parsanol 1.3.74-arm-linux → 1.3.75-arm-linux

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 5f4fd6971980959c2416856fa2afa23b0ba0cb6ed62930a759a52493b2d0e64d
-  data.tar.gz: b6d3f736b93a5992acbb8c702a3fd1c5da0a346adc9547f204add487988de24d
+  metadata.gz: 4199b78dc8200a8a90795c96bace54501aadb60429056af97862088d582940f5
+  data.tar.gz: fa8a6a0f2517ea649f9a958ab359ffc6b631074a771f8664d4b1d02ee9c2fa9d
 SHA512:
-  metadata.gz: fc2a5ea8e093812e8cf9d705bde54cfed2c8842843bd4b5734f4489912ff366b54e90e502ea9d36fa5e5ade81f56dddeae11bffb9a577b72e9cffdd0227ae727
-  data.tar.gz: 3ec2f56416d260345e4ab2a9c9170de1b72f49decd06ec5c2e58cbac747b0f8167bc43b74db35fcc121e3203826d6c00555a32303da50194f3f490144b582e6e
+  metadata.gz: 660a66463a938caf83903dd6478fc855bccc58897ef03620e53298844424cc88ba52f2a2c0fcf11223d7ae5315a1e284f36d73dc0b8d895e6b185801dd18cec8
+  data.tar.gz: bf43ca3a6d4f8cb67251a2476e4bc89766f9c0b4bc3c39b5cf68289f4163156a9a47754857745f278c85258c8d281dab79bc14c760183c0b3b17586621d8c3dc

data/lib/parsanol/atoms/constant.rb

@@ -0,0 +1,26 @@
+# frozen_string_literal: true
+
+module Parsanol
+  module Atoms
+    # Matches empty and yields a constant value (coradoc-markdown
+    # Output parity, rs#137 follow-up): the native wire carries the
+    # value as wire data, so constant-yielding grammar shapes parse
+    # on both engines with identical trees.
+    class Constant < Base
+      attr_reader :value
+
+      def initialize(value)
+        super()
+        @value = value
+      end
+
+      def try(_source, _context, _consume_all)
+        [true, @value]
+      end
+
+      def to_s_inner(_prec)
+        "constant(#{@value.inspect})"
+      end
+    end
+  end
+end

data/lib/parsanol/atoms/lookbehind.rb

@@ -0,0 +1,32 @@
+# frozen_string_literal: true
+
+module Parsanol
+  module Atoms
+    # Inspects the byte window behind the current position: the
+    # previous +count+ bytes must equal +pattern+ (positive) or
+    # differ (negative). Consumes nothing, yields nil — the
+    # precedes?/does_not_precede? guard shape (coradoc-markdown
+    # parity, rs#137 follow-up).
+    class Lookbehind < Base
+      attr_reader :count, :pattern, :positive
+
+      def initialize(count, pattern, positive: true)
+        super()
+        @count = count
+        @pattern = pattern
+        @positive = positive
+      end
+
+      def try(source, _context, _consume_all)
+        start = source.bytepos
+        behind = start >= @count ? source.input.byteslice(start - @count, @count) : nil
+        matched = !behind.nil? && behind == @pattern.b
+        [matched == @positive, nil]
+      end
+
+      def to_s_inner(_prec)
+        "lookbehind(#{@count}, #{@pattern.inspect}, #{@positive ? '+' : '-'})"
+      end
+    end
+  end
+end

data/lib/parsanol/atoms.rb

@@ -30,6 +30,8 @@ module Parsanol
     require "parsanol/atoms/trivia"
     require "parsanol/atoms/named"
     require "parsanol/atoms/lookahead"
+    require "parsanol/atoms/constant"
+    require "parsanol/atoms/lookbehind"
     require "parsanol/atoms/cut"
     require "parsanol/atoms/alternative"
     require "parsanol/atoms/sequence"

data/lib/parsanol/native/parser.rb

@@ -191,6 +191,13 @@ module Parsanol
           when ::Parsanol::Atoms::Lookahead
             [:lookahead, atom.positive,
              atom_structure(atom.bound_parslet, visited)]
+          when ::Parsanol::Atoms::Constant
+            # The value differentiates otherwise-identical constant
+            # atoms — without it every Constant hashes alike and the
+            # grammar cache serves the first wire forever.
+            [:constant, atom.value.inspect]
+          when ::Parsanol::Atoms::Lookbehind
+            [:lookbehind, atom.count, atom.pattern, atom.positive]
           else
             [:unknown, atom.class.name]
           end

