Sorbet Cheatsheet Part 7

Published Nov 19, 2020

Changes from last time:

  • Added T.type_alias
  • Added T.self_type
  • Added T.attached_class
  • Should the examples be in named methods or inlined into the module bodies? I think the method boilerplate might be more noise than signal, but maybe it’s just the sig { void } that’s the problem. I guess I don’t need those because I’m using typed: true.
  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
# Every file should have a "typed sigil" that tells Sorbet how strict to be
# during static type checking.
#
# Strictness levels (lax to strict):
#
# ignore: Sorbet won't even read the file.  This means its contents are not
# visible during type checking.  Avoid this.
#
# false: Sorbet will only report errors related to constant resolution.  This
# is the default if no sigil is included.
#
# true: Sorbet will report all static type errors.  This is the sweet spot of
# safety for effort.
#
# strict: Sorbet will require that all methods, constants, and instance
# variables have static types.
#
# strong: Sorbet will no longer allow anything to be T.untyped, even
# explicitly.  Almost nothing satisfies this.

# typed: true

# Include the runtime type-checking library.  This lets you write inline sigs
# and have them checked at runtime (instead of running Sorbet as RBI-only).
# These runtime checks happen even for files with `ignore` or `false` sigils.
require 'sorbet-runtime'

module BasicSigs
  # Bring in the type definition helpers.  You'll almost always need this.
  extend T::Sig

  # Sigs are defined with `sig` and a block.  Define the return value type with
  # `returns`.
  #
  # This method returns a value whose class is `String`.  These are the most
  # common types, and Sorbet calls them "class types".
  sig { returns(String) }
  def greet
    'Hello, World!'
  end

  # Define parameter value types with `params`.
  sig { params(n: Integer).returns(String) }
  def greet_repeat(n)
    (1..n).map { greet }.join("\n")
  end

  # Define keyword parameters the same way.
  sig { params(n: Integer, sep: String).returns(String) }
  def greet_repeat(n, sep: "\n")
    (1..n).map { greet }.join(sep)
  end

  # Notice that positional/keyword and required/optional make no difference
  # here.  They're all defined the same way in `params`.

  # For lots of parameters, it's nicer to use do..end and a multiline block
  # instead of curly braces.
  sig do
    params(
      str: String,
      num: Integer,
      sym: Symbol,
    ).returns(String)
  end
  def uhh(str:, num:, sym:)
    'What would you even do with these?'
  end

  # For a method whose return value is useless, use `void`.
  sig { params(name: String).void }
  def say_hello(name)
    puts "Hello, #{name}!"
  end

  # Splats! Also known as "rest parameters", "*args", "**kwargs", and others.
  #
  # Type the value that a _member_ of `args` or `kwargs` will have, not `args`
  # or `kwargs` itself.
  sig { params(args: Integer, kwargs: String).void }
  def no_op(*args, **kwargs)
    if kwargs[:op] == 'minus'
      args.each { |i| puts(i - 1) }
    else
      args.each { |i| puts(i + 1) }
    end
  end

  # Most initializers should be `void`.
  sig { params(name: String).void }
  def initialize(name:)
    # Instance variables must have annotated types to participate in static
    # type checking.

    # The value in `T.let` is checked statically and at runtime.
    @upname = T.let(name.upcase, String)

    # Sorbet can infer this one!
    @name = name
  end

  # Constants also need annotated types.
  SORBET = T.let('A delicious frozen treat', String)

  # Class variables too.
  @@the_answer = T.let(42, Integer)

  # Sorbet knows about the `attr_*` family.
  sig { returns(String) }
  attr_reader :upname

  sig { params(Integer).returns(Integer) }
  attr_writer :write_only

  # You say the reader part and Sorbet will say the writer part.
  sig { returns(String) }
  attr_accessor :name
end

module StandardHelpers
  extend T::Sig
  # Sorbet provides some helpers for typing the Ruby standard library.

