idml 0.6.0 → 0.6.2

data/TODO.pdf/21-coordinate-transform.md

@@ -1,5 +1,10 @@
 # TODO PDF 21: Coordinate transform utility
 
+## Status: DONE — Centralized in `Idml::Geometry` (Y-flip + matrix math)
+and `Render::Placement` (item → PDF rect, with memoized
+`geometric_bounds`). All renderers use `Placement.box`. See
+`lib/idml/geometry.rb`, `lib/idml/render/placement.rb`, TODO 39.
+
 ## Goal
 
 Centralize IDML-to-PDF coordinate system conversion in a dedicated

data/TODO.pdf/22-color-resolution.md

@@ -1,5 +1,10 @@
 # TODO PDF 22: Color resolution from Resources
 
+## Status: DONE — `Render::ColorResolver` resolves `Color/...` and
+`Swatch/...` references from `Parts::Graphic` into RGB/CMYK tuples.
+`[None]` and `[Registration]` handled specially. See
+`lib/idml/render/color_resolver.rb`.
+
 ## Goal
 
 Resolve IDML color references (e.g., "Color/Red", "Color/c0m100y100k0")

data/TODO.pdf/23-fontmetrics-ps-name-fix.md

@@ -1,5 +1,10 @@
 # TODO PDF 23: FontMetrics PostScript name fix
 
+## Status: OBSOLETE — The original `FontMetrics` class (ttfunk-backed)
+was replaced by `TextEngine::PdfrbFontMetrics`, which delegates to
+pdfrb's typed Fonts API. PS name extraction is now pdfrb's
+responsibility. See TODOs 63, 65.
+
 ## Goal
 
 Fix the `FontMetrics#postscript_name` method to prefer platform 3

data/TODO.pdf/24-antipattern-coverage.md

@@ -1,5 +1,10 @@
 # TODO PDF 24: Anti-pattern coverage for render modules
 
+## Status: DONE — `spec/anti_patterns_spec.rb` auto-discovers every
+`.rb` file under `lib/` and asserts no `nokogiri`, `rexml`,
+`require_relative`, `instance_variable_set`, `.send(`, `respond_to?`,
+`double()`, hand-rolled serializers, etc. Currently 279 files, 0 offenses.
+
 ## Goal
 
 Ensure all new render and text-engine modules pass the anti-pattern

data/TODO.pdf/25-cli-render.md

@@ -1,5 +1,9 @@
 # TODO PDF 25: CLI render subcommand
 
+## Status: DONE — `idml render input.idml output.pdf` exists with
+`--font-path`, `--pdfa`, `--tagged`, `--no-subset`, `--compress`, and
+`--verbose` flags. See `lib/idml/cli.rb` and TODO 60.
+
 ## Goal
 
 Add a `render` subcommand to the `idml` CLI that converts an IDML

data/TODO.pdf/26-renderer-registry.md

@@ -1,5 +1,10 @@
 # TODO PDF 26: Page-item renderer registry (OCP)
 
+## Status: DONE — `PageItemRenderer::RENDERER_MAP` is a class-level
+Hash mapping element class → renderer class. Adding a new page-item
+type = appending to RENDERER_MAP, not editing a switch statement.
+See `lib/idml/render/page_item_renderer.rb`, `lib/idml/render/renderers/`.
+
 ## Goal
 
 Replace the `case/when` dispatch in `SpreadRenderer` with a registration-

data/TODO.pdf/27-text-engine-integration.md

@@ -1,5 +1,11 @@
 # TODO PDF 27: Text engine integration
 
+## Status: DONE — `TextFrameRenderer` runs each story through the
+text engine: `StyleResolver` extracts runs → `Shaper` shapes glyphs →
+`LineBreaker` wraps → `Justifier` aligns → `VerticalLayout` positions.
+Falls back to pdfrb's `text_rich` when metrics are missing. See
+`lib/idml/render/renderers/text_frame_renderer.rb` and TODO 67.
+
 ## Goal
 
