Skip to content

Latest commit

 

History

History
1624 lines (1285 loc) · 43.5 KB

style-guide-ruby.md

File metadata and controls

1624 lines (1285 loc) · 43.5 KB
Raw

The Ruby Style Guide

This Ruby style guide recommends best practices so that real-world Ruby programmers can write code that can be maintained by other real-world Ruby programmers. A style guide that reflects real-world usage gets used, and a style guide that holds to an ideal that has been rejected by the people it is supposed to help risks not getting used at all – no matter how good it is.

The guide is separated into several sections of related rules. We've tried to add the rationale behind the rules (if it's omitted we've assumed that is pretty obvious).

The guidelines didn't come out of nowhere. They are largely based on @bbatsov's ruby-style-guide, modified according to the sober judgement and good taste of our senior engineers. The guide reflects feedback and suggestions from members of the Ruby community and various highly regarded Ruby programming resources, such as "Programming Ruby 1.9" and "The Ruby Programming Language".

The guide is still a work in progress, and we strongly invite your feedback -- guidelines to add, decisions you disagree with, illustrations of good or bad practice.

A foolish consistency is the hobgoblin of little minds -- Ralph Waldo Emerson

This is a guide, not a rulebook. Break the rules with good taste, but not capriciously. For every one of the guidelines, there is a sound argument against it -- otherwise we wouldn't need to spell it out. Nobody needs a style guide to say "don't write your documentation in pig latin". Reasonable people can disagree whether drink = if Time.now.hour > 19 then :beer else :soda end is clearer than drink = (Time.now.hour > 19) ? :beer : :soda1, but at infochimps we've settled on the second. Having one uniform choice carries more benefit than whatever fine distinction separates two close alternatives.

You can generate a PDF or an HTML copy of this guide using Transmuter.

Table of Contents

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, 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
[1, 2, 3].each{|e| puts e }

  • Indent when as deep as case.

    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

# good
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

# good
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 and to break up a method into logical paragraphs.

    • However, multiple long paragraphs in a method is a smell -- consider moving each into a small self-documenting method.

      def some_method
  data = initialize(options)

  data.manipulate!

  data.result
end

def some_method
  result
end

  • If the parameters of a method call span multiple lines, move all of them down to read in parallel.

    # 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

# bad (hanging way out in space):
def send_mail(source)
  Mailer.deliver(to =>      '[email protected]',
                 from =>    '[email protected]',
                 subject => 'Important message',
                 body =>    source.text)
end

# bad (inconsistent indentation):
def send_mail(source)
  Mailer.deliver(to => '[email protected]',
    from => '[email protected]',
    subject => 'Important message',
    body => source.text)
end

# good (can easily read down the list of keys and list of values)
def send_mail(source)
  Mailer.deliver(
    to      => '[email protected]',
    from    => '[email protected]',
    subject => 'Important message',
    body    => source.text)
end

  • Align parallel constructions or assignments: it makes the code easier to read, and highlights parallel functionality.

    # bad
def send_mail(source)
  Mailer.deliver(
    to => '[email protected]',
    from => '[email protected]',
    subject => 'Important message',
    body => source.text)
end

# good
def send_mail(source)
  Mailer.deliver(
    to      => '[email protected]',
    from    => '[email protected]',
    subject => 'Important message',
    body    => source.text)
end

# good
helicity    = hemiconducer.cromulence ** 2
reluctance  = hemiconducer.reluctor.reluctance
phase       = moon.phase - average_marzelvane_phase

  • Don't be over-DRY -- a repeated clause, written in parallel, is easier to read than a trivial loop:

    # bad
[ :flux, :phrasal, :nimbus ].each do |capacitor_type|
  counterrotate_capacitor capacitor_type, moon.phase, reluctance
end

# good
counterrotate_capacitor :flux,    moon.phase, reluctance
counterrotate_capacitor :phrasal, moon.phase, reluctance
counterrotate_capacitor :nimbus,  moon.phase, reluctance

    let good taste be your guide, but when lines of code are equivalent prefer the small multiple over the explicit loop.

