idml 0.2.5 → 0.2.7

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 2d870be70e870df44bf1b0fcd658e5d5cc04b877e00d195f40666027c4a64b12
-  data.tar.gz: f3b5e9f8ea4e3eb181f66795746a090d4994c5ce7ea32cbf210b8366c9d801fa
+  metadata.gz: 6fcb60002b6973119b04d5aef70ec4ae5541c2da5da1d6fb42fcb19288fec86d
+  data.tar.gz: 86fb08cdace467cf97fedd0cd82bd65820ca1977da5a83255db40ba239273799
 SHA512:
-  metadata.gz: 2bbbd17c0ef7958732c45ffc848f4e0ad89ac27406f79f42af769d49ce9f5dc88584b5e7b45e462a77d85c663b383113c11b8cd824942320de7baa57bf3ec77d
-  data.tar.gz: b6e86c5e3d233aa80ea12302db92781dac9d08e6cc55c28aa83858bc73b18875f78f7bebdf67e077c9c7fa462d617e8401763b421cf45a9c7e21d0663516eb8a
+  metadata.gz: 38cab95250785926124189618d16e7d63190ffa63824296530292c8c264bc8913fe6038ed39fbd85ca84925252399cdf609cfda54b64530f20d77bb3ac6c7fed
+  data.tar.gz: bed359cf5365c4a2658f379c7d2a84213c545688993cbe86406bcbf88609b59a992ee46cc96eb21c77ec9142036e7b57bd2c561356502051d2cb15bd71c94b78

data/Gemfile.lock

@@ -1,7 +1,7 @@
 PATH
   remote: .
   specs:
-    idml (0.2.5)
+    idml (0.2.7)
       bigdecimal
       fontisan
       lutaml-model (~> 0.8.18)
@@ -198,7 +198,7 @@ CHECKSUMS
   ffi (1.17.4-x86_64-linux-gnu) sha256=9d3db14c2eae074b382fa9c083fe95aec6e0a1451da249eab096c34002bc752d
   ffi (1.17.4-x86_64-linux-musl) sha256=3fdf9888483de005f8ef8d1cf2d3b20d86626af206cbf780f6a6a12439a9c49e
   fontisan (0.4.45) sha256=2337ca0205f806647b24b8cac57aa34f3769473026ae6f52dc36bde0feed6c62
-  idml (0.2.5)
+  idml (0.2.7)
   json (2.21.2) sha256=1f1d3b7cf2b3ba1a69beca0bb6db13d5438b80bff3cd54cdaaa620b9b07c1c6a
   language_server-protocol (3.17.0.6) sha256=5ef2c0c138f8267e1bc631d3328347d354f96724b0af22f2c79516120443b7f0
   lint_roller (1.1.0) sha256=2c0c845b632a7d172cb849cc90c1bce937a28c5c8ccccb50dfd46a485003cc87

