Source Code Layout

Nearly everybody is convinced that every style but their own is ugly and unreadable. Leave out the "but their own" and they're probably right...
-- Jerry Coffin (on indentation)

• Use UTF-8 as the source file encoding.

• Use two spaces per indentation level.

  # good
  def some_method
    do_something
  end
  
  # bad - four spaces
  def some_method
      do_something
  end

• Use Unix-style line endings. (*BSD/Solaris/Linux/OSX users are covered by default, Windows users have to be extra careful.)

  • If you're using Git you might want to add the following configuration setting to protect your project from Windows line endings creeping in:

    $ git config --global core.autocrlf true

• Use spaces around operators, after commas, colons and semicolons, around { and before }. Whitespace might be (mostly) irrelevant to the Ruby interpreter, but its proper use is the key to writing easily readable code.

  sum = 1 + 2
  a, b = 1, 2
  1 > 2 ? true : false; puts 'Hi'
  [1, 2, 3].each { |e| puts e }

  The only exception is when using the exponent operator:

  # bad
  e = M * c ** 2
  
  # good
  e = M * c**2

• No spaces after (, [ or before ], ).

  some(arg).other
  [1, 2, 3].length

• Indent when as deep as case. I know that many would disagree with this one, but it's the style established in both the "The Ruby Programming Language" and "Programming Ruby".

  case
  when song.name == 'Misty'
    puts 'Not again!'
  when song.duration > 120
    puts 'Too long!'
  when Time.now.hour > 21
    puts "It's too late"
  else
    song.play
  end
  
  kind = case year
         when 1850..1889 then 'Blues'
         when 1890..1909 then 'Ragtime'
         when 1910..1929 then 'New Orleans Jazz'
         when 1930..1939 then 'Swing'
         when 1940..1950 then 'Bebop'
         else 'Jazz'
         end

• Use empty lines between defs.

• Align the parameters of a method call if they span over multiple lines.

  # starting point (line is too long)
  def send_mail(source)
    Mailer.deliver(to: '[email protected]', from: '[email protected]', subject: 'Important message', body: source.text)
  end
  
  # good (normal indent)
  def send_mail(source)
    Mailer.deliver(
      to: '[email protected]',
      from: '[email protected]',
      subject: 'Important message',
      body: source.text)
  end
  
  # better
  def send_mail(source)
    Mailer.deliver(to: '[email protected]',
                   from: '[email protected]',
                   subject: 'Important message',
                   body: source.text)
  end

• Use RDoc or TomDoc and its conventions for API documentation. Don't put an empty line between the comment block and the def.

• Keep lines fewer than 80 characters.

• Avoid trailing whitespace.

Syntax

• Use def with parentheses when there are arguments. Omit the parentheses when the method doesn't accept any arguments.

  def some_method
    # body omitted
  end
  
  def some_method_with_arguments(arg1, arg2)
    # body omitted
  end

• Never use for, unless you know exactly why. Most of the time iterators should be used instead. for is implemented in terms of each (so you're adding a level of indirection), but with a twist - for doesn't introduce a new scope (unlike each) and variables defined in its block will be visible outside it.

  arr = [1, 2, 3]
  
  # bad
  for elem in arr do
    puts elem
  end
  
  # good
  arr.each { |elem| puts elem }

• Never use then for multi-line if/unless.

  # bad
  if some_condition then
    # body omitted
  end
  
  # good
  if some_condition
    # body omitted
  end

• Favor the ternary operator(?:) over if/then/else/end constructs. It's more common and obviously more concise.

  # bad
  result = if some_condition then something else something_else end
  
  # good
  result = some_condition ? something : something_else

• Use one expression per branch in a ternary operator. This also means that ternary operators must not be nested. Prefer if/else constructs in these cases.

  # bad
  some_condition ? (nested_condition ? nested_something : nested_something_else) : something_else
  
  # good
  if some_condition
    nested_condition ? nested_something : nested_something_else
  else
    something_else
  end

• Never use if x: ... - it is removed in Ruby 1.9. Use the ternary operator instead.