  • Indent protected, public, private and module_function at the same level as the enclosing declaration.

    • good:

      module Validations

  def valid?(context = nil)
    current_context, self.validation_context = validation_context, context
    errors.clear
    run_validations!
  ensure
    self.validation_context = current_context
  end

  # ...

protected

  def run_validations!
    run_callbacks :validate
    errors.empty?
  end
end

    • bad (easy to read past):

      module Validations
  def valid?
    # ....
  end

  protected

  def run_validations!
    # ...
  end
end

    • bad (inconsistent indentation, editors will screw up):

      module Validations
  def valid?
    # ....
  end

  protected

    def run_validations!
      # ...
    end
end

  • Keep lines fewer than 120 characters.

  • Remove trailing whitespace.

  • Convert all tabs to spaces.

Documentation

Good code is its own best documentation. As you're about to add a comment, ask yourself, "How can I improve the code so that this comment isn't needed?" Improve the code and then document it to make it even clearer.
-- Steve McConnell

  • Use YARD and its conventions for API documentation.

  • Don't put an empty line between the comment block and the def.

  • Avoid superfluous comments:

    # bad: adds nothing to my understanding

# convert the model to xml
def to_xml
  # ...
end

# good: method name tells me everything I need to know.

def to_xml
  # ...
end

  • For text files and complex documentation blocks, use Markdown (not textile, plain text or RDoc).

Internal comments

  • Write self-documenting code and ignore the rest of this section. Seriously!

  • Internal comments are often a smell that a method should be broken up:

    # bad: defocused

def adjust_hemiconducer_circuit(sagacity)
  # align the marzelvanes
  ... 4 lines ...
  # calculate moon phase and reluctance
  ... 7 lines ..
  # counterrotate the plenum and flux capacitors
  ... 12 lines ...
end

# good: story doesn't get in the way of the plot

def adjust_hemiconducer_circuit(sagacity)
  align_marzelvanes(sagacity)

  moon       = Moon.current
  reluctance = Reluctor.find_reluctance(:sagacity => sagacity)

  counterrotate_capacitor :plenum, moon.phase, reluctance
  counterrotate_capacitor :nimbus, moon.phase, reluctance
end

  • Comments longer than a word are capitalized and use punctuation. Use one space after periods.

  • Here again: avoid superfluous comments.

    # bad
counter += 1 # increments counter by one

  • Keep existing comments up-to-date. No comment is better than an outdated comment.

  • Avoid writing comments to explain bad code. Refactor the code to make it self-explanatory. (Do or do not - there is no try.)

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. Group complex sub-clauses in parentheses.

    # bad
result = if some_condition then something else something_else end

# good
result = some_condition ? something : something_else

# bad (needs parens)
drinks =  age > 18  ? ['beer', 'wine'] : ['shirley temple']

# good
drinks = (age > 18) ? ['beer', 'wine'] : ['shirley temple']

  • 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; ... (with a semicolon). Also never use if x: ... (with a colon) - it is not only bad form but it has been removed in Ruby 1.9. Use the ternary operator instead.

    # bad
result = if some_condition: something else something_else end

# bad
result = if some_condition; something else something_else end

# good
result = some_condition ? something : something_else

  • 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!
    • us and/or implies "don't worry about this

  • 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 not some_condition

# good
do_something unless some_condition

# another good option
some_condition or do_something

  • Prefer not to ! -- it's more visible. Always parenthesize a not clause in a compound expression.

  • 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 a simple condition in 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

  • 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 or missing hash values. (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?

# bad
nuke[:launch] ||= default_launch_state

# good
nuke[:launch] = default_launch_state unless nuke.has_key?(:launch)

  • 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).

  • In an expression with multiple parentheses, tastefully insert whitespace to let the reader more easily pair open/close parens and see groups.

    # bad (a petty consistency is
# good
quad_1 = (-b + Math.sqrt( b**2 + (4 * a * c) )) / (2 * a)
quad_1 = (-b + Math.sqrt( b**2 - (4 * a * c) )) / (2 * a)

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

  • Use _ for unused block parameters.

