parsanol 1.3.74 → 1.3.75

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: f6eac7cb7780cfacc7d2a1989e6cea0ee26edc13f3e2d133cefc4f363a8f3e2e
-  data.tar.gz: 6629c13a4cdbf3fea6378028ae736e49a547688443769278e9583bf508462087
+  metadata.gz: ca6a45b629235f8dc4976c046341a91d159a699f2784c564b08f73f31ed7cb49
+  data.tar.gz: f0fcbe3350657d533a298e67384eb54861fd5289a4ec049d98425d6db689012b
 SHA512:
-  metadata.gz: f5f374af54ca2375e94cd04d9771088b10b383ec046731ff57f2d36da85961d29ba6c2c6af2063e88836980a5bb0c1c4d8fbaaa98af98112e4cbf9b5da4dec39
-  data.tar.gz: 6f6f2ee120b8a45fe28a403504cd198225ae4833d4e8925758cc16160a0f64f4c2b005b622fec7bd0ad3ded616f402550edf0fa4893c0858434cbe825d0b48c3
+  metadata.gz: 720729fb5363a237a40bf29ca97c21d977c45a0b3f30c58025abc32b940edfbc2ee8bcbe791697180d9272ea21370df0b59469a984b0cc815177f0989ca2e044
+  data.tar.gz: 66a5dd4694fd3ee9d8791e2fa94b68c6103762b2cd2f43322b508eb6dbee2d7bf1cb4c345403de1eecbfa9ce50ed5f5161441f9f6f2f958a550bdf6e7c893c05

data/Cargo.lock

@@ -307,9 +307,9 @@ checksum = "9f7c3e4beb33f85d45ae3e3a1792185706c8e16d043238c593331cc7cd313b50"
 
 [[package]]
 name = "parsanol"
-version = "0.14.0"
+version = "0.15.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "25d42f7d6475885046f668bdabb6f2f8c13807148811b9d1b618fd00ec07a858"
+checksum = "01c2f6b018d28cdf676aa117f37747b227476b3f5fad5b577b51abb01609e6b5"
 dependencies = [
  "ahash",
  "getrandom 0.3.4",
@@ -326,9 +326,9 @@ dependencies = [
 
 [[package]]
 name = "parsanol-derive"
-version = "0.14.0"
+version = "0.15.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8ebddcf7d0b843657a5699f3cdeb7805bc934fcb8df24c6ab460d4dd345562d1"
+checksum = "0aff970b9bcf36471356270512814045b011f30bb93f46382529ee94b3d8e11a"
 dependencies = [
  "proc-macro2",
  "quote",

data/ext/parsanol_cdylib/Cargo.toml

@@ -12,5 +12,5 @@ crate-type = ["cdylib"]
 doc = false
 
 [dependencies]
-# 0.14.0 carries the TriviaCapture atom (native capture parity)
-parsanol = { version = "0.14.0" }
+# 0.15.0 carries the Constant and Lookbehind atoms
+parsanol = { version = "0.15.0" }

data/ext/parsanol_native/Cargo.toml

@@ -29,9 +29,9 @@ magnus = { version = "0.9" }
 
 # parsanol parser library (crates.io; 0.12 carries the rs#174 native
 # backtracking fix: dynamic-dependent memoization + outcome cache)
-# parsanol parser library (crates.io; 0.14.0 carries the TriviaCapture
-# atom: native trivia-capture parity, rs#172/#152)
-parsanol = { version = "0.14.0", features = ["ruby"] }
+# parsanol parser library (crates.io; 0.15.0 carries the Constant and
+# Lookbehind atoms: the coradoc-markdown native gap, rs#137 follow-up)
+parsanol = { version = "0.15.0", features = ["ruby"] }
 
 # Logging
 log = "0.4"

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: ruby
 authors:
 - Ribose Inc.
@@ -55,6 +55,7 @@ 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
@@ -65,6 +66,7 @@ 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