rubocop 1.71.1 → 1.91.0

data/lib/rubocop/cop/mixin/project_index_help.rb

@@ -0,0 +1,188 @@
+# frozen_string_literal: true
+
+module RuboCop
+  module Cop
+    # Common helpers for cops that consult the project-wide static-analysis index
+    # via `Cop::Base#project_index`.
+    #
+    # Mixed-in cops gain the `external_dependency_checksum` override that invalidates
+    # the `ResultCache` whenever the indexed project files change on disk.
+    # To run index-backed analysis, cops should simply check whether `project_index` is non-nil;
+    # the runner only exposes a non-nil index when the user opted in via `AllCops/UseProjectIndex`
+    # and the underlying gem is available.
+    module ProjectIndexHelp
+      BUILTIN_DOCUMENT_URI = 'rubydex:built-in'
+      FILE_URI_PREFIX = 'file://'
+      # Matches the spurious leading slash before a Windows drive letter that
+      # remains after stripping `file://` from a `file:///C:/...` URI.
+      WINDOWS_DRIVE_PREFIX = %r{\A/(?=[A-Za-z]:[/\\])}.freeze
+
+      class << self
+        # The signature is a property of the index, not of the cop, and computing
+        # it stats every indexed file, so all cops sharing an index share one
+        # computation. A single-entry cache (instead of a hash keyed by index)
+        # avoids retaining stale graphs in long-lived processes.
+        attr_accessor :cached_index_signature
+      end
+
+      def external_dependency_checksum
+        return nil unless project_index
+
+        @external_dependency_checksum ||= Digest::SHA1.hexdigest(
+          project_index_signature.join("\n")
+        )
+      end
+
+      private
+
+      # Returns the definitions among `definitions` that live in a file other than the
+      # one being inspected, ordered by path and line. Definitions without a `file://`
+      # URI (e.g. Rubydex's built-in declarations) are ignored.
+      def definitions_in_other_files(definitions)
+        current = processed_source.file_path
+
+        definitions
+          .select { |definition| definition.location.uri.start_with?(FILE_URI_PREFIX) }
+          .reject { |definition| File.identical?(definition.location.to_file_path, current) }
+          .sort_by do |definition|
+          [definition.location.to_file_path,
+           definition.location.start_line]
+        end
+      end
+
+      def prior_definition_in_other_file(definitions)
+        definitions_in_other_files(definitions).first
+      end
+
+      # Resolves a constant node the way Ruby does: the first segment through
+      # the lexical nesting and every following segment inside the previous
+      # one. (Qualified names cannot be passed to `resolve_constant` as a
+      # whole, since it only applies the nesting to the full name. Later
+      # segments use `find_member` — the namespace and its ancestors — since
+      # `resolve_constant` cannot take an already-resolved namespace as the
+      # scope, only names as they appear in source.)
+      def resolve_constant_in_index(const_node)
+        segments = const_node.const_name.split('::')
+        nesting = const_node.absolute? ? [] : lexical_nesting_of(const_node)
+
+        declaration = project_index.resolve_constant(segments.first, nesting)
+        segments.drop(1).each do |segment|
+          return nil unless declaration.is_a?(Rubydex::Namespace)
+
+          declaration = declaration.find_member(segment)
+        end
+
+        declaration
+      end
+
+      # The lexical nesting the node's constants resolve through, outermost
+      # first. Only scopes whose *body* contains the node count: a class or
+      # module's identifier and superclass expression are evaluated before its
+      # scope exists, so ancestors reached through them are excluded.
+      def lexical_nesting_of(node)
+        nesting = []
+        child = node
+
+        node.each_ancestor do |ancestor|
+          if ancestor.type?(:class, :module) && child.equal?(ancestor.body)
+            nesting << ancestor.identifier.const_name
+          end
+          child = ancestor
+        end
+
+        nesting.reverse
+      end
+
+      # The declaration of `declaration`'s singleton class, or nil when no
+      # singleton method is defined on it anywhere in the project.
+      def indexed_singleton_of(declaration)
+        project_index["#{declaration.name}::<#{declaration.name.split('::').last}>"]
+      end
+
+      # A namespace without any singleton method has no singleton-class
+      # declaration of its own, so the lookup starts from the first ancestor
+      # that has one; its `find_member` covers the rest of the chain.
+      def indexed_singleton_member(namespace, member_name)
+        namespace.ancestors.each do |ancestor|
+          singleton = indexed_singleton_of(ancestor)
+          return singleton.find_member(member_name) if singleton
+        end
+
+        nil
+      end
+
+      # Whether an ancestor of `scope` other than `scope` itself defines
+      # `member_name`.
+      def inherited_index_member?(scope, member_name)
+        scope.ancestors.any? do |ancestor|
+          ancestor.name != scope.name && ancestor.member(member_name)
+        end
+      end
+
+      # Whether every link of the declaration's ancestry is resolved in the
+      # index: no definition of any ancestor has an unresolved superclass or
+      # mixin reference. Unresolved links silently vanish from `ancestors`,
+      # so this is the completeness check that must pass before reasoning
+      # about what an ancestry does or does not define. `extend` mixins only
+      # affect the singleton class; pass `ignore_extend: true` when reasoning
+      # about instance-side ancestry only.
+      def fully_resolved_index_ancestry?(declaration, ignore_extend: false)
+        declaration.ancestors.all? do |ancestor|
+          ancestor.definitions.all? do |definition|
+            resolved_superclass_reference?(definition) &&
+              resolved_mixin_references?(definition, ignore_extend: ignore_extend)
+          end
+        end
+      end
+
+      def resolved_superclass_reference?(definition)
+        return true unless definition.is_a?(Rubydex::ClassDefinition)
+
+        !definition.superclass.is_a?(Rubydex::UnresolvedConstantReference)
+      end
+
+      def resolved_mixin_references?(definition, ignore_extend: false)
+        return true unless definition.respond_to?(:mixins)
+
+        definition.mixins.none? do |mixin|
+          next false if ignore_extend && mixin.is_a?(Rubydex::Extend)
+
+          mixin.constant_reference.is_a?(Rubydex::UnresolvedConstantReference)
+        end
+      end
+
+      def same_file?(path, other)
+        return true if File.identical?(path, other)
+
+        normalized = [path, other].map { |p| File.expand_path(p).tr('\\', '/') }
+        normalized.uniq.one? || (Platform.windows? && normalized[0].casecmp?(normalized[1]))
+      end
+
+      def project_index_signature
+        index, signature = ProjectIndexHelp.cached_index_signature
+        return signature if index.equal?(project_index)
+
+        compute_project_index_signature.tap do |computed|
+          ProjectIndexHelp.cached_index_signature = [project_index, computed]
+        end
+      end
+
+      def compute_project_index_signature
+        project_index.documents.filter_map do |doc|
+          uri = doc.uri
+          next if uri == BUILTIN_DOCUMENT_URI
+
+          path = uri.delete_prefix(FILE_URI_PREFIX).sub(WINDOWS_DRIVE_PREFIX, '')
+          mtime, size = begin
+            stat = File.stat(path)
+            [stat.mtime.to_f, stat.size]
+          rescue StandardError
+            [0, 0]
+          end
+
+          "#{path}:#{mtime}:#{size}"
+        end.sort
+      end
+    end
+  end
+end