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

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

Ruby 1.9-only sugar

For all internal projects and most external projects we have abandoned ruby 1.8 compatibility. The exceptions are Wukong, all cookbooks, ironfan, configilere, and most of gorillib. Outside of those:

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

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

# if project is 1.9-only and keys are only symbols
hash = { one: 1, two: 2 }

  • You should use the new lambda literal syntax unless 1.8-compatibility is required. Omit parentheses if there are no arguments to the block. Avoid situations where understanding the subtle differences between Procs and lambda is required, i.e. returning. Prefer -> over lambda over Proc.new.

    # bad
adder = lambda{|a, b| a + b }
adder.call(1, 2)
id_generator = Proc.new{ [Time.now.to_f, $!, rand].join('-') }

# good
adder = ->(a, b){a + b }
adder.(1, 2)
id_generator = ->{ [Time.now.to_f, $!, rand].join('-') }

Naming

The only real difficulties in programming are cache invalidation and naming things.
-- Phil Karlton ... and off by one errors -- apocryphal

  • Use underscore_case for methods and variables.

  • Use CamelCase for classes and modules. Even acronyms like HTTP, RFC, XML should be camelcased: HttpRequest, XmlDoc.

  • The CamelCase and underscore_case should always agree:

    # bad (missing underscore)
twitteruser = TwitterUser.new('bob')

# good
twitter_user = TwitterUser.new('bob')

  • Use ALL_CAPS for other constants.

  • Do not use lowerCamelCase in any context.

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

  • Method names should end with an exclamation point when they are:

    • potentially "dangerous": launch_nukes!, db_table.drop!
    • have suprising side effects: modify self or the arguments, exit!
    • in rare cases (notably ActiveRecord), an exclamation point indicates the 'loud' (exception-raising) version of a method that normally returns false on failure. Use this sparingly -- it's not assertive, and makes it easy to ignore necessary error handling.

  • When using Array methods with short blocks, name a generic argument |el| -- do not use |e| or other variants. If they are not generic, use precise names (|user|).

  • When using Hash methods with short blocks, name generic arguments |key, val| -- do not use |k, v| or other variants. If they are not generic, use precise names (|user_id, user_name|). If it is a named-record mapping, name the arguments something_name and something_info:

    players = { cal: { id: 8, team: 'BAL' }, pedro: { id: 45, team: 'BOS' }, }
players.each do |player_name, player_info|
  # ...
end

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

  • When defining binary operators, name the argument other.

    def +(other)
  # body omitted
end

  • Prefer map over collect, find over detect, select over find_all, inject over reduce. No preference between size and length. This is not a hard requirement; if the use of the alias enhances readability, it's ok to use it. The rhyming methods are inherited from Smalltalk and are not common in other programming languages. The reason the use of select is encouraged over find_all is that it goes together nicely with reject and its name is pretty self-explanatory.

Annotations

  • Annotations should 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?

  • Don't use other custom annotation keywords.

Classes

  • When designing class hierarchies make sure that they conform to the Liskov Substitution Principle -- roughly, subclasses should behave like their ancestors.

    • In particular, don't put Abstract Factory methods on a superclass if they don't make sense on a subclass.

  • Try to make your classes as [SOLID](http://en.wikipedia.org/wiki/SOLID_(object-oriented_design\)) as possible.

  • 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 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.

    • If the attribute is not visible, use class instance variables
    • If the attribute is visible, use class_attribute (from eg. gorillib). Be careful to not modify parent class_attributes in place.

  • Assign proper visibility levels to methods (private, protected) in accordance with their intended usage. Don't go off leaving everything public (which is the default).

  • Indent the public, protected, and private methods as much the enclosing scope they apply to. Leave one blank line above and below them.

    class SomeClass
  def public_method
    # ...
  end

private

  def private_method
    # ...
  end
end

  • Use def self.method to define class 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

  • Don't suppress exceptions.

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

  • 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

  • Do not rescue the Exception class. StandardError is the most reasonable catch-all, but you should rescue the most precise error you can.

