lrama 0.5.3 → 0.5.4

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 32eaf486545ca4143a227a34cee2bfdb94928e002a85fcb2e098dba3dedb71c7
-  data.tar.gz: b80a7402d9f6caf9f0e8b5396e4bd6b80d84487a35c8cd4b4d711e51ffb56601
+  metadata.gz: f2e675399217ba6b1c8cc57aaa36bbf863ed8b22dc3e22777c88d0d0aaf1cb26
+  data.tar.gz: 940b0da60c6b25edb1c10ed80539b17617e033d7ccc3a5c9c6036959c356ae37
 SHA512:
-  metadata.gz: 1d118074af8984d85e27485a0c9fb2f8c7eab9a55b771ea3eff17e26d84fe56316404f6316ba5a9c83e61591c83ff9048c6625a904529aa686417f6f295499d9
-  data.tar.gz: e3591cc5f70e6fe8a04af5070a7b5d6ee60e98718c4c6dc6db1fbef2b3be9f5239155da91bb51ba10611b5017925f1bf71aa1259fd5550b5af6d0ba458e2ab96
+  metadata.gz: 26eb48911cf5ba12b3087390bb7f7053165ff092a8919f7969e31b3e784a0085e876140d419cf4b463426cbfc4ae5f3625504a6c1433f9f9842ee9c32490082f
+  data.tar.gz: 8bce42121ffb5d45076dc3a1664e61209fec0f1a06a08d50d3a4279eb5158911b35165e21f8de0ac87c19e7cab1d2b7b0cac16bcaba0606ec61ea3b1c6195e0b

data/.github/workflows/test.yaml

@@ -22,6 +22,15 @@ jobs:
           bundler-cache: true
       - run: bundle install
       - run: bundle exec rspec
+  check-misc:
+    runs-on: ubuntu-20.04
+    steps:
+      - uses: actions/checkout@v3
+      # Copy from https://github.com/ruby/ruby/blob/089227e94823542acfdafa68541d330eee42ffea/.github/workflows/check_misc.yml#L27
+      - name: Check for trailing spaces
+        run: |
+          git grep -I -n '[	 ]$' -- '*.rb' '*.[chy]' '*.rs' && exit 1 || :
+          git grep -n '^[	 ][	 ]*$' -- '*.md' && exit 1 || :
   steep-check:
     runs-on: ubuntu-20.04
     strategy:
@@ -58,7 +67,7 @@ jobs:
       - run: mkdir -p tool/lrama
         working-directory: ../ruby
       - name: Copy Lrama to ruby/tool
-        run: cp -r exe lib template ../ruby/tool/lrama
+        run: cp -r LEGAL.md MIT exe lib template ../ruby/tool/lrama
         working-directory:
       - run: tree tool/lrama
         working-directory: ../ruby

data/README.md

@@ -49,10 +49,20 @@ Enter the formula:
 
 ## Versions and Branches
 
+### v0_5 (`master` branch)
+
+This branch is for Ruby 3.3. `lrama_0_5` branch is created from this branch, once Ruby 3.3 is released.
+
 ### v0_4 (`lrama_0_4` branch)
 
 This branch generates "parse.c" compatible with Bison 3.8.2 for ruby 3.0, 3.1, 3.2. The first version migrated to ruby is ["0.4.0"](https://github.com/ruby/ruby/pull/7798) therefore keep this branch for Bison compatible branch.
 