data/lib/rubocop/cop/mixin/range_help.rb

@@ -4,9 +4,10 @@ module RuboCop
   module Cop
     # Methods that calculate and return Parser::Source::Ranges
     module RangeHelp
-      private
-
       BYTE_ORDER_MARK = 0xfeff # The Unicode codepoint
+      NOT_GIVEN = Module.new
+
+      private
 
       def source_range(source_buffer, line_number, column, length = 1)
         if column.is_a?(Range)
@@ -33,6 +34,18 @@ module RuboCop
         range_between(node.loc.begin.end_pos, node.loc.end.begin_pos)
       end
 
+      # A range containing the first to the last argument
+      # of a method call or method definition.
+      # def foo(a, b:)
+      #         ^^^^^
+      # bar(1, 2, 3, &blk)
+      #     ^^^^^^^^^^^^^
+      # baz { |x, y:, z:| }
+      #        ^^^^^^^^^
+      def arguments_range(node)
+        node.first_argument.source_range.join(node.last_argument.source_range)
+      end
+
       def range_between(start_pos, end_pos)
         Parser::Source::Range.new(processed_source.buffer, start_pos, end_pos)
       end
@@ -51,7 +64,6 @@ module RuboCop
         Parser::Source::Range.new(buffer, begin_pos, end_pos)
       end
 