    # bad
begin
  # an exception occurs here
rescue
  # exception handling
end

# still bad
begin
  # an exception occurs here
rescue Exception
  # exception handling
end

# better
begin
  # an exception occurs here
rescue StandardError => e
  # exception handling
end

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

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

# good
begin
  # some code
rescue ArgumentError => e
  # some handling
rescue StandardError => 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

  • Custom exceptions should inherit from the most specific appropriate class in the standard library. Override the to_s method, not the message method.

  • In most cases however, raise an exception with specific information. Put single quotes around values (so that blank values are visible). Avoid using < and > in error messages.

    raise ArgumentError, "Animals must enter ark two-by-two: got '#{animal_1}', '#{animal_2}'" unless (animal_1.species == animal_2.species)
raise ArgumentError, "Ark cannot hold imaginary animals: got '#{animal_1}', '#{animal_2}'" if animal_1.imaginary? || animal_2.imaginary?

Collections

  • Prefer %w to the literal array syntax when you need an array of tokens. Always use %w[ ... ], and leave a space at the start and end.

    # bad
STATES = ['draft', 'open', 'closed']

# bad (wrong brackets)
STATES = %w(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 or array-with-join over string concatenation:

    # bad
email_with_name = user.name + ' <' + user.email + '>'

# good (string is short)
email_with_name = "#{user.name} <#{user.email}>"

  • 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'

  • Use {} around instance variables being interpolated into a string.

    # bad (too clever)
def to_s
  "#@first_name #@last_name"
end

# good
def to_s
  "#{@first_name} #{@last_name}"
end

  • Avoid using String#+ when you need to construct large data chunks. Instead, use String#<< or build up an array and us #join. 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

# ok, because there's no conditional logic
def to_s
  [ scheme, ':',
    '//', authority,
    path,
    '?', query_string,
    '#', fragment
  ].join
end

# good
def to_s
  uri_str = ''
  uri_string << "#{scheme}:"       unless scheme.nil?
  uri_string << "//#{authority}"   unless authority.nil?
  uri_string << path.to_s
  uri_string << "?#{query_string}" unless query_string.nil?
  uri_string << "##{fragment}"     unless fragment.nil?
  uri_string
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 (but note that the regexp must precede the string):

    # 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. Never use \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 (use square brackets always, since it is an array):

    STATES = %w[ draft open closed ]

  • Use %Q{} for single-line strings which require both interpolation and embedded double-quotes. Do not use %{} (no Q) and do not use %Q() (wrong brackets). For multi-line strings, prefer heredocs.

    # bad (no interpolation needed)
%Q{<div class="text">Some text</div>}
# should be '<div class="text">Some text</div>'
# or %q{<div class="text">Some text</div>}

# bad (no need for fanciness)
%Q{This is #{quality} style}
# should be "This is #{quality} style"

# bad (multiple lines)
%Q{<div>\n<span class="big">#{exclamation}</span>\n</div>}
# should be a heredoc.

# good (requires interpolation, has quotes, single line)
%Q{<tr><td class="name">#{name}</td>}

  • Use %r{} for regular expressions matching one or more '/' characters.

    # bad
%r(\s+)

# good
%r{^/(.*)$}
%r{^/blog/2011/(.*)$}

  • Avoid %x, %s, %W and plain %{}.

  • Prefer {} as delimiters for all string-like literals, and [] for %w.

Metaprogramming

Use metaprogramming sparely. Metaprogramming should only occur in frameworks, not applications -- it is justified to abstract a widely-repeated pattern of long use, and rarely otherwise.

  • Our canonical language is Ruby 1.9.2+ plus Gorillib's extensions.

    • Do not otherwise mess around (monkey patch) with core classes.
    • Do not bring in other code from ActiveSupport, extlib or the like. Exception: if you are writing a Rails app, chef plugin, or other fully framework-immersed code, use that framework's features freely.

  • Provide light predictable magic or no magic at all:

    • separate sugar from fuctionality.

      Good: The collects method doesn't do anything but dispatch to other methods.