 Wire the text engine (Shaper → LineBreaker → Justifier → VerticalLayout)

data/TODO.pdf/28-path-geometry-models.md

@@ -1,5 +1,11 @@
 # TODO PDF 28: Path geometry models
 
+## Status: DONE — `Idml::Elements::{PathPointType, PathPointArray,
+GeometryPathType, PathGeometry, Properties}` all carry the RNC
+attributes. Page items expose `geometric_bounds` via memoized helpers.
+See `lib/idml/elements/{path_point_type,path_point_array,
+geometry_path_type,path_geometry,properties}.rb`.
+
 ## Goal
 
 Create typed element models for IDML's path geometry hierarchy:

data/TODO.pdf/29-character-style-extraction.md

@@ -1,5 +1,10 @@
 # TODO PDF 29: Character style extraction
 
+## Status: DONE — `Render::StyleResolver` walks PSR → CSR and emits
+`StyledRun` structs (text + font + size + color + style attributes).
+Per-run font resolution is wired in `TextFrameRenderer#font_for_run`
+via `font_map`. See `lib/idml/render/style_resolver.rb` and TODO 93.
+
 ## Goal
 
 Extract font family, size, color, and style attributes from

data/TODO.pdf/30-shape-geometry-rendering.md

@@ -1,5 +1,11 @@
 # TODO PDF 30: Shape geometry rendering
 
+## Status: DONE — RectangleRenderer / PolygonRenderer / GraphicLineRenderer
+derive bounds from `PathGeometry` and emit pdfrb path calls.
+ItemTransform + Y-flip handled by `Placement.box`. See
+`lib/idml/render/renderers/{rectangle,polygon,graphic_line}_renderer.rb`
+and TODOs 21, 39.
+
 ## Goal
 
 Render Rectangle, Polygon, and GraphicLine shapes using their actual

data/TODO.pdf/31-stroke-rendering.md

@@ -1,5 +1,11 @@
 # TODO PDF 31: Stroke rendering
 
+## Status: DONE — `Render::StrokeStyle.apply(canvas, item)` saves
+graphics state, applies line_cap / line_join / miter_limit / dash
+patterns from DashedStrokeStyle / DottedStrokeStyle / StripedStrokeStyle,
+and yields. Used by every shape renderer. See
+`lib/idml/render/stroke_style.rb` and TODO 70.
+
 ## Goal
 
 Render stroke attributes on page items: stroke color, weight, end cap,

data/TODO.pdf/32-master-spread-rendering.md

@@ -1,5 +1,11 @@
 # TODO PDF 32: Master spread rendering
 
+## Status: DONE — `SpreadRenderer#render_master_items` resolves each
+Page's `AppliedMaster` via `Package#master_spread_by_id` and renders
+the master's items first as a background layer. Layer visibility
+applied via `LayerFilter`. See `lib/idml/render/spread_renderer.rb`,
+TODO 44.
+
 ## Goal
 
 Render master spread page items as background on each page. Master

data/TODO.pdf/33-font-resolution-from-idml.md

@@ -1,5 +1,12 @@
 # TODO PDF 33: Font resolution from IDML
 
+## Status: DONE — `Parts::Fonts` parses FontFamily → Font entries
+(`postscript_name` accessor). `FontReferenceResolver.build(package)`
+constructs the family → PSName lookup. `FontSetup` resolves to a file
+path via pdfrb's FontResolver. See TODOs 40, 93 and
+`lib/idml/render/font_reference_resolver.rb`,
+`lib/idml/render/font_setup.rb`.
+
 ## Goal
 
 Resolve fonts referenced in IDML documents using Resources/Fonts.xml.

data/TODO.pdf/34-story-threading.md

@@ -1,5 +1,10 @@
 # TODO PDF 34: Story threading
 
+## Status: DONE — `Render::StoryThreader` walks the
+PreviousTextFrame / NextTextFrame chain, distributes story text
+across linked frames, and guards against cycles. See
+`lib/idml/story_threader.rb`.
+
 ## Goal
 
 Follow the PreviousTextFrame/NextTextFrame chain to thread stories

data/TODO.pdf/35-pdf-metadata.md

@@ -1,5 +1,11 @@
 # TODO PDF 35: PDF metadata
 
+## Status: DONE — `Render::MetadataBuilder` reads XMP from
+`META-INF/metadata.xml` via `Parts::XmpMeta` (typed dc:/xmp:/rdf:
+namespace composition), merges with defaults (Producer, CreationDate),
+and writes via `PdfrbWriter#set_info`. See
+`lib/idml/render/metadata_builder.rb`, `lib/idml/parts/xmp.rb`, TODO 75.
+
 ## Goal
 
