-
Notifications
You must be signed in to change notification settings - Fork 27
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Merge pull request #175 from Shopify/buggy-obsolete-memoization-cop
Extract new `BuggyObsoleteStrictMemoization` cop
- Loading branch information
Showing
9 changed files
with
317 additions
and
31 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
83 changes: 83 additions & 0 deletions
83
lib/rubocop/cop/sorbet/buggy_obsolete_strict_memoization.rb
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,83 @@ | ||
# frozen_string_literal: true | ||
|
||
require "rubocop" | ||
|
||
module RuboCop | ||
module Cop | ||
module Sorbet | ||
# Checks for the a mistaken variant of the "obsolete memoization pattern" that used to be required | ||
# for older Sorbet versions in `#typed: strict` files. The mistaken variant would overwrite the ivar with `nil` | ||
# on every call, causing the memoized value to be discarded and recomputed on every call. | ||
# | ||
# This cop will correct it to read from the ivar instead of `nil`, which will memoize it correctly. | ||
# | ||
# The result of this correction will be the "obsolete memoization pattern", which can further be corrected by | ||
# the `Sorbet/ObsoleteStrictMemoization` cop. | ||
# | ||
# See `Sorbet/ObsoleteStrictMemoization` for more details. | ||
# | ||
# @safety | ||
# If the computation being memoized had side effects, calling it only once (instead of once on every call | ||
# to the affected method) can be observed, and might be a breaking change. | ||
# | ||
# @example | ||
# # bad | ||
# sig { returns(Foo) } | ||
# def foo | ||
# # This `nil` is likely a mistake, causing the memoized value to be discarded and recomputed on every call. | ||
# @foo = T.let(nil, T.nilable(Foo)) | ||
# @foo ||= some_computation | ||
# end | ||
# | ||
# # good | ||
# sig { returns(Foo) } | ||
# def foo | ||
# # This will now memoize the value as was likely intended, so `some_computation` is only ever called once. | ||
# # ⚠️If `some_computation` has side effects, this might be a breaking change! | ||
# @foo = T.let(@foo, T.nilable(Foo)) | ||
# @foo ||= some_computation | ||
# end | ||
# | ||
# @see Sorbet/ObsoleteStrictMemoization | ||
class BuggyObsoleteStrictMemoization < RuboCop::Cop::Base | ||
include RuboCop::Cop::MatchRange | ||
include RuboCop::Cop::Alignment | ||
include RuboCop::Cop::LineLengthHelp | ||
include RuboCop::Cop::RangeHelp | ||
extend AutoCorrector | ||
|
||
include TargetSorbetVersion | ||
|
||
MSG = "This might be a mistaken variant of the two-stage workaround that used to be needed for memoization in "\ | ||
"`#typed: strict` files. See https://sorbet.org/docs/type-assertions#put-type-assertions-behind-memoization." | ||
|
||
# @!method buggy_legacy_memoization_pattern?(node) | ||
def_node_matcher :buggy_legacy_memoization_pattern?, <<~PATTERN | ||
(begin | ||
... # Ignore any other lines that come first. | ||
(ivasgn $_ivar # First line: @_ivar = ... | ||
(send # T.let(_ivar, T.nilable(_ivar_type)) | ||
(const {nil? cbase} :T) :let | ||
$nil | ||
(send (const {nil? cbase} :T) :nilable _ivar_type))) # T.nilable(_ivar_type) | ||
(or-asgn (ivasgn _ivar) _initialization_expr)) # Second line: @_ivar ||= _initialization_expr | ||
PATTERN | ||
|
||
def on_begin(node) | ||
buggy_legacy_memoization_pattern?(node) do |ivar, nil_node| | ||
add_offense(nil_node) do |corrector| | ||
corrector.replace( | ||
range_between(nil_node.source_range.begin_pos, nil_node.source_range.end_pos), | ||
ivar, | ||
) | ||
end | ||
end | ||
end | ||
|
||
def relevant_file?(file) | ||
super && sorbet_enabled? | ||
end | ||
end | ||
end | ||
end | ||
end |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
165 changes: 165 additions & 0 deletions
165
spec/rubocop/cop/sorbet/buggy_obsolete_strict_memoization_spec.rb
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,165 @@ | ||
# frozen_string_literal: true | ||
|
||
require "spec_helper" | ||
|
||
RSpec.describe(RuboCop::Cop::Sorbet::BuggyObsoleteStrictMemoization, :config) do | ||
