uniword 1.3.0 → 1.4.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: c28f3b26ea127104f901589ebdb8752163be6f9e963c11eca94a26c86bdbd0df
-  data.tar.gz: d8fbe7de4f8f005d78d10acd4f6c1f7459d06385b419c487732e0398470d2c23
+  metadata.gz: ee9e7988ad5a17537188aa93af92867049b6c71fc99e00b24b09cde8f81eda15
+  data.tar.gz: 57908a45bf89d01a4b0630736678a5a2ff8ea891fe25650229f83b7b94a88b69
 SHA512:
-  metadata.gz: 807f859be7a5b6b45512ea946e4a4b1bd12a8e9df975e9dd9e06dd75c0842d916e5f811e9ac10062873426402b0a3526fa5301a0634e15062fb3f65696f3bd03
-  data.tar.gz: 3d6d287a816d2c2cc42a59b5da669dc92026774b0a2e27ff02e33a916367a63336e390338e556d1688b6cb7bbf78676dd5f200b47e8fdcd5c1509e0ebd238f90
+  metadata.gz: e7b6660cbc36ccff3170559c4c8eebaeed56a2face7fba85ae5ec1c42ad09ba22270fe9d0b5cabc82be78a5cfc29dfa67cb8caf32afc976bd0c7d8530dd358db
+  data.tar.gz: b1d88957cf0b8c3b647246afb81d3e61e3a9e6c327115994b69f0a0c7070754380dafbd3e9a7ef7cf0379821d741de190f0fce58a04db8f86b1dcecd3095286e

data/CHANGELOG.md

@@ -8,6 +8,96 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [1.4.0] - 2026-07-20
+
+### Added
+
+- DOCX parts the library does not model (e.g. `docProps/meta.xml`,
+  glossary documents, VBA projects, customXml outside the standard
+  layout, header/comment `.rels` sidecars) now round-trip
+  byte-for-byte instead of being dropped on load. Unclaimed ZIP
+  entries are preserved as raw parts (`Docx::RawPart`,
+  `Package#raw_parts`) with their source content types, re-emitted
+  verbatim on save; relationships targeting them are kept by the
+  reconciler's dangling-relationship sweep, and the write-time
+  package integrity gate treats them as first-class parts.
+
+### Changed
+
+- Internal refactor (no user-visible behavior change): `Ooxml::PartRegistry`
+  is now bidirectional and loader-aware. `Ooxml::PartDefinition` describes
+  how each part loads (loader strategy key, parsing model, load priority,
+  dynamic path resolution) and where it lives on
+  `Wordprocessingml::DocumentRoot` and `Docx::Package`. Registry
+  registrations now cover every part the library reads (content types,
+  package/document/settings/theme/footnote/endnote rels, theme media).
+  `Docx::Package.from_zip_content` loads parts by iterating registry
+  entries through the new `Docx::PartLoader` strategy registry
+  (`XmlModelLoader`, `CustomXmlLoader`, `HeaderFooterLoader`,
+  `ChartLoader`, `ImageLoader`, `EmbeddingLoader`, `ThemeMediaLoader`)
+  instead of a hand-written per-part sequence, and the two
+  document↔package copy lists
+  (`DocumentFactory.copy_package_parts_to_document`,
+  `Docx::PackageDefaults.copy_document_parts_to_package`) are single
+  registry-driven loops — a part can no longer exist in only one
+  direction.
+- Internal refactor (no user-visible behavior change): the remaining
+  hash-based part families are now model objects. New
+  `Docx::ImagePart` (binary data, target, content type, relationship
+  id, source path) backs `DocumentRoot#image_parts`, which is now an
+  rId-keyed `Docx::PartCollection` that normalizes legacy
+  `{ data:, target:, content_type: }` hash assignments and keeps
+  hash-style reads working (`[]`, `each`, `each_value`, `keys`,
+  `size`, `empty?`, `[]=`, `delete`, `key?`/`has_key?`).
+  `Docx::PartCollection` wraps hash entries through the new
+  polymorphic `Docx::Part.from_hash` hook. The write-only
+  `Package#chart_parts` and `Package#bibliography_sources` copies are
+  gone — `DocumentRoot#chart_parts` and
+  `DocumentRoot#bibliography_sources` are the single homes, and the
+  registry's `:chart`/`:bibliography` definitions no longer carry
+  document/package copy attributes. Builders, part loaders,
+  serializers and reconcilers read and write the model objects
+  directly.
+- Internal refactor (no user-visible behavior change): "will this part
+  be emitted?" is now answered by `Ooxml::PartRegistry.emitted_paths`
+  instead of a hand-written mirror in the reconciler. Part objects
+  (`Docx::Part`, `Docx::CustomXmlItem`) report their own emitted
+  package paths (`package_paths`), and family definitions
+  (`:header`/`:footer`/`:chart`/`:image`/`:bibliography`) gained the
+  collection attributes needed to enumerate them.
+- Internal refactor (no user-visible behavior change):
+  `Uniword::Ooxml::ElementOrder` is now the single safe
+  `element_order` mutation point (`mutable_order`, `append`,
+  `insert_at`, `insert_once`), replacing per-model workarounds for
+  lutaml-model's frozen parse arrays in footnotes, endnotes, the TOC
+  generator, and the reconciler helpers
+
+### Fixed
+
+- `uniword verify` no longer reports false OPC-006 "relationship
+  target not found" errors for relationship targets containing dot
+  segments (e.g. `../docProps/meta.xml`); target resolution now
+  normalizes `.`/`..` per OPC Part 2, matching the write-time gate.
+- `uniword verify --xsd` no longer drowns in false errors for MCE
+  extension content every real document carries: the XSD layer
+  preprocesses each part per its own `mc:Ignorable` declaration,
+  stripping ignorable-marked attributes (w14:paraId/textId) and the
+  `mc:Ignorable` attribute itself before validation. Extension elements
+  are deliberately kept (removing them exposed baseline Word-vs-schema
+  quirks as false errors); their findings remain classified as
+  MCE-attributable
+
+### Changed
+
+- `DocumentRoot#remove_style` now returns the removed `Style` object or
+  nil (previously a Boolean), consistent with `#rename_style`; the
+  theme/style/scheme application methods on `DocumentRoot` moved into
+  `Wordprocessingml::DocumentStyling` (internal extraction, signatures
+  unchanged)
+- `Ooxml::Relationships::ImageRelationship.new` now requires an
+  explicit `id:` (the unreachable random rId default is gone;
+  `Docx::IdAllocator` remains the single rId authority)
+
 ## [1.3.0] - 2026-07-19
 