-      NOT_GIVEN = Module.new
       def range_with_surrounding_space(range_positional = NOT_GIVEN, # rubocop:disable Metrics/ParameterLists
                                        range: NOT_GIVEN, side: :both, newlines: true,
                                        whitespace: false, continuations: false,
@@ -104,14 +116,13 @@ module RuboCop
         end
       end
 
-      # rubocop:disable Metrics/ParameterLists
+      # rubocop:disable-next Metrics/ParameterLists
       def final_pos(src, pos, increment, continuations, newlines, whitespace)
         pos = move_pos(src, pos, increment, true, /[ \t]/)
         pos = move_pos_str(src, pos, increment, continuations, "\\\n")
         pos = move_pos(src, pos, increment, newlines, /\n/)
         move_pos(src, pos, increment, whitespace, /\s/)
       end
-      # rubocop:enable Metrics/ParameterLists
 
       def move_pos(src, pos, step, condition, regexp)
         offset = step == -1 ? -1 : 0

data/lib/rubocop/cop/mixin/reparsed_equivalence.rb

@@ -0,0 +1,171 @@
+# frozen_string_literal: true
+
+module RuboCop
+  module Cop
+    # Verifies that a candidate correction of the inspected source is a
+    # syntactic no-op: the corrected source must parse successfully and produce
+    # the same AST as the original.
+    #
+    # This turns "is this piece of syntax redundant?" questions into a
+    # parser-backed check instead of a hand-maintained enumeration of grammar
+    # rules: instead of modeling which constructs make (for example) a line
+    # continuation or a pair of parentheses significant, a cop can apply the
+    # correction and let the parser answer.
+    #
+    # Cops implement `#apply_reparse_correction(corrector, item)` and call
+    # `verified_by_reparse(items)`, which returns the items whose corrections
+    # are verified. Items sharing a reparse scope (the innermost method
+    # definition or class/module body, which parse standalone and cannot
+    # capture outer local variables) are verified together with a single
+    # reparse, falling back to per-item verification when the batch does not
+    # hold. Scope groups are keyed by node identity, since structurally
+    # identical definitions in different places must not share a group.
+    #
+    # Two hooks customize the comparison:
+    #
+    # * `normalize_reparsed_ast(node)` - loosen strict tree equality where a
+    #   cop's correction is equivalence-preserving beyond parse identity (such
+    #   laws must be justified per cop; the default compares trees as-is).
+    # * `preprocess_reparsed_source(source)` - rewrite both sides before
+    #   parsing (e.g. to neutralize `__LINE__` when a correction legitimately
+    #   shifts line numbers).
+    #
+    # `__FILE__` is handled by parsing with the original path, and fragments
+    # containing `__LINE__` are verified by reparsing both sides so that
+    # fragment line offsets cancel out.
+    #
+    # Note that comments are invisible to the AST, so corrections that may
+    # touch comment text must be guarded separately by the caller.
+    module ReparsedEquivalence
+      # Above this size, verification does not reparse. What happens to such
+      # items depends on `oversized`; in practice only machine-generated files
+      # come near the limit.
+      MAX_VERIFICATION_FRAGMENT_SIZE = 64 * 1024
+
+      private
+
+      # Returns the items whose corrections are verified. `oversized` controls
+      # items whose reparse scope exceeds `MAX_VERIFICATION_FRAGMENT_SIZE`:
+      # `:report` accepts them unverified (for cops whose offense logic stands
+      # on its own and uses verification as a safety gate), `:verify` reparses
+      # regardless of size (for cops where verification is the offense logic).
+      def verified_by_reparse(items, oversized: :report)
+        scope_groups(items).flat_map do |scope, group|
+          if (oversized == :report && verification_too_large?(scope)) ||
+             (group.size > 1 && corrections_verify?(scope, group))
+            group
+          else
+            group.select { |item| corrections_verify?(scope, [item]) }
+          end
+        end
+      end
+
+      # Whether the exact correction for `item` produces source that still
+      # parses, without requiring an equivalent AST. For corrections that
