canon 0.3.122 → 0.3.124

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 5c8f274f503852b87469eb653b74b16cff4a24313a642ba665f9d00f4ddc5ca2
-  data.tar.gz: 1ff259b65b606b03400e5c729ba2384643877a48fdca3e7ff12f6ba5adfae62d
+  metadata.gz: 8c444720f0582084ebebf230399b6a0e1e275d8543c4b4099c90fe46a42a7178
+  data.tar.gz: 76be8f3de7a5e02dbe6273c5e9ae03d9ecde088a727227ccf7af38f0f623143f
 SHA512:
-  metadata.gz: 936d7dd4ca541dc44435294f26ac3f04886668ccfd5c919ef6cd50c033b75d633e33e4fcdbe9cdd809e9400f2c882af7baaf00d1316e6c832b46a87a8c0e6ec7
-  data.tar.gz: ae0c2379e7aefd1445bd108aa6533c8570858418e02e35653339618d597bfedab7254af12955a5cd7c8717e29a93591bfd458f6b3920fc343cf8af0a8189b0f8
+  metadata.gz: 269545a7a9c1d496db323de5fb5a31a67739938e612c812a5ad332fdc28c3e119cd9d39cda0806d218c4f6ef34807bc63dbe879e1973c260c64353dc7efffdaf
+  data.tar.gz: 63ddcc8bb82d0d0548e26cd1ee2436e6d03a94d746c959fd3c36a2e21de7bc307f517113d86744942ed12d304592265f1ece8ffe4a87c7c96bf7748d4c996ec8

data/docs/HANDOFF.adoc

@@ -5,18 +5,21 @@ published: false
 
 A snapshot of the open arcs across the org repos — what is known, where
 to dig, and what has already been proven — so any team can pick an issue
-up cold. Statuses as of 2026-11-01.
+up cold. Statuses as of 2026-11-05.
 
 == State of the world
 
-Shipped since the last board (2026-10-28): leptris 1.9.398.0 (the
-#1866 adopt + parent-of-attribute fixes and the #1846/#1854/#1855 perf
-slices), yeptris 0.6.52.0/0.6.53.0 (marshal reader #322, multibyte
-breaks #782), moxml **0.6.0** (the Opal surface deleted by the owner's
-phase-4 call, lutaml/moxml#516 — the forward Opal path is the moxml-ts
-wasm binding, lutaml/moxml#489), lutaml-model 0.8.155 (admits both
-moxml lines: `>= 0.5.139, < 0.7`), canon 0.3.118 (floors `leptris ~>
-1.9.398`, `yeptris ~> 0.6.52`, `moxml ~> 0.6.0`; Opal lane retired).
+Shipped since the last board (2026-11-01): leptris 1.9.401→**1.9.408**
+(the 1.9.407 plan ns_prefix deep-copy fix — the use-after-free behind
+lutaml-model#1126's poisoning signatures; release PR leptris#1973),
+yeptris **0.6.55**, moxml **0.6.2** (the engine XHTML display mode,
+moxml#390 — canon's fixture-ready lane is now engine-served end to
+end), lutaml-model **0.8.169** (the native emission lane fix series:
+plan poisoning #1126, the render contract #1136, the valued
+child_deferred gate #1130, xmlns retention #1131), canon 0.3.123
+(floors `leptris ~> 1.9.401`, `moxml ~> 0.6.2`, `yeptris ~> 0.6.55`;
+the subset-xpath engine grew parent/self/attribute steps, PR
+canon#308).
 
