lrama 0.5.2 → 0.5.4

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

data/lib/lrama/grammar/auxiliary.rb

@@ -0,0 +1,7 @@
+module Lrama
+  class Grammar
+    # Grammar file information not used by States but by Output
+    class Auxiliary < Struct.new(:prologue_first_lineno, :prologue, :epilogue_first_lineno, :epilogue, keyword_init: true)
+    end
+  end
+end

data/lib/lrama/grammar/code.rb

@@ -0,0 +1,123 @@
+require "forwardable"
+
+module Lrama
+  class Grammar
+    class Code < Struct.new(:type, :token_code, keyword_init: true)
+      extend Forwardable
+
+      def_delegators "token_code", :s_value, :line, :column, :references
+
+      # $$, $n, @$, @n is translated to C code
+      def translated_code
+        case type
+        when :user_code
+          translated_user_code
+        when :initial_action
+          translated_initial_action_code
+        end
+      end
+
+      # * ($1) error
+      # * ($$) *yyvaluep
+      # * (@1) error
+      # * (@$) *yylocationp
+      def translated_printer_code(tag)
+        t_code = s_value.dup
+
+        references.reverse.each do |ref|
+          first_column = ref.first_column
+          last_column = ref.last_column
+
+          case
+          when ref.value == "$" && ref.type == :dollar # $$
+            # Omit "<>"
+            member = tag.s_value[1..-2]
+            str = "((*yyvaluep).#{member})"
+          when ref.value == "$" && ref.type == :at # @$
+            str = "(*yylocationp)"
+          when ref.type == :dollar # $n
+            raise "$#{ref.value} can not be used in %printer."
+          when ref.type == :at # @n
+            raise "@#{ref.value} can not be used in %printer."
+          else
+            raise "Unexpected. #{self}, #{ref}"
+          end
+
+          t_code[first_column..last_column] = str
+        end
+
+        return t_code
+      end
+      alias :translated_error_token_code :translated_printer_code
+
+
+      private
+
+      # * ($1) yyvsp[i]
+      # * ($$) yyval
+      # * (@1) yylsp[i]
+      # * (@$) yyloc
+      def translated_user_code
+        t_code = s_value.dup
+
+        references.reverse.each do |ref|
+          first_column = ref.first_column
+          last_column = ref.last_column
+
+          case
+          when ref.value == "$" && ref.type == :dollar # $$
+            # Omit "<>"
+            member = ref.tag.s_value[1..-2]
+            str = "(yyval.#{member})"
+          when ref.value == "$" && ref.type == :at # @$
+            str = "(yyloc)"
+          when ref.type == :dollar # $n
+            i = -ref.position_in_rhs + ref.value
+            # Omit "<>"
+            member = ref.tag.s_value[1..-2]
+            str = "(yyvsp[#{i}].#{member})"
+          when ref.type == :at # @n
+            i = -ref.position_in_rhs + ref.value
+            str = "(yylsp[#{i}])"
+          else
+            raise "Unexpected. #{self}, #{ref}"
+          end
+
+          t_code[first_column..last_column] = str
+        end
+
+        return t_code
+      end
+
+      # * ($1) error
+      # * ($$) yylval
+      # * (@1) error
+      # * (@$) yylloc
+      def translated_initial_action_code
+        t_code = s_value.dup
+
+        references.reverse.each do |ref|
+          first_column = ref.first_column
+          last_column = ref.last_column
+
+          case
+          when ref.value == "$" && ref.type == :dollar # $$
+            str = "yylval"
+          when ref.value == "$" && ref.type == :at # @$
+            str = "yylloc"
+          when ref.type == :dollar # $n
+            raise "$#{ref.value} can not be used in initial_action."
+          when ref.type == :at # @n
+            raise "@#{ref.value} can not be used in initial_action."
+          else
+            raise "Unexpected. #{self}, #{ref}"
+          end
+
+          t_code[first_column..last_column] = str
+        end
+
+        return t_code
+      end
+    end
+  end
+end

data/lib/lrama/grammar/error_token.rb

@@ -0,0 +1,9 @@
+module Lrama
+  class Grammar
+    class ErrorToken < Struct.new(:ident_or_tags, :code, :lineno, keyword_init: true)
+      def translated_code(member)
+        code.translated_error_token_code(member)
+      end
+    end
+  end
+end

data/lib/lrama/grammar/precedence.rb

@@ -0,0 +1,11 @@
+module Lrama
+  class Grammar
+    class Precedence < Struct.new(:type, :precedence, keyword_init: true)
+      include Comparable
+
+      def <=>(other)
+        self.precedence <=> other.precedence
+      end
+    end
+  end
+end

data/lib/lrama/grammar/printer.rb

@@ -0,0 +1,9 @@
+module Lrama
+  class Grammar
+    class Printer < Struct.new(:ident_or_tags, :code, :lineno, keyword_init: true)
+      def translated_code(member)
+        code.translated_printer_code(member)
+      end
+    end
+  end
+end