+## Supported Ruby version
+
+Lrama is executed with BASERUBY when building ruby from source code. Therefore Lrama needs to support BASERUBY, currently 2.5, or later version.
+
+This also requires Lrama to be able to run with only default gems and bundled gems.
+
 ## Build Ruby
 
 1. Install Lrama
@@ -62,7 +72,7 @@ This branch generates "parse.c" compatible with Bison 3.8.2 for ruby 3.0, 3.1, 3
 
 1. Update `Lrama::VERSION`
 2. Release as a gem by `rake release`
-3. Update Lrama in ruby/ruby by `cp -r LEGAL.md MIT exe lib ruby/tool/lrama`
+3. Update Lrama in ruby/ruby by `cp -r LEGAL.md MIT exe lib template ruby/tool/lrama`
 4. Create new release on [GitHub](https://github.com/ruby/lrama/releases)
 
 ## License

data/doc/TODO.md

@@ -44,10 +44,14 @@
 * Reporting
   * [ ] Bison style
     * [ ] Wrap not selected reduce with "[]". See basic.output file generated by Bison.
+* Counterexamples
+  * [x] Nonunifying Counterexamples
+  * [ ] Unifying Counterexamples
+  * [ ] Performance improvement using reverse_transitions and reverse_productions
 * Error Tolerance
   * [x] Corchuelo et al. algorithm with N = 1 (this means the next token when error is raised)
   * [x] Add new decl for error token semantic value initialization (%error-token)
-  * [ ] Use YYMALLOC & YYFREE
+  * [x] Use YYMALLOC & YYFREE
 * Lex state
 * CI
   * [x] Setup CI

data/lib/lrama/command.rb

@@ -67,7 +67,7 @@ module Lrama
       bison_list = %w[states itemsets lookaheads solved counterexamples cex all none]
       others = %w[verbose]
       list = bison_list + others
-      not_supported = %w[counterexamples cex none]
+      not_supported = %w[cex none]
       h = { grammar: true }
 
       report.each do |r|
@@ -121,13 +121,13 @@ module Lrama
       # Output Files:
       opt.on('-h', '--header=[FILE]') {|v| @header = true; @header_file = v }
       opt.on('-d') { @header = true }
-      opt.on('-r', '--report=THINGS') {|v| @report = v.split(',') }
+      opt.on('-r', '--report=THINGS', Array) {|v| @report = v }
       opt.on('--report-file=FILE')    {|v| @report_file = v }
       opt.on('-v') {  } # Do nothing
       opt.on('-o', '--output=FILE')   {|v| @outfile = v }
 
       # Hidden
-      opt.on('--trace=THINGS') {|v| @trace = v.split(',') }
+      opt.on('--trace=THINGS', Array) {|v| @trace = v }
 
       # Error Recovery
       opt.on('-e') {|v| @error_recovery = true }

data/lib/lrama/counterexamples/derivation.rb

@@ -0,0 +1,63 @@
+module Lrama
+  class Counterexamples
+    class Derivation
+      attr_reader :item, :left, :right
+      attr_writer :right
+
+      def initialize(item, left, right = nil)
+        @item = item
+        @left = left
+        @right = right
+      end
+
+      def to_s
+        "#<Derivation(#{item.display_name})>"
+      end
+      alias :inspect :to_s
+
+      def render_strings_for_report
+        result = []
+        _render_for_report(self, 0, result, 0)
+        result.map(&:rstrip)
+      end
+
+      def render_for_report
+        render_strings_for_report.join("\n")
+      end
+
+      private
+
+      def _render_for_report(derivation, offset, strings, index)
+        item = derivation.item
+        if strings[index]
+          strings[index] << " " * (offset - strings[index].length)
+        else
+          strings[index] = " " * offset
+        end
+        str = strings[index]
+        str << "#{item.rule_id}: #{item.symbols_before_dot.map(&:display_name).join(" ")} "
+
+        if derivation.left
+          len = str.length
+          str << "#{item.next_sym.display_name}"
+          length = _render_for_report(derivation.left, len, strings, index + 1)
+          # I want String#ljust!
+          str << " " * (length - str.length)
+        else
+          str << " • #{item.symbols_after_dot.map(&:display_name).join(" ")} "
+          return str.length
+        end
+
+        if derivation.right&.left
+          length = _render_for_report(derivation.right.left, str.length, strings, index + 1)
+          str << "#{item.symbols_after_dot[1..-1].map(&:display_name).join(" ")} "
+          str << " " * (length - str.length) if length > str.length
+        elsif item.next_next_sym
+          str << "#{item.symbols_after_dot[1..-1].map(&:display_name).join(" ")} "
+        end
+
+        return str.length
+      end
+    end
+  end
+end

