rubocop 1.53.0 → 1.90.0

data/lib/rubocop/cop/style/exponential_notation.rb

@@ -8,7 +8,7 @@ module RuboCop
       #
       # * `scientific` which enforces a mantissa between 1 (inclusive) and 10 (exclusive).
       # * `engineering` which enforces the exponent to be a multiple of 3 and the mantissa
-      #   to be between 0.1 (inclusive) and 10 (exclusive).
+      #   to be between 0.1 (inclusive) and 1000 (exclusive).
       # * `integral` which enforces the mantissa to always be a whole number without
       #   trailing zeroes.
       #
@@ -59,9 +59,10 @@ module RuboCop
       #
       class ExponentialNotation < Base
         include ConfigurableEnforcedStyle
+
         MESSAGES = {
-          scientific: 'Use a mantissa in [1, 10[.',
-          engineering: 'Use an exponent divisible by 3 and a mantissa in [0.1, 1000[.',
+          scientific: 'Use a mantissa >= 1 and < 10.',
+          engineering: 'Use an exponent divisible by 3 and a mantissa >= 0.1 and < 1000.',
           integral: 'Use an integer as mantissa, without trailing zero.'
         }.freeze
 
@@ -87,7 +88,7 @@ module RuboCop
           true
         end
 
-        def integral(node)
+        def integral?(node)
           mantissa, = node.source.split('e')
           /^-?[1-9](\d*[1-9])?$/.match?(mantissa)
         end
@@ -101,7 +102,7 @@ module RuboCop
           when :engineering
             !engineering?(node)
           when :integral
-            !integral(node)
+            !integral?(node)
           else
             false
           end

data/lib/rubocop/cop/style/fetch_env_var.rb

@@ -6,10 +6,23 @@ module RuboCop
       # Suggests `ENV.fetch` for the replacement of `ENV[]`.
       # `ENV[]` silently fails and returns `nil` when the environment variable is unset,
       # which may cause unexpected behaviors when the developer forgets to set it.
-      # On the other hand, `ENV.fetch` raises KeyError or returns the explicitly
+      # On the other hand, `ENV.fetch` raises `KeyError` or returns the explicitly
       # specified default value.
       #
-      # @example
+      # @example DefaultToNil: true (default)
+      #   # bad
+      #   ENV['X']
+      #   x = ENV['X']
+      #
+      #   # good
+      #   ENV.fetch('X', nil)
+      #   x = ENV.fetch('X', nil)
+      #
+      #   # also good
+      #   !ENV['X']
+      #   ENV['X'].some_method # (e.g. `.nil?`)
+      #
+      # @example DefaultToNil: false
       #   # bad
       #   ENV['X']
       #   x = ENV['X']
@@ -25,7 +38,9 @@ module RuboCop
       class FetchEnvVar < Base
         extend AutoCorrector
 
-        MSG = 'Use `ENV.fetch(%<key>s)` or `ENV.fetch(%<key>s, nil)` instead of `ENV[%<key>s]`.'
+        MSG_WITH_NIL = 'Use `ENV.fetch(%<key>s, nil)` instead of `ENV[%<key>s]`.'
+        MSG_WITHOUT_NIL = 'Use `ENV.fetch(%<key>s)` instead of `ENV[%<key>s]`.'
+        RESTRICT_ON_SEND = [:[]].freeze
 
         # @!method env_with_bracket?(node)
         def_node_matcher :env_with_bracket?, <<~PATTERN
@@ -36,7 +51,7 @@ module RuboCop
           env_with_bracket?(node) do |name_node|
             break unless offensive?(node)
 
-            message = format(MSG, key: name_node.source)
+            message = format(offense_message, key: name_node.source)
             add_offense(node, message: message) do |corrector|
               corrector.replace(node, new_code(name_node))
             end
@@ -45,19 +60,27 @@ module RuboCop
 
         private
 
+        def default_to_nil?
+          cop_config.fetch('DefaultToNil', true)
+        end
+
+        def offense_message
+          default_to_nil? ? MSG_WITH_NIL : MSG_WITHOUT_NIL
+        end
+
         def allowed_var?(node)
           env_key_node = node.children.last
-          env_key_node.str_type? && cop_config['AllowedVars'].include?(env_key_node.value)
+          env_key_node.str_type? && cop_config['AllowedVariables'].include?(env_key_node.value)
         end
 
         def used_as_flag?(node)
           return false if node.root?
-          return true if used_if_condition_in_body(node)
+          return true if used_if_condition_in_body?(node)
 
           node.parent.send_type? && (node.parent.prefix_bang? || node.parent.comparison_method?)
         end
 