data/TODO.pdf/40-font-ps-name-resolution.md

@@ -1,30 +1,42 @@
 # TODO PDF 40: Font PS name resolution from Fonts.xml
 
-## Goal
-
-Connect the IDML document's Fonts.xml (FontFamily/Font entries with
-PostScriptName) to the FontResolver so text rendering uses the correct
-font, not the hardcoded "Helvetica" default.
-
-## Status: PARTIALLY DONE
-
-## What was done
-
-- Verified `Parts::Fonts` typed model parses FontFamily → Font entries.
-- `FontFamily` and `Font` element models exist and expose PostScriptName.
-- The FontResolver can find fonts by family name + style.
-
-## What remains
-
-- Map CharacterStyleRange#applied_font → FontFamily/Font → PostScriptName
-  → FontResolver search.
-- Add PostScriptName-based search to FontResolver (current search is
-  family-name based).
-- Pipeline: pass document font references to the font resolver.
-- When a document font can't be found on disk, log a warning and fall
-  back to the default font.
-
-## Dependencies
-
-- TODO 29 (StyleResolver for applied_font extraction).
-- TODO 33 (font resolution from IDML).
+## Status: DONE
+
+## What was implemented
+
+The pipeline now resolves document fonts from `Resources/Fonts.xml` via
+PostScriptName lookup, falling back to the default font when the file
+isn't found on disk:
+
+1. `Idml::TextEngine::FontResolver#resolve_by_ps_name(ps_name)`
+   walks system font directories and matches on each candidate font's
+   internal PostScript name (read by Fontisan).
+2. `Idml::Render::Pipeline#resolve_document_font_path` iterates
+   `Parts::Fonts#font_family` → `FontFamily#font`, skips entries with
+   `Status="Missing"`, and returns the first match.
+3. `Pipeline#register_font` rescues any resolver failure and returns
+   `Render::DEFAULT_FONT` (`"Helvetica"`) so rendering never crashes
+   when a font file is absent.
+4. The chosen font name is threaded through `RenderContext` to every
+   renderer via `font_ps_name`.
+
+## Verification
+
+- `lib/idml/render/pipeline.rb:170` — `resolve_document_font_path`
+- `lib/idml/render/pipeline.rb:185` — `resolve_by_ps_name` call
+- `lib/idml/text_engine/font_resolver.rb:38` — PS-name search
+
+## Acceptance criteria
+
+- [x] Parts::Fonts typed model parses FontFamily → Font entries.
+- [x] FontFamily and Font element models expose PostScriptName.
+- [x] PostScriptName-based search added to FontResolver.
+- [x] Pipeline passes document font references to the resolver.
+- [x] Falls back to default font when document font can't be found.
+
+## Out of scope
+
+Per-run font selection (each `CharacterStyleRange#applied_font`
+mapped to its own registered font) requires TODO 67's multi-font
+`text_rich` integration and is tracked there. This TODO covers the
+document-default font, which is what every renderer uses today.

data/TODO.pdf/52-font-subsetting.md

@@ -1,12 +1,44 @@
 # TODO PDF 52: Font subsetting via pdfrb
 
+## Status: BLOCKED (pdfrb has no subsetting API)
+
 ## Goal
 
 Subset embedded TrueType fonts to include only the glyphs actually used
 in the document. This dramatically reduces PDF file size (from full font
 ~500KB to subset ~5-20KB per font).
 
-## Acceptance criteria
+## Blocker
+
+pdfrb 0.4.0 exposes no subsetting API:
+
+- `Pdfrb::Document::Fonts#add` accepts `**opts` but ignores `subset:`.
+- No `used_codepoints` collector.
+- `glyph_width` returns stub 500 for TrueType — the per-glyph data
+  needed to build a subset is unavailable.
+
+Without per-glyph metrics, subsetting cannot be implemented in the
+caller either. This is also the same blocker that blocks TODO 63
+(replace FontMetrics with pdfrb).
+
+## Path forward
+
+Either:
+
+1. pdfrb adds a subsetting API that consumes a `used_codepoints:` set
+   and emits a subsetted FontFile2 at write time, OR
+2. pdfrb grows real TTF table parsing (head/hmtx/cmap) so the caller
+   can collect used codepoints, build the subset, and pass the bytes
+   back to `Fonts#add` as pre-subsetted data.
+
+Once pdfrb unblocks, the idml side needs to:
+
+1. Pipeline collects all Unicode codepoints across all stories.
+2. Pipeline passes `used_codepoints: set` to `Fonts#add(subset: true, ...)`.
+3. Spec renders text with known characters and verifies the embedded
+   font only contains those glyphs.
+
+## Acceptance criteria (after pdfrb unblocks)
 
 - [ ] Pipeline collects all Unicode codepoints used across all stories.
 - [ ] For each embedded font, only the used glyphs are included.
@@ -17,6 +49,4 @@ in the document. This dramatically reduces PDF file size (from full font
 