data/lib/lrama/counterexamples/example.rb

@@ -0,0 +1,124 @@
+module Lrama
+  class Counterexamples
+    class Example
+      attr_reader :path1, :path2, :conflict, :conflict_symbol
+
+      # path1 is shift conflict when S/R conflict
+      # path2 is always reduce conflict
+      def initialize(path1, path2, conflict, conflict_symbol, counterexamples)
+        @path1 = path1
+        @path2 = path2
+        @conflict = conflict
+        @conflict_symbol = conflict_symbol
+        @counterexamples = counterexamples
+      end
+
+      def type
+        @conflict.type
+      end
+
+      def path1_item
+        @path1.last.to.item
+      end
+
+      def path2_item
+        @path2.last.to.item
+      end
+
+      def derivations1
+        @derivations1 ||= _derivations(path1)
+      end
+
+      def derivations2
+        @derivations2 ||= _derivations(path2)
+      end
+
+      private
+
+      def _derivations(paths)
+        derivation = nil
+        current = :production
+        lookahead_sym = paths.last.to.item.end_of_rule? ? @conflict_symbol : nil
+
+        paths.reverse.each do |path|
+          item = path.to.item
+
+          case current
+          when :production
+            case path
+            when StartPath
+              derivation = Derivation.new(item, derivation)
+              current = :start
+            when TransitionPath
+              derivation = Derivation.new(item, derivation)
+              current = :transition
+            when ProductionPath
+              derivation = Derivation.new(item, derivation)
+              current = :production
+            end
+
+            if lookahead_sym && item.next_next_sym && item.next_next_sym.first_set.include?(lookahead_sym)
+              state_item = @counterexamples.transitions[[path.to, item.next_sym]]
+              derivation2 = find_derivation_for_symbol(state_item, lookahead_sym)
+              derivation.right = derivation2
+              lookahead_sym = nil
+            end
+
+          when :transition
+            case path
+            when StartPath
+              derivation = Derivation.new(item, derivation)
+              current = :start
+            when TransitionPath
+              # ignore
+              current = :transition
+            when ProductionPath
+              # ignore
+              current = :production
+            end
+          else
+            raise "BUG: Unknown #{current}"
+          end
+
+          break if current == :start
+        end
+
+        derivation
+      end
+
+      def find_derivation_for_symbol(state_item, sym)
+        queue = []
+        queue << [state_item]
+
+        while (sis = queue.shift)
+          si = sis.last
+          next_sym = si.item.next_sym
+
+          if next_sym == sym
+            derivation = nil
+
+            sis.reverse.each do |si|
+              derivation = Derivation.new(si.item, derivation)
+            end
+
+            return derivation
+          end
+
+          if next_sym.nterm? && next_sym.first_set.include?(sym)
+            @counterexamples.productions[si].each do |next_item|
+              next if next_item.empty_rule?
+              next_si = StateItem.new(si.state, next_item)
+              next if sis.include?(next_si)
+              queue << (sis + [next_si])
+            end
+
+            if next_sym.nullable
+              next_si = @counterexamples.transitions[[si, next_sym]]
+              queue << (sis + [next_si])
+            end
+          end
+        end
+      end
+    end
+  end
+end

