Use ftype & give windows a glob_gitignore escape

Author: robotdana

.github/workflows/ci.yml

@@ -7,7 +7,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        ruby: [2.7, '3.0', 3.1, 3.2, 3.3, ruby-head, jruby-9.4, jruby-head]
+        ruby: [2.7, '3.0', 3.1, 3.2, ruby-head, jruby-9.4, jruby-head]
         platform: [ubuntu, windows, macos]
     continue-on-error: ${{ endsWith(matrix.ruby, 'head') }}
     runs-on: ${{matrix.platform}}-latest

.spellr_wordlists/english.txt

@@ -137,6 +137,7 @@ tsx
 ttributes
 txt
 unanchorable
+unc
 unexpandable
 unfuck
 unnegated

bin/benchmark

@@ -268,6 +268,16 @@ benchmark('add-trailing-slash') do
     @string_trailing_slash = 'D:/'
     @string_no_trailing_slash = '/bin'
 
+    raise unless "/str" == (@string_slash.end_with?('/') ? @string_slash : "#{@string_slash}/") + "str"
+    raise unless "D:/str" == (@string_trailing_slash.end_with?('/') ? @string_trailing_slash : "#{@string_trailing_slash}/") + "str"
+    raise unless "/bin/str" == (@string_no_trailing_slash.end_with?('/') ? @string_no_trailing_slash : "#{@string_no_trailing_slash}/") + "str"
+    raise unless "/str" == ::File.join(@string_slash, "str")
+    raise unless "D:/str" == ::File.join(@string_trailing_slash, "str")
+    raise unless "/bin/str" == ::File.join(@string_no_trailing_slash, "str")
+    raise unless "/str" == @string_slash.sub(/(?<!\/)\z/, '/') + "str"
+    raise unless "D:/str" == @string_trailing_slash.sub(/(?<!\/)\z/, '/') + "str"
+    raise unless "/bin/str" == @string_no_trailing_slash.sub(/(?<!\/)\z/, '/') + "str"
+
     x.report(:string_slash_end_with) { (@string_slash.end_with?('/') ? @string_slash : "#{@string_slash}/") + "str" }
     x.report(:string_trailing_slash_end_with) { (@string_trailing_slash.end_with?('/') ? @string_trailing_slash : "#{@string_trailing_slash}/") + "str" }
     x.report(:string_no_trailing_slash_end_with) { (@string_no_trailing_slash.end_with?('/') ? @string_no_trailing_slash : "#{@string_no_trailing_slash}/") + "str" }
@@ -276,6 +286,10 @@ benchmark('add-trailing-slash') do
     x.report(:string_trailing_slash_file_join) { ::File.join(@string_trailing_slash, "str") }
     x.report(:string_no_trailing_slash_file_join) { ::File.join(@string_no_trailing_slash, "str") }
 
+    x.report(:string_slash_sub) { @string_slash.sub(/(?<!\/)\z/, '/') + "str" }
+    x.report(:string_trailing_slash_sub) { @string_trailing_slash.sub(/(?<!\/)\z/, '/') + "str" }
+    x.report(:string_no_trailing_slash_sub) { @string_no_trailing_slash.sub(/(?<!\/)\z/, '/') + "str" }
+
     x.compare!
   end
 end

lib/path_list/candidate.rb

@@ -18,6 +18,7 @@ def initialize(full_path, directory = nil, shebang = nil)
 
       @child_candidates = nil
       @children = nil
+      @ftype = nil
     end
 
     # @return [String] full path downcased
@@ -54,33 +55,16 @@ def children
       end
     end
 
-    # :nocov:
-    if ::RUBY_PLATFORM == 'java' && ::RbConfig::CONFIG['host_os'].match?(/mswin|mingw/)
-      # @return [Boolean] whether this path is a directory (false for symlinks to directories)
-      def directory?
-        return @directory unless @directory.nil?
-
-        @directory = if ::File.symlink?(@full_path)
-          warn 'Symlink lstat'
-          warn lstat.inspect
-          false
-        else
-          lstat&.directory? || false
-        end
-      end
-      # :nocov:
-    else
-      # @return [Boolean] whether this path is a directory (false for symlinks to directories)
-      def directory?
-        return @directory unless @directory.nil?
+    # @return [Boolean] whether this path is a directory (false for symlinks to directories)
+    def directory?
+      return @directory unless @directory.nil?
 