+      # intentionally change the tree.
+      def correction_parses?(item)
+        corrector = Corrector.new(processed_source)
+        apply_reparse_correction(corrector, item)
+
+        parse(corrector.process, processed_source.path).valid_syntax?
+      rescue ::Parser::ClobberingError
+        false
+      end
+
+      # Hook: loosen strict tree equality for corrections that are
+      # equivalence-preserving beyond parse identity.
+      def normalize_reparsed_ast(node)
+        node
+      end
+
+      # Hook: rewrite both sides before parsing.
+      def preprocess_reparsed_source(source)
+        source
+      end
+
+      def scope_groups(items)
+        groups = {}.compare_by_identity
+        items.each { |item| (groups[reparse_scope(item_range(item))] ||= []) << item }
+
+        groups
+      end
+
+      def item_range(item)
+        item.respond_to?(:source_range) ? item.source_range : item
+      end
+
+      # The innermost scope that both contains `range` and parses standalone.
+      # Method definitions and class/module bodies neither capture outer local
+      # variables nor continue an outer expression; blocks and single
+      # statements do not qualify, since an outer local would reparse as a
+      # method call.
+      def reparse_scope(range)
+        node = processed_source.ast
+        scope = nil
+
+        while node.is_a?(::Parser::AST::Node)
+          scope = node if node.type?(:any_def, :class, :module, :sclass)
+          node = node.children.find do |child|
+            child.is_a?(::Parser::AST::Node) && child.source_range&.contains?(range)
+          end
+        end
+
+        scope
+      end
+
+      def verification_too_large?(scope)
+        (scope ? scope.source_range.size : processed_source.raw_source.size) >
+          MAX_VERIFICATION_FRAGMENT_SIZE
+      end
+
+      def corrections_verify?(scope, items)
+        corrector = Corrector.new(processed_source)
+        items.each { |item| apply_reparse_correction(corrector, item) }
+        corrected = corrector.process
+
+        if scope
+          fragment = corrected_scope_fragment(scope, corrected)
+          parses_equivalently?(scope.source, scope, fragment)
+        else
+          parses_equivalently?(processed_source.raw_source, processed_source.ast, corrected)
+        end
+      rescue ::Parser::ClobberingError
+        false
+      end
+
+      # The corrections' edits are all contained within the scope, so the
+      # corrected fragment can be cut out of the corrected source by adjusting
+      # for the edits' length delta.
+      def corrected_scope_fragment(scope, corrected)
+        delta = corrected.length - processed_source.raw_source.length
+        scope_range = scope.source_range
+
+        corrected[scope_range.begin_pos...(scope_range.end_pos + delta)]
+      end
+
+      # Both sides are parsed with the original path so that `__FILE__`
+      # resolves identically. The original side is reparsed (instead of using
+      # the already parsed node) when preprocessing altered it or it contains
+      # `__LINE__`, whose values would otherwise differ from a fragment parsed
+      # at a different line offset.
+      def parses_equivalently?(original_source, original_ast, corrected_source)
+        original_normalized = normalized_original(original_source, original_ast)
+        return false unless original_normalized
+
+        rewritten = parse(preprocess_reparsed_source(corrected_source), processed_source.path)
+        rewritten.valid_syntax? && normalize_reparsed_ast(rewritten.ast) == original_normalized
+      end
+
+      def normalized_original(original_source, original_ast)
+        original = preprocess_reparsed_source(original_source)
+        if original != original_source || original_source.include?('__LINE__')
+          reparsed = parse(original, processed_source.path).ast
+          reparsed && normalize_reparsed_ast(reparsed)
+        else
+          @normalized_originals ||= {}.compare_by_identity
+          @normalized_originals[original_ast] ||= normalize_reparsed_ast(original_ast)
+        end
+      end
+    end
+  end
+end

data/lib/rubocop/cop/mixin/space_after_punctuation.rb

@@ -8,8 +8,8 @@ module RuboCop
       MSG = 'Space missing after %<token>s.'
 