 Embed document metadata (Title, Author, Subject, Keywords, Creator,

data/TODO.pdf/36-performance-caching.md

@@ -1,5 +1,10 @@
 # TODO PDF 36: Performance caching
 
+## Status: DONE — `PdfrbFontMetrics` is reused across the whole render
+(a single pdfrb `Fonts` collection is shared). `Placement.box` memoizes
+`geometric_bounds` per page item. `ColorResolver` and `Package#part`
+memoize per instance. See TODO 86 for benchmarks.
+
 ## Goal
 
 Cache expensive operations: FontMetrics file parsing, FontResolver

data/TODO.pdf/37-round-trip-specs.md

@@ -1,5 +1,10 @@
 # TODO PDF 37: Round-trip specs for element models
 
+## Status: DONE — Every typed element class has round-trip specs under
+`spec/idml/elements/` and `spec/idml/parts/` (parse → serialize →
+xml-equivalent). The `sample-with-image` fixture's 14 parts are
+validated against the RNC schemas; see also TODO 20 in TODO.complete.
+
 ## Goal
 
 Add round-trip specs for every page-item element model: `from_xml(xml)`

data/TODO.pdf/38-pdf-output-validation.md

@@ -1,5 +1,10 @@
 # TODO PDF 38: PDF output validation
 
+## Status: DONE — pdfrb 0.6.0 emits spec-compliant xref (20-byte
+entries), valid object streams, and matching `endobj`s. Pipeline specs
+under `spec/idml/render/render_pdfrb_pipeline_spec.rb` round-trip
+the PDF through pdfrb's parser to confirm validity. See TODOs 54, 92.
+
 ## Goal
 
 Validate that the PDF files produced by the Pipeline are structurally

data/TODO.pdf/41-graphic-line-rendering.md

@@ -1,5 +1,10 @@
 # TODO PDF 41: GraphicLine rendering
 
+## Status: DONE — `Render::GraphicLineRenderer` (registered in
+`PageItemRenderer::RENDERER_MAP`) emits move-to / line-to via pdfrb
+Canvas. StrokeColor and StrokeWeight via `StrokeStyle.apply`. See
+`lib/idml/render/renderers/graphic_line_renderer.rb`.
+
 ## Goal
 
 Implement `GraphicLineRenderer` to render IDML `<GraphicLine>` elements as

data/TODO.pdf/42-group-transform-application.md

@@ -1,5 +1,10 @@
 # TODO PDF 42: Group transform application
 
+## Status: DONE — `Render::GroupRenderer` saves graphics state,
+applies the Group's ItemTransform via `canvas.concat_matrix`, renders
+children, and restores. Nested groups recurse correctly. See
+`lib/idml/render/renderers/group_renderer.rb`.
+
 ## Goal
 
 Apply a Group's ItemTransform to all child page items during rendering.

data/TODO.pdf/43-text-frame-insets.md

@@ -1,5 +1,11 @@
 # TODO PDF 43: Text frame insets
 
+## Status: DONE — `Elements::TextFramePreference` carries the inset
+attributes (InsetTop / InsetLeft / InsetBottom / InsetRight).
+`TextFrameRenderer` subtracts them from the frame bounds before
+layout. See `lib/idml/elements/text_frame_preference.rb`,
+`lib/idml/render/renderers/text_frame_renderer.rb`.
+
 ## Goal
 
 Respect `TextFramePreference` inset margins (InsetTop, InsetLeft,

data/TODO.pdf/44-layer-support.md

@@ -1,5 +1,11 @@
 # TODO PDF 44: Layer support
 
+## Status: DONE — `Render::LayerFilter.from_designmap(designmap)`
+constructs the layer visibility map from the designmap `<Layer>`
+entries. `SpreadRenderer` consults `LayerFilter#visible?(item)` for
+both master and spread items. See `lib/idml/render/layer_filter.rb`,
+`lib/idml/elements/layer.rb`.
+
 ## Goal
 