 ### Added

data/lib/uniword/builder/chart_builder.rb

@@ -133,16 +133,14 @@ module Uniword
       # @param root [Wordprocessingml::DocumentRoot] target document
       # @return [String] relationship id of the chart part
       def register_chart_part(root)
-        root.chart_parts ||= {}
         root.allocator ||= Docx::IdAllocator.new
 
         target = "charts/chart#{root.chart_parts.size + 1}.xml"
         r_id = root.allocator.alloc_rid(target: target, type: CHART_REL_TYPE)
 
-        root.chart_parts[r_id] = {
-          xml: build_xml,
-          target: target,
-        }
+        root.chart_parts[r_id] = Docx::ChartPart.new(
+          r_id: r_id, target: target, content: build_xml,
+        )
         r_id
       end
 

data/lib/uniword/builder/image_builder.rb

@@ -60,13 +60,13 @@ module Uniword
                        end
 
         if root
-          root.image_parts ||= {}
-          root.image_parts[r_id] = {
-            path: path,
+          root.image_parts[r_id] = Docx::ImagePart.new(
+            r_id: r_id,
             data: File.binread(path),
             content_type: content_type,
-            target: "media/#{File.basename(path)}",
-          }
+            target: target,
+            source_path: path,
+          )
         end
         r_id
       end

data/lib/uniword/document_factory.rb

@@ -169,38 +169,24 @@ module Uniword
         detector.detect(path)
       end
 
-
       # Copy package parts to document for round-trip preservation
       #
+      # Registry-driven: every PartDefinition naming both a package
+      # and a document attribute (and not excluded via
+      # +copy_to_document: false+) is mirrored; a part can no longer
+      # exist in only one direction.
+      #
       # @param package [Docx::Package] The source package
       # @param document [Wordprocessingml::DocumentRoot] The target document
       # @return [void]
       def copy_package_parts_to_document(package, document)
         return unless document.is_a?(Uniword::Wordprocessingml::DocumentRoot)
 