  # bad
  result = if some_condition: something else something_else end
  
  # good
  result = some_condition ? something : something_else

• Never use if x; .... Use the ternary operator instead.

• Use when x then ... for one-line cases. The alternative syntax when x: ... is removed in Ruby 1.9.

• Never use when x; .... See the previous rule.

• Use &&/|| for boolean expressions, and/or for control flow. (Rule of thumb: If you have to use outer parentheses, you are using the wrong operators.)

  # boolean expression
  if some_condition && some_other_condition
    do_something
  end
  
  # control flow
  document.saved? or document.save!

• Avoid multi-line ?: (the ternary operator), use if/unless instead.

• Favor modifier if/unless usage when you have a single-line body. Another good alternative is the usage of control flow and/or.

  # bad
  if some_condition
    do_something
  end
  
  # good
  do_something if some_condition
  
  # another good option
  some_condition and do_something

• Favor unless over if for negative conditions (or control flow or).

  # bad
  do_something if !some_condition
  
  # good
  do_something unless some_condition
  
  # another good option
  some_condition or do_something

• Never use unless with else. Rewrite these with the positive case first.

  # bad
  unless success?
    puts 'failure'
  else
    puts 'success'
  end
  
  # good
  if success?
    puts 'success'
  else
    puts 'failure'
  end

• Don't use parentheses around the condition of an if/unless/while, unless the condition contains an assignment (see "Using the return value of =" below).

  # bad
  if (x > 10)
    # body omitted
  end
  
  # good
  if x > 10
    # body omitted
  end
  
  # ok
  if (x = self.next_value)
    # body omitted
  end

• Favor modifier while/until usage when you have a single-line body.

  # bad
  while some_condition
    do_something
  end
  
  # good
  do_something while some_condition

• Favor until over while for negative conditions.

  # bad
  do_something while !some_condition
  
  # good
  do_something until some_condition

• Omit parentheses around parameters for methods that are part of an internal DSL (e.g. Rake, Rails, RSpec), methods that are with "keyword" status in Ruby (e.g. attr_reader, puts) and attribute access methods. Use parentheses around the arguments of all other method invocations.

  class Person
    attr_reader :name, :age
  
    # omitted
  end
  
  temperance = Person.new('Temperance', 30)
  temperance.name
  
  puts temperance.age
  
  x = Math.sin(y)
  array.delete(e)

• Prefer {...} over do...end for single-line blocks. Avoid using {...} for multi-line blocks (multiline chaining is always ugly). Always use do...end for "control flow" and "method definitions" (e.g. in Rakefiles and certain DSLs). Avoid do...end when chaining.

  names = ["Bozhidar", "Steve", "Sarah"]
  
  # good
  names.each { |name| puts name }
  
  # bad
  names.each do |name|
    puts name
  end
  
  # good
  names.select { |name| name.start_with?("S") }.map { |name| name.upcase }
  
  # bad
  names.select do |name|
    name.start_with?("S")
  end.map { |name| name.upcase }

  Some will argue that multiline chaining would look OK with the use of {...}, but they should ask themselves - it this code really readable and can't the blocks contents be extracted into nifty methods.

• Avoid return where not required.

  # bad
  def some_method(some_arr)
    return some_arr.size
  end
  
  # good
  def some_method(some_arr)
    some_arr.size
  end

• Use spaces around the = operator when assigning default values to method parameters:

  # bad
  def some_method(arg1=:default, arg2=nil, arg3=[])
    # do something...
  end
  
  # good
  def some_method(arg1 = :default, arg2 = nil, arg3 = [])
    # do something...
  end

  While several Ruby books suggest the first style, the second is much more prominent in practice (and arguably a bit more readable).

• Avoid line continuation (\) where not required. In practice, avoid using line continuations at all.

  # bad
  result = 1 - \
           2
  
  # good (but still ugly as hell)
  result = 1 \
           - 2

• Using the return value of = (an assignment) is ok, but surround the assignment with parenthesis.