  # Use T::Boolean to catch both `true` and `false`.
  #
  # For the curious, this is equivalent to
  #     T.type_alias { T.any(TrueClass, FalseClass) }
  sig { params(str: String).returns(T::Boolean) }
  def confirmed?(str)
    str == 'yes'
  end

  # Reminder that the value `nil` is an instance of NilClass.
  sig { params(val: NilClass).void }
  def only_nil(val:); end

  # To avoid modifying standard library classes, Sorbet provides wrappers to
  # support common generics.
  #
  # Here's the full list:
  #   * T::Array
  #   * T::Enumerable
  #   * T::Enumerator
  #   * T::Hash
  #   * T::Range
  #   * T::Set
  sig { params(config: T::Hash[Symbol, String]).returns(T::Array[String]) }
  def merge_values(config)
    keyset = [:old_key, :new_key]
    config.each_pair.flat_map do |key, value|
      keyset.include?(key) ? value : nil
    end
  end

  # Sometimes (usually dependency injection), a method will accept a reference
  # to a class rather than an instance of the class.  Use `T.class_of(Dep)` to
  # accept the `Dep` class itself (or something that inherits from it).
  class Dep; end

  sig { params(dep: T.class_of(Dep)).returns(Dep) }
  def dependency_injection(dep:)
    dep.new
  end

  # Blocks, procs, and lambdas, oh my!  All of these are typed with `T.proc`.
  #
  # Limitations:
  # 1. All parameters are assumed to be required positional parameters.
  # 2. The only runtime check is that the value is a `Proc`.  The argument
  #    types are only checked statically.
  sig do
    params(
      data: T::Array[String],
      blk: T.proc.params(val: String).returns(Integer),
    ).returns(Integer)
  end
  def count(data, &blk)
    data.sum(&blk)
  end

  sig { returns(Integer) }
  def count_usage
    count(["one", "two", "three"]) { |word| word.length + 1 }
  end

  # If the method takes an implicit block, Sorbet will infer `T.untyped` for
  # it.  Use the explicit block syntax if the types are important.
  sig { params(str: String).returns(T.untyped) }
  def implicit_block(str)
    yield(str)
  end

  # If you're writing a DSL and will execute the block in a different context,
  # use `bind`.
  sig { params(num: Integer, blk: T.proc.bind(Integer).void).void }
  def number_fun(num, &blk)
    num.instance_eval(&blk)
  end

  sig { void }
  def number_fun_usage(num)
    number_fun(10) { puts digits.join }
  end

  # If the block doesn't take any parameters, don't include `params`.
  sig { params(blk: T.proc.returns(Integer)).returns(Integer) }
  def doubled_block(&blk)
    2 * blk.call
  end
end

module Combinators
  extend T::Sig
  # These methods let you define new types from existing types.

  # Use `T.any` when you have a value that can be one of many types.  These are
  # sometimes known as "union types" or "sum types".
  sig { params(num: T.any(Integer, Float)).returns(Integer) }
  def hundreds(num)
    num.round(-2)
  end

  # `T.nilable(Type)` is a convenient alias for `T.any(Type, NilClass)`.
  sig { params(val: T.nilable(String)).returns(Integer) }
  def strlen(val)
    val.nil? ? -1 : val.length
  end

  # Use `T.all` when you have a value that must be satisfy multiple types.
  # These are sometimes known as "intersection types".  They're most useful for
  # interfaces (described later), but can also be useful for helper modules.

  module Reversible
    extend T::Sig
    sig { void }
    def reverse
      # Pretend this is actually implemented
    end
  end

  module Sortable
    extend T::Sig
    sig { void }
    def sort
      # Pretend this is actually implemented
    end
  end

  class List
    include Reversible
    include Sortable
  end

  sig { params(list: T.all(Reversible, Sortable)).void }
  def rev_sort(list)
    # reverse from Reversible
    list.reverse
    # sort from Sortable
    list.sort
  end

  def rev_sort_usage
    rev_sort(List.new)
  end

  # Sometimes, actually spelling out the type every time becomes more confusing
  # than helpful.  Use type aliases to make them easier to work with.
  JSONLiteral = T.type_alias { T.any(Float, String, T::Boolean, NilClass) }

  sig { params(val: JSONLiteral).returns(String) }
  def stringify(val)
    val.to_s
  end
end

module DataClasses
  # Use `T::Struct` to create a new class with type-checked fields.  It
  # combines the best parts of the standard Struct and OpenStruct, and then
  # adds static typing on top.
  #
  # Types constructed this way are sometimes known as "product types".

  class Matcher < T::Struct
    # Use `prop` to define a field with both a reader and writer.
    prop :count, Integer
    # Use `const` to only define the reader and skip the writer.
    const :pattern, Regexp
    # You can still set a default value with `default`.
    const :message, String, default: 'Found one!'