 == lutaml/canon
 
@@ -24,11 +27,12 @@ moxml lines: `>= 0.5.139, < 0.7`), canon 0.3.118 (floors `leptris ~>
 |===
 |Item |Status |Known + first moves
 
-|Floors raised (0.3.118)
+|Floors raised (0.3.123)
 |Shipped
-|`leptris ~> 1.9.398`, `yeptris ~> 0.6.52`, `moxml ~> 0.6.0` — all
-verified against exactly those engines (2744 examples, 0 failures, 25
-pending). Constraints live in the gemspec only; Gemfile/Gemfile.local
+|`leptris ~> 1.9.401`, `yeptris ~> 0.6.55`, `moxml ~> 0.6.2` — all
+verified against exactly those engines (2760 examples, 0 failures, 25
+pending, verified 2026-11-05). Constraints live in the gemspec only;
+Gemfile/Gemfile.local
 are untracked verification lanes.
 
 |Opal lane
@@ -39,11 +43,13 @@ The `RUBY_ENGINE == "opal"` branches in lib stay for the moxml-ts wasm
 migration (moxml#489).
 
 |Subset-xpath mini engine (`lib/canon/xml/xpath_engine.rb`)
-|Reported, owner decision pending
-|It silently returns `[]` for `..` and `@attr` step forms (XPath 1.0
-§2.2 names the owner element). No `lib/` caller uses those forms today
-(swept 2026-10-28); decide between implementing the steps and
-documenting/raising on unsupported forms.
+|Shipped (canon#308, 0.3.123)
+|The `..` and `@attr` step forms are implemented under a linear
+context model: element steps filter one level down (block semantics),
+parent/self/attribute steps operate on the matched nodes themselves,
+and attribute nodes render per strict C14N law (only when the owner
+element is in the node-set). The `//`-rooted and `./`-rooted relative
+forms resolve from the document root, matching the subset context.
 |===
 
 == lutaml/moxml
@@ -76,6 +82,18 @@ plan-path-FFI perf slices
 |Open (umbrella)
 |Re-baseline against 1.9.398 with the interleaved A/B harness before
 opening new slices — the old ~3-4x numbers predate the closed slices.
+
+|Plan descriptor memory safety + the 1.9.408 line
+|Shipped (1.9.407/1.9.408); #1972 open
+|1.9.407 fixes the `leptris_plan_build` use-after-free: attribute
+rows' `ns_prefix` was the one caller-anchored string NOT deep-copied,
+so the built descriptor read freed FFI anchor memory at serialize
+time (the lutaml-model#1126 empty-output / garbage-prefix
+signatures). There is no capability constant for a pure bugfix, so
+consumers gate on the version floor. 1.9.408 is the next release line
+(release PR #1973, merged 2026-11-05). Open on top: PR #1972
+(xsd: namespace-aware particle typing — substitution members ride
+attributeGroup refs).
 |===
 
 == leptris/yeptris
@@ -108,6 +126,16 @@ platforms)
 |Open, partially improved
 |Remaining: scalar-route derivation cost. `benchmarks/PERF-LEDGER.md`
 has the trusted-into entries.
+
+|yeptris#951 — plain fold emits raw lead-stripped 0x85 at a
+¥-preceded NEL run + the PS-as-break law
+|CLOSED (2026-11-05)
+|Diff 1 (the raw continuation bytes in a k=2 NEL fold) fixed by the
+engine team; Diff 2 settled as a comparator-claimed documented
+divergence — libyaml's `READ_LINE` breaks AT LS/PS but spells them as
+their raw 3 bytes (the fold law never applies since the join test
+expects `\n`), while yeptris folds the run to one `\n`. The suite is
+silent either way.
 |===
 
 == leptris/yeptris-ruby
@@ -130,17 +158,39 @@ If the CBOR face reopens, copy the registry law and its pin wholesale.
 |Item |Status |Known + first moves
 
 |moxml 0.6 admission
-|Shipped (0.8.155)
+|Shipped (0.8.156; line now at 0.8.169)
 |`>= 0.5.139, < 0.7`; the Opal boot lane pins moxml 0.5.x locally in
 `opal.yml` (Gemfile-append + plain `bundle update moxml` — bundler 4
 removed `update --version`) until the moxml-ts wasm migration.
 