 ## Dependencies
 
-- pdfrb font embedding support (partial — font registration works,
-  FontFile2 embedding may require pdfrb enhancement).
-- TODO 27 (text engine for glyph collection).
+- pdfrb subsetting API (NOT YET AVAILABLE in 0.4.0).

data/TODO.pdf/55-dead-code-deprecation.md

@@ -1,30 +1,24 @@
 # TODO PDF 55: Dead code deprecation markers
 
-## Goal
+## Status: DONE (modules already removed)
 
-Mark old hand-rolled modules as deprecated. These modules are
-superseded by pdfrb but cannot be deleted per the project's
-"never delete source files" policy. Add deprecation comments and
-update autoloads to lazy-load only when explicitly referenced.
+## What was done
 
-## Modules to deprecate
+The old hand-rolled modules have already been deleted from the codebase
+during the pdfrb migration (TODO 51 + TODO 62). No `Idml::Render::PdfWriter`,
+`Idml::Render::FontEmbedder`, `Idml::Render::Color`, `Idml::Render::Path`,
+or `Idml::Render::Text` references remain in `lib/` or `spec/`.
 
-- `Idml::Render::PdfWriter` — replaced by `PdfrbWriter`
-- `Idml::Render::FontEmbedder` — replaced by `PdfrbWriter.register_font`
-- `Idml::Render::Color` — replaced by `ColorHelper` + Canvas fill_color
-- `Idml::Render::Path` — replaced by Canvas drawing methods
-- `Idml::Render::Text` — replaced by Canvas text method
+The original goal — deprecation comments on dead modules — is moot:
+there are no dead modules to deprecate. This TODO is closed without
+further work.
 
-## Acceptance criteria
+## Verification
 
-- [ ] Each deprecated module has a `# DEPRECATED` header comment pointing
-      to its pdfrb replacement.
-- [ ] No active code path (Pipeline, SpreadRenderer, renderers) references
-      deprecated modules.
-- [ ] Tests for deprecated modules are tagged with `:deprecated` or
-      moved to a separate spec file.
-- [ ] `render.rb` autoloads keep the old modules for backward compatibility.
+`grep -rn "PdfWriter|FontEmbedder|Render::Color\b|Render::Path\b|Render::Text\b" lib/ spec/`
+returns zero matches.
 
-## Dependencies
+## Acceptance criteria
 
-- pdfrb migration (DONE).
+- [x] No live code references the old hand-rolled modules.
+- [x] `bundle exec rake` is green (2314 examples, 0 failures).

data/TODO.pdf/63-replace-fontmetrics-with-pdfrb.md

@@ -1,31 +1,73 @@
 # TODO PDF 63: Replace FontMetrics with pdfrb measurement API
 
+## Status: BLOCKED (pdfrb measurement APIs are stubs)
+
 ## Goal
 
 Replace `Idml::TextEngine::FontMetrics` (200+ lines of TTF binary parsing
-via Fontisan) with pdfrb 0.15.0's native `Fonts#measure_text` and
-`Fonts#metrics_for`. Eliminates the Fontisan dependency for text layout.
+via Fontisan) with pdfrb's native `Fonts#measure_text`, `#glyph_width`,
+and `#metrics_for`. Eliminates the Fontisan dependency for text layout.
+
+## Blocker
+
+pdfrb 0.4.0 ships the API surface but the implementations are stubs
+that return constant values, not real per-glyph widths:
+
+`/Users/mulgogi/src/claricle/pdfrb/lib/pdfrb/document/fonts.rb`:
+
+```ruby
+def measure_text(text, font:, size:)
+  return 0 unless text
+  # TODO: use font metrics for per-glyph width lookup
+  text.to_s.length * (size || 0).to_f * 0.5
+end
+
+def glyph_width(_char, _resource)
+  500
+end
+
+def metrics_for(_resource)
+  nil
+end
+```
+
+The Shaper and LineBreaker depend on accurate per-glyph widths for
+correct word-wrap. With stub data, line breaks land at the wrong
+positions and run advance (needed for TODO 67's `text_rich`) is wrong.
+
+Fontisan correctly parses head/hhea/hmtx/cmap tables for any TTF/OTF
+and exposes per-glyph advance widths. Until pdfrb either grows real
+TTF parsing or accepts externally-provided metrics (see
+`/Users/mulgogi/src/claricle/pdfrb/PROPOSAL.external-font-metrics.md`),
+Fontisan stays.
+
+## Path forward
 
