moxml 0.5.95 → 0.5.96

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: de8e16de76cef81c34b0a4e097d5aa16196fdbc03d9bfe2cbee1e765312f1599
-  data.tar.gz: ab5cc00b4c1a4a209065620c89b7541e107ed3d0a80e1e91e4455cb10c17fd98
+  metadata.gz: 011415dfeb5b083b1cb88c8a7b4619ed3cc896c51b78500fe849c02bbab055a4
+  data.tar.gz: a8b99c779efcca214047df53e1891b07961f0d3ab154e3f3538e83bfdbfb64f5
 SHA512:
-  metadata.gz: bf21f0b903d142e6ebce40004d4e3bb657ec812396a2e8e6bd49c2eb06b6763fa6917f18c6918a134784d279360092c127fad56b2a75449f21d7e2dc94ae2e3c
-  data.tar.gz: 0bc9b8912c18c946101b76c88ce2c5c1d7f3ab2d8f643d06a8760d78332e1fd0d426844caff21efd8bb5241af9b35471478223fff9754b78405c3dda28d1990b
+  metadata.gz: d1eb8163f7e725c42bba78d058abcece733204a414df60278d4fbfc477808160e8800eaa9fd5de5c011c1ce849d3f9e630eb667ebe15808cc6a614120a92fcd5
+  data.tar.gz: 5d4108fc8881b6c8073dd8de95c0023b3aaeca379e258fa5b8345a6c6016f0ee814b9d2c6b047ea1d91c724391e4fc8b6127090781c33665dcba59c536f8bb52

data/lib/moxml/adapter/rexml.rb

@@ -25,12 +25,12 @@ module Moxml
         def parse(xml, options = {}, _context = nil)
           xml = "" if xml.nil?
 
-          # Handle frozen strings by creating a mutable copy
-          processed_xml = if xml.frozen?
-                            xml.dup.force_encoding("UTF-8").encode("UTF-8")
-                          else
-                            xml.force_encoding("UTF-8").encode("UTF-8")
-                          end
+          # Transcode from the input's own encoding tag; encode returns
+          # a fresh string (or self, already UTF-8) — the caller's
+          # buffer is never retagged in place. The old non-frozen
+          # branch force_encoding!-mutated the CALLER's string,
+          # corrupting it (invalid UTF-8 tag) for every later parse.
+          processed_xml = xml.encode("UTF-8")
 
           # Preprocess entities to avoid double-escaping on output
           processed_xml = preprocess_entities(processed_xml)

data/lib/moxml/entity.rb

@@ -112,6 +112,7 @@ module Moxml
             else
               xml.encode("UTF-8")
             end
+      str = rewrite_declared_encoding(str)
       # Fast path: no `&` means no entity references to mark — skip
       # the regex scan and string allocation entirely. The vast
       # majority of XML payloads contain no entity references.
@@ -139,6 +140,24 @@ module Moxml
       preprocess_with_marker_flag(xml)[0]
     end
 
+    # Engines honor the DECLARED encoding and re-decode the bytes they
+    # are handed — moxml always feeds UTF-8 (see the normalization
+    # above), so a declaration naming another encoding must be
+    # rewritten to match, or the already-transcoded bytes decode a
+    # second time into mojibake ("café" -> "café"). The anchored match
+    # fails on the first two bytes for documents without a declaration.
+    DECLARED_ENCODING_RE = /\A(<\?xml[^>]*?encoding=)['"][^'"]*['"]/
+    private_constant :DECLARED_ENCODING_RE
+
+    def rewrite_declared_encoding(str)
+      # The passthrough contract (Entity spec: plain text returns the
+      # SAME object) requires no allocation when no declaration names
+      # another encoding — both gates are allocation-free.
+      return str unless str.start_with?("<?xml") && str.include?("encoding=")
+
+      str.sub(DECLARED_ENCODING_RE) { "#{Regexp.last_match(1)}\"UTF-8\"" }
+    end
+
     # Resolve numeric (&#NN; / &#xNN;) and the five standard named
     # (&amp; &lt; &gt; &quot; &apos;) XML entity references in one
     # single pass. The resulting characters are data; no further