  # good - shows intented use of assignment
  if (v = array.grep(/foo/)) ...
  
  # bad
  if v = array.grep(/foo/) ...
  
  # also good - shows intended use of assignment and has correct precedence.
  if (v = self.next_value) == "hello" ...

• Use ||= freely to initialize variables.

  # set name to Bozhidar, only if it's nil or false
  name ||= 'Bozhidar'

• Don't use ||= to initialize boolean variables. (Consider what would happen if the current value happened to be false.)

  # bad - would set enabled to true even if it was false
  enabled ||= true
  
  # good
  enabled = true if enabled.nil?

• Avoid using Perl-style special variables (like $0-9, `$``, etc. ). They are quite cryptic and their use in anything but one-liner scripts is discouraged.

• Never put a space between a method name and the opening parenthesis.

  # bad
  f (3 + 2) + 1
  
  # good
  f(3 + 2) + 1

• If the first argument to a method begins with an open parenthesis, always use parentheses in the method invocation. For example, write f((3 + 2) + 1).

• Always run the Ruby interpreter with the -w option so it will warn you if you forget either of the rules above!

• When the keys of your hash are symbols use the Ruby 1.9 hash literal syntax.

  # bad
  hash = { :one => 1, :two => 2 }
  
  # good
  hash = { one: 1, two: 2 }

• Use the new lambda literal syntax.

  # bad
  lambda = lambda { |a, b| a + b }
  lambda.call(1, 2)
  
  # good
  lambda = ->(a, b) { a + b }
  lambda.(1, 2)

• Use _ for unused block parameters.

  # bad
  result = hash.map { |k, v| v + 1 }
  
  # good
  result = hash.map { |_, v| v + 1 }

Naming

The only real difficulties in programming are cache invalidation and naming things.
-- Phil Karlton

• Use snake_case for methods and variables.

• Use CamelCase for classes and modules. (Keep acronyms like HTTP, RFC, XML uppercase.)

• Use SCREAMING_SNAKE_CASE for other constants.

• The names of predicate methods (methods that return a boolean value) should end in a question mark. (i.e. Array#empty?).

• The names of potentially "dangerous" methods (i.e. methods that modify self or the arguments, exit! (doesn't run the finalizers like exit does), etc.) should end with an exclamation mark if there exists a safe version of that dangerous method.

  # bad - there is not matching 'safe' method
  class Person
    def update!
    end
  end
  
  # good
  class Person
    def update
    end
  end
  
  # good
  class Person
    def update!
    end
  
    def update
    end
  end

• Define the non-bang (safe) method in terms of the bang (dangerous) one if possible.

  class Array
    def flatten_once!
      res = []
    
      each do |e|
        [*e].each { |f| res << f }
      end
  
      replace(res)
    end
  
    def flatten_once
      dup.flatten_once!
    end
  end

• When using reduce with short blocks, name the arguments |a, e| (accumulator, element).

• When defining binary operators, name the argument other.

  def +(other)
    # body omitted
  end

Annotations

• Annotations should usually be written on the line immediately above the relevant code.

• The annotation keyword is followed by a colon and a space, then a note describing the problem.

• If multiple lines are required to describe the problem, subsequent lines should be indented two spaces after the #.

  def bar
    # FIXME: This has crashed occasionally since v3.2.1. It may
    #   be related to the BarBazUtil upgrade.
    baz(:quux)
  end

• In cases where the problem is so obvious that any documentation would be redundant, annotations may be left at the end of the offending line with no note. This usage should be the exception and not the rule.

  def bar
    sleep 100 # OPTIMIZE
  end

• Use TODO to note missing features or functionality that should be added at a later date.

• Use FIXME to note broken code that needs to be fixed.

• Use OPTIMIZE to note slow or inefficient code that may cause performance problems.

• Use HACK to note code smells where questionable coding practices were used and should be refactored away.

• Use REVIEW to note anything that should be looked at to confirm it is working as intended. For example: REVIEW: Are we sure this is how the client does X currently?