-        @directory = lstat&.directory? || false
-      end
+      @directory = ftype == 'directory'
     end
 
     # @return [Boolean] whether this path exists
     def exists?
-      lstat ? true : false
+      ftype != 'error'
     end
 
     alias_method :original_inspect, :inspect # leftovers:keep
@@ -112,7 +96,7 @@ def shebang
           ''
         end
       rescue ::IOError, ::SystemCallError
-        @lstat ||= nil
+        @ftype ||= 'error'
         ''
       ensure
         file&.close
@@ -121,12 +105,29 @@ def shebang
 
     private
 
-    def lstat
-      return @lstat if defined?(@lstat)
+    # :nocov:
+    # https://github.com/jruby/jruby/issues/8018
+    # ftype follows symlinks on jruby on windows.
+    if ::RUBY_PLATFORM == 'java' && ::RbConfig::CONFIG['host_os'].match?(/mswin|mingw/)
+      refine ::File do
+        # :nodoc:
+        def ftype(path)
+          if ::File.symlink?(path)
+            'link'
+          else
+            super
+          end
+        end
+      end
+    end
+    # :nocov:
+
+    def ftype
+      return @ftype if @ftype
 
-      @lstat = ::File.lstat(@full_path)
+      @ftype = ::File.ftype(@full_path)
     rescue ::SystemCallError
-      @lstat = nil
+      @ftype = 'error'
     end
   end
 end

lib/path_list/pattern_parser/glob_gitignore.rb

@@ -20,10 +20,13 @@ class PatternParser
     # - Patterns containing with `/../` are resolved relative to the `root:` directory
     # - Patterns beginning with `*` (or `!*`) will match any descendant of the `root:` directory
     # - Other patterns match children (not descendants) of the `root:` directory
-    # - Additionally, on windows
-    #   - \ is treated as a path separator, not an escape character.
-    #     There is no cross-platform escape character when using :glob_gitignore format.
-    #   - Patterns beginning with `c:/`, `d:\`, or `!c:/`, or etc are absolute.
+    # - Additionally, on windows:
+    #   - either / or \ (slash or backslash) can be used as path separators.
+    #   - therefore \ (backslash) isn't available to be used as an escape character
+    #   - instead ` (grave accent) is used as an escape character
+    #   - patterns beginning with `c:/`, `d:\`, or `!c:/`, or etc are absolute.
+    #   - a path beginning with / or \ is a shortcut for the current working directory drive.
+    # - there is no cross platform escape character, this is intended to match the current shell
     # @example
     #   PathList.only(ARGV, format: :glob_gitignore)
     #   PathList.only(
@@ -51,19 +54,34 @@ class GlobGitignore < Gitignore
       def initialize(pattern, polarity, root)
         pattern = +'' if pattern.start_with?('#')
         negated_sigil = '!' if pattern.delete_prefix!('!')
+        pattern = normalize_slash(pattern)
         if pattern.start_with?('*')
-          pattern = "#{negated_sigil}#{pattern.tr(::File::ALT_SEPARATOR.to_s, ::File::SEPARATOR)}"
+          pattern = "#{negated_sigil}#{pattern}"
         elsif pattern.match?(EXPANDABLE_PATH)
-          dir_only! if pattern.match?(%r{[/\\]\s*\z}) # expand_path will remove it
+          dir_only! if pattern.match?(%r{/\s*\z}) # expand_path will remove it
 
           pattern = "#{negated_sigil}#{CanonicalPath.full_path_from(pattern, root)}"
           root = nil
           @anchored = true
         else
-          pattern = "#{negated_sigil}/#{pattern.tr(::File::ALT_SEPARATOR.to_s, ::File::SEPARATOR)}"
+          pattern = "#{negated_sigil}/#{pattern}"
         end
 
-        super(pattern, polarity, root)
+        super(normalize_escape(pattern), polarity, root)
+      end
+
+      private
+
+      def normalize_slash(pattern)
+        return pattern unless ::File::ALT_SEPARATOR
+
+        pattern.tr('\\', '/')
+      end
+
+      def normalize_escape(pattern)
+        return pattern unless ::File::ALT_SEPARATOR
+
+        pattern.tr('`', '\\')
       end
     end
   end