-        def used_if_condition_in_body(node)
+        def used_if_condition_in_body?(node)
           if_node = node.ancestors.find(&:if_type?)
 
           return false unless (condition = if_node&.condition)
@@ -124,7 +147,11 @@ module RuboCop
         end
 
         def new_code(name_node)
-          "ENV.fetch(#{name_node.source}, nil)"
+          if default_to_nil?
+            "ENV.fetch(#{name_node.source}, nil)"
+          else
+            "ENV.fetch(#{name_node.source})"
+          end
         end
       end
     end

data/lib/rubocop/cop/style/file_null.rb

@@ -0,0 +1,91 @@
+# frozen_string_literal: true
+
+module RuboCop
+  module Cop
+    module Style
+      # Use `File::NULL` instead of hardcoding the null device (`/dev/null` on Unix-like
+      # OSes, `NUL` or `NUL:` on Windows), so that code is platform independent.
+      # Only looks for full string matches, substrings within a longer string are not
+      # considered.
+      #
+      # However, only files that use the string `'/dev/null'` are targeted for detection.
+      # This is because the string `'NUL'` is not limited to the null device.
+      # This behavior results in false negatives when the `'/dev/null'` string is not used,
+      # but it is a trade-off to avoid false positives. `NULL:`
+      # Unlike `'NUL'`, `'NUL:'` is regarded as something like `C:` and is always detected.
+      #
+      # NOTE: Uses inside arrays and hashes are ignored.
+      #
+      # @safety
+      #   It is possible for a string value to be changed if code is being run
+      #   on multiple platforms and was previously hardcoded to a specific null device.
+      #
+      #   For example, the following string will change on Windows when changed to
+      #   `File::NULL`:
+      #
+      #   [source,ruby]
+      #   ----
+      #   path = "/dev/null"
+      #   ----
+      #
+      # @example
+      #   # bad
+      #   '/dev/null'
+      #   'NUL'
+      #   'NUL:'
+      #
+      #   # good
+      #   File::NULL
+      #
+      #   # ok - inside an array
+      #   null_devices = %w[/dev/null nul]
+      #
+      #   # ok - inside a hash
+      #   { unix: "/dev/null", windows: "nul" }
+      class FileNull < Base
+        extend AutoCorrector
+
+        REGEXP = %r{\A(/dev/null|NUL:?)\z}i.freeze
+        MSG = 'Use `File::NULL` instead of `%<source>s`.'
+
+        def on_new_investigation
+          return unless (ast = processed_source.ast)
+
+          @contain_dev_null_string_in_file = ast.each_descendant(:str).any? do |str|
+            content = str.str_content
+
+            valid_string?(content) && content.downcase == '/dev/null' # rubocop:disable Style/FileNull
+          end
+        end
+
+        def on_str(node)
+          value = node.value
+          return unless valid_string?(value)
+          return if acceptable?(node)
+          return if value.downcase == 'nul' && !@contain_dev_null_string_in_file # rubocop:disable Style/FileNull
+          return unless REGEXP.match?(value)
+
+          add_offense(node, message: format(MSG, source: value)) do |corrector|
+            corrector.replace(node, 'File::NULL')
+          end
+        end
+
+        private
+
+        def valid_string?(value)
+          !value.empty? && value.valid_encoding?
+        end
+
+        def acceptable?(node)
+          # Using a hardcoded null device is acceptable when inside an array or
+          # inside a hash to ensure behavior doesn't change. A `str` that is part of
+          # an interpolated or concatenated string (`dstr`) is not a standalone null
+          # device either, and replacing it would corrupt the surrounding string.
+          return false unless node.parent
+
+          node.parent.type?(:array, :pair, :dstr)
+        end
+      end
+    end
+  end
+end