 Parse Layer definitions from designmap.xml. Each page item's

data/TODO.pdf/45-image-clipping.md

@@ -1,5 +1,9 @@
 # TODO PDF 45: Image clipping
 
+## Status: DONE — `SpreadRenderer#apply_image_clip` emits a pdfrb
+clip rectangle from the parent frame's `PathGeometry` before
+`draw_image_matrix`. See `lib/idml/render/spread_renderer.rb`.
+
 ## Goal
 
 Clip placed images to their parent Rectangle's bounds. Currently

data/TODO.pdf/46-spread-to-page-separation.md

@@ -1,5 +1,10 @@
 # TODO PDF 46: Spread-to-page separation
 
+## Status: DONE — `Pipeline#render_spread` iterates each Page in the
+Spread separately, calls `writer.add_page` with per-page dimensions,
+and tracks the global `page_index` for hyperlink/destination wiring.
+See `lib/idml/render/pipeline.rb` and TODO 19.
+
 ## Goal
 
 Render each IDML Page as a separate PDF page (currently one PDF page

data/TODO.pdf/47-pdf-bookmarks.md

@@ -1,5 +1,11 @@
 # TODO PDF 47: PDF bookmarks/outlines
 
+## Status: DONE — `Render::BookmarkResolver` walks the designmap
+bookmark entries and yields `(title, page_index)` tuples;
+`PdfrbWriter#add_bookmark` builds the Outlines tree. See
+`lib/idml/bookmark_resolver.rb`, `lib/idml/render/pdfrb_writer.rb`,
+and TODO 79.
+
 ## Goal
 
 Generate PDF outline (bookmark) entries from the document's XML

data/TODO.pdf/48-table-rendering.md

@@ -1,5 +1,12 @@
 # TODO PDF 48: Table rendering
 
+## Status: DONE — `Render::TableRenderer` handles both the
+schema-faithful layout (Table > {Cell, Row} siblings, real IDML) and
+the legacy nested model. SchemaLayout computes per-cell rects from
+`Row#single_row_height` and `Cell#col_row`. Inline Tables discovered
+via `CSR#table`. See `lib/idml/render/renderers/table_renderer.rb`,
+`lib/idml/elements/{table,cell,row}.rb`, and TODOs 82, 84.
+
 ## Goal
 
 Render IDML tables (TableRow, TableCell elements within Stories) as

data/TODO.pdf/49-gradient-fill-rendering.md

@@ -1,5 +1,11 @@
 # TODO PDF 49: Gradient fill rendering
 
+## Status: DONE — `Render::GradientResolver` maps IDML Gradient +
+GradientStop entries to pdfrb shading dictionaries
+(`Shadings#add_axial` / `add_radial`). RectangleRenderer / PolygonRenderer
+route gradient fills through this resolver. See
+`lib/idml/gradient_resolver.rb` and TODOs 66, 68.
+
 ## Goal
 
 Render gradient fills on page items. IDML supports linear and radial

data/TODO.pdf/50-pdf-a-compliance.md

@@ -1,5 +1,11 @@
 # TODO PDF 50: PDF/A compliance mode
 
+## Status: DONE — `Pipeline#apply_compliance` attaches the PDF/A-2a
+XMP packet (`Render::PdfaPacket`) and embeds the sRGB ICC profile
+(`Render::IccProfile`). Pipeline option `compliance: :pdfa2a`. CLI
+exposes `--pdfa`. See `lib/idml/render/{pdfa_packet,icc_profile}.rb`,
+TODOs 60, 77, 81.
+
 ## Goal
 
 Add a `--pdf-a` option that produces PDF/A-1a, PDF/A-2a, or PDF/A-3a

data/TODO.pdf/68-radial-gradient-shadings.md

@@ -1,5 +1,9 @@
 # TODO PDF 68: Radial gradient shadings via pdfrb 0.4.0
 
+## Status: DONE — RectangleRenderer dispatches on `Gradient#type`:
+`"Radial"` → `Shadings#add_radial`, otherwise → `Shadings#add_axial`.
+Spec covers both branches.
+
 ## Goal
 
