canon 0.3.45 → 0.3.47

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: a17cd55bc56c7287d027191b11cd33a6bc0999b29abce17bb8c368758379e4ad
-  data.tar.gz: 7b7c89255ad4134219b512265b6b26c08a1352742dfc1500a821a830fa3cd2d3
+  metadata.gz: 8180f10b0965070c408985ca44868fc14ae2256f24dac607492470aea7e5d076
+  data.tar.gz: e50498b713b972237a67af3b3f5c2cd5fb0cf1cbc3fef0dbb57f628810ce2341
 SHA512:
-  metadata.gz: d4673a7e48e7fb4aa901ece0c6fc3c097e3c046af4aee63b959b218d8809a32ac9e8f4ea3db990ff969f521d9842e826e03f4815da8610fba7138070cc420f8e
-  data.tar.gz: f5771189aed0ef86eb98b32b9bd94261221688aece84f6f7b44861c66afcde65d80b3b0a2cb599bb96c18aaa575da4ff86d1bb385b48ff7a3ff34c858342a632
+  metadata.gz: 52c5eb3fcb8ab3e09c7de5b8b7120f2baad191467aeccab6154254cd0007128f76ae2f5dbd26102abf19c3cf9f289f3c0445db88917562d7b7edafe9aef7ae5b
+  data.tar.gz: 373032bcba62f9685c41f09bb0d3dce294c50fa8785252e3737231a67f9a0fdfdb5372c85f3b28da5983faf39d4cdf89723dfae8397969511d2688afecaf3418

data/lib/canon/diff/diff_line_builder.rb

@@ -21,19 +21,22 @@ module Canon
       # @param text1 [String] The first document (preprocessed)
       # @param text2 [String] The second document (preprocessed)
       # @return [Array<DiffLine>] The assembled diff lines
-      def self.build(diff_nodes, text1, text2)
+      def self.build(diff_nodes, text1, text2, lines1: nil, lines2: nil)
         return [] if diff_nodes.nil? || diff_nodes.empty?
         return [] if text1.nil? || text2.nil?
 
-        new(diff_nodes, text1, text2).build
+        new(diff_nodes, text1, text2, lines1, lines2).build
       end
 
-      def initialize(diff_nodes, text1, text2)
+      def initialize(diff_nodes, text1, text2, lines1 = nil, lines2 = nil)
         @diff_nodes = diff_nodes
         @text1 = text1
         @text2 = text2
-        @lines1 = text1.split("\n")
-        @lines2 = text2.split("\n")
+        # The caller may hand in the already-split lines (the
+        # formatter splits these documents too); re-splitting both
+        # documents per pipeline stage was ~2x the line strings.
+        @lines1 = lines1 || text1.split("\n")
+        @lines2 = lines2 || text2.split("\n")
         # Build reverse indices for efficient content lookup in gap handling.
         # Maps content string to array of line indices where that content appears.
         @line_to_indices1 = build_line_index(@lines1)

data/lib/canon/diff/diff_node_enricher.rb

@@ -22,21 +22,22 @@ module Canon
       # @param text1 [String] The first document (preprocessed)
       # @param text2 [String] The second document (preprocessed)
       # @return [Array<DiffNode>] The same DiffNodes, enriched in place
-      def self.build(diff_nodes, text1, text2)
+      def self.build(diff_nodes, text1, text2, lines1: nil, lines2: nil)
         return diff_nodes if diff_nodes.nil? || diff_nodes.empty?
         return diff_nodes if text1.nil? || text2.nil?
 
-        new(diff_nodes, text1, text2).enrich
+        new(diff_nodes, text1, text2, lines1, lines2).enrich
       end
 
-      def initialize(diff_nodes, text1, text2)
+      def initialize(diff_nodes, text1, text2, lines1 = nil, lines2 = nil)
         @diff_nodes = diff_nodes
         @text1 = text1
         @text2 = text2
         @line_map1 = SourceLocator.build_line_map(text1)
         @line_map2 = SourceLocator.build_line_map(text2)
-        @lines1 = text1.split("\n")
-        @lines2 = text2.split("\n")
+        # Shared split — see DiffLineBuilder's note.
+        @lines1 = lines1 || text1.split("\n")
+        @lines2 = lines2 || text2.split("\n")
         # Track occurrences for text_content dimension to find correct element instance
         @text_occurrence1 = Hash.new(0)
         @text_occurrence2 = Hash.new(0)

data/lib/canon/diff_formatter/by_line/xml_formatter.rb

@@ -44,12 +44,22 @@ module Canon
           # Compute line number width BEFORE formatting
           compute_line_num_width(doc1, doc2)
 