data/lib/lrama/counterexamples/path.rb

@@ -0,0 +1,69 @@
+module Lrama
+  class Counterexamples
+    class Path
+      def initialize(from_state_item, to_state_item)
+        @from_state_item = from_state_item
+        @to_state_item = to_state_item
+      end
+
+      def from
+        @from_state_item
+      end
+
+      def to
+        @to_state_item
+      end
+
+      def to_s
+        "#<Path(#{type})>"
+      end
+      alias :inspect :to_s
+    end
+
+    class StartPath < Path
+      def initialize(to_state_item)
+        super nil, to_state_item
+      end
+
+      def type
+        :start
+      end
+
+      def transition?
+        false
+      end
+
+      def production?
+        false
+      end
+    end
+
+    class TransitionPath < Path
+      def type
+        :transition
+      end
+
+      def transition?
+        true
+      end
+
+      def production?
+        false
+      end
+    end
+
+    class ProductionPath < Path
+      def type
+        :production
+      end
+
+      def transition?
+        false
+      end
+
+      def production?
+        true
+      end
+    end
+  end
+end

data/lib/lrama/counterexamples/state_item.rb

@@ -0,0 +1,6 @@
+module Lrama
+  class Counterexamples
+    class StateItem < Struct.new(:state, :item)
+    end
+  end
+end

data/lib/lrama/counterexamples/triple.rb

@@ -0,0 +1,21 @@
+module Lrama
+  class Counterexamples
+    # s: state
+    # itm: item within s
+    # l: precise lookahead set
+    class Triple < Struct.new(:s, :itm, :l)
+      alias :state :s
+      alias :item :itm
+      alias :precise_lookahead_set :l
+
+      def state_item
+        StateItem.new(state, item)
+      end
+
+      def inspect
+        "#{state.inspect}. #{item.display_name}. #{l.map(&:id).map(&:s_value)}"
+      end
+      alias :to_s :inspect
+    end
+  end
+end