-        document.styles_configuration = package.styles if package.styles
-        document.numbering_configuration = package.numbering if package.numbering
-        document.settings = package.settings if package.settings
-        document.font_table = package.font_table if package.font_table
-        document.web_settings = package.web_settings if package.web_settings
-        document.theme = package.theme if package.theme
-        document.core_properties = package.core_properties if package.core_properties
-        document.app_properties = package.app_properties if package.app_properties
-        document.document_rels = package.document_rels if package.document_rels
-        document.theme_rels = package.theme_rels if package.theme_rels
-        document.package_rels = package.package_rels if package.package_rels
-        document.content_types = package.content_types if package.content_types
-        document.custom_properties = package.custom_properties if package.custom_properties
-        document.custom_xml_items = package.custom_xml_items if package.custom_xml_items
-        document.footnotes = package.footnotes if package.footnotes
-        document.endnotes = package.endnotes if package.endnotes
-        document.comments = package.comments if package.comments
-        document.settings_rels = package.settings_rels if package.settings_rels
-        if package.footnotes_rels
-          document.footnotes_rels = package.footnotes_rels
-        end
-        if package.endnotes_rels
-          document.endnotes_rels = package.endnotes_rels
+        Ooxml::PartRegistry.copied_to_document.each do |definition|
+          value = package.method(definition.package_attribute).call
+          next if value.nil?
+
+          document.method(:"#{definition.document_attribute}=").call(value)
         end
       end
 

data/lib/uniword/docx/custom_xml_item.rb

@@ -65,6 +65,14 @@ module Uniword
           .path_for(index: index)
       end
 
+      # Package paths this item emits (item part plus its properties
+      # part when present).
+      #
+      # @return [Array<String>] emitted package paths
+      def package_paths
+        [path, props_path].compact
+      end
+
       # Hash-style read compatibility.
       #
       # @param key [Symbol, String] one of :index, :xml_content,

data/lib/uniword/docx/image_part.rb

@@ -0,0 +1,83 @@
+# frozen_string_literal: true
+
+module Uniword
+  module Docx
+    # An image part (word/media/*) held by the package: binary image
+    # data plus the packaging metadata needed to emit it (relationship
+    # target, relationship id, content type).
+    #
+    # The image content type is per-file (resolved from the file
+    # extension), so it is always carried explicitly — the registry's
+    # :image definition contributes only the relationship type.
+    #
+    # Replaces the raw +{ data:, target:, content_type:, path: }+ hash
+    # entries formerly stored in +DocumentRoot#image_parts+. Hash-style
+    # read access (+part[:data]+, +part[:target]+, +part[:content_type]+,
+    # +part[:path]+) is kept for backward compatibility.
+    #
+    # @example
+    #   part = ImagePart.new(
+    #     r_id: "rId5", target: "media/image1.png",
+    #     data: binary_data, content_type: "image/png",
+    #   )
+    #   part.data         # => binary image data
+    #   part.content_type # => "image/png"
+    class ImagePart < Part
+      # @return [String, nil] path of the source file the image was
+      #   read from (builder-added images only; nil for loaded parts)
+      attr_accessor :source_path
+
+      # @param r_id [String, nil] relationship id
+      # @param target [String, nil] e.g. "media/image1.png"
+      # @param data [String, nil] binary image data
+      # @param content_type [String, nil] e.g. "image/png"
+      # @param source_path [String, nil] on-disk source path
+      # rubocop:disable Metrics/ParameterLists
+      def initialize(r_id: nil, target: nil, data: nil, content_type: nil,
+                     source_path: nil, **rest)
+        super(
+          definition: Ooxml::PartRegistry.find_by_key(:image),
+          r_id: r_id, target: target, content: data,
+          content_type: content_type, **rest
+        )
+        @source_path = source_path
+      end
+      # rubocop:enable Metrics/ParameterLists
+
+      # Wrap a legacy raw-hash entry ({ data:, target:, content_type:,
+      # path: }) into an ImagePart.
+      #
+      # @param hash [Hash] legacy hash entry
+      # @return [ImagePart]
+      def self.from_hash(hash)
+        new(target: hash[:target], data: hash[:data],
+            content_type: hash[:content_type], source_path: hash[:path])
+      end
+
+      # Binary image data (alias for content).
+      #
+      # @return [String, nil]
+      def data
+        content
+      end
+
+      # Set the binary image data (alias for content=).
+      #
+      # @param value [String, nil] binary image data
+      # @return [void]
+      def data=(value)
+        self.content = value
+      end
+
+      # Hash-style read compatibility (+:data+ and the legacy +:path+
+      # source path in addition to the Part keys).
+      def [](key)
+        case key.to_sym
+        when :data then content
+        when :path then source_path
+        else super
+        end
+      end
+    end
+  end
+end