+|The native emission lane fix series (the uniword corpus, 2026-11-05)
+|Shipped (0.8.165-0.8.169)
+|Four arcs, all closed: **#1126** (plans poison every later instance
+of a class) root-caused to an ENGINE use-after-free —
+`leptris_plan_build` kept a caller-anchored attribute `ns_prefix`
+pointer into freed FFI anchor memory; fixed leptris 1.9.407.
+**#1136** (render_default: false ignored — ghost `w:default="0"`
+attributes) fixed by applying the DOM lane's `rule.render?` contract
+at all three native emission sites (6fd1c0fc, 0.8.168). **#1130**
+(nested same-element recursion drops data — the bibliography's
+Author → author collapse to `<b:Author/>`) fixed by the compatibility
+gate refusing valued `:child_deferred` rows, which ride the DOM lane
+(PR #1135, 01ab5a3a, 0.8.169). **#1140** (dotx glossary TOC document
+displaces body elements across sdtContent) is the #1130 shape —
+bisect-proven fixed by the same gate; the OTHER five skip-listed dotx
+glossary shapes fail identically with the native lane OFF (DOM lane,
+engine without PLAN_OP_RAW), so they are uniword's pre-existing
+TODO.refactor/14 drift, not this lane. Also shipped: #1131 (xmlns
+declarations retained on parsed models, dbd7eb36).
+
 |update-baselines red on every main push
-|PR lutaml-model#1076
-|checkout's persisted Authorization extraheader collides with
+|Shipped (0be7a7fc + a784f452)
+|checkout's persisted Authorization extraheader collided with
 create-pull-request's token ("Duplicate header: Authorization", HTTP
-400) on the action's `git remote prune`. Fix: `persist-credentials:
-false` on that job's checkout (the documented pairing).
+400) on the action's `git remote prune`. Fixed with
+`persist-credentials: false` on that job's checkout, plus the
+`contents: write` + `pull-requests: write` grants (the git push and
+the API PR creation are separate permissions).
 
 |metanorma/docbook dependent leg (ubuntu/macos)
 |PR metanorma/docbook#15
@@ -194,10 +244,11 @@ the remaining step is the vendor commit into pubid's `data/parg/`
 (proven mechanical: baking origin/main at 1.3.121 yields `edition 2`
 headers with bindings intact — iso 84, bsi 92, ashrae 37). The
 owner's `worktrees/grammar-rebake` is on it. standoc's severity
-question (#1318/#1324) is re-pinned in metanorma-standoc#1333:
+question (#1318/#1324/#1325) is RESOLVED — re-pinned on standoc main
+directly by the owner (the dup PR #1333 was closed as superseded):
 relaton 3.0.2 degrades `ISO 0a` to RELATON_4 at severity 3, which a
-severity-2 filter legitimately drops; that PR's CI goes green the
-moment the vendor lands.
+severity-2 filter legitimately drops, and the suppression semantics
+are at-or-above N.
 |===
 
 == Mechanics every team hits

data/lib/canon/pretty_printer/html.rb

@@ -36,14 +36,10 @@ module Canon
       end
 
       def format(html_string)
-        if @fixture_ready
-          # The fixture-ready display mode is the one Nokogiri lane
-          # left (libxml FORMAT|AS_XHTML save flags — moxml#390 filed
-          # for an engine XHTML mode); everything else is engine-
-          # served.
-          Canon::NokogiriLoader.require!("HTML fixture-ready display")
-          return format_fixture_ready(html_string)
-        end
+        # The fixture-ready display lane is engine-served end to end
+        # (moxml 0.6.2's XHTML serialization mode — moxml#390): no
+        # Nokogiri requirement remains in this mode.
+        return format_fixture_ready(html_string) if @fixture_ready
 
         if xhtml?(html_string)
           format_as_xhtml(html_string)
@@ -91,14 +87,8 @@ module Canon
         end
       end
 
