parsanol 1.3.123 → 1.3.125

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 0006f48489613bd59db08d6d334afaac8ca42c814d7cba0745f155c60eb1c0cb
-  data.tar.gz: 9f8cacd841900d3563591c8d5669b850cb451add9ec7750f45a4940d4d59ee7b
+  metadata.gz: 9300ad7831ee208e1424f8a65c035db1db73f335bedfa2e3cb77027e7ae048ba
+  data.tar.gz: ec7e0cc9a73e967e957eba039f52b66e80c63cf53ca4048e69c6303b2d09f23b
 SHA512:
-  metadata.gz: 5420477cdd73fef4a4da6671219e6b363d105032712d933479e9cacb7c3d96ba6ea14c02851f9dd4704b06f14f40ec3f23f0fd2e8a8d722f0a703f80662fc953
-  data.tar.gz: b1fd5130113ee6f86919c0c9f348b0fb337f9b806e0216fb3144d9fd52dd20610ebd6ea961eba569b92ac6cbbccccf5ee06e08e127970c262d79143f058eac32
+  metadata.gz: c57b6923afc7be098b06844fd17fc42957c84ca7829363b8ce90fc04947c64ff7d30d16acef7aa8352f15c2b0b6f406095b6c97acd4aed4806057d4e96c31bf7
+  data.tar.gz: 200c2f64e60befe15181fbb92735d65428f02e3c01245fc66732da1e38083dfdc5143968d5b86f8d3957d4572c42a74af114eea7e0ce9917abe9379885a43aa9

data/lib/parsanol/parg/artifact.rb

@@ -241,6 +241,7 @@ module Parsanol
               end
               "test #{hash['input'].inspect}: expected captures " \
                 "#{mismatched.transform_values(&:inspect).inspect}, got #{actuals.inspect}"
+            when "bind" then bind_mismatch(entry_name, shape, hash)
             end
           rescue Parsanol::ParseFailed
             unless hash["kind"] == "reject"
@@ -250,6 +251,20 @@ module Parsanol
         end
       end
 
+      # parsanol-ruby#414: the whole bound map must equal the expectation
+      # exactly (not the partial `example` match) — the grammar author
+      # pins what the bindings produce, including absent keys and
+      # phantom rows.
+      def bind_mismatch(entry_name, shape, hash)
+        bound = apply_bindings(entry_name, shape)
+        actual = Bindings.canonical(bound)
+        expected = Bindings.canonical(hash["expect"])
+        return nil if actual == expected
+
+        "test #{hash['input'].inspect}: expected bound #{expected.inspect}, " \
+          "got #{actual.inspect}"
+      end
+
       # Best-effort recovery (TODO.sota/1, R1): on success returns
       # {"ok": true, "shape": tree}; on ParseFailed returns
       # {"ok": false, "error": {"offset", "message"}, "captures":

data/lib/parsanol/parg/document.rb

@@ -11,9 +11,9 @@ module Parsanol
       # `capture` = phase 1 (grammar layer): asserts the raw capture tree,
       # no bindings section needed. `example` = phase 2 (bindings layer):
       # asserts the bound, preprocessed attribute hash.
-      TEST_KINDS = %i[accept reject example capture shape].freeze
+      TEST_KINDS = %i[accept reject example capture shape bind].freeze
       # Test kinds whose expectations are `{ key: value }` pairs.
-      PAIR_KINDS = %w[example capture].freeze
+      PAIR_KINDS = %w[example capture bind].freeze
 
       Test = Struct.new(:entry, :kind, :input, :expect)
 

data/lib/parsanol/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Parsanol
-  VERSION = "1.3.123"
+  VERSION = "1.3.125"
 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: 1.3.123
+  version: 1.3.125
 platform: ruby
 authors:
 - Ribose Inc.