data/lib/rubocop/cop/style/file_open.rb

@@ -0,0 +1,84 @@
+# frozen_string_literal: true
+
+module RuboCop
+  module Cop
+    module Style
+      # Checks for `File.open` without a block, which can leak file descriptors.
+      #
+      # When `File.open` is called without a block, the caller is responsible
+      # for closing the file descriptor. If it is not explicitly closed, it
+      # will only be closed when the garbage collector runs, which may lead
+      # to resource exhaustion. Using the block form ensures the file is
+      # automatically closed when the block exits.
+      #
+      # This cop only registers an offense when the result of `File.open` is
+      # assigned to a variable or has a method chained on it, as those are the
+      # clearest indicators that the block form should be used instead. When
+      # `File.open` is used as a return value or passed as an argument, the
+      # caller is likely managing the file descriptor intentionally.
+      #
+      # @safety
+      #   This cop is unsafe because it relies on syntax heuristics and cannot
+      #   verify whether the file descriptor is safely managed. For example, it
+      #   still flags intentional one-shot reads (`File.open("f").read`) where
+      #   the file descriptor is closed by the garbage collector.
+      #
+      # @example
+      #   # bad
+      #   f = File.open('file')
+      #
+      #   # bad
+      #   File.open('file').read
+      #
+      #   # good
+      #   File.open('file') do |f|
+      #     f.read
+      #   end
+      #
+      #   # good
+      #   File.open('file', &:read)
+      #
+      #   # good - pass an open file object to an API that manages its lifecycle
+      #   process(io: File.open('file'))
+      #
+      #   # good - return an open file object for the caller to manage
+      #   def json_key_io
+      #     File.open('file')
+      #   end
+      #
+      #   # good - use File.read for one-shot reads
+      #   File.read('file')
+      #
+      class FileOpen < Base
+        MSG = '`File.open` without a block may leak a file descriptor; use the block form.'
+        RESTRICT_ON_SEND = %i[open].freeze
+
+        # @!method file_open?(node)
+        def_node_matcher :file_open?, <<~PATTERN
+          (send (const {nil? cbase} :File) :open ...)
+        PATTERN
+
+        def on_send(node)
+          return unless file_open?(node)
+          return if node.block_argument?
+          return unless offensive_usage?(node)
+
+          add_offense(node)
+        end
+        alias on_csend on_send
+
+        private
+
+        def offensive_usage?(node)
+          return true unless node.value_used?
+
+          node.parent.lvasgn_type? || receiver_of_chained_call?(node)
+        end
+
+        def receiver_of_chained_call?(node)
+          node.parent.receiver == node
+        end
+      end
+    end
+  end
+end

data/lib/rubocop/cop/style/file_read.rb

@@ -6,8 +6,7 @@ module RuboCop
       # Favor `File.(bin)read` convenience methods.
       #
       # @example
-      #   ## text mode
-      #   # bad
+      #   # bad - text mode
       #   File.open(filename).read
       #   File.open(filename, &:read)
       #   File.open(filename) { |f| f.read }
@@ -23,9 +22,7 @@ module RuboCop
       #   # good
       #   File.read(filename)
       #
-      # @example
-      #   ## binary mode
-      #   # bad
+      #   # bad - binary mode
       #   File.open(filename, 'rb').read
       #   File.open(filename, 'rb', &:read)
       #   File.open(filename, 'rb') do |f|