       def on_new_investigation
-        each_missing_space(processed_source.tokens) do |token|
-          add_offense(token.pos, message: format(MSG, token: kind(token))) do |corrector|
+        each_missing_space(processed_source.tokens) do |token, kind|
+          add_offense(token.pos, message: format(MSG, token: kind)) do |corrector|
             PunctuationCorrector.add_space(corrector, token)
           end
         end
@@ -19,11 +19,12 @@ module RuboCop
 
       def each_missing_space(tokens)
         tokens.each_cons(2) do |token1, token2|
-          next unless kind(token1)
+          kind = kind(token1, token2)
+          next unless kind
           next unless space_missing?(token1, token2)
           next unless space_required_before?(token2)
 
-          yield token1
+          yield token1, kind
         end
       end
 

data/lib/rubocop/cop/mixin/statement_modifier.rb

@@ -5,6 +5,7 @@ module RuboCop
     # Common functionality for modifier cops.
     module StatementModifier
       include LineLengthHelp
+      include AllowedPattern
 
       private
 
@@ -34,17 +35,58 @@ module RuboCop
         condition.each_node.any?(&:lvasgn_type?)
       end
 
+      # The rendered modifier form is compared against the bare maximum, not
+      # `acceptable_line_length?`: `Layout/LineLength`'s exemptions describe
+      # long lines the user tolerates, not permission to manufacture new ones,
+      # so collapsing to modifier form must never push a line past the maximum.
       def modifier_fits_on_single_line?(node)
         return true unless max_line_length
 
-        length_in_modifier_form(node) <= max_line_length
+        line_length(line_in_modifier_form(node)) <= max_line_length
       end
 
-      def length_in_modifier_form(node)
+      def line_in_modifier_form(node)
         keyword_element = node.loc.keyword
         code_before = keyword_element.source_line[0...keyword_element.column]
-        expression = to_modifier_form(node)
-        line_length("#{code_before}#{expression}#{code_after(node)}")
+
+        "#{code_before}#{to_modifier_form(node)}#{code_after(node)}"
+      end
+
+      # Whether an existing line is acceptable to `Layout/LineLength`,
+      # exemptions (the cop disabled at that line, an allowed pattern, an
+      # allowed cop directive, an allowed URI) included. Only for judging
+      # lines already in the source as too long - see
+      # `modifier_fits_on_single_line?` for why the exemptions must not apply
+      # when deciding whether a modifier line may be created.
+      def acceptable_line_length?(line, line_number)
+        return true unless max_line_length
+        return true if line_length(line) <= max_line_length
+        return true unless line_length_enabled_at_line?(line_number)
+        return true if matches_allowed_pattern?(line)
+
+        if allow_cop_directives? && directive_on_source_line?(line_number - 1)
+          return line_length_without_directive(line) <= max_line_length
+        end
+
+        allowed_by_uri?(line)
+      end
+
+      def allowed_by_uri?(line)
+        return false unless allow_uri?
+
+        uri_range = find_excessive_range(line, :uri)
+        !uri_range.nil? && allowed_position?(line, uri_range)
+      end
+
+      def line_length_enabled_at_line?(line)
+        processed_source.comment_config.cop_enabled_at_line?('Layout/LineLength', line)
+      end
+
+      # `Layout/LineLength`'s allowed patterns, so a modifier line that the
+      # user has configured that cop to accept is not flagged for length here.
+      def allowed_patterns
+        line_length_config = config.for_cop('Layout/LineLength')
+        line_length_config['AllowedPatterns'] || line_length_config['IgnoredPatterns'] || []
       end
 
       def to_modifier_form(node)
@@ -75,7 +117,7 @@ module RuboCop
       end
 
       def first_line_comment(node)
-        comment = processed_source.comments.find { |c| same_line?(c, node) }
+        comment = processed_source.comment_at_line(node.first_line)
         return unless comment
 
         comment_source = comment.source
@@ -100,12 +142,6 @@ module RuboCop
         node.parent.send_type?
       end
 
