canon 0.3.29 → 0.3.30

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: ff4b9d6b5a57176d2d382f7b0d730047161920dfb4cb4004b58d60bba9c0edab
-  data.tar.gz: ed8dfddca4d957fbc3fbb1e62dbd1b03032c661e82c06513304d2abcf4e288cc
+  metadata.gz: 308416a45e76807d58f00f40a88d5d6f53617d2a9dac468da8add04fa8175a2f
+  data.tar.gz: 2fa30986100df7b8238263546513705513635bd6c39f1552f0e0f8774f510f5f
 SHA512:
-  metadata.gz: 778faacee321541b7d8ceef7c3497969080243acd01fc01bde3902deb68b7664e5ed966a230fc969766cfb32aefe76c81be9d7ae8a86d6b0f0ea870a6ed32cf1
-  data.tar.gz: a3a42291c8100b20362c273c34778abdb4f22eec057c616e54daa60b77e4272c9fe5742fa352874ffc1145c0800e89f6c00800d8938bfcf2af364cff9c71ba9f
+  metadata.gz: 0ffdb20ffc91117fdf24afb405f1684aea837316c1481ed5cc7e435bff4bf5d409876d9ca5851e187d3cd128b192ce3516b49114e6ac5d56a327dd87330f0ec9
+  data.tar.gz: aaeac239ac4707d8034eb92c8205753e3ec94ff203bf0b4023c33d0b7a92d57742e632ee29998de5a2ed041e45ea7bf2453b2479bb24ed9b9164dacba8db4ab1

data/CLAUDE.md

@@ -143,6 +143,10 @@ Engine parity is complete through libleptris 1.9.8 / leptris-ruby 1.9.33 / moxml
 
 Engine A/B testing: `CANON_XML_BACKEND=nokogiri bundle exec rspec` (or `=moxml` to force leptris when it isn't the resolved default). The default suite must stay green under BOTH values; the only expected pendings are the upstream-tracked ones. The benchmark header (`rake performance:quick`) reports the active engine.
 
+### C14N Engines
+
+`Canon::Xml::C14n.canonicalize` stays on canon's Ruby C14N 1.1 processor by default. leptris' native C-side C14N 1.1 is ~50x faster on large documents and byte-identical on the main corpus, but diverges from the spec on edge cases (attribute ordering, prefix preservation, `>`/tab escaping, document-level PIs — leptris#1015, all pinned in `spec/canon/xml/c14n_engine_parity_spec.rb`). `CANON_C14N_BACKEND=leptris` opts in; when the parity gate runs clean the default flips. `canonicalize_subset` and `with_comments: true` always use the Ruby processor.
+
 ### YAML Engines
 
 `Canon::YamlBackend` selects the YAML engine: `:psych` (default) or `:yeptris` (FFI over libyeptris, the YAML counterpart of the leptris XML stack — `CANON_YAML_BACKEND=yeptris` opts in; the optional `yeptris` Gemfile group must be enabled). `Canon::YamlParsing` is the single gateway for string loads; `Canon::JsonParsing` mirrors it for JSON (the same yeptris engine serves both — `Yeptris::YAML.load` auto-detects JSON and routes to the native C-API materializer; JSON falls back to stdlib unless `YamlBackend.yeptris_native?`, since the FFI ladder is ~29x slower than the stdlib C extension). `YAML.dump` stays on Psych everywhere — canonical output bytes are canon's product and the writers differ. **JSON defaults to the strict yeptris surface whenever the native materializer is installed** (yeptris 0.1.13.4 ships platform gems — zero compilation, zero env; parity spec-pinned to `JSON.parse` upstream). The YAML default stays `:psych` until the yeptris Psych-parity gaps close (yeptris-ruby#30 sexagesimal scalars, #31 >64-bit integers — #29 empty documents was fixed in 0.1.12); `spec/canon/yaml_engine_parity_spec.rb` is the executable gate, with upstream-tracked cases pending. Never `require "yeptris/psych"` — it rebinds the global `::Psych` constant for the whole process; only the namespaced `Yeptris::YAML` API is used.

data/lib/canon/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Canon
-  VERSION = "0.3.29"
+  VERSION = "0.3.30"
 end

data/lib/canon/xml/c14n.rb

@@ -10,6 +10,10 @@ module Canon
       # @param with_comments [Boolean] Include comments in canonical form
       # @return [String] Canonical form in UTF-8
       def self.canonicalize(xml, with_comments: false)
+        if (native = native_canonicalize(xml, with_comments))
+          return native
+        end
+
         # Build XPath data model
         root_node = DataModel.from_xml(xml)
 
@@ -18,6 +22,46 @@ module Canon
         processor.process(root_node)
       end
 
+      # leptris' C-side C14N 1.1 — ~50x faster than the Ruby processor
+      # but NOT yet byte-identical on canon's edge-case corpus
+      # (attribute ordering, `>` escaping, document-level PIs, one
+      # prefix case — leptris#1015; the parity spec pins
+      # each). Opt-in via CANON_C14N_BACKEND=leptris until those
+      # close; comments mode keeps the Ruby path regardless (the
+      # native seam exposes no with-comments form).
+      def self.native_canonicalize(xml, with_comments)
+        return nil if with_comments
+        return nil if RUBY_ENGINE == "opal"
+        return nil unless ENV["CANON_C14N_BACKEND"].to_s.casecmp("leptris").zero?
+        return nil unless Canon::XmlBackend.moxml? &&
+          Canon::XmlParsing.moxml_adapter_name == :leptris
+        return nil unless native_c14n_available?
+
+        doc = Canon::XmlParsing.moxml_context.parse(xml, readonly: true,
+                                                         strict: false)
+        begin
+          doc.native.canonicalize(::Leptris::XML::FFI::C14N_1_1, nil,
+                                  mode: ::Leptris::XML::FFI::C14N_MODE_CANONICAL)
+        ensure
+          doc.free
+        end
+      rescue StandardError
+        # Malformed inputs are the Ruby path's domain (recovery parse
+        # + parse_errors surfacing), not the native lane's.
+        nil
+      end
+
+      def self.native_c14n_available?
+        return @native_c14n_available unless @native_c14n_available.nil?
+
+        @native_c14n_available = defined?(::Leptris::XML::FFI::C14N_1_1) &&
+          ::Leptris::XML::FFI.constants.include?(:C14N_MODE_CANONICAL)
+      end
+
+      def self.reset_native_probe!
+        @native_c14n_available = nil
+      end
+
       # Canonicalize a document subset selected by XPath expression.
       #
       # Implements W3C C14N 1.1 subset canonicalization:

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: canon
 version: !ruby/object:Gem::Version
-  version: 0.3.29
+  version: 0.3.30
 platform: ruby
 authors:
 - Ribose Inc.
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2026-09-09 00:00:00.000000000 Z
+date: 2026-09-12 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: diff-lcs