    # This is otherwise a normal class, so you can still define methods.

    # You'll still need to bring `sig` in if you want to use it though.
    extend T::Sig

    sig { void }
    def reset
      self.count = 0
    end
  end

  sig { params(text: String, matchers: T::Array[Matcher]).void }
  def awk(text, matchers)
    matchers.each(&:reset)
    text.lines.each do |line|
      matchers.each do |matcher|
        if matcher.pattern =~ line
          puts matcher.message
          matcher.count += 1
        end
      end
    end
  end

  # Gotchas and limitations

  # 1. `const` fields are not truly immutable.  They don't have a writer
  #    method, but may be changed in other ways.
  class ChangeMe < T::Struct
    const :list, T::Array[Integer]
  end

  def whoops!(change_me)
    change_me = ChangeMe.new(list: [1, 2, 3, 4])
    change_me.list.reverse!
    change_me.list == [4, 3, 2, 1]
  end

  # 2. `T::Struct` inherits its equality method from `BasicObject`, which uses
  #    identity equality (also known as "reference equality").
  class Position < T::Struct
    const :x, Integer
    const :y, Integer
  end

  def never_equal!
    p1 = Position.new(x: 1, y: 2)
    p2 = Position.new(x: 1, y: 2)
    p1 != p2
  end

  # Define your own `#==` method to check the fields, if that's what you want.
  class Position < T::Struct
    # Note: reopened class
    def ==(other)
      self.class == other.class && self.x == other.x && self.y == other.y
    end
  end

  # Use `T::Enum` to define a fixed set of values that are easy to reference.
  # This is especially useful when you don't care what the values _are_ as much
  # as you care that the set of possibilities is closed and static.
  class Crayon < T::Enum
    extend T::Sig

    # Start initialization with `enum`.
    enums do
      # Define each member with `new`.  Each of these is an instance of the
      # `Crayon` class.
      Red = new
      Orange = new
      Yellow = new
      Green = new
      Blue = new
      Violet = new
      Brown = new
      Black = new
      # The default value of the enum is its name in all-lowercase.  To change
      # that, pass a value to `new`.
      Gray90 = new('light-gray')
    end

    # Define any aliases outside the initialization block.
    Purple = Violet

    sig { returns(String) }
    def to_hex
      case self
      when Red   then '#ff0000'
      when Green then '#00ff00'
      # ...
      else            '#ffffff'
      end
    end
  end

  sig { params(crayon: Crayon, path: T::Array[Point]).void }
  def draw(crayon:, path:)
    path.each do |point|
      puts "(#{point.x}, #{point.y}) = " + crayon.to_hex
    end
  end
end

module FlowSensitivity
  extend T::Sig
  # Sorbet understands Ruby's control flow constructs and uses that information
  # to get more accurate types when your code branches.

  # You'll see this most often when doing nil checks.
  sig { params(name: T.nilable(String)).returns(String) }
  def greet_loudly(name)
    if name.nil?
      'HELLO, YOU!'
    else
      # Sorbet knows that `name` must be a String here, so it's safe to call
      # `#upcase`.
      "HELLO, #{name.upcase}!"
    end
  end

  # The nils are a special case of refining `T.any`.
  sig { params(id: T.any(Integer, T::Array[Integer])).returns(T::Array[String]) }
  def database_lookup(id)
    if id.is_a?(Integer)
      # `ids` must be an Integer here.
      [id.to_s]
    else
      # `ids` must be a T::Array[Integer] here.
      id.map(&:to_s)
    end
  end

  # Sorbet recognizes these methods that narrow type definitions:
  # * is_a?
  # * kind_of?
  # * nil?
  # * Class#===
  # * Class#<
  # * block_given?
  #
  # Because there so common, it also recognizes these Rails extensions:
  # * blank?
  # * present?
  #
  # Be careful to maintain Sorbet assumptions if you redefine these methods!

