parsanol 1.3.124-arm-linux → 1.3.126-arm-linux

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 21ca385359c1b4ae077a428395141e481cd3b3c043ba31219e4c62f922aa0a34
-  data.tar.gz: 6c98020dba568700d3edbd031e49236d45762f96b1fff3f285b89bfcbed9b55c
+  metadata.gz: b875944f9e4504b567fcf811a42831cf31c1804945a681c9ce87342e468f7fc7
+  data.tar.gz: 2f1708d2394539c82e9d9d2b5a3c0a56135112165583361480370fa58d43c0c2
 SHA512:
-  metadata.gz: 0dc25924bbf9cc5f9c04ee887722b99af2b0b70eca0e52bbc9436f885d134168211c6dc5670f246e28584d53433da0b14a6e77604bdc216a6cbd14bc41d18164
-  data.tar.gz: 8a3d9a0ca46171b618424baf1147f4263252f9060cdbf21cad65a97cc3c88a6864a0127505b94a727509157006c06eaa48bfd7df9443c498558b1adb0f313fd4
+  metadata.gz: 157335ceb6abe3e2c568c4ce7d76f12f591a68459f5fb9d75c99de6a37dd4caa5d95ae83d48d10748328c2a3d3e9ab26a91253e6711f8fea6c6d07ee3df5212b
+  data.tar.gz: 8fd8137a8797d4d3c13359a517ae7d34797dfd4561052f292946bd499c274a93c4350231adb5954a3089b8fb700b126cb5533068562d5521c00e4891c1ed0671

data/lib/parsanol/parg/compiler.rb

@@ -632,8 +632,8 @@ module Parsanol
               markers[literal] = node.a.to_sym if literal && !literal.strip.empty?
               collect.call(@document.rules[node.a]) if @document.rules.key?(node.a)
             when :alt, :seq then node.a.each { |b| collect.call(b) }
-            when :rep, :opt, :cap then collect.call(node.a)
-            when :pred then collect.call(node.b)
+            when :rep, :opt then collect.call(node.a)
+            when :cap, :pred then collect.call(node.b)
             end
           end
           collect.call(@document.skip_declaration)
@@ -652,7 +652,11 @@ module Parsanol
           case n.kind
           when :lit then literal = n.a
           when :seq, :alt then n.a.each { |c| walk.call(c) unless literal }
-          when :rep, :opt, :cap, :pred then walk.call(n.a || n.b) unless literal
+          # :cap stores the capture NAME in `a` (a plain String) and the
+          # captured node in `b`; :pred stores its flag in `a`. Walking
+          # `a` on either crashes — descend through `b`.
+          when :rep, :opt then walk.call(n.a) unless literal
+          when :cap, :pred then walk.call(n.b) unless literal
           end
         end
         walk.call(node)

data/lib/parsanol/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Parsanol
-  VERSION = "1.3.124"
+  VERSION = "1.3.126"
 end

data/lib/parsanol/vm.rb

@@ -990,8 +990,12 @@ module Parsanol
         @memoize = memoize
       end
 
-      # The RE match's [char width, byte advance]: single-class
-      # patterns match one character; multi-character patterns (%i
+      # The RE match's [span length, byte advance], both in BYTES: the
+      # interpreter's packed value spans are byte-based (STR packs the
+      # literal's byte length; ANY and the RUN ops pack the char's
+      # UTF-8 byte width; materialize byteslices the span), so the RE
+      # span must count bytes too. Single-class patterns match one
+      # character (its byte width); multi-character patterns (%i
       # literals reach RE with no byte table) consume their full
       # length (ruby#357: advancing one character skipped the
       # pattern's remainder and let the parse re-match inside it).
@@ -1002,12 +1006,17 @@ module Parsanol
           [w, w]
         else
           len = scanner.skip(pattern)
-          [input.byteslice(pos, len).length, len]
+          [len, len]
         end
       end
 
-      # An RE instruction's match at pos: [matched?, char width, byte
-      # advance]. Single-class patterns match one character;
+      # An RE instruction's match at pos: [matched?, span length, byte
+      # advance], both lengths in BYTES (see #re_match_width: the
+      # interpreter's spans are byte-based; a character count here
+      # truncated every slice containing a multibyte character —
+      # ruby#407: "Class[éin]" lost é's second continuation byte
+      # because the span stored the run's char count and materialize
+      # bytesliced it). Single-class patterns match one character;
       # multi-character patterns (%i literals reach RE with no byte
       # table) consume their FULL length (ruby#357: advancing one
       # character skipped the pattern's remainder and let the parse
@@ -1030,7 +1039,7 @@ module Parsanol
           [true, w, w]
         else
           len = scanner.skip(pattern)
-          [true, input.byteslice(pos, len).length, len]
+          [true, len, len]
         end
       end
 

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: parsanol
 version: !ruby/object:Gem::Version
-  version: 1.3.124
+  version: 1.3.126
 platform: arm-linux
 authors:
 - Ribose Inc.