data/lib/parsanol/native/serializer.rb

@@ -18,6 +18,10 @@ module Parsanol
     def initialize
       @atoms = []
       @atom_cache = {} # object_id => atom_id for deduplication
+      # Strong retention: cache keys are object_ids; without holding the
+      # atoms, the GC can free one and a later allocation reuses its id,
+      # aliasing to a stale wire atom (the #138 recycled-id class).
+      @retained_atoms = []
     end
 
     # Main serialization method
@@ -94,6 +98,20 @@ module Parsanol
                      # reject skip-carrying artifacts loudly (unknown
                      # variant), which is the declared gate.
                      serialize_trivia(atom)
+                   when Parsanol::Atoms::Constant
+                     # rs#137 follow-up: coradoc-markdown's Output
+                     # atoms — empty match yielding a wire constant.
+                     { "Constant" => { "value" => serialize_constant_value(atom.value) } }
+                   when Parsanol::Atoms::Lookbehind
+                     # rs#137 follow-up: the precedes?/does_not_precede?
+                     # guards — inspect the bytes behind the position.
+                     {
+                       "Lookbehind" => {
+                         "count" => atom.count,
+                         "pattern" => atom.pattern,
+                         "positive" => atom.positive,
+                       },
+                     }
                    when Parsanol::Atoms::Ignored
                      serialize_ignored(atom)
                    else
@@ -104,6 +122,7 @@ module Parsanol
       # Now reserve an atom_id and cache
       atom_id = @atoms.size
       @atom_cache[cache_key] = atom_id
+      @retained_atoms << atom
       @atoms << serialized
 
       atom_id
@@ -179,6 +198,7 @@ module Parsanol
       # This prevents infinite recursion when a rule references itself
       atom_id = @atoms.size
       @atom_cache[cache_key] = atom_id
+      @retained_atoms << atom
 
       # Add a placeholder that will be replaced
       @atoms << nil
@@ -292,6 +312,22 @@ module Parsanol
             "parse this grammar with mode: :ruby"
     end
 
+    def serialize_constant_value(value)
+      case value
+      when nil then "Nil" # unit variant serializes as the bare tag
+      when true, false then { "Bool" => value }
+      when Integer then { "Int" => value }
+      when String then { "Str" => value }
+      when Array then { "Array" => value.map { |v| serialize_constant_value(v) } }
+      when Hash
+        { "Hash" => value.map { |k, v| [k.to_s, serialize_constant_value(v)] } }
+      else
+        raise UnsupportedGrammar,
+              "constant atoms carry nil/bool/int/string/array/hash values " \
+              "(got #{value.class}); restructure the grammar for native parsing"
+      end
+    end
+
     def serialize_trivia_capture(atom)
       inner_id = serialize_atom(atom.wrapped_atom)
       {

data/lib/parsanol/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Parsanol
-  VERSION = "1.3.74"
+  VERSION = "1.3.75"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: parsanol
 version: !ruby/object:Gem::Version
-  version: 1.3.74
+  version: 1.3.75
 platform: arm-linux
 authors:
 - Ribose Inc.
@@ -37,6 +37,7 @@ files:
 - lib/parsanol/atoms/base.rb
 - lib/parsanol/atoms/can_flatten.rb
 - lib/parsanol/atoms/capture.rb
+- lib/parsanol/atoms/constant.rb
 - lib/parsanol/atoms/context.rb
 - lib/parsanol/atoms/context_optimized.rb
 - lib/parsanol/atoms/custom.rb
@@ -47,6 +48,7 @@ files:
 - lib/parsanol/atoms/ignored.rb
 - lib/parsanol/atoms/infix.rb
 - lib/parsanol/atoms/lookahead.rb
+- lib/parsanol/atoms/lookbehind.rb
 - lib/parsanol/atoms/named.rb
 - lib/parsanol/atoms/re.rb
 - lib/parsanol/atoms/repetition.rb