• Use other custom annotation keywords if it feels appropriate, but be sure to document them in your project's README or similar.

Classes

• Always supply a proper to_s method for classes that represent domain objects.

  class Person
    attr_reader :first_name, :last_name
  
    def initialize(first_name, last_name)
      @first_name = first_name
      @last_name = last_name
    end
  
    def to_s
      "#@first_name #@last_name"
    end
  end

• Use the attr family of functions to define trivial accessors or mutators.

  # bad
  class Person
    def initialize(first_name, last_name)
      @first_name = first_name
      @last_name = last_name
    end
  
    def first_name
      @first_name
    end
  
    def last_name
      @last_name
    end
  end
  
  # good
  class Person
    attr_reader :first_name, :last_name
  
    def initialize(first_name, last_name)
      @first_name = first_name
      @last_name = last_name
    end
  end

• Consider using Struct.new, which defines the trivial accessors, constructor and comparison operators for you.

  # good
  class Person
    attr_reader :first_name, :last_name
  
    def initialize(first_name, last_name)
      @first_name = first_name
      @last_name = last_name
    end
  end
  
  # better
  class Person < Struct.new (:first_name, :last_name)
  end

• Consider adding factory methods to provide additional sensible ways to create instances of a particular class.

  class Person
    def self.create(options_hash)
      # body omitted
    end
  end

• Prefer duck-typing over inheritance.

  # bad
  class Animal
    # abstract method
    def speak
    end
  end
  
  # extend superclass
  class Duck < Animal
    def speak
      puts 'Quack! Quack'
    end
  end
  
  # extend superclass
  class Dog < Animal
    def speak
      puts 'Bau! Bau!'
    end
  end
  
  # good
  class Duck
    def speak
      puts 'Quack! Quack'
    end
  end
  
  class Dog
    def speak
      puts 'Bau! Bau!'
    end
  end

• Avoid the usage of class (@@) variables due to their "nasty" behavior in inheritance.

  class Parent
    @@class_var = 'parent'
  
    def self.print_class_var
      puts @@class_var
    end
  end
  
  class Child < Parent
    @@class_var = 'child'
  end
  
  Parent.print_class_var # => will print "child"

  As you can see all the classes in a class hierarchy actually share one class variable. Class instance variables should usually be preferred over class variables.

• Use def self.method to define singleton methods. This makes the methods more resistant to refactoring changes.

  class TestClass
    # bad
    def TestClass.some_method
      # body omitted
    end
  
    # good
    def self.some_other_method
      # body omitted
    end
  
    # Also possible and convenient when you
    # have to define many singleton methods.
    class << self
      def first_method
        # body omitted
      end
  
      def second_method_etc
        # body omitted
      end
    end
  end

Exceptions

• Signal exceptions using the fail keyword. Use raise only when catching an exception and re-raising it (because here you're not failing, but explicitly and purposefully raising an exception).

  begin
   fail "Oops";
  rescue => error
    raise if error.message != "Oops"
  end

• Never return form an ensure block. If you explicitly return from a method inside an ensure block, the return will take precedence over any exception being raised, and the method will return as if no exception had been raised at all. In effect, the exception will be silently thrown away.

  def foo
    begin
      fail
    ensure
      return "very bad idea"
    end
  end

• Don't suppress exceptions.

  # bad
  begin
    # an exception occurs here
  rescue SomeError
    # the rescue clause does absolutely nothing
  end
  
  # bad
  do_something rescue nil

• Don't use exceptions for flow of control.

  # bad
  begin
    n / d
  rescue ZeroDivisionError
    puts "Cannot divide by 0!"
  end
  
  # good
  if d.zero?
    puts "Cannot divide by 0!"
  else
    n / d
  end

• Put more specific exceptions higher up the rescue chain, otherwise they'll never be rescued from.