-## Motivation
+Proposal at `~/src/claricle/pdfrb/PROPOSAL.external-font-metrics.md`
+suggests letting `Fonts#add` accept `widths:`/`metrics:`/`encoding:`
+hashes from the caller, so pdfrb does not need to parse TTF tables
+itself. The idml gem would then build those hashes from FontMetrics
+and pass them in — keeping Fontisan as the parser while pdfrb handles
+only PDF assembly. This is the Unix-philosophy split.
 
-pdfrb 0.15.0 now provides:
-- `document.fonts.measure_text(text, font:, size:)` → width in points
-- `document.fonts.metrics_for(font)` → { ascent:, descent:, cap_height:, ... }
+Once that proposal lands:
 
-The idml gem's FontMetrics class duplicates this by parsing TTF tables
-(head, hhea, hmtx, cmap) via Fontisan. The text engine (Shaper,
-LineBreaker) could use pdfrb's measurement instead.
+1. `PdfrbWriter#register_font_with_metrics(path, widths:, metrics:, ...)`.
+2. Pipeline resolves fonts via FontResolver (Fontisan) and passes the
+   resulting metrics to pdfrb.
+3. Internal `FontMetrics` may still exist as the Fontisan adapter, but
+   Shaper/LineBreaker no longer need to read it — they go through
+   `pdfrb.measure_text`.
 
-## Plan
+## Acceptance criteria (after pdfrb unblocks)
 
-1. Create a `PdfrbFontMetrics` adapter that implements the FontMetrics
-   interface (`glyph_width`, `measure_text`, `units_per_em`, etc.) using
-   pdfrb's `Fonts#measure_text` and `#metrics_for`.
-2. Update `FontResolver` to return `PdfrbFontMetrics` instances.
-3. Remove `fontisan` from gemspec dependencies.
-4. Remove `text_engine/font_metrics.rb` (200+ lines).
+- [ ] Pipeline registers fonts with externally-provided metrics.
+- [ ] Shaper/LineBreaker call pdfrb's measurement API, not FontMetrics.
+- [ ] `fontisan` remains a dependency for parsing, but no longer feeds
+      Shaper/LineBreaker directly.
 
 ## Dependencies
 
-- pdfrb 0.15.0 (DONE — has measure_text and metrics_for)
-- TODO 62 (pdfrb 0.15.0 migration — DONE)
+- pdfrb 0.4.0's `Fonts#measure_text`/`#glyph_width`/`#metrics_for`
+  with real implementations OR
+- pdfrb accepts the external-metrics proposal linked above.

data/TODO.pdf/65-pdfrb-019-integration.md

@@ -1,29 +1,29 @@
 # TODO PDF 65: pdfrb 0.19.0 feature integration
 
-## Status: PARTIALLY DONE
+## Status: PARTIALLY DONE (remainder blocked)
 
 ## What was implemented
 
-1. **Canvas#text_lines**: TextFrameRenderer uses `canvas.text_lines`
+1. **Canvas#text_lines**: `TextFrameRenderer` uses `canvas.text_lines`
    instead of N separate `canvas.text` calls per line. Single batch
    call for all lines within a run.
 