let(:specs_without_sorbet) do | ||
[ | ||
Gem::Specification.new("foo", "0.0.1"), | ||
Gem::Specification.new("bar", "0.0.2"), | ||
] | ||
end | ||
|
||
before(:each) do | ||
allow(Bundler).to(receive(:locked_gems)).and_return( | ||
Struct.new(:specs).new([ | ||
*specs_without_sorbet, | ||
Gem::Specification.new("sorbet-static", "0.5.10210"), | ||
]), | ||
) | ||
allow(cop).to(receive(:configured_indentation_width).and_return(2)) | ||
end | ||
|
||
describe "non-offending cases" do | ||
it "the new memoization pattern doesn't register any offense" do | ||
expect_no_offenses(<<~RUBY) | ||
def foo | ||
@foo ||= T.let(Foo.new, T.nilable(Foo)) | ||
end | ||
RUBY | ||
end | ||
|
||
describe "the correct obsolete memoization pattern" do | ||
it " doesn't register any offense" do | ||
expect_no_offenses(<<~RUBY) | ||
def foo | ||
@foo = T.let(@foo, T.nilable(Foo)) | ||
@foo ||= Foo.new | ||
end | ||
RUBY | ||
end | ||
|
||
describe "with fully qualified ::T" do | ||
it " doesn't register any offense" do | ||
expect_no_offenses(<<~RUBY) | ||
def foo | ||
@foo = ::T.let(@foo, ::T.nilable(Foo)) | ||
@foo ||= Foo.new | ||
end | ||
RUBY | ||
end | ||
end | ||
|
||
describe "with a complex type" do | ||
it "doesn't register any offense" do | ||
expect_no_offenses(<<~RUBY) | ||
def client_info_hash | ||
@client_info_hash = T.let(@client_info_hash, T.nilable(T::Hash[Symbol, T.untyped])) | ||
@client_info_hash ||= client_info.to_hash | ||
end | ||
RUBY | ||
end | ||
end | ||
|
||
describe "with multiline initialization expression" do | ||
it "doesn't register any offense" do | ||
expect_no_offenses(<<~RUBY) | ||
def foo | ||
@foo = T.let(@foo, T.nilable(Foo)) | ||
@foo ||= multiline_method_call( | ||
foo, | ||
bar, | ||
baz, | ||
) | ||
end | ||
RUBY | ||
end | ||
|
||
describe "with a gap between the two lines" do | ||
it "doesn't register any offense" do | ||
expect_no_offenses(<<~RUBY) | ||
def foo | ||
@foo = T.let(@foo, T.nilable(Foo)) | ||
@foo ||= multiline_method_call( | ||
foo, | ||
bar, | ||
baz, | ||
) | ||
end | ||
RUBY | ||
end | ||
end | ||
end | ||
|
||
describe "with non-empty lines between the two lines" do | ||
it "doesn't register any offense" do | ||
expect_no_offenses(<<~RUBY) | ||
def foo | ||
@foo = T.let(@foo, T.nilable(Foo)) | ||
some_other_computation | ||
@foo ||= multiline_method_call( | ||
foo, | ||
bar, | ||
baz, | ||
) | ||
end | ||
RUBY | ||
end | ||
end | ||
|
||
context "when its not the first line in a method" do | ||
it "doesn't register any offense" do | ||
expect_no_offenses(<<~RUBY) | ||
def foo | ||
some | ||
other | ||
code | ||
@foo = T.let(@foo, T.nilable(Foo)) | ||
@foo ||= Foo.new | ||
end | ||
RUBY | ||
end | ||
end | ||
end | ||
end | ||
|
||
describe "a mistaken variant of the obsolete memoization pattern" do | ||
context "not using Sorbet" do | ||
# If the project is not using Sorbet, the obsolete memoization pattern might be intentional. | ||
it "does not register an offence" do | ||
allow(Bundler).to(receive(:locked_gems)).and_return( | ||
Struct.new(:specs).new(specs_without_sorbet), | ||
) | ||
|
||
expect_no_offenses(<<~RUBY) | ||
sig { returns(Foo) } | ||
def foo | ||
@foo = T.let(@foo, T.nilable(Foo)) | ||
@foo ||= Foo.new | ||
end | ||
RUBY | ||
end | ||
end | ||
|
||
it "registers an offence and autocorrects" do | ||
# This variant would have been a mistake, which would have caused the memoized value to be discarded | ||
# and recomputed on every call. We can fix it up into the working version. | ||
|
||
expect_offense(<<~RUBY) | ||
def foo | ||
@foo = T.let(nil, T.nilable(Foo)) | ||
^^^ This might be a mistaken variant of the two-stage workaround that used to be needed for memoization in `#typed: strict` files. See https://sorbet.org/docs/type-assertions#put-type-assertions-behind-memoization. | ||
@foo ||= Foo.new | ||
end | ||
RUBY | ||
|
||
expect_correction(<<~RUBY) | ||
def foo | ||
@foo = T.let(@foo, T.nilable(Foo)) | ||
@foo ||= Foo.new | ||
end | ||
RUBY | ||
end | ||
end | ||
end |
Oops, something went wrong.