data/lib/moxml/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Moxml
-  VERSION = "0.5.95"
+  VERSION = "0.5.96"
 end

data/spec/moxml/c14n/character_encoding_spec.rb

@@ -0,0 +1,57 @@
+# frozen_string_literal: true
+
+require "spec_helper"
+require "moxml/c14n"
+
+# TODO.c14n/13: encoding edge cases. moxml normalizes every parse to
+# UTF-8 bytes and rewrites the declaration to match — engines honor
+# the DECLARED encoding and would otherwise re-decode the already
+# transcoded bytes into mojibake.
+CAF_E_UTF8 = [60, 114, 62, 99, 97, 102, 195, 169, 60, 47, 114, 62].freeze
+LATIN1_XML = (+"<?xml version=\"1.0\" encoding=\"ISO-8859-1\"?><r>caf\xE9</r>").force_encoding("ISO-8859-1")
+BOM_XML = (+"\xEF\xBB\xBF<r/>").force_encoding("UTF-8")
+NFC_XML = (+"<r>caf\xC3\xA9</r>").force_encoding("UTF-8")
+
+# oga's parse layer leaks the declaration as a PI node and the BOM as
+# text (pre-existing, documented in the W3C examples spec).
+OGA_LEAK = %i[oga].freeze
+
+RSpec.describe "Moxml character encoding at the parse boundary" do
+  Moxml::Adapter::AVAILABLE_ADAPTERS.each do |adapter_name|
+    context "with the #{adapter_name} adapter" do
+      let(:ctx) { Moxml.new(adapter_name) }
+
+      unless OGA_LEAK.include?(adapter_name)
+        it "transcodes declared Latin-1 to UTF-8 bytes" do
+          doc = ctx.parse(LATIN1_XML.dup)
+          expect(Moxml::C14n.canonicalize(doc).bytes).to eq(CAF_E_UTF8)
+        end
+
+        it "strips a leading UTF-8 BOM" do
+          doc = ctx.parse(BOM_XML.dup)
+          expect(Moxml::C14n.canonicalize(doc)).to eq("<r></r>")
+        end
+      end
+
+      it "passes NFC content through without normalization" do
+        doc = ctx.parse(NFC_XML.dup)
+        expect(Moxml::C14n.canonicalize(doc).bytes).to eq(CAF_E_UTF8)
+      end
+
+      it "does not mutate the caller's input string" do
+        input = LATIN1_XML.dup
+        expect(input).not_to be_frozen
+        ctx.parse(input)
+        expect(input.encoding).to eq(Encoding::ISO_8859_1)
+        expect(input.valid_encoding?).to be(true)
+        expect(input.bytes).to eq(LATIN1_XML.bytes)
+      end
+
+      it "serializes transcoded documents as UTF-8 declarations" do
+        doc = ctx.parse(LATIN1_XML.dup)
+        expect(doc.to_xml).to include("UTF-8")
+        expect(doc.to_xml).not_to include("ISO-8859-1")
+      end
+    end
+  end
+end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: moxml
 version: !ruby/object:Gem::Version
-  version: 0.5.95
+  version: 0.5.96
 platform: ruby
 authors:
 - Ribose Inc.
@@ -461,6 +461,7 @@ files:
 - spec/moxml/c14n/adapter_delegation_spec.rb
 - spec/moxml/c14n/api_spec.rb
 - spec/moxml/c14n/c14n_spec.rb
+- spec/moxml/c14n/character_encoding_spec.rb
 - spec/moxml/c14n/comments_pis_spec.rb
 - spec/moxml/c14n/inclusive10_spec.rb
 - spec/moxml/c14n/namespace_edge_cases_spec.rb