      # Given a class, creates a method to create-or-retrieve
#
# @example creating helper methods
#   class Kitchen
#     collects(Utensil)
#   end
#   my_kitchen = Kitchen.new
#   my_kitchen.utensil 'Sauce Pot', :gallons => 5
#   my_kitchen.utensil('Sauce Pot') #=> #<Utensil name="Sauce Pot" gallons=5>
#
def collects(klass)
  field_name = klass.name.underscore
  define_method(field_name) do |obj_name, *args, &block|
    obj = registry(klass).find_or_create(obj_name)
    obj.configure(*args, &block) if args.present? || block_given?
    obj
  end
end

    • Options are often a smell. Encode the common case in the sugar method; the preceding principle ensures the user can answer a necessarily complex use case with necessarily explicit code. The resulting verbosity is a good thing: the reader is left in no doubt that something unusual is being done.

    • Be assertive always -- but especially when providing sugar. Light type conversion and multiple behaviors in service of readability is great, but don't provide multiple ways of doing the same thing.

      Bad: this disastrously flexible interface can't conceivably be documented, let alone tested.

      #
# Get or update path to the input file. 
#
# @param [Array,Pathname,String] filename - path to input file. You can pass
#   in pretty much anything and it will be converted.
#
def input_file(filename=nil)
  return @input_file if filename.blank?
  @input_file =
    case filename
    when Array       then File.join(filename)
    when Pathname    then filename
    when %r{file://} then Addressable.parse(filename).path
    when             then File.expand_path(filename)
    else raise "Don't know how to interpret filename '#{filename}'"
    end
end

      Good: makes bold, predictable choices. Want to use file:// references? Well tough titty toenails, do it yourself.

      #
# Get or update path to the input file.
#
# @example
#   input_file   '~/skrilla.csv'
#   input_file   #=> '/Users/flip/skrilla.csv'
#
# @param [#to_s] filename - path to input file. You may use shell shorthand
#   like '~/script.tsv' and './accounts.csv' -- they will be `expand_path`ed
#   into absolute paths.
#
# @return path to input_file, nil if unset.
#
def input_file(filename=nil)
  if filename then @input_file = File.expand_path(filename.to_s) ; end
  @input_file
end

  • The block form of class_eval is preferable to the string-interpolated form. define_method is preferable to class_eval{ def ... }

  • When using class_eval (or other eval) with string interpolation:

    • Supply __FILE__ and __LINE__ so that your backtraces make sense:

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

    • add a comment block showing its appearance if interpolated:

      # 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 define respond_to_missing? (1.9 only; it means you don't have to also define respond_to?)

    • call super at the end of your statement

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

    • delegate to assertive, non-magical methods, named for that prefix:

      # bad
def method_missing?(meth, *args, &block)
  if /^find_by_(?<prop>.*)/ =~ meth.to_s
    # ... 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

  • avoid Aliasing/Redefining methods. Instead, generate a Module and inject it.

    • calling super is a good idea.
    • if you're going to monkey patch, make sure that the method isn’t already there.

Setup

  • Include a README.md that includes all installation steps.
  • Use Markdown, not textile, plain text or RDoc.
  • Include a Procfile if there are processes to start.
  • Include a Gemfile. If the project is a standalone app, you should check Gemfile.lock in to the repo. If it is not deployed independently, you should exclude Gemfile.lock.
  • Do not version an .rvmrc file into a repo.

Misc

  • Write ruby -w safe code.

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

  • Avoid methods longer than 10 LOC (lines of code). Ideally, most methods will be shorter than 5 LOC. Empty lines do not contribute to the relevant LOC.

  • Avoid parameter lists longer than three or four parameters.

  • 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_attribute :bar
end

Foo.bar = 1

  • Avoid alias when alias_method will do.

  • Use Configliere for parsing command line options.

  • Code in a functional way, avoiding mutation when that makes sense.

  • Avoid needless metaprogramming.

  • Do not mutate arguments unless that is the purpose of the method.

  • Avoid more than three levels of block nesting.

  • Be consistent. In an ideal world, be consistent with these guidelines.

  • Use common sense.

Footnotes

  1. http://bit.ly/coffescript-says-no-to-ternary "Indeed, the authors of coffeescript outlawed the ternary (t ? a : b) operator" ↩