data/lib/lrama/counterexamples.rb

@@ -0,0 +1,285 @@
+require "set"
+
+require "lrama/counterexamples/derivation"
+require "lrama/counterexamples/example"
+require "lrama/counterexamples/path"
+require "lrama/counterexamples/state_item"
+require "lrama/counterexamples/triple"
+
+module Lrama
+  # See: https://www.cs.cornell.edu/andru/papers/cupex/cupex.pdf
+  #      4. Constructing Nonunifying Counterexamples
+  class Counterexamples
+    attr_reader :transitions, :productions
+
+    def initialize(states)
+      @states = states
+      setup_transitions
+      setup_productions
+    end
+
+    def to_s
+      "#<Counterexamples>"
+    end
+    alias :inspect :to_s
+
+    def compute(conflict_state)
+      conflict_state.conflicts.flat_map do |conflict|
+        case conflict.type
+        when :shift_reduce
+          shift_reduce_example(conflict_state, conflict)
+        when :reduce_reduce
+          reduce_reduce_examples(conflict_state, conflict)
+        end
+      end.compact
+    end
+
+    private
+
+    def setup_transitions
+      # Hash [StateItem, Symbol] => StateItem
+      @transitions = {}
+      # Hash [StateItem, Symbol] => Set(StateItem)
+      @reverse_transitions = {}
+
+      @states.states.each do |src_state|
+        trans = {}
+
+        src_state.transitions.each do |shift, next_state|
+          trans[shift.next_sym] = next_state
+        end
+
+        src_state.items.each do |src_item|
+          next if src_item.end_of_rule?
+          sym = src_item.next_sym
+          dest_state = trans[sym]
+
+          dest_state.kernels.each do |dest_item|
+            next unless (src_item.rule == dest_item.rule) && (src_item.position + 1 == dest_item.position)
+            src_state_item = StateItem.new(src_state, src_item)
+            dest_state_item = StateItem.new(dest_state, dest_item)
+
+            @transitions[[src_state_item, sym]] = dest_state_item
+
+            key = [dest_state_item, sym]
+            @reverse_transitions[key] ||= Set.new
+            @reverse_transitions[key] << src_state_item
+          end
+        end
+      end
+    end
+
+    def setup_productions
+      # Hash [StateItem] => Set(Item)
+      @productions = {}
+      # Hash [State, Symbol] => Set(Item). Symbol is nterm
+      @reverse_productions = {}
+
+      @states.states.each do |state|
+        # LHS => Set(Item)
+        h = {}
+
+        state.closure.each do |item|
+          sym = item.lhs
+
+          h[sym] ||= Set.new
+          h[sym] << item
+        end
+
+        state.items.each do |item|
+          next if item.end_of_rule?
+          next if item.next_sym.term?
+
+          sym = item.next_sym
+          state_item = StateItem.new(state, item)
+          key = [state, sym]
+
+          @productions[state_item] = h[sym]
+
+          @reverse_productions[key] ||= Set.new
+          @reverse_productions[key] << item
+        end
+      end
+    end
+
+    def shift_reduce_example(conflict_state, conflict)
+      conflict_symbol = conflict.symbols.first
+      shift_conflict_item = conflict_state.items.find { |item| item.next_sym == conflict_symbol }
+      path2 = shortest_path(conflict_state, conflict.reduce.item, conflict_symbol)
+      path1 = find_shift_conflict_shortest_path(path2, conflict_state, shift_conflict_item)
+
+      Example.new(path1, path2, conflict, conflict_symbol, self)
+    end
+
+    def reduce_reduce_examples(conflict_state, conflict)
+      conflict_symbol = conflict.symbols.first
+      path1 = shortest_path(conflict_state, conflict.reduce1.item, conflict_symbol)
+      path2 = shortest_path(conflict_state, conflict.reduce2.item, conflict_symbol)
+
+      Example.new(path1, path2, conflict, conflict_symbol, self)
+    end
+
+    def find_shift_conflict_shortest_path(reduce_path, conflict_state, conflict_item)
+      state_items = find_shift_conflict_shortest_state_items(reduce_path, conflict_state, conflict_item)
+      build_paths_from_state_items(state_items)
+    end
+
+    def find_shift_conflict_shortest_state_items(reduce_path, conflict_state, conflict_item)
+      target_state_item = StateItem.new(conflict_state, conflict_item)
+      result = [target_state_item]
+      reversed_reduce_path = reduce_path.to_a.reverse
+      # Index for state_item
+      i = 0
+
+      while (path = reversed_reduce_path[i])
+        # Index for prev_state_item
+        j = i + 1
+        _j = j
+
+        while (prev_path = reversed_reduce_path[j])
+          if prev_path.production?
+            j += 1
+          else
+            break
+          end
+        end
+
+        state_item = path.to
+        prev_state_item = prev_path&.to
+
+        if target_state_item == state_item || target_state_item.item.start_item?
+          result.concat(reversed_reduce_path[_j..-1].map(&:to))
+          break
+        end
+
+        if target_state_item.item.beginning_of_rule?
+          queue = []
+          queue << [target_state_item]
+
+          # Find reverse production