data/lib/rubocop/cop/style/file_touch.rb

@@ -0,0 +1,75 @@
+# frozen_string_literal: true
+
+module RuboCop
+  module Cop
+    module Style
+      # Checks for usage of `File.open` in append mode with empty block.
+      #
+      # Such a usage only creates a new file, but it doesn't update
+      # timestamps for an existing file, which might have been the intention.
+      #
+      # For example, for an existing file `foo.txt`:
+      #
+      #   ruby -e "puts File.mtime('foo.txt')"
+      #   # 2024-11-26 12:17:23 +0100
+      #
+      #   ruby -e "File.open('foo.txt', 'a') {}"
+      #
+      #   ruby -e "puts File.mtime('foo.txt')"
+      #   # 2024-11-26 12:17:23 +0100 -> unchanged
+      #
+      # If the intention was to update timestamps, `FileUtils.touch('foo.txt')`
+      # should be used instead.
+      #
+      # @safety
+      #   Autocorrection is unsafe for this cop because unlike `File.open`,
+      #   `FileUtils.touch` updates an existing file's timestamps.
+      #
+      # @example
+      #   # bad
+      #   File.open(filename, 'a') {}
+      #   File.open(filename, 'a+') {}
+      #
+      #   # good
+      #   FileUtils.touch(filename)
+      #
+      class FileTouch < Base
+        extend AutoCorrector
+
+        MSG = 'Use `FileUtils.touch(%<argument>s)` instead of `File.open` in ' \
+              'append mode with empty block.'
+
+        RESTRICT_ON_SEND = %i[open].freeze
+
+        APPEND_FILE_MODES = %w[a a+ ab a+b at a+t].to_set.freeze
+
+        # @!method file_open?(node)
+        def_node_matcher :file_open?, <<~PATTERN
+          (send
+            (const {nil? cbase} :File) :open
+            $(...)
+            (str %APPEND_FILE_MODES))
+        PATTERN
+
+        def on_send(node)
+          filename = file_open?(node)
+          parent = node.parent
+
+          return unless filename
+          return unless parent && empty_block?(parent)
+
+          message = format(MSG, argument: filename.source)
+          add_offense(parent, message: message) do |corrector|
+            corrector.replace(parent, "FileUtils.touch(#{filename.source})")
+          end
+        end
+
+        private
+
+        def empty_block?(node)
+          node.block_type? && !node.body
+        end
+      end
+    end
+  end
+end

data/lib/rubocop/cop/style/file_write.rb

@@ -17,8 +17,7 @@ module RuboCop
       # ----
       #
       # @example
-      #   ## text mode
-      #   # bad
+      #   # bad - text mode
       #   File.open(filename, 'w').write(content)
       #   File.open(filename, 'w') do |f|
       #     f.write(content)
@@ -27,9 +26,7 @@ module RuboCop
       #   # good
       #   File.write(filename, content)
       #
-      # @example
-      #   ## binary mode
-      #   # bad
+      #   # bad - binary mode
       #   File.open(filename, 'wb').write(content)
       #   File.open(filename, 'wb') do |f|
       #     f.write(content)
@@ -107,28 +104,33 @@ module RuboCop
 
         def replacement(mode, filename, content, write_node)
           replacement = "#{write_method(mode)}(#{filename.source}, #{content.source})"
+          heredocs = removed_heredocs(filename, content, write_node)
+          return replacement if heredocs.empty?
 
-          if heredoc?(write_node)
-            first_argument = write_node.body.first_argument
+          [replacement, *heredocs.map { |heredoc| heredoc_range(heredoc).source }].join("\n")
+        end
 
-            <<~REPLACEMENT.chomp
-              #{replacement}
-              #{heredoc_range(first_argument).source}
-            REPLACEMENT
-          else
-            replacement
-          end
+        # Heredocs opened in the arguments keep working in the replacement, but their
+        # bodies are lost when they lie within the replaced range, so they need to be
+        # restored after the replacement.
+        def removed_heredocs(filename, content, write_node)
+          [filename, content].flat_map { |argument| find_heredocs(argument) }
+                             .select { |heredoc| removed?(heredoc, write_node) }
+                             .sort_by { |heredoc| heredoc.loc.heredoc_body.begin_pos }
         end
 
-        def heredoc?(write_node)
-          write_node.block_type? && (first_argument = write_node.body.first_argument) &&
-            first_argument.respond_to?(:heredoc?) && first_argument.heredoc?
+        def heredoc_range(heredoc)
+          range_between(heredoc.loc.heredoc_body.begin_pos, heredoc.loc.heredoc_end.end_pos)
         end
 