  # bad
  begin
    # some code
  rescue Exception => e
    # some handling
  rescue StandardError => e
    # some handling
  end
  
  # good
  begin
    # some code
  rescue StandardError => e
    # some handling
  rescue Exception => e
    # some handling
  end

• Release external resources obtained by your program in an ensure block.

  f = File.open("testfile")
  begin
    # .. process
  rescue
    # .. handle error
  ensure
    f.close unless f.nil?
  end

• Favor the use of exceptions for the standard library over introducing new exception classes.

Collections

• Prefer literal array and hash creation notation (unless you need to pass parameters to their constructors, that is).

  # bad
  arr = Array.new
  hash = Hash.new
  
  # good
  arr = []
  hash = {}

• Prefer %w to the literal array syntax when you need an array of strings.

  # bad
  STATES = ['draft', 'open', 'closed']
  
  # good
  STATES = %w(draft open closed)

• Avoid the creation of huge gaps in arrays.

  arr = []
  arr[100] = 1 # now you have an array with lots of nils

• Use Set instead of Array when dealing with unique elements. Set implements a collection of unordered values with no duplicates. This is a hybrid of Array's intuitive inter-operation facilities and Hash's fast lookup.

• Use symbols instead of strings as hash keys.

  # bad
  hash = { 'one' => 1, 'two' => 2, 'three' => 3 }
  
  # good
  hash = { one: 1, two: 2, three: 3 }

• Avoid the use of mutable object as hash keys.

• Use the new 1.9 literal hash syntax in preference to the hashrocket syntax.

  # bad
  hash = { :one => 1, :two => 2, :three => 3 }
  
  # good
  hash = { one: 1, two: 2, three: 3 }

• Rely on the fact that hashes in 1.9 are ordered.

• Never modify a collection while traversing it.

Strings

• Prefer string interpolation instead of string concatenation:

  # bad
  email_with_name = user.name + ' <' + user.email + '>'
  
  # good
  email_with_name = "#{user.name} <#{user.email}>"

• Consider padding string interpolation code with space. It more clearly sets the code apart from the string.

  "#{ user.last_name }, #{ user.first_name }"

• Prefer single-quoted strings when you don't need string interpolation or special symbols such as \t, \n, ', etc.

  # bad
  name = "Bozhidar"
  
  # good
  name = 'Bozhidar'

• Don't use {} around instance variables being interpolated into a string.

  class Person
    attr_reader :first_name, :last_name
  
    def initialize(first_name, last_name)
      @first_name = first_name
      @last_name = last_name
    end
  
    # bad
    def to_s
      "#{@first_name} #{@last_name}"
    end
  
    # good
    def to_s
      "#@first_name #@last_name"
    end
  end

• Avoid using String#+ when you need to construct large data chunks. Instead, use String#<<. Concatenation mutates the string instance in-place and is always faster than String#+, which creates a bunch of new string objects.

  # good and also fast
  html = ''
  html << '<h1>Page title</h1>'
  
  paragraphs.each do |paragraph|
    html << "<p>#{paragraph}</p>"
  end

Regular Expressions

• Don't use regular expressions if you just need plain text search in string: string['text']

• For simple constructions you can use regexp directly through string index.

  match = string[/regexp/]             # get content of matched regexp
  first_group = string[/text(grp)/, 1] # get content of captured group
  string[/text (grp)/, 1] = 'replace'  # string => 'text replace'

• Use non capturing groups when you don't use captured result of parenthesis.

  /(first|second)/   # bad
  /(?:first|second)/ # good

• Avoid using $1-9 as it can be hard to track what they contain. Named groups can be used instead.

  # bad
  /(regexp)/ =~ string
  ...
  process $1
  
  # good
  /(?<meaningful_var>regexp)/ =~ string
  ...
  process meaningful_var

• Character classes have only few special characters you should care about: ^, -, \, ], so don't escape . or brackets in [].