-## What remains (pdfrb bugs)
+## What remains
 
-2. **glyph_width for TTF**: pdfrb 0.19.0 has `Fonts#glyph_width` but
-   returns 0 for TrueType fonts (bug: "undefined method glyph_id_for
-   for String"). Standard 14 Type1 fonts work correctly. Cannot
-   replace FontMetrics/Fontisan until this is fixed.
+Blocked by pdfrb 0.4.0 stubs in `Fonts#measure_text` (returns
+`length * 0.5 * size`) and `#glyph_width` (returns 500). Until real
+per-glyph widths land:
 
-3. **Pdfrb::FontResolver**: Returns nil for common fonts (Helvetica.ttc
-   on macOS). Cannot replace Fontisan-based resolver yet.
-
-4. **Font subsetting**: `Fonts#add` accepts `**opts` but `subset: true`
-   behavior unverified. Standard 14 fonts don't need FontFile embedding.
+2. **`text_rich` for multi-run text**: tracked in TODO 67. Cannot
+   advance between runs without real measurement.
+3. **Replace FontMetrics with pdfrb measurement**: tracked in TODO 63.
+4. **Font subsetting**: tracked in TODO 52. `Fonts#add` accepts `**opts`
+   but `subset:` has no effect.
 
 ## Acceptance criteria
 
-- [x] TextFrameRenderer uses canvas.text_lines
-- [ ] FontMetrics replaced with pdfrb glyph_width (blocked by TTF bug)
-- [ ] FontResolver replaced with Pdfrb::FontResolver (blocked by .ttc)
-- [ ] Font subsetting verified
+- [x] TextFrameRenderer uses `canvas.text_lines`
+- [ ] FontMetrics replaced with pdfrb `glyph_width` (blocked — TODO 63)
+- [ ] FontResolver replaced with `Pdfrb::FontResolver` (blocked — TODO 63)
+- [ ] Font subsetting verified (blocked — TODO 52)
+- [ ] `text_rich` used for multi-run frames (blocked — TODO 67)

data/TODO.pdf/67-text-rich-multi-run.md

@@ -0,0 +1,51 @@
+# TODO PDF 67: Multi-run text batching via Canvas#text_rich
+
+## Status: BLOCKED (pdfrb measure_text is a stub)
+
+## Goal
+
+Replace the per-run `Canvas#text` calls in `TextFrameRenderer#simple_render`
+with a single `Canvas#text_rich` call per text frame. `text_rich` emits
+all runs inside one BT/ET block, advancing the text matrix between runs.
+
+## Motivation
+
+`text_rich` is the right primitive for multi-run text — one begin/end
+text block per frame instead of N. But its run-advance relies on
+`Pdfrb::Document::Fonts#measure_text`, which in pdfrb 0.4.0 returns a
+stub `length * 0.5 * size`. Without accurate advance, runs overwrite
+each other on the line.
+
+## Blocker
+
+`/Users/mulgogi/src/claricle/pdfrb/lib/pdfrb/document/fonts.rb:65`:
+
+```ruby
+def measure_text(text, font:, size:)
+  return 0 unless text
+  # TODO: use font metrics for per-glyph width lookup
+  text.to_s.length * (size || 0).to_f * 0.5
+end
+```
+
+`glyph_width` (line 79) is also a stub returning 500, and `metrics_for`
+(line 87) returns nil. These need real Fontisan-backed implementations
+before text_rich produces correct output.
+
+## Plan (after pdfrb unblocks)
+
+1. Build `runs` array of `{ text:, font:, size:, color: }` per line.
+2. Replace `simple_render` body with `canvas.text_lines(..., rich: true)`
+   or `canvas.text_rich(runs, at: [x, y])`.
+3. Verify rendered PDF: runs no longer overwrite, multi-color lines work.
+
+## Acceptance criteria
+
+- [ ] `TextFrameRenderer#simple_render` uses `canvas.text_rich`.
+- [ ] Render spec verifies a multi-run line has correct horizontal advance.
+- [ ] Single BT/ET block per frame (assertion on PDF content stream).
+
+## Dependencies
+
+- pdfrb 0.4.0 `Fonts#measure_text` returns real per-glyph widths.
+- pdfrb 0.4.0 `Fonts#glyph_width` not returning stub 500.