+          # The pipeline splits both documents once here; every stage
+          # downstream reuses the arrays (three redundant full-doc
+          # splits per render before).
+          @pipeline_lines1 = doc1.split("\n")
+          @pipeline_lines2 = doc2.split("\n")
+
           # Phase 1: Enrich DiffNodes with character positions
-          Canon::Diff::DiffNodeEnricher.build(@differences, doc1, doc2)
+          Canon::Diff::DiffNodeEnricher.build(@differences, doc1, doc2,
+                                              lines1: @pipeline_lines1,
+                                              lines2: @pipeline_lines2)
 
           # Phase 2: Assemble DiffLines from enriched DiffNodes
           diff_lines = Canon::Diff::DiffLineBuilder.build(@differences, doc1,
-                                                          doc2)
+                                                          doc2,
+                                                          lines1: @pipeline_lines1,
+                                                          lines2: @pipeline_lines2)
 
           # Layers 3-5: Build report through pipeline
           report = Canon::Diff::DiffReportBuilder.build(
@@ -178,8 +188,8 @@ module Canon
         def format_report(report, doc1, doc2)
           return "" if report.contexts.empty?
 
-          lines1 = doc1.split("\n")
-          lines2 = doc2.split("\n")
+          lines1 = @pipeline_lines1 || doc1.split("\n")
+          lines2 = @pipeline_lines2 || doc2.split("\n")
 
           output = []
 

data/lib/canon/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Canon
-  VERSION = "0.3.45"
+  VERSION = "0.3.47"
 end

data/lib/canon/xml/c14n.rb

@@ -23,19 +23,24 @@ module Canon
       end
 
       # leptris' C-side C14N 1.1 — 23x faster than the Ruby processor
-      # through canon's own API (1MB document). libleptris 1.9.164
-      # closed four of the five leptris#1015 families (attribute
-      # ordering, prefix loss and rebinding, `>` and TAB escaping —
-      # unpinned in the parity spec since gem 1.9.174.0). A local
-      # default-flip attempt surfaced three MORE divergences the
-      # edge corpus did not cover, so the lane stays opt-in
-      # (CANON_C14N_BACKEND=leptris) until they close (leptris#1096):
-      # redundant namespace redeclarations are kept (Ruby removes
-      # them), xmlns:xml with the standard URI is kept (Ruby omits
-      # it), and document-level processing instructions are dropped.
-      # The Ruby lane additionally validates relative namespace URIs
-      # at parse time — a future flip must carry that check into the
-      # native wrapper. Comments mode keeps the Ruby path regardless.
+      # through canon's own API (1MB document) and byte-identical on
+      # the parity corpus, but OPT-IN (CANON_C14N_BACKEND=leptris)
+      # pending three families verified against libxml2 ground truth
+      # (leptris#1096 follow-up): (1) whitespace-only text between
+      # elements is preserved — the Ruby lane's parse drops it and
+      # compact bytes are canon's product for every pretty-printed
+      # document; (2) whitespace-only processing-instruction data is
+      # retained (libxml2 drops it); (3) document-level nodes carry
+      # no "\n" separators (the Ruby lane, like libxml2, inserts
+      # them). libleptris 1.9.176/177 (gem 1.9.177.0) closed the
+      # earlier families — attribute ordering, prefix loss and
+      # rebinding, `>`/TAB escaping, redundant namespace
+      # redeclarations, xmlns:xml omission — and document-level PIs
+      # now serialize in document order (the Ruby processor's
+      # root-first order is non-conformant for prolog PIs).
+      # Relative namespace URIs raise exactly as the Ruby lane does.
+      # Comments mode keeps the Ruby path (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"
@@ -44,6 +49,8 @@ module Canon
           Canon::XmlParsing.moxml_adapter_name == :leptris
         return nil unless native_c14n_available?
 
+        validate_relative_namespaces!(xml)
+
         doc = Canon::XmlParsing.moxml_context.parse(xml, readonly: true,
                                                          strict: false)
         begin
@@ -58,6 +65,25 @@ module Canon
         nil
       end
 
+      # Canon::Error parity with the Ruby lane, whose parse rejects
+      # relative namespace URIs. A declaration-shaped scan; a
+      # relative-URI-shaped string inside CDATA is the only false
+      # positive this can produce.
+      RELATIVE_NS_DECL = /xmlns(?::[\w.-]+)?="([^"]*)"/
+      URI_SCHEME = %r{\A[a-zA-Z][a-zA-Z0-9+.-]*:}
+
+      def self.validate_relative_namespaces!(xml)
+        return unless xml.is_a?(String)
+
+        xml.scan(RELATIVE_NS_DECL) do |(uri)|
+          next if uri.nil? || uri.empty?
+
+          unless uri.match?(URI_SCHEME)
+            raise Canon::Error, "Relative namespace URI not allowed: #{uri}"
+          end
+        end
+      end
+
       def self.native_c14n_available?
         return @native_c14n_available unless @native_c14n_available.nil?
 

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: canon
 version: !ruby/object:Gem::Version
-  version: 0.3.45
+  version: 0.3.47
 platform: ruby
 authors:
 - Ribose Inc.