• Be careful with ^ and $ as they match start/end of line, not string endings. If you want to match the whole string use: \A and \z (not to be confused with \Z which is the equivalent of /\n?\z/).

  string = "some injection\nusername"
  string[/^username$/]   # matches
  string[/\Ausername\z/] # don't match

• Use x modifier for complex regexps. This makes them more readable and you can add some useful comments. Just be careful as spaces are ignored.

  regexp = %r{
    start         # some text
    \s            # white space char
    (group)       # first group
    (?:alt1|alt2) # some alternation
    end
  }x

• For complex replacements sub/gsub can be used with block or hash.

Percent Literals

• Use %w freely.

  STATES = %w(draft open closed)

• Use %() for single-line strings which require both interpolation and embedded double-quotes. For multi-line strings, prefer heredocs.

  # bad (no interpolation needed)
  %(<div class="text">Some text</div>)
  # should be '<div class="text">Some text</div>'
  
  # bad (no double-quotes)
  %(This is #{quality} style)
  # should be "This is #{quality} style"
  
  # bad (multiple lines)
  %(<div>\n<span class="big">#{exclamation}</span>\n</div>)
  # should be a heredoc.
  
  # good (requires interpolation, has quotes, single line)
  %(<tr><td class="name">#{name}</td>)

• Use %r only for regular expressions matching more than one '/' character.

  # bad
  %r(\s+)
  
  # still bad
  %r(^/(.*)$)
  # should be /^\/(.*)$/
  
  # good
  %r(^/blog/2011/(.*)$)

• Avoid %q, %Q, %x, %s, and %W.

• Prefer () as delimiters for all % literals.

Metaprogramming

• Do not mess around in core classes when writing libraries. (Do not monkey patch them.)

• The block form of class_eval is preferable to the string-interpolated form.

  • when you use the string-interpolated form, always supply __FILE__ and __LINE__, so that your backtraces make sense:

    class_eval "def use_relative_model_naming?; true; end", __FILE__, __LINE__

  • define_method is preferable to class_eval{ def ... }

• When using class_eval (or other eval) with string interpolation, add a comment block showing its appearance if interpolated (a practice I learned from the rails code):

  # from activesupport/lib/active_support/core_ext/string/output_safety.rb
  UNSAFE_STRING_METHODS.each do |unsafe_method|
    if 'String'.respond_to?(unsafe_method)
      class_eval <<-EOT, __FILE__, __LINE__ + 1
        def #{unsafe_method}(*args, &block)       # def capitalize(*args, &block)
          to_str.#{unsafe_method}(*args, &block)  #   to_str.capitalize(*args, &block)
        end                                       # end
  
        def #{unsafe_method}!(*args)              # def capitalize!(*args)
          @dirty = true                           #   @dirty = true
          super                                   #   super
        end                                       # end
      EOT
    end
  end

• avoid using method_missing for metaprogramming. Backtraces become messy; the behavior is not listed in #methods; misspelled method calls might silently work (nukes.luanch_state = false). Consider using delegation, proxy, or define_method instead. If you must use method_missing,

  • be sure to also define respond_to?

  • only catch methods with a well-defined prefix, such as find_by_* -- make your code as assertive as possible.

  • call super at the end of your statement

  • delegate to assertive, non-magical methods:

    # bad
    def method_missing?(meth, *args, &block)
      if /^find_by_(?<prop>.*)/ =~ meth
        # ... lots of code to do a find_by
      else
        super
      end
    end
    
    # good
    def method_missing?(meth, *args, &block)
      if /^find_by_(?<prop>.*)/ =~ meth
        find_by(prop, *args, &block)
      else
        super
      end
    end
    
    # best of all, though, would to define_method as each findable attribute is declared  

Misc

• Write ruby -w safe code.

• Avoid hashes as optional parameters. Does the method do too much?

• If you really have to, add "global" methods to Kernel and make them private.

• Use class instance variables instead of global variables.

  #bad
  $foo_bar = 1
  
  #good
  class Foo
    class << self
      attr_accessor :bar
    end
  end
  
  Foo.bar = 1

• Avoid alias when alias_method will do.

• Use OptionParser for parsing complex command line options and ruby -s for trivial command line options.