canon 0.3.6 → 0.3.7

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 04012234db38cae77206ba33f856aca1262a0ad0111e7d233d5b814e90c22187
-  data.tar.gz: 3700758afcb5619f59ba7e2f896963bc63695d534f60af7d922dfce8c26076ec
+  metadata.gz: b1ce43e1e3cb461ac60687f0e9cdd815916f85cd40a87c8d70f0a071c56b1c04
+  data.tar.gz: 3eff7d2db22b5fe57674e64aa5f6b5ebb1d4b82d6f3afde5c45444f57f2c8e55
 SHA512:
-  metadata.gz: c6452b462325e65c3f38f31438656b1cf60790646413314d8e835f6a27f78026bc10de2474065627b26a602962c424fd6858e1bc97ed3bca43e8a009b1779291
-  data.tar.gz: 749b78e282b4b2c7d440f9dff4a2361ee8d7742afcdab54b490bf836ae9417afc5f677e2039143020a6e50eceaf12d42950372c7a5f62dfcd335223243962fef
+  metadata.gz: 7428800d8125c5173b61338ce8929f1c3700684ccba7cae25847a1d5cc533c99eea946a46cd5f567841b978597670e26d546849b853b11bd88cb6f72c023c647
+  data.tar.gz: 77cdd82d177e2d476aece71503a70487ee31f6593f30d401e748199944fbc9230756c5bba08b73411dae62621c5d08c16c1fdb4ec427a5e96f0d43955cea0414

data/lib/canon/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Canon
-  VERSION = "0.3.6"
+  VERSION = "0.3.7"
 end

data/lib/canon/xml/data_model.rb

@@ -205,9 +205,16 @@ inherited_namespaces: nil)
       # scopes cannot flow down during the stream).
 
       def self.from_moxml_xml(xml_string, preserve_whitespace:)
-        doc = Canon::XmlParsing.moxml_context.parse(xml_string, readonly: true)
+        # strict: false keeps the Nokogiri-engine contract — malformed
+        # input yields a recovered document plus diagnostics instead of
+        # an exception (moxml 0.5.15 defaults strict to true; issue
+        # #147 rides recover errors on Document#parse_errors).
+        doc = Canon::XmlParsing.moxml_context.parse(xml_string,
+                                                    readonly: true,
+                                                    strict: false)
         check_moxml_relative_namespace_uris(doc)
         result = build_from_moxml(doc, preserve_whitespace: preserve_whitespace)
+        result.parse_errors = doc.parse_errors if doc.parse_errors.any?
         # Canon owns this document's full lifecycle: the canon tree holds
         # no engine references once built, so release the native tree now
         # instead of waiting for the GC finalizer (moxml#134; no-op on

data/lib/canon/xml/sax.rb

@@ -10,16 +10,13 @@ module Canon
     # API to that protocol. OCP: a new engine is a new driver plus one line
     # here; the builder never changes.
     #
-    # Engine choice: the leptris SAX flip is armed but blocked twice over.
-    # The batch-attribute corruption (leptris-ruby#95) is answered by a
-    # correctness-first path in leptris 1.9.36 — attributes are correct,
-    # but that path DROPS namespace declarations from SAX events
-    # entirely (raw `["<root xmlns:a=.../>", []]` — xmlns pairs vanish),
-    # blinding namespace comparison. Until xmlns reporting is restored
-    # (and then the fast path returns with the C fix), CRuby keeps the
-    # Nokogiri driver; the flip is the one-line change below. With
-    # leptris absent (moxml resolving nokogiri) the driver calls Nokogiri
-    # directly — the wrapper layer only adds overhead there.
+    # Engine choice follows the resolved adapter — independent of the
+    # DOM engine default: leptris SAX (corruption leptris#625 and the
+    # missing xmlns reporting leptris-ruby#99 both fixed) builds canon
+    # trees at parity with or faster than Nokogiri SAX, CI-gated. The
+    # DOM engine default stays Nokogiri until the record conversion
+    # passes the same gate (see XmlBackend). With leptris absent the
+    # driver calls Nokogiri directly.
     module Sax
       autoload :NokogiriDriver, "canon/xml/sax/nokogiri_driver"
       autoload :MoxmlDriver, "canon/xml/sax/moxml_driver"
@@ -27,7 +24,7 @@ module Canon
       class << self
         # Drive `builder` with the runtime's SAX engine.
         def parse(xml_string, builder)
-          if RUBY_ENGINE == "opal"
+          if RUBY_ENGINE == "opal" || Canon::XmlParsing.moxml_adapter_name != :nokogiri
             MoxmlDriver.new(builder).parse(xml_string)
           else
             NokogiriDriver.new(builder).parse(xml_string)

data/lib/canon/xml_backend.rb

@@ -9,16 +9,14 @@ module Canon
   # - HTML support — always Nokogiri on CRuby; moxml has no HTML adapter and
   #   leptris no HTML parser (see Canon::Html::NokogiriSupport).
   #
-  # SSOT: moxml owns adapter preference (Moxml::Config prefers leptris when
-  # installed and capable; see XmlParsing.moxml_adapter_name). Canon's
-  # default engine is Nokogiri on CRuby until leptris WINS somewhere
-  # measurably: conformance parity is complete (engine_parity_spec 12/12),
-  # but the leptris SAX driver is blocked by a data-corruption bug
-  # (leptris-ruby#95) and the record-based conversion runs ~0.55x Nokogiri
-  # (CI performance gate: -42..45% on from_xml) pending lighter record
-  # emission upstream. Flipping the default is the one-line change in
-  # #detect; CANON_XML_BACKEND=moxml opts in today. Under Opal the engine
-  # is moxml (rexml adapter).
+  # SSOT: moxml owns adapter preference (Moxml::Config prefers leptris
+  # when installed; see XmlParsing.moxml_adapter_name). The SAX engine
+  # follows the resolved adapter — leptris (correct since leptris#625
+  # and leptris-ruby#99; CI perf-gated). The DOM engine stays Nokogiri
+  # until the record conversion passes the same gate (moxml#143's fix
+  # landed; still -40%/-37% from_xml on the CI runner). Conformance
+  # parity is complete (engine_parity_spec 12/12, zero pendings).
+  # CANON_XML_BACKEND=moxml opts into the leptris DOM engine today.
   #
   # HTML stays on Nokogiri on CRuby (Canon::Html::NokogiriSupport) and the
   # pretty-printers keep the Nokogiri pipeline — pretty-printed bytes are
@@ -63,6 +61,11 @@ module Canon
       def detect
         return :moxml if RUBY_ENGINE == "opal"
 
+        # The SAX engine flipped to leptris (leptris#625/#99 fixed,
+        # CI-gated — see Canon::Xml::Sax), but the DOM conversion still
+        # trails Nokogiri on the CI runner (-40%/-37% from_xml after
+        # moxml#143's emission fix landed). Nokogiri stays the DOM
+        # default until that gate passes; flipping is this one line.
         :nokogiri
       end
     end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: canon
 version: !ruby/object:Gem::Version
-  version: 0.3.6
+  version: 0.3.7
 platform: ruby
 authors:
 - Ribose Inc.
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2026-08-29 00:00:00.000000000 Z
+date: 2026-08-30 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: diff-lcs