data/lib/lrama/grammar/reference.rb

@@ -0,0 +1,22 @@
+# type: :dollar or :at
+# ex_tag: "$<tag>1" (Optional)
+
+module Lrama
+  class Grammar
+    class Reference < Struct.new(:type, :value, :ex_tag, :first_column, :last_column, :referring_symbol, :position_in_rhs, keyword_init: true)
+      def tag
+        if ex_tag
+          ex_tag
+        else
+          # FIXME: Remove this class check
+          if referring_symbol.is_a?(Symbol)
+            referring_symbol.tag
+          else
+            # Lrama::Lexer::Token (User_code) case
+            nil
+          end
+        end
+      end
+    end
+  end
+end

data/lib/lrama/grammar/rule.rb

@@ -0,0 +1,39 @@
+module Lrama
+  class Grammar
+    class Rule < Struct.new(:id, :lhs, :rhs, :code, :nullable, :precedence_sym, :lineno, keyword_init: true)
+      # TODO: Change this to display_name
+      def to_s
+        l = lhs.id.s_value
+        r = rhs.empty? ? "ε" : rhs.map {|r| r.id.s_value }.join(", ")
+
+        "#{l} -> #{r}"
+      end
+
+      # Used by #user_actions
+      def as_comment
+        l = lhs.id.s_value
+        r = rhs.empty? ? "%empty" : rhs.map(&:display_name).join(" ")
+
+        "#{l}: #{r}"
+      end
+
+      # opt_nl: ε     <-- empty_rule
+      #       | '\n'  <-- not empty_rule
+      def empty_rule?
+        rhs.empty?
+      end
+
+      def precedence
+        precedence_sym&.precedence
+      end
+
+      def initial_rule?
+        id == 0
+      end
+
+      def translated_code
+        code&.translated_code
+      end
+    end
+  end
+end

data/lib/lrama/grammar/symbol.rb

@@ -0,0 +1,87 @@
+# Symbol is both of nterm and term
+# `number` is both for nterm and term
+# `token_id` is tokentype for term, internal sequence number for nterm
+#
+# TODO: Add validation for ASCII code range for Token::Char
+
+module Lrama
+  class Grammar
+    class Symbol < Struct.new(:id, :alias_name, :number, :tag, :term, :token_id, :nullable, :precedence, :printer, :error_token, keyword_init: true)
+      attr_accessor :first_set, :first_set_bitmap
+      attr_writer :eof_symbol, :error_symbol, :undef_symbol, :accept_symbol
+
+      def term?
+        term
+      end
+
+      def nterm?
+        !term
+      end
+
+      def eof_symbol?
+        !!@eof_symbol
+      end
+
+      def error_symbol?
+        !!@error_symbol
+      end
+
+      def undef_symbol?
+        !!@undef_symbol
+      end
+
+      def accept_symbol?
+        !!@accept_symbol
+      end
+
+      def display_name
+        alias_name || id.s_value
+      end
+
+      # name for yysymbol_kind_t
+      #
+      # See: b4_symbol_kind_base
+      def enum_name
+        case
+        when accept_symbol?
+          name = "YYACCEPT"
+        when eof_symbol?
+          name = "YYEOF"
+        when term? && id.type == Token::Char
+          name = number.to_s + display_name
+        when term? && id.type == Token::Ident
+          name = id.s_value
+        when nterm? && (id.s_value.include?("$") || id.s_value.include?("@"))
+          name = number.to_s + id.s_value
+        when nterm?
+          name = id.s_value
+        else
+          raise "Unexpected #{self}"
+        end
+
+        "YYSYMBOL_" + name.gsub(/\W+/, "_")
+      end
+
+      # comment for yysymbol_kind_t
+      def comment
+        case
+        when accept_symbol?
+          # YYSYMBOL_YYACCEPT
+          id.s_value
+        when eof_symbol?
+          # YYEOF
+          alias_name
+        when (term? && 0 < token_id && token_id < 128)
+          # YYSYMBOL_3_backslash_, YYSYMBOL_14_
+          alias_name || id.s_value
+        when id.s_value.include?("$") || id.s_value.include?("@")
+          # YYSYMBOL_21_1
+          id.s_value
+        else
+          # YYSYMBOL_keyword_class, YYSYMBOL_strings_1
+          alias_name || id.s_value
+        end
+      end
+    end
+  end
+end

data/lib/lrama/grammar/union.rb

@@ -0,0 +1,10 @@
+module Lrama
+  class Grammar
+    class Union < Struct.new(:code, :lineno, keyword_init: true)
+      def braces_less_code
+        # Remove braces
+        code.s_value[1..-2]
+      end
+    end
+  end
+end