parsanol 1.3.76-aarch64-linux → 1.3.77-aarch64-linux

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 0defff0d1dc2a81a83d59ded0f844998e86fc389e5cfd9cc7f1f10839224482b
-  data.tar.gz: 3a5783cd54d979027a4d8efd763c90c23629fe9c6479bdee95f9ab4453aafdc4
+  metadata.gz: 237536c5543feeee187f725cefadecb004938ea5fffc48f138075bf87bb8fe81
+  data.tar.gz: 2b77dca5437b84a811cf14ef76ff4a75ddabf60104c6f96ee33ae9b55d2bd47c
 SHA512:
-  metadata.gz: aabd3e3e025c9760715b1543d95300e31fe460dff7fb42b50534729735767e82a0956f9ecd0253e4e14baea0616501b56984813f7ecf6d398765a7f38dc95e4a
-  data.tar.gz: 0c1cfa9abc73e1b7e2ccebeaefddd8541a853a0d0576267b3be4c2ee2d007dd39adec09d05d8469995414a04a9dbaffee92aafe410c02fb4fbf5217b8881c458
+  metadata.gz: ad8e95dfe39e5105e088dc75bf23559dbed8d89d425ea6a7c5108f9527564b7fcf0768f17b1dd207efd88922f400d74b0a200d11a95b9c9d6ce79c2096e8c4db
+  data.tar.gz: 829a4d2ee8d777e54eabcdb05ce4d111adab5be847cfb60ad2a945ab4130ad40c045544684a9131f174f78e0b20479c8e1aeddc713dc533f7852ec71f03132d5

data/lib/parsanol/3.2/parsanol_native.so

@@ -1,4 +1,4 @@
 [diffend] Oversized file quarantined before diffing.
 name: data/lib/parsanol/3.2/parsanol_native.so
-size: 39822280 bytes
-sha256: aa14d9984fc880b02e6e5126508898f004bbdbab054c1f4683ebbb0cc5bf311e
+size: 39947696 bytes
+sha256: 3f642d14ad0b823fc3a133014ef6929d3a597c5ab9c3a832d3fa996caf7bbabc

data/lib/parsanol/3.3/parsanol_native.so

@@ -1,4 +1,4 @@
 [diffend] Oversized file quarantined before diffing.
 name: data/lib/parsanol/3.3/parsanol_native.so
-size: 40057336 bytes
-sha256: 1e6652fb486af4fb4565c875abff07df0af4f6731d1e5b1869fff471f04d9189
+size: 40182696 bytes
+sha256: 67674d19fb132d8ef5ee0aaf763d4d3df4ee823a01469e7adc1213430f6a1fc3

data/lib/parsanol/3.4/parsanol_native.so

@@ -1,4 +1,4 @@
 [diffend] Oversized file quarantined before diffing.
 name: data/lib/parsanol/3.4/parsanol_native.so
-size: 40057288 bytes
-sha256: 29e86620fa72fa17f6e52a438f180745a168145d640707aca366e83531362d31
+size: 40182568 bytes
+sha256: 821e0e98b25e93c7a140b795af49dd7b7ce45b7e90401142e88d255ccaf7a3c5

data/lib/parsanol/4.0/parsanol_native.so

@@ -1,4 +1,4 @@
 [diffend] Oversized file quarantined before diffing.
 name: data/lib/parsanol/4.0/parsanol_native.so
-size: 40067328 bytes
-sha256: fc606510fde2c18139ff4b80d9b6501225b847119248000df76fbed895971f17
+size: 40192624 bytes
+sha256: 1f5c4793a904141f174de9bd33ce872e5cef2b2637a0f69590e0da37a4ab37ad

data/lib/parsanol/atoms/lookbehind.rb

@@ -2,30 +2,60 @@
 
 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).
+    # Inspects the text behind the current position and consumes
+    # nothing, yielding nil.
+    #
+    # Two forms (rs#137/#163):
+    # - Literal: the previous +count+ bytes must equal +pattern+
+    #   (the precedes?/does_not_precede? fixed-window guard).
+    # - Regex: +source+ is searched in the preceding text and must
+    #   end at the position — the CommonMark flanking form, which is
+    #   class-based, multibyte and variable-length.
     class Lookbehind < Base
-      attr_reader :count, :pattern, :positive
+      attr_reader :count, :pattern, :regex_source, :positive
 
       def initialize(count, pattern, positive: true)
         super()
         @count = count
         @pattern = pattern
+        @regex_source = nil
+        @compiled = nil
         @positive = positive
       end
 