 RectangleRenderer routes gradient fills to either `Shadings#add_axial`

data/TODO.pdf/69-transparency-blend-modes.md

@@ -1,5 +1,11 @@
 # TODO PDF 69: Transparency and blend modes via pdfrb 0.4.0
 
+## Status: DONE — `Render::Blending.wrap(canvas, setting, &block)`
+maps IDML `BlendingSetting` (BlendMode + Opacity) to pdfrb's
+`canvas.with_transparency(opacity:, blend_mode:)`. Used by every
+shape renderer. See `lib/idml/render/blending.rb`,
+`lib/idml/elements/blending_setting.rb`.
+
 ## Goal
 
 Honor IDML `<BlendingSetting BlendMode="..." Opacity="..."/>` on page

data/TODO.pdf/94-text-frame-vertical-layout.md

@@ -0,0 +1,63 @@
+# TODO PDF 94: TextFrameRenderer should delegate to VerticalLayout
+
+## Status: OPEN — gap identified in 2026-08-08 audit
+
+## Problem
+
+`TextEngine::VerticalLayout` exists (lib/idml/text_engine/vertical_layout.rb)
+with full support for insets, leading, paragraph spacing, and indents.
+`TextFrameRenderer#render_run_lines` reinvents that math inline and
+ignores every paragraph-level attribute:
+
+- No TextFramePreference insets (InsetTop / InsetLeft / InsetBottom /
+  InsetRight) subtracted from the layout box.
+- No SpaceBefore / SpaceAfter applied between paragraphs.
+- No FirstLineIndent / LeftIndent / RightIndent applied within paragraphs.
+- No AutoLeading honored — always uses `size * 1.2`.
+- No DropCapLines / DropCapCharacters.
+
+The current `engine_render` walks CSR runs as a flat list, so it
+can't even see paragraph boundaries (PSR is the paragraph boundary;
+CSR is a run within a paragraph).
+
+## What needs to happen
+
+1. **Group runs by paragraph** in `StyleResolver.extract_runs` —
+   return a list of `Paragraph` structs, each carrying:
+   - `runs:` — array of `StyledRun`
+   - `space_before:`, `space_after:` — from PSR
+   - `first_line_indent:`, `left_indent:`, `right_indent:` — from PSR
+   - `auto_leading:` — from PSR
+   - `alignment:` — already wired today
+2. **Replace `render_run_lines`** with a call to
+   `VerticalLayout.layout(lines:, frame:, font_size:, leading:,
+   space_before:, space_after:, first_line_indent:, left_indent:)`.
+3. **Subtract insets** from the frame box in `frame_box` —
+   read TextFramePreference InsetTop / InsetLeft / InsetBottom /
+   InsetRight and pass them to VerticalLayout::Frame.
+4. **Emit positioned glyphs** from VerticalLayout via
+   `canvas.text` per line (already the case today, but with
+   VerticalLayout-computed positions).
+
+## Architecture rationale
+
+- **DRY**: removes duplicate line-positioning logic.
+- **MECE**: VerticalLayout owns vertical placement; TextFrameRenderer
+  owns run extraction + canvas emission.
+- **OCP**: new paragraph attributes (drop caps, gyoudori) extend
+  VerticalLayout, not TextFrameRenderer.
+
+## Acceptance criteria
+
+- [ ] `StyleResolver.extract_runs` returns paragraphs (not flat runs)
+      when called with paragraph-aware mode.
+- [ ] TextFramePreference insets shrink the layout box.
+- [ ] PSR SpaceBefore adds space before the paragraph's first line.
+- [ ] PSR FirstLineIndent shifts the first line right.
+- [ ] PSR AutoLeading overrides the 1.2 factor.
+- [ ] All existing specs still pass.
+- [ ] New specs for each paragraph attribute.
+
+## Dependencies
+
+- None — pure refactor on top of existing typed models.