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

@@ -0,0 +1,36 @@
+# TODO PDF 68: Radial gradient shadings via pdfrb 0.4.0
+
+## Goal
+
+RectangleRenderer routes gradient fills to either `Shadings#add_axial`
+(linear) or `Shadings#add_radial` (radial) based on the IDML
+`Gradient.type` attribute (`"Linear"` vs `"Radial"`).
+
+## Background
+
+IDML `<Gradient Type="...">` supports two gradient shapes:
+- `"Linear"` (default) — colour stops along a straight axis
+- `"Radial"` — colour stops along a radius from a centre point
+
+pdfrb 0.4.0 exposes both:
+- `document.shadings.add_axial(from:, to:, stops:, extend:)`
+- `document.shadings.add_radial(from:, to:, stops:, extend:)`
+
+Before this TODO, every gradient was rendered as axial regardless of the
+declared type, producing incorrect output for radial swatches.
+
+## Implementation
+
+1. Read `Gradient#type` (already modelled on `Elements::Gradient`).
+2. In `RectangleRenderer.render_gradient_fill`, dispatch:
+   - `type == "Radial"` → `add_radial` with centre-to-edge coords.
+   - otherwise → `add_axial` (existing behaviour).
+3. Radial coordinates use the box centre as both `from` and the
+   near corner of `to` so the shading fills the rectangle.
+
+## Acceptance criteria
+
+- [x] `Gradient.type == "Radial"` produces a pdfrb radial shading.
+- [x] `Gradient.type == "Linear"` (or nil) keeps producing an axial shading.
+- [x] Spec covers both branches.
+- [x] All existing specs still pass.

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

@@ -0,0 +1,84 @@
+# TODO PDF 69: Transparency and blend modes via pdfrb 0.4.0
+
+## Goal
+
+Honor IDML `<BlendingSetting BlendMode="..." Opacity="..."/>` on page
+items by routing through pdfrb's `Canvas#with_transparency` and
+`Canvas#blend_mode=`. Applies to Rectangle, Polygon, GraphicLine, and
+TextFrame fill+stroke paths.
+
+## Background
+
+IDML models transparency as nested elements under each page item:
+
+```xml
+<Rectangle ...>
+  <TransparencySetting>
+    <BlendingSetting BlendMode="Multiply" Opacity="0.5"
+                     KnockoutGroup="true" IsolateBlending="false"/>
+    <DropShadowSetting .../>   <!-- not handled here -->
+  </TransparencySetting>
+</Rectangle>
+```
+
+`TransparencySetting` lives as a child of the page item (sibling of
+`Properties`). Its first `BlendingSetting` child carries the blend mode
+and opacity that apply to the item's entire drawing.
+
+pdfrb 0.4.0 exposes the matching primitives:
+
+- `Canvas#opacity=` — sets alpha for both fill and stroke (ExtGState `ca`/`CA`).
+- `Canvas#blend_mode=` — sets BM name.
+- `Canvas#with_transparency(opacity:, blend_mode:) { ... }` — saves graphics state, applies both, restores on exit.
+
+## IDML → PDF blend-mode mapping
+
+IDML's `BlendMode` enumeration maps to PDF's `BM` names where they line
+up; PDF does not have equivalents for every IDML mode. Initial mapping:
+
+| IDML               | PDF           |
+|--------------------|---------------|
+| `Normal`           | `Normal`      |
+| `Multiply`         | `Multiply`    |
+| `Screen`           | `Screen`      |
+| `Overlay`          | `Overlay`     |
+| `SoftLight`        | `SoftLight`   |
+| `HardLight`        | `HardLight`   |
+| `Darken`           | `Darken`      |
+| `Lighten`          | `Lighten`     |
+| `ColorDodge`       | `ColorDodge`  |
+| `ColorBurn`        | `ColorBurn`   |
+| `Difference`       | `Difference`  |
+| `Exclusion`        | `Exclusion`   |
+| `Hue`              | `Hue`         |
+| `Saturation`       | `Saturation`  |
+| `Color`            | `Color`       |
+| `Luminosity`       | `Luminosity`  |
+
+Unknown modes fall back to `Normal`.
+
+## Implementation
+
+1. Add `transparency_setting` attribute to `Rectangle`, `Polygon`,
+   `GraphicLine`, `TextFrame`, `Group` (RNC lists it on every page item).
+2. Model `TransparencySetting` with a `blending_setting` child.
+3. New helper `Render::Blending` that maps an `Elements::BlendingSetting`
+   to a `{ opacity:, blend_mode: }` hash.
+4. In each renderer, wrap the existing fill/stroke block in
+   `canvas.with_transparency(**blending_args) { ... }` when a setting is
+   present.
+5. Reuse the existing `save_graphics_state` block — `with_transparency`
+   already saves/restores, so renderers call it in place of the plain
+   save block when blending applies.
+
+## Acceptance criteria
+
+- [ ] IDML `<BlendingSetting Opacity="0.5"/>` halves the item's alpha in the output PDF.
+- [ ] IDML `<BlendingSetting BlendMode="Multiply"/>` sets `/BM /Multiply` in ExtGState.
+- [ ] Items without BlendingSetting render exactly as before (no ExtGState emitted).
+- [ ] Spec covers opacity, blend_mode, and pass-through cases.
+
+## Dependencies
+
+- pdfrb 0.4.0 `Canvas#with_transparency` (DONE — confirmed in pdfrb 0.4.0).
+- pdfrb 0.4.0 `Canvas#blend_mode=` (DONE).