-      # Fixture-ready serialisation: parse with Nokogiri::HTML5 (so we
-      # get permissive recovery on real-world Word / XHTML5 / HTML5
-      # input shapes), then write through libxml's XML writer with
-      # +FORMAT+ + +AS_XHTML+ + +NO_DECLARATION+.  +FORMAT+ inserts
-      # indentation; +AS_XHTML+ produces well-shaped output (void
-      # elements self-closed, non-void paired); +NO_DECLARATION+
-      # suppresses the +<?xml ...?>+ prefix.
-      # Engine parse + significance-aware strip + the engine's
+      # Fixture-ready serialisation, engine-served end to end:
+      # engine HTML parse + significance-aware strip + moxml 0.6.2's
       # XHTML serialization mode (moxml#390 delivered to_xhtml /
       # to_html(xhtml: true)): void elements self-closed, non-void
       # paired — the AS_XHTML contract, engine-served.

data/lib/canon/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Canon
-  VERSION = "0.3.122"
+  VERSION = "0.3.124"
 end

data/lib/canon/xml/xpath_engine.rb

@@ -14,9 +14,14 @@ module Canon
     # - Predicates: [1] (position), [@attr], [@attr='value']
     # - Wildcards: *
     # - Union: expr1 | expr2
+    # - Self and parent steps: ., ..
+    # - Attribute steps: @attr (selects the attribute node; per C14N
+    #   law it renders only when its owner element is in the node-set,
+    #   so @attr/.. is the productive form — XPath 1.0 section 2.2
+    #   names the owner element)
     #
     # Not supported (not needed for C14N subset):
-    # - Axes other than child and descendant-or-self
+    # - Axes other than child, parent, self, descendant-or-self
     # - Functions (last(), position(), etc.)
     # - Variables
     #
@@ -56,6 +61,10 @@ module Canon
         elsif xpath.start_with?("/")
           # Absolute path
           evaluate_absolute(xpath[1..])
+        elsif xpath.start_with?("./")
+          # Document-rooted self step: the subset context is the
+          # document, so ./rest resolves from the root like /rest
+          evaluate_absolute(xpath[2..])
         else
           # Relative path — treat as descendant
           evaluate_descendant(xpath)
@@ -68,35 +77,47 @@ module Canon
         steps = parse_steps(path)
         return [] if steps.empty?
 
-        # Start from root's children
-        current_nodes = @root.children
-        apply_steps(current_nodes, steps)
+        # Start from root's children; the first element step filters
+        # this entry set directly
+        apply_steps(@root.children, steps, descend_first: false)
       end
 
       def evaluate_descendant(path)
         steps = parse_steps(path)
         return [] if steps.empty?
 
+        # ./ is transparent on the relative law: the subset context is
+        # the document, so ./rest searches descendants like rest
+        steps.shift while steps.first&.dig(:type) == :self
+        return [] if steps.empty?
+
         # Collect all descendant element nodes
         all_elements = []
         collect_elements(@root, all_elements)
 
-        # For each element, try to match the full path starting there
-        result = []
-        all_elements.each do |element|
-          first_step = steps.first
-          next unless step_matches?(element, first_step)
+        first_step = steps.first
+        remaining = steps[1..]
 
-          if steps.length == 1
-            result << element
+        context =
+          case first_step[:type]
+          when :attribute
+            # //@attr — every matching attribute in the tree
+            all_elements.flat_map do |element|
+              attribute_matches(element, first_step)
+            end
+          when :parent
+            return [] # no meaningful parents above descendants
           else
-            remaining = steps[1..]
-            matched = apply_steps(element.children, remaining)
-            result.concat(matched)
+            all_elements.select do |element|
+              step_matches?(element, first_step)
+            end
           end
-        end
+        return context.uniq if remaining.empty?
 
-        result.uniq
+        # The remaining steps run from the matches: element steps
+        # descend (block semantics), axis steps apply to the nodes
+        # themselves
+        apply_steps(context, remaining, descend_first: true).uniq
       end
 
       def collect_elements(node, result)
@@ -108,20 +129,54 @@ module Canon
         end
       end
 
-      def apply_steps(nodes, steps)
-        return nodes if steps.empty?
+      # Linear context model: each step transforms the current node
+      # set. Element steps filter one level down from the context
+      # (block semantics — the first step filters the entry set
+      # itself); parent/self/attribute steps are context steps and
+      # operate on the nodes themselves.
+      def apply_steps(nodes, steps, descend_first:)
+        steps.reduce(nodes) do |current, step|
+          break [] if current.empty?
+
+          if element_step?(step)
+            level = descend_first ? descend(current) : current
+            current = level.select { |n| step_matches?(n, step) }.uniq
+          else
+            current = axis_apply(current, step)
+          end
+          descend_first = true
+          current
+        end
+      end
 