lib/path_list/pattern_parser/glob_gitignore/expandable_path.rb

@@ -3,14 +3,28 @@
 class PathList
   class PatternParser
     class GlobGitignore
-      EXPANDABLE_PATH = if File.expand_path('/') == '/'
-        %r{(?:\A(?:[~/]|\.{1,2}(?:/|\z))|(?:[^\\]|\A)(?:\\{2})*/\.\./)}
-        # :nocov:
-      else
-        # this isn't actually nocov, but it's cov is because i reload the file
-        %r{(?:\A(?:[~/\\]|[a-zA-Z]:[/\\]|[/\\]{2}|\.{1,2}(?:[/\\]|\z))|[\\/]\.\.[/\\])}
-        # :nocov:
-      end
+      # :nocov:
+      # this isn't actually nocov, but it's cov is because i reload the file
+      EXPANDABLE_PATH = %r{(?:
+        \A(?:
+          [~/] # start with slash or tilde
+          |
+          \.{1,2}(?:/|\z) # start with dot or dot dot followed by slash or nothing
+          #{
+            if ::File.expand_path('/') != '/' # only if drive letters are applicable
+              "
+              |
+              [a-zA-Z]:/ # drive letter
+              |
+              // # UNC path
+              "
+            end
+          }
+        )
+        |
+        (?:[^\\]|\A)(?:\\{2})*/\.\./) # unescaped slash dot dot slash
+      }x.freeze
+      # :nocov:
     end
   end
 end

spec/candidate_spec.rb

@@ -25,15 +25,15 @@
   end
 
   describe '#parent' do
-    before { allow(File).to receive_messages(exist?: nil, lstat: nil, directory?: nil) }
+    before { allow(File).to receive_messages(exist?: nil, ftype: nil, directory?: nil) }
 
     it 'returns a candidate for the parent with preset directory value' do
       expect(candidate.parent).to be_like described_class.new('/path/from/root', true)
       expect(candidate.parent).to have_attributes(
         directory?: true
       )
       expect(File).not_to have_received(:directory?)
-      expect(File).not_to have_received(:lstat)
+      expect(File).not_to have_received(:ftype)
     end
 
     context 'when the path is /' do
@@ -54,35 +54,35 @@
       before { create_file_list 'foo' }
 
       it 'is memoized when true' do
-        allow(File).to receive(:lstat).and_call_original
+        allow(File).to receive(:ftype).and_call_original
 
         expect(candidate.exists?).to be true
-        expect(File).to have_received(:lstat).once
+        expect(File).to have_received(:ftype).once
         expect(candidate.exists?).to be true
-        expect(File).to have_received(:lstat).once
+        expect(File).to have_received(:ftype).once
       end
     end
 
     context 'when the file does not exist' do
       let(:full_path) { './foo' }
 
       it 'is memoized when false' do
-        allow(File).to receive(:lstat).and_call_original
+        allow(File).to receive(:ftype).and_call_original
 
         expect(candidate.exists?).to be false
-        expect(File).to have_received(:lstat).with('./foo').once
+        expect(File).to have_received(:ftype).with('./foo').once
         expect(candidate.exists?).to be false
-        expect(File).to have_received(:lstat).with('./foo').once
+        expect(File).to have_received(:ftype).with('./foo').once
       end
 
       it 'is false when there is an error' do
-        allow(File).to receive(:lstat).and_call_original
-        allow(File).to receive(:lstat).with(full_path).and_raise(Errno::EACCES)
+        allow(File).to receive(:ftype).and_call_original
+        allow(File).to receive(:ftype).with(full_path).and_raise(Errno::EACCES)
 
         expect(candidate.exists?).to be false
-        expect(File).to have_received(:lstat).with('./foo').once
+        expect(File).to have_received(:ftype).with('./foo').once
         expect(candidate.exists?).to be false
-        expect(File).to have_received(:lstat).with('./foo').once
+        expect(File).to have_received(:ftype).with('./foo').once
       end
     end
   end
@@ -139,9 +139,6 @@
       create_symlink('foo' => 'foo_target')
 
       candidate = described_class.new(File.expand_path('foo'))
-      expect(File.symlink?('./foo')).to be true
-      expect(File.stat('./foo')).to have_attributes(directory?: false, symlink?: true)
-      expect(candidate.send(:lstat)).to have_attributes(directory?: false, symlink?: true)
       expect(candidate).not_to be_directory
     end
   end

spec/path_list_spec.rb

@@ -40,8 +40,8 @@
     it 'copes with being given fs root' do
       whatever_file_we_get = subject.each('/').first
       expect(whatever_file_we_get).not_to start_with('/')
-      # use lstat because it could be a symlink to nowhere and File.exist? will be sad
-      expect(File.lstat("/#{whatever_file_we_get}")).to be_a File::Stat
+      # use symlink? because it could be a symlink to nowhere and File.exist? would return false
+      expect { File.symlink?("/#{whatever_file_we_get}") || File.exist?("/#{whatever_file_we_get}") }.not_to raise_error
     end
 
     it 'copes with being given nonsense root' do