-        def heredoc_range(first_argument)
-          range_between(
-            first_argument.loc.heredoc_body.begin_pos, first_argument.loc.heredoc_end.end_pos
-          )
+        def find_heredocs(node)
+          [node, *node.each_descendant(:any_str)].select do |child|
+            child.respond_to?(:heredoc?) && child.heredoc?
+          end
+        end
+
+        def removed?(heredoc, write_node)
+          heredoc.loc.heredoc_end.end_pos <= write_node.source_range.end_pos
         end
       end
     end

data/lib/rubocop/cop/style/float_division.rb

@@ -7,9 +7,12 @@ module RuboCop
       # It is recommended to either always use `fdiv` or coerce one side only.
       # This cop also provides other options for code consistency.
       #
+      # For `Regexp.last_match` and nth reference (e.g., `$1`), it assumes that the value
+      # is a string matched by a regular expression, and allows conversion with `#to_f`.
+      #
       # @safety
       #   This cop is unsafe, because if the operand variable is a string object
-      #   then `.to_f` will be removed and an error will occur.
+      #   then `#to_f` will be removed and an error will occur.
       #
       #   [source,ruby]
       #   ----
@@ -65,19 +68,31 @@ module RuboCop
 
         # @!method right_coerce?(node)
         def_node_matcher :right_coerce?, <<~PATTERN
-          (send _ :/ (send _ :to_f))
+          (send _ :/ #to_f_method?)
         PATTERN
         # @!method left_coerce?(node)
         def_node_matcher :left_coerce?, <<~PATTERN
-          (send (send _ :to_f) :/ _)
+          (send #to_f_method? :/ _)
         PATTERN
         # @!method both_coerce?(node)
         def_node_matcher :both_coerce?, <<~PATTERN
-          (send (send _ :to_f) :/ (send _ :to_f))
+          (send #to_f_method? :/ #to_f_method?)
         PATTERN
         # @!method any_coerce?(node)
         def_node_matcher :any_coerce?, <<~PATTERN
-          {(send _ :/ (send _ :to_f)) (send (send _ :to_f) :/ _)}
+          {(send _ :/ #to_f_method?) (send #to_f_method? :/ _)}
+        PATTERN
+        # @!method to_f_method?(node)
+        def_node_matcher :to_f_method?, <<~PATTERN
+          (send !nil? :to_f)
+        PATTERN
+
+        # @!method regexp_last_match?(node)
+        def_node_matcher :regexp_last_match?, <<~PATTERN
+          {
+            (send (const {nil? cbase} :Regexp) :last_match int)
+            (:nth_ref _)
+          }
         PATTERN
 
         def on_send(node)
@@ -100,6 +115,9 @@ module RuboCop
         private
 
         def offense_condition?(node)
+          return false if regexp_last_match?(node.receiver.receiver) ||
+                          regexp_last_match?(node.first_argument.receiver)
+
           case style
           when :left_coerce
             right_coerce?(node)

data/lib/rubocop/cop/style/for.rb

@@ -8,6 +8,9 @@ module RuboCop
       # parameter. An `each` call with a block on a single line is always
       # allowed.
       #
+      # NOTE: `each` is preferred in idiomatic Ruby because `for` leaks
+      # its loop variable into the surrounding scope.
+      #
       # @example EnforcedStyle: each (default)
       #   # bad
       #   def foo
@@ -44,7 +47,6 @@ module RuboCop
       #
       class For < Base
         include ConfigurableEnforcedStyle
-        include RangeHelp
         extend AutoCorrector
 
         EACH_LENGTH = 'each'.length
@@ -66,6 +68,8 @@ module RuboCop
           return unless suspect_enumerable?(node)
 
           if style == :for
+            return unless node.receiver
+
             add_offense(node, message: PREFER_FOR) do |corrector|
               EachToForCorrector.new(node).call(corrector)
               opposite_style_detected
@@ -76,11 +80,12 @@ module RuboCop
         end
 
         alias on_numblock on_block
+        alias on_itblock on_block
 
         private
 
         def suspect_enumerable?(node)
-          node.multiline? && node.send_node.method?(:each) && !node.send_node.arguments?
+          node.multiline? && node.method?(:each) && !node.send_node.arguments?
         end
       end
     end