data/TODO.pdf/95-paragraph-spacing-indents.md

@@ -0,0 +1,50 @@
+# TODO PDF 95: Paragraph spacing and indents (SpaceBefore, SpaceAfter, FirstLineIndent)
+
+## Status: OPEN — gap identified in 2026-08-08 audit
+
+## Problem
+
+`ParagraphStyleRange` carries 284 attributes from the RNC schema.
+Today the renderer honors exactly one of them: `Justification`
+(via `StyleResolver::ALIGNMENT_MAP`).
+
+The following paragraph-level layout attributes are typed on the
+model but never read by the renderer:
+
+| Attribute               | Effect on layout                                  |
+|-------------------------|---------------------------------------------------|
+| `SpaceBefore`           | vertical gap above the paragraph                  |
+| `SpaceAfter`            | vertical gap below the paragraph                  |
+| `FirstLineIndent`       | left offset for the first line only               |
+| `LeftIndent`            | left offset for all lines                         |
+| `RightIndent`           | right offset (shrinks frame width)                |
+| `AutoLeading`           | multiplier (1.2 default) for line height          |
+| `DropCapLines`          | drop-cap height in lines                          |
+| `DropCapCharacters`     | drop-cap width in characters                      |
+| `LastLineIndent`        | right offset for the last line                    |
+| `MinimumWordSpacing`    | justification clamp                               |
+| `MaximumWordSpacing`    | justification clamp                               |
+| `DesiredWordSpacing`    | justification target                              |
+
+## What needs to happen
+
+1. Extend `StyledRun` (or wrap it in a `Paragraph` struct per TODO 94)
+   to carry these attributes from PSR.
+2. Wire `Justifier` to honor `Minimum/Maximum/DesiredWordSpacing`
+   when distributing slack in justified text.
+3. Wire `VerticalLayout` to honor `SpaceBefore` / `SpaceAfter` /
+   `AutoLeading` (already accepts them, but the renderer doesn't pass them).
+4. Wire `FirstLineIndent` / `LeftIndent` / `RightIndent` in VerticalLayout.
+
+## Acceptance criteria
+
+- [ ] `psr.space_before = 12` produces 12pt gap before the paragraph.
+- [ ] `psr.first_line_indent = 24` indents only the first line by 24pt.
+- [ ] `psr.left_indent = 18` indents all lines by 18pt.
+- [ ] `psr.right_indent = 18` shrinks the line-wrap width by 18pt.
+- [ ] `psr.auto_leading = 1.5` produces 1.5× line height.
+- [ ] Word-spacing clamps apply in justified mode.
+
+## Dependencies
+
+- TODO 94 (VerticalLayout delegation).

data/TODO.pdf/96-text-frame-insets-actual.md

@@ -0,0 +1,39 @@
+# TODO PDF 96: TextFramePreference insets honored in layout
+
+## Status: OPEN — gap identified in 2026-08-08 audit
+
+## Problem
+
+`Elements::TextFramePreference` carries the four text insets:
+
+- `InsetTop`
+- `InsetLeft`
+- `InsetBottom`
+- `InsetRight`
+
+`TextFrame#text_frame_preference` exposes the typed child. But
+`TextFrameRenderer#frame_box` reads only `Placement.box(frame, ...)`
+— it doesn't subtract insets from the box before laying out text.
+
+Result: text starts at the frame's geometric edge, ignoring the
+inset margin the document declared. Visually wrong for any frame
+that uses non-zero insets (which is most real-world IDML documents).
+
+## What needs to happen
+
+1. `frame_box` reads `frame.text_frame_preference` and extracts
+   `inset_top`, `inset_left`, `inset_bottom`, `inset_right`.
+2. Subtracts left/right from `box[:width]`, adds left to `box[:x]`.
+3. Subtracts top/bottom from `box[:height]`, adds bottom to `box[:y]`.
+4. Passes the resulting inset-aware box to VerticalLayout.
+
+## Acceptance criteria
+
+- [ ] TextFrame with InsetTop=20 lays first line 20pt below frame top.
+- [ ] TextFrame with InsetLeft=15 shifts first glyph 15pt right of frame left.
+- [ ] TextFrame with no TextFramePreference child renders as today (zero insets).
+- [ ] Spec: parse fixture TextFrame, verify insets honored.
+
+## Dependencies
+
+- TODO 94 (VerticalLayout delegation uses the inset-adjusted box).