-        step = steps.first
-        remaining = steps[1..]
+      def element_step?(step)
+        step[:type] == :element
+      end
 
-        matched = nodes.select { |n| step_matches?(n, step) }
+      def descend(nodes)
+        nodes.flat_map do |n|
+          next [] if n.is_a?(Nodes::AttributeNode)
 
-        if remaining.empty?
-          matched
-        else
-          matched.flat_map do |node|
-            apply_steps(node.children, remaining)
-          end
+          n.children
+        end
+      end
+
+      # Context steps: navigate from the nodes themselves.
+      def axis_apply(nodes, step)
+        case step[:type]
+        when :parent
+          nodes.filter_map(&:parent).uniq
+        when :self
+          nodes
+        when :attribute
+          nodes.grep(Nodes::ElementNode)
+            .flat_map { |el| attribute_matches(el, step) }.uniq
+        end
+      end
+
+      def attribute_matches(element, step)
+        element.attribute_nodes.select do |a|
+          attr_name_matches?(a, step[:name])
         end
       end
 
@@ -132,6 +187,17 @@ module Canon
           predicates_match?(node, step[:predicates])
       end
 
+      def attr_name_matches?(attr, name)
+        return true if name == "*"
+
+        if name.include?(":")
+          prefix, local = name.split(":", 2)
+          attr.prefix == prefix && attr.name == local
+        else
+          attr.name == name
+        end
+      end
+
       def name_matches?(node, name)
         return true if name == "*"
 
@@ -179,7 +245,7 @@ module Canon
       # Parse a path string into an array of steps.
       #
       # @param path [String] XPath path (without leading /)
-      # @return [Array<Hash>] Array of { name:, predicates: }
+      # @return [Array<Hash>] Array of { type:, name:, predicates: }
       def parse_steps(path)
         steps = []
         scanner = StringScanner.new(path)
@@ -189,14 +255,27 @@ module Canon
           break if scanner.eos?
 
           # Skip /
-          scanner.scan(%r{/})
+          scanner.skip(%r{/})
+          scanner.skip(/\s+/)
+
+          if scanner.scan("..")
+            steps << { type: :parent, name: nil, predicates: [] }
+            next
+          elsif scanner.scan(".")
+            steps << { type: :self, name: nil, predicates: [] }
+            next
+          elsif scanner.scan(/@/)
+            name = scanner.scan(/[a-zA-Z_][\w.-]*/)
+            steps << { type: :attribute, name: name, predicates: [] }
+            next
+          end
 
           name = scan_name(scanner)
           break if name.nil?
 
           predicates = scan_predicates(scanner)
 
-          steps << { name: name, predicates: predicates }
+          steps << { type: :element, name: name, predicates: predicates }
         end
 
         steps

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: canon
 version: !ruby/object:Gem::Version
-  version: 0.3.122
+  version: 0.3.124
 platform: ruby
 authors:
 - Ribose Inc.
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2026-11-03 00:00:00.000000000 Z
+date: 2026-11-08 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: diff-lcs
@@ -44,28 +44,28 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 1.9.401
+        version: 1.9.412
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 1.9.401
+        version: 1.9.412
 - !ruby/object:Gem::Dependency
   name: moxml
   requirement: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.6.0
+        version: 0.6.4
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.6.0
+        version: 0.6.4
 - !ruby/object:Gem::Dependency
   name: paint
   requirement: !ruby/object:Gem::Requirement
@@ -142,14 +142,14 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.6.55
+        version: 0.6.60
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.6.55
+        version: 0.6.60
 description: |-
   Canon provides canonicalization and pretty-printing for various serialization
   formats (XML, HTML, JSON, YAML), producing standardized forms suitable for