-      def max_line_length
-        return unless config.cop_enabled?('Layout/LineLength')
-
-        config.for_cop('Layout/LineLength')['Max']
-      end
-
       def comment_disables_cop?(comment)
         regexp_pattern = "# rubocop : (disable|todo) ([^,],)* (all|#{cop_name})"
         Regexp.new(regexp_pattern.gsub(' ', '\s*')).match?(comment)

data/lib/rubocop/cop/mixin/string_help.rb

@@ -10,7 +10,7 @@ module RuboCop
       def on_str(node)
         # Constants like __FILE__ are handled as strings,
         # but don't respond to begin.
-        return unless node.loc.respond_to?(:begin) && node.loc.begin
+        return unless node.loc?(:begin)
         return if part_of_ignored_node?(node)
 
         if offense?(node)

data/lib/rubocop/cop/mixin/target_ruby_version.rb

@@ -29,7 +29,7 @@ module RuboCop
         min = required_minimum_ruby_version || 0
         max = required_maximum_ruby_version || Float::INFINITY
 
-        min <= version && max >= version
+        version.between?(min, max)
       end
     end
   end

data/lib/rubocop/cop/mixin/trailing_comma.rb

@@ -4,6 +4,7 @@ module RuboCop
   module Cop
     # Common methods shared by Style/TrailingCommaInArguments,
     # Style/TrailingCommaInArrayLiteral and Style/TrailingCommaInHashLiteral
+    # rubocop:disable-next Metrics/ModuleLength
     module TrailingComma
       include ConfigurableEnforcedStyle
       include RangeHelp
@@ -57,6 +58,8 @@ module RuboCop
           ', unless each item is on its own line'
         when :consistent_comma
           ', unless items are split onto multiple lines'
+        when :diff_comma
+          ', unless that item immediately precedes a newline'
         else
           ''
         end
@@ -68,6 +71,8 @@ module RuboCop
           multiline?(node) && no_elements_on_same_line?(node)
         when :consistent_comma
           multiline?(node) && !method_name_and_arguments_on_same_line?(node)
+        when :diff_comma
+          multiline?(node) && last_item_precedes_newline?(node)
         else
           false
         end
@@ -90,19 +95,21 @@ module RuboCop
         node.multiline? && !allowed_multiline_argument?(node)
       end
 
+      # rubocop:disable-next Metrics/AbcSize
       def method_name_and_arguments_on_same_line?(node)
-        return false unless node.call_type?
+        return false if !node.call_type? || node.last_line != node.last_argument.last_line
+        return true if node.last_argument.hash_type? && node.last_argument.braces?
 
-        line = node.loc.selector.nil? ? node.loc.line : node.loc.selector.line
+        line = node.loc.selector&.line || node.loc.line
 
-        line == node.last_argument.last_line && node.last_line == node.last_argument.last_line
+        line == node.last_argument.last_line
       end
 
       # A single argument with the closing bracket on the same line as the end
       # of the argument is not considered multiline, even if the argument
       # itself might span multiple lines.
       def allowed_multiline_argument?(node)
-        elements(node).one? && !Util.begins_its_line?(node.loc.end)
+        elements(node).one? && !Util.begins_its_line?(node_end_location(node))
       end
 
       def elements(node)
@@ -122,14 +129,24 @@ module RuboCop
 
       def no_elements_on_same_line?(node)
         items = elements(node).map(&:source_range)
-        items << node.loc.end
+        items << node_end_location(node)
         items.each_cons(2).none? { |a, b| on_same_line?(a, b) }
       end
 
+      def node_end_location(node)
+        node.loc.end || node.source_range.end.adjust(begin_pos: -1)
+      end
+
       def on_same_line?(range1, range2)
         range1.last_line == range2.line
       end
 
+      def last_item_precedes_newline?(node)
+        after_last_item = node.children.last.source_range.end.join(node.source_range.end)
+
+        after_last_item.source.start_with?(/,?\s*(#.*)?\n/)
+      end
+
       def avoid_comma(kind, comma_begin_pos, extra_info)
         range = range_between(comma_begin_pos, comma_begin_pos + 1)
         article = kind.include?('array') ? 'an' : 'a'
@@ -170,7 +187,7 @@ module RuboCop
 
       def heredoc?(node)
         return false unless node.is_a?(RuboCop::AST::Node)
-        return true if node.loc.respond_to?(:heredoc_body)
+        return true if node.loc?(:heredoc_body)
 
         return heredoc_send?(node) if node.send_type?