data/lib/rubocop/cop/style/format_string.rb

@@ -4,36 +4,49 @@ module RuboCop
   module Cop
     module Style
       # Enforces the use of a single string formatting utility.
-      # Valid options include Kernel#format, Kernel#sprintf and String#%.
+      # Valid options include `Kernel#format`, `Kernel#sprintf`, and `String#%`.
       #
-      # The detection of String#% cannot be implemented in a reliable
+      # The detection of `String#%` cannot be implemented in a reliable
       # manner for all cases, so only two scenarios are considered -
       # if the first argument is a string literal and if the second
       # argument is an array literal.
       #
+      # Autocorrection will be applied when the argument is a literal or uses a known
+      # built-in conversion method such as `to_d`, `to_f`, `to_h`, `to_i`, `to_r`, `to_s`,
+      # and `to_sym` on variables, provided that their return value is not an array.
+      # For example, when using `to_s`,
+      # `'%s' % [1, 2, 3].to_s` can be autocorrected without any incompatibility:
+      #
+      # [source,ruby]
+      # ----
+      # '%s' % [1, 2, 3]        #=> '1'
+      # format('%s', [1, 2, 3]) #=> '[1, 2, 3]'
+      # '%s' % [1, 2, 3].to_s   #=> '[1, 2, 3]'
+      # ----
+      #
       # @example EnforcedStyle: format (default)
       #   # bad
-      #   puts sprintf('%10s', 'hoge')
-      #   puts '%10s' % 'hoge'
+      #   puts sprintf('%10s', 'foo')
+      #   puts '%10s' % 'foo'
       #
       #   # good
-      #   puts format('%10s', 'hoge')
+      #   puts format('%10s', 'foo')
       #
       # @example EnforcedStyle: sprintf
       #   # bad
-      #   puts format('%10s', 'hoge')
-      #   puts '%10s' % 'hoge'
+      #   puts format('%10s', 'foo')
+      #   puts '%10s' % 'foo'
       #
       #   # good
-      #   puts sprintf('%10s', 'hoge')
+      #   puts sprintf('%10s', 'foo')
       #
       # @example EnforcedStyle: percent
       #   # bad
-      #   puts format('%10s', 'hoge')
-      #   puts sprintf('%10s', 'hoge')
+      #   puts format('%10s', 'foo')
+      #   puts sprintf('%10s', 'foo')
       #
       #   # good
-      #   puts '%10s' % 'hoge'
+      #   puts '%10s' % 'foo'
       #
       class FormatString < Base
         include ConfigurableEnforcedStyle
@@ -42,6 +55,9 @@ module RuboCop
         MSG = 'Favor `%<prefer>s` over `%<current>s`.'
         RESTRICT_ON_SEND = %i[format sprintf %].freeze
 
+        # Known conversion methods whose return value is not an array.
+        AUTOCORRECTABLE_METHODS = %i[to_d to_f to_h to_i to_r to_s to_sym].freeze
+
         # @!method formatter(node)
         def_node_matcher :formatter, <<~PATTERN
           {
@@ -53,7 +69,7 @@ module RuboCop
 
         # @!method variable_argument?(node)
         def_node_matcher :variable_argument?, <<~PATTERN
-          (send {str dstr} :% {send_type? lvar_type?})
+          (send {str dstr} :% #autocorrectable?)
         PATTERN
 
         def on_send(node)
@@ -70,6 +86,12 @@ module RuboCop
 
         private
 
+        def autocorrectable?(node)
+          return true if node.lvar_type?
+
+          node.send_type? && !AUTOCORRECTABLE_METHODS.include?(node.method_name)
+        end
+
         def message(detected_style)
           format(MSG, prefer: method_name(style), current: method_name(detected_style))
         end
@@ -122,10 +144,22 @@ module RuboCop
         end
 
         def format_single_parameter(arg)
+          # `format(fmt, *args)` is equivalent to `fmt % args`, so unwrap the splat
+          # and render the argument it splats.
+          return format_single_parameter(arg.children.first) if arg.splat_type?
+
           source = arg.source
           return "{ #{source} }" if arg.hash_type?
 
-          arg.send_type? && arg.operator_method? && !arg.parenthesized? ? "(#{source})" : source
+          requires_parentheses?(arg) ? "(#{source})" : source
+        end
+
+        # An argument that binds looser than `%` (a ternary, range, assignment, or
+        # operator call) must be parenthesized to keep its meaning.
+        def requires_parentheses?(arg)
+          return true if arg.assignment? || arg.type?(:if, :and, :or, :range)
+
+          arg.send_type? && arg.operator_method? && !arg.parenthesized?
         end
       end
     end