data/TODO.pdf/97-table-cell-rendering-fidelity.md

@@ -0,0 +1,62 @@
+# TODO PDF 97: Table cell rendering — fill, stroke, insets, vertical justification
+
+## Status: OPEN — gap identified in 2026-08-08 audit
+
+## Problem
+
+`TableRenderer#render_cell_text` is a stub:
+
+```ruby
+def self.render_cell_text(canvas, cell, x, y, height, context)
+  text = cell.text_content
+  return if text.nil? || text.empty?
+
+  runs = [{
+    text: text,
+    font: context.font_ps_name,    # ignores cell CSR
+    size: DEFAULT_SIZE,            # 10.0 hardcoded
+  }]
+  baseline = y + (height / 2)      # ignores VerticalJustification
+  canvas.text_rich(runs, at: [x + INSET, baseline])  # INSET=4.0 hardcoded
+end
+```
+
+The Cell model (`Elements::Cell`) has:
+- `top_inset`, `left_inset`, `bottom_inset`, `right_inset` — text insets
+- `fill_color`, `fill_tint` — cell background
+- `vertical_justification` — Top/Center/Bottom/Justify
+- `paragraph_style_range` — typed runs with their own fonts/sizes
+
+None of these are honored today. Cells render as plain rectangles
+with default font at center-baseline with 4pt fixed inset.
+
+## What needs to happen
+
+1. **Cell background fill**: when `cell.fill_color` is set and not
+   "Color/None", resolve via ColorResolver and emit a `canvas.fill`
+   before the cell border.
+2. **Cell stroke**: emit `canvas.stroke` per cell using the cell's
+   top/left/bottom/right stroke weights from `Cell` (currently
+   Table-level TopBorderWeight only).
+3. **Per-cell insets**: read Cell's TextTopInset/TextLeftInset/
+   TextBottomInset/TextRightInset instead of the fixed `INSET = 4.0`.
+4. **Vertical justification**: position text per
+   `cell.vertical_justification` (Top/Center/Bottom/Justify).
+5. **Per-cell CSR styling**: walk `cell.paragraph_style_range` →
+   `character_style_range` and emit one `text_rich` call with each
+   run's font + size, instead of collapsing all text into a single
+   DEFAULT_SIZE run.
+6. **RotationAngle**: cells can be rotated (Cell#rotation_angle);
+   emit `canvas.concat_matrix` if non-zero.
+
+## Acceptance criteria
+
+- [ ] Cell with FillColor="Color/Blue" renders a blue background.
+- [ ] Cell with TextTopInset=8 starts text 8pt below cell top.
+- [ ] Cell with VerticalJustification="Top" puts text at cell top.
+- [ ] Cell with bold CSR run renders bold (per-run font).
+- [ ] Cell with PointSize=18 CSR uses 18pt text, not 10pt.
+
+## Dependencies
+
+- None — pure feature gap on existing models.

data/TODO.pdf/98-rowspan-columnspan.md

@@ -0,0 +1,35 @@
+# TODO PDF 98: RowSpan / ColumnSpan (merged cells) in TableRenderer
+
+## Status: OPEN — gap identified in 2026-08-08 audit
+
+## Problem
+
+`Elements::Cell` already carries `row_span` and `column_span`
+attributes (mapped from the RNC). The renderer's `SchemaLayout`
+iterates `col_count × row_count` cells and looks up each cell by
+`[col, row]` — it doesn't merge across spans.
+
+For a 2×2 table where cell `0:0` has `ColumnSpan=2`, the layout
+today emits a single cell-width rectangle at column 0 and an empty
+cell at column 1, instead of one rectangle spanning both columns.
+
+## What needs to happen
+
+1. `SchemaLayout#each_cell` reads `cell.column_span` and `cell.row_span`
+   and computes the rect width = `cell_w * column_span`, height =
+   `cell_h * row_span`.
+2. Skip cells that are "covered" by a spanning cell (i.e. when
+   `[col, row]` falls within another cell's span). Track a
+   coverage set per iteration.
+3. Spanning cell's text is rendered once, centered on the merged rect.
+
+## Acceptance criteria
+
+- [ ] Cell with ColumnSpan=2 renders a rect of width = 2 × column width.
+- [ ] Cell with RowSpan=3 renders a rect of height = 3 × row height.
+- [ ] Covered cells (within a span) don't render borders.
+- [ ] Spec: 2x2 grid with cell 0:0 spanning both columns.
+
+## Dependencies
+
+- TODO 97 (cell rendering fidelity).