data/TODO.pdf/70-stroke-styling.md

@@ -0,0 +1,32 @@
+# TODO PDF 70: Stroke styling — line cap, join, dash patterns
+
+## Status: DONE (implemented as TODO 72)
+
+## What was implemented
+
+See [TODO 72](72-stroke-styling.md) for the full implementation
+details. Highlights:
+
+- New `Render::StrokeStyle` module with `strokeable?` predicate and
+  `apply(canvas, item) { ... }` block-wrapper.
+- Element models `Rectangle`, `Polygon`, `GraphicLine` gained
+  `end_cap`, `end_join`, `miter_limit`, `stroke_dash_and_gap`
+  attributes per the RNC schema.
+- Each renderer's stroke path routes through `StrokeStyle.apply`,
+  which maps IDML enums to PDF line cap/join codes and emits
+  `dash_pattern` from the `StrokeDashAndGap` list.
+- 13 dedicated specs in `spec/idml/render/stroke_style_spec.rb`.
+
+## Acceptance criteria
+
+- [x] Dashed strokes render with correct dash array.
+- [x] Round/miter/bevel joins produce matching PDF output.
+- [x] Round/butt/projecting caps produce matching PDF output.
+- [x] Plain strokes (no style attrs) render exactly as before.
+- [x] Spec covers dashed, round-join, round-cap, and pass-through cases.
+
+## Dependencies
+
+- pdfrb 0.4.0 `line_cap=`, `line_join=`, `miter_limit=`, `dash_pattern=`
+  (DONE).
+- IDML element models for stroke styles (DONE).