  # Have you've ever written this line of code?
  #
  #     raise StandardError, "Can't happen"
  #
  # Sorbet can help you prove that statically (this is known as
  # "exhaustiveness") with `T.absurd`.  It's extra cool when combined with
  # `T::Enum`!

  class Size < T::Enum
    enums do
      Byte = new('B')
      Kibibyte = new('KiB')
      Mebibyte = new('MiB')
      # "640K ought to be enough for anybody"
    end

    sig { returns(Integer) }
    def bytes
      case self
        when Byte     then 1 <<  0
        when Kibibyte then 1 << 10
        when Mebibyte then 1 << 20
        else
          # Sorbet knows you've checked all the cases, so there's no possible
          # value that `self` could have here.
          #
          # But if you _do_ get here somehow, this will raise at runtime.
          T.absurd(self)

          # If you're missing a case, Sorbet can even tell you which one it is!
      end
    end
  end

  # Sorbet knows that no code can execute after a `raise` statement because it
  # "never returns".
  sig { params(num: T.nilable(Integer)).returns(Integer) }
  def decrement(num)
    raise ArgumentError, '¯\_(ツ)_/¯' unless num

    num - 1
  end

  # You can annotate your own error-raising methods with `T.noreturn`.
  class CustomError < StandardError; end
  sig { params(message: String).returns(T.noreturn) }
  def oh_no(message = 'A bad thing happened')
    puts message
    raise CustomError, message
  end

  # It also applies to infinite loops.
  sig { returns(T.noreturn) }
  def loading
    loop do
      %q(-\|/).each_char do |c|
        print "\r#{c} reticulating splines..."
        sleep 1
      end
    end
  end

  # You may run into a situation where Sorbet "loses" your type refinement.
  # Remember that almost everything you do in Ruby is a method call that could
  # return a different value next time you call it.  Sorbet doesn't assume that
  # any methods are pure (even those from `attr_reader` and `attr_accessor`).
  sig { returns(T.nilable(Integer)) }
  def answer
    rand > 0.5 ? 42 : nil
  end

  sig { void }
  def bad_typecheck
    if answer.nil?
      0
    else
      # But answer might return `nil` if we call it again!
      answer + 1
    end
  end

  sig { void }
  def good_typecheck
    ans = answer
    if ans.nil?
      0
    else
      # Now Sorbet knows that `ans` is non-nil.
      ans + 1
    end
  end
end

module ReceiverTypes
  #

  # TODO: T.self_type
  # TODO: T.attached_class
end

module Inheritance
  extend T::Sig

  # If you have a method that always returns the type of its receiver, use
  # `T.self_type`.  This is common in fluent interfaces and DSLs.
  #
  # Warning: This feature is still experimental!
  class Logging
    extend T::Sig

    sig { returns(T.self_type) }
    def log
      pp self
      self
    end
  end

  class Data < Logging
    extend T::Sig

    sig { params(x: Integer, y: String).void }
    def initialize(x: 0, y: '')
      @x = x
      @y = y
    end

    # You don't _have_ to use `T.self_type` if there's only one relevant class.
    sig { params(x: Integer).returns(Data) }
    def setX(x)
      @x = x
      self
    end

    sig { params(y: String).returns(Data) }
    def setY(y)
      @y = y
      self
    end
  end

  # Tada!
  sig { void }
  def chaining(data: Data)
    data.setX(1).log.setY('a')
  end

  # If it's a class method (a.k.a. singleton method), use `T.attached_class`.
  # No warning here.  This one is stable!
  class Box
    extend T::Sig

    sig { params(contents: String, weight: Integer).void }
    def initialize(contents, weight)
      @contents = contents
      @weight = weight
    end

    sig { params(contents: String).returns(T.attached_class) }
    def self.pack(contents)
      new(contents, contents.chars.uniq.length)
    end
  end

  class CompanionCube < Box
    extend T::Sig

    sig { returns(String) }
    def pick_up
      "♥#{@contents}🤍"
    end
  end

  sig { returns(String) }
  def befriend
    CompanionCube.pack('').pick_up
  end

  # TODO: abstract!
  # TODO: interface!
  # TODO: abstract. / override.
  # TODO: mixes_in_class_methods
  # TODO: final!
  # TODO: sealed!
end

module Debugging
  # TODO: T.reveal_type
end

module EscapeHatches
  # TODO: T.untyped
  # TODO: T.cast
  # TODO: T.unsafe
  # TODO: T.must
  # TODO: T.assert_type!
end

# The following types are not officially documented but are still useful.

module ValueSet
  # TODO: T.enum
end

module Generics
  # TODO: type_parameters / T.type_parameter
  # TODO: T::Generic
  # TODO: type_member
end