first
  [1, 2].at(silly) # => TypeError

In the second case, the expected behavior is restored if the Silly class is modified to implement a #respond_to?(:to_int).

The point is that it really is not sensible to implement an object that provides an interface but does not let the world know about it by either 1) defining the method properly, or 2) defining #respond_to? to indicate that the object provides the interface.

If real-world code exists that depends on this silly implementation (i.e. cannot be coded in a more realistic way), then we can revisit the utility of specs that require #respond_to? to be called. Otherwise, these specs are too tied to the implementation and impose an unrealistic burden on implementations that may exhibit perfectly compatible behavior but not call #respond_to?.

2. Language

For the language specs, there is nothing as convenient or as concrete as a particular method to spec. Review the discussion of the organization of the language specs. The general conventions apply here: use simple English to describe the behavior of the language entities and only add detail as needed. Use a single describe block initially and add distinguishing describe blocks as necessary. Use it rather than specify whenever possible.