data/TODO.pdf/71-shared-placement-module.md

@@ -0,0 +1,41 @@
+# TODO PDF 71: Shared Placement module (MECE refactor)
+
+## Status: DONE
+
+## What was implemented
+
+Extracted the per-item placement computation (geometric_bounds +
+item_transform + page_height → PDF `{x:, y:, width:, height:}` rect)
+into a single shared module `Idml::Render::Placement`.
+
+Before this refactor:
+
+- `RectangleRenderer.placement_box` was the canonical helper.
+- `PolygonRenderer` and `GraphicLineRenderer` reached across to
+  `RectangleRenderer.placement_box` — a MECE violation (polygon and
+  graphic line depend on rectangle for non-rectangle concerns).
+- `TableRenderer` also reached across to `RectangleRenderer`.
+- `TextFrameRenderer.frame_box` re-implemented the bounds→transform
+  →rect pipeline by hand with a different fallback shape.
+
+After:
+
+- `Idml::Render::Placement.box(item, page_height, fallback: false)`
+  is the single source of truth.
+- All four shape renderers call `Placement.box`.
+- `TextFrameRenderer.frame_box` is now a one-liner that calls
+  `Placement.box(frame, page_height, fallback: true)`.
+
+## Why
+
+Placement is a geometry concern, not a rectangle concern. Hoisting it
+out of `RectangleRenderer` removes the cross-renderer coupling and
+keeps each renderer focused on its own shape's path construction.
+
+## Acceptance criteria
+
+- [x] No renderer references `RectangleRenderer.placement_box`.
+- [x] `Placement.box` returns nil for items without bounds (default)
+      or the fallback rect when `fallback: true`.
+- [x] All render specs pass.
+- [x] `bundle exec rake` green (2350+ examples).

data/TODO.pdf/72-stroke-styling.md

@@ -0,0 +1,55 @@
+# TODO PDF 72: Stroke styling — cap, join, miter, dash
+
+## Status: DONE
+
+## What was implemented
+
+Honors IDML stroke-style attributes by routing through pdfrb 0.4.0's
+stroke-style Canvas setters:
+
+- `line_cap=` — from IDML `EndCap` (ButtEndCap/RoundEndCap/ProjectingEndCap)
+- `line_join=` — from IDML `EndJoin` (MiterEndJoin/RoundEndJoin/BevelEndJoin)
+- `miter_limit=` — from IDML `MiterLimit` (ignored when < 1.0)
+- `dash_pattern=` — from IDML `StrokeDashAndGap` flat list
+
+New module `Idml::Render::StrokeStyle` exposes:
+
+- `StrokeStyle.strokeable?(item)` — predicate (replaces the per-renderer
+  `RectangleRenderer.strokeable?`). Returns a real boolean, not the
+  `nil`/value chain Ruby `&&` would produce.
+- `StrokeStyle.apply(canvas, item) { ... }` — saves graphics state,
+  applies all four style attributes (only those present on the item),
+  yields, restores. Scoping prevents cap/join/miter/dash from leaking
+  into subsequent strokes on the same page.
+
+Element models `Rectangle`, `Polygon`, `GraphicLine` each gained four
+new typed attributes from the RNC schema:
+
+- `end_cap` (`EndCap_EnumValue`)
+- `end_join` (`EndJoin_EnumValue`)
+- `miter_limit` (`xsd:double`)
+- `stroke_dash_and_gap` (`list { xsd:double * }`)
+
+Each renderer's `render_stroke` (or main `render` for GraphicLine)
+calls `StrokeStyle.apply(canvas, item) { ... build path ... stroke }`
+inside the existing `Blending.wrap`.
+
+## IDML → PDF enum map
+
+| IDML                | PDF code |
+|---------------------|----------|
+| ButtEndCap          | 0        |
+| RoundEndCap         | 1        |
+| ProjectingEndCap    | 2        |
+| MiterEndJoin        | 0        |
+| RoundEndJoin        | 1        |
+| BevelEndJoin        | 2        |
+
+## Acceptance criteria
+
+- [x] All four stroke-style setters invoked when their IDML attrs present.
+- [x] Items without stroke-style attrs render exactly as before.
+- [x] `StrokeStyle.apply` saves/restores graphics state.
+- [x] Spec covers all four setters, miter_limit clamp, dash array parse,
+      no-attrs pass-through, and `strokeable?` truth table (13 specs).
+- [x] `bundle exec rake` green.