+      # Class-based flanking guard: +source+ must match the text
+      # ending at the position.
+      def self.regex(source, positive: true)
+        atom = new(0, "", positive: positive)
+        atom.instance_variable_set(:@regex_source, source)
+        atom
+      end
+
+      def regex?
+        !@regex_source.nil?
+      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
+        if regex?
+          @compiled ||= Regexp.new("(?:#{@regex_source})\\z")
+          matched = @compiled.match?(source.input.byteslice(0, start))
+        else
+          behind =
+            if start >= @count
+              source.input.byteslice(start - @count, @count)
+            end
+          matched = !behind.nil? && behind == @pattern.b
+        end
         [matched == @positive, nil]
       end
 
       def to_s_inner(_prec)
-        "lookbehind(#{@count}, #{@pattern.inspect}, #{@positive ? '+' : '-'})"
+        if regex?
+          "lookbehind(/#{@regex_source}/, #{@positive ? '+' : '-'})"
+        else
+          "lookbehind(#{@count}, #{@pattern.inspect}, #{@positive ? '+' : '-'})"
+        end
       end
     end
   end

data/lib/parsanol/native/libparsanol.so

@@ -1,4 +1,4 @@
 [diffend] Oversized file quarantined before diffing.
 name: data/lib/parsanol/native/libparsanol.so
-size: 35606288 bytes
-sha256: e284d8864ede849a2f17d431a648f57335a7808ded52b645e924c1552b8e9a57
+size: 35749424 bytes
+sha256: 17585f657e7684ce70ff38a9bc4c85c928aa67caeaf7f7b7b5fdb32105c75980

data/lib/parsanol/native/serializer.rb

@@ -103,15 +103,17 @@ module Parsanol
                      # 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,
-                       },
-                     }
+                     # rs#137/#163: the behind-window test — a fixed
+                     # literal window or an end-anchored regex (the
+                     # CommonMark flanking form).
+                     look = if atom.regex?
+                              { "Regex" => { "source" => atom.regex_source } }
+                            else
+                              { "Literal" => { "count" => atom.count,
+                                               "pattern" => atom.pattern } }
+                            end
+                     { "Lookbehind" => { "look" => look,
+                                         "positive" => atom.positive } }
                    when Parsanol::Atoms::Ignored
                      serialize_ignored(atom)
                    else

data/lib/parsanol/parg/artifact.rb

@@ -65,8 +65,16 @@ module Parsanol
       def parse(entry_name, input, mode: :native)
         case mode
         when :native
-          if !envelope["dynamic"] && Native.available?
-            return Native.parse(JSON.generate(entry(entry_name).fetch("grammar")), input)
+          if Native.available? && !envelope["dynamic"]
+            # Fast lane (parsanol-ruby#162): the per-instance atom
+            # tree rides the DSL serialization/registration caches
+            # (GRAMMAR_CACHE / HANDLE_CACHE keyed on the stable root
+            # atom), so an artifact parses at compiled speed after
+            # the first call. The old path re-serialized the whole
+            # grammar JSON and re-registered it per parse (~8x).
+            # Dynamic artifacts carry ruby-tier atoms their wire
+            # cannot express — they keep the interpreter.
+            return root_atom(entry_name).parse(input, mode: :native)
           end
 
           root_atom(entry_name).parse(input)

data/lib/parsanol/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Parsanol
-  VERSION = "1.3.76"
+  VERSION = "1.3.77"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: parsanol
 version: !ruby/object:Gem::Version
-  version: 1.3.76
+  version: 1.3.77
 platform: aarch64-linux
 authors:
 - Ribose Inc.