+          while (sis = queue.shift)
+            si = sis.last
+
+            # Reach to start state
+            if si.item.start_item?
+              sis.shift
+              result.concat(sis)
+              target_state_item = si
+              break
+            end
+
+            if !si.item.beginning_of_rule?
+              key = [si, si.item.previous_sym]
+              @reverse_transitions[key].each do |prev_target_state_item|
+                next if prev_target_state_item.state != prev_state_item.state
+                sis.shift
+                result.concat(sis)
+                result << prev_target_state_item
+                target_state_item = prev_target_state_item
+                i = j
+                queue.clear
+                break
+              end
+            else
+              key = [si.state, si.item.lhs]
+              @reverse_productions[key].each do |item|
+                state_item = StateItem.new(si.state, item)
+                queue << (sis + [state_item])
+              end
+            end
+          end
+        else
+          # Find reverse transition
+          key = [target_state_item, target_state_item.item.previous_sym]
+          @reverse_transitions[key].each do |prev_target_state_item|
+            next if prev_target_state_item.state != prev_state_item.state
+            result << prev_target_state_item
+            target_state_item = prev_target_state_item
+            i = j
+            break
+          end
+        end
+      end
+
+      result.reverse
+    end
+
+    def build_paths_from_state_items(state_items)
+      paths = state_items.zip([nil] + state_items).map do |si, prev_si|
+        case
+        when prev_si.nil?
+          StartPath.new(si)
+        when si.item.beginning_of_rule?
+          ProductionPath.new(prev_si, si)
+        else
+          TransitionPath.new(prev_si, si)
+        end
+      end
+
+      paths
+    end
+
+    def shortest_path(conflict_state, conflict_reduce_item, conflict_term)
+      # queue: is an array of [Triple, [Path]]
+      queue = []
+      visited = {}
+      start_state = @states.states.first
+      raise "BUG: Start state should be just one kernel." if start_state.kernels.count != 1
+
+      start = Triple.new(start_state, start_state.kernels.first, Set.new([@states.eof_symbol]))
+
+      queue << [start, [StartPath.new(start.state_item)]]
+
+      while true
+        triple, paths = queue.shift
+
+        next if visited[triple]
+        visited[triple] = true
+
+        # Found
+        if triple.state == conflict_state && triple.item == conflict_reduce_item && triple.l.include?(conflict_term)
+          return paths
+        end
+
+        # transition
+        triple.state.transitions.each do |shift, next_state|
+          next unless triple.item.next_sym && triple.item.next_sym == shift.next_sym
+          next_state.kernels.each do |kernel|
+            next if kernel.rule != triple.item.rule
+            t = Triple.new(next_state, kernel, triple.l)
+            queue << [t, paths + [TransitionPath.new(triple.state_item, t.state_item)]]
+          end
+        end
+
+        # production step
+        triple.state.closure.each do |item|
+          next unless triple.item.next_sym && triple.item.next_sym == item.lhs
+          l = follow_l(triple.item, triple.l)
+          t = Triple.new(triple.state, item, l)
+          queue << [t, paths + [ProductionPath.new(triple.state_item, t.state_item)]]
+        end
+
+        break if queue.empty?
+      end
+
+      return nil
+    end
+
+    def follow_l(item, current_l)
+      # 1. follow_L (A -> X1 ... Xn-1 • Xn) = L
+      # 2. follow_L (A -> X1 ... Xk • Xk+1 Xk+2 ... Xn) = {Xk+2} if Xk+2 is a terminal
+      # 3. follow_L (A -> X1 ... Xk • Xk+1 Xk+2 ... Xn) = FIRST(Xk+2) if Xk+2 is a nonnullable nonterminal
+      # 4. follow_L (A -> X1 ... Xk • Xk+1 Xk+2 ... Xn) = FIRST(Xk+2) + follow_L (A -> X1 ... Xk+1 • Xk+2 ... Xn) if Xk+2 is a nullable nonterminal
+      case
+      when item.number_of_rest_symbols == 1
+        current_l
+      when item.next_next_sym.term?
+        Set.new([item.next_next_sym])
+      when !item.next_next_sym.nullable
+        item.next_next_sym.first_set
+      else
+        item.next_next_sym.first_set + follow_l(item.new_by_next_position, current_l)
+      end
+    end
+  end
+end

data/lib/lrama/digraph.rb

@@ -33,7 +33,7 @@ module Lrama
       @h[x] = d
       @result[x] = @base_function[x] # F x = F' x
 
-      @relation[x] && @relation[x].each do |y|
+      @relation[x]&.each do |y|
         traverse(y) if @h[y] == 0
         @h[x] = [@h[x], @h[y]].min
         @result[x] |= @result[y] # F x = F x + F y
@@ -43,9 +43,8 @@ module Lrama
         while true do
           z = @stack.pop
           @h[z] = Float::INFINITY
-          @result[z] = @result[x] # F (Top of S) = F x
-
           break if z == x
+          @result[z] = @result[x] # F (Top of S) = F x
         end
       end
     end