ea 0.4.0 → 0.5.1

data/TODO.complete/51-remove-respond-to.md

@@ -0,0 +1,25 @@
+# TODO.complete/51: Remove `respond_to?` from lib code
+
+## Status: open
+
+The user's rule: "never use respond_to (poor typing)". Five call
+sites violate it, all added in recent sessions:
+
+- lib/ea/svg/ea_emitter/element/stereotype_icon_renderer.rb:61
+- lib/ea/lint/rules/missing_stereotype.rb:29
+- lib/ea/query/builder.rb:66
+- lib/ea/ocl/evaluator.rb:41, 52
+
+## Plan
+
+Replace `respond_to?(:x)` with explicit type checks via `is_a?`:
+
+- For `stereotype_refs` checks: `classifier.is_a?(Ea::Model::Classifier)`
+  (Classifier is the only type that owns stereotype_refs).
+- For `each` checks: `collection.is_a?(Array)` (collections are
+  always Arrays in our model; nil-safe via early return).
+
+## Acceptance
+
+- `grep -rn "respond_to?" lib/ea/` returns 0 hits.
+- Existing specs still pass.

data/TODO.complete/52-xsd-decouple-fixtures.md

@@ -0,0 +1,23 @@
+# TODO.complete/52: Decouple XSD Generator from fixture paths
+
+## Status: open
+
+`Ea::Export::Xsd::Generator` hardcodes:
+    DEFAULT_MAPPING_PATH = "spec/fixtures/mdg/ea_config/gml/GMLClassMapping.xml"
+
+Lib code should never reference `spec/` paths. The CLI layer should
+load fixtures (it knows about the dev environment) and pass the
+parsed data to the Generator.
+
+## Plan
+
+1. Generator's `initialize` defaults to empty `ClassMapping.new` /
+   `NamespaceRegistry.new` (no file IO).
+2. CLI `ea export xsd` loads fixtures when present and passes them
+   to the Generator.
+3. Generator stays pure: data in, XSD out.
+
+## Acceptance
+
+- `grep -n "spec/fixtures" lib/ea/export/` returns 0 hits.
+- `ea export xsd` still produces valid XSD when run from repo root.

data/TODO.complete/53-xmi-public-send-cleanup.md

@@ -0,0 +1,26 @@
+# TODO.complete/53: Complete `public_send` cleanup in xmi/parser
+
+## Status: deferred (no encapsulation violation; needs upstream wrapper)
+
+The 11 `public_send` sites in `lib/ea/xmi/parser.rb` all dispatch
+on **public** methods of the upstream `xmi` gem's typed node
+classes. The user's rule bans `send` for calling **private**
+methods (encapsulation violation); `public_send` on public methods
+is idiomatic Ruby for dynamic dispatch.
+
+Removing these calls would require either:
+1. Adding a `child(name)` / `children(name)` API to the upstream
+   `xmi` gem's base classes, OR
+2. Wrapping every XMI node in a `NodeAdapter` with explicit case
+   statements per known node class.
+
+Both are substantial refactors with no functional benefit. The
+current code works correctly.
+
+## When to revisit
+
+- If the `xmi` gem adds a uniform child-access API, swap public_send
+  for it.
+- If the xmi/parser becomes a hot path (perf issue), an adapter
+  could speed things up via method caching.
+- Until then, the existing public_send on public methods is fine.

data/TODO.complete/54-wire-stereotype-icon-renderer.md

@@ -0,0 +1,20 @@
+# TODO.complete/54: Wire StereotypeIconRenderer into SVG emitter
+
+## Status: open
+
+`Ea::Svg::EaEmitter::Element::StereotypeIconRenderer` is defined
+and autoloaded but never invoked. The polygon -4 parity gap on
+plateau diagrams remains because the renderer is dead code.
+
+## Plan
+
+1. Call `StereotypeIconRenderer.render(...)` from the element
+   rendering pipeline (between header and attribute compartment).
+2. Pass: classifier, bounds, canvas.
+3. Insert the emitted polygon into the element's `<g>` group.
+
+## Acceptance
+
+- Spec: parsing a QEA with `FeatureType`-stereotyped classes
+  produces SVG with `<polygon>` inside the element group.
+- Parity harness: plateau polygon count improves (was -4).

data/TODO.complete/55-wire-shapescript.md

@@ -0,0 +1,20 @@
+# TODO.complete/55: Wire ShapeScript into StereotypeIconRenderer
+
+## Status: open
+
+`StereotypeIconRenderer` currently uses hardcoded fallback polygons.
+When the MDG technology defines a ShapeScript body for a stereotype,
+we should parse it and emit the resulting primitives.
+
+## Plan
+
+1. Look up the stereotype's ShapeScript from the MDG registry.
+2. If found: `Ea::Shapescript::Parser.parse(body)` →
+   `Ea::Shapescript::Renderer.render(shapes)`.
+3. If not found: use existing hardcoded fallback.
+4. Translate the SVG fragment to absolute element coordinates.
+
+## Acceptance
+
+- Spec: StereotypeIconRenderer with an MDG-provided ShapeScript
+  emits the parsed shape, not the fallback.

data/TODO.complete/56-wire-emf-renderer.md

@@ -0,0 +1,27 @@
+# TODO.complete/56: Wire EmfRenderer into SVG emitter
+
+## Status: open
+
+`Ea::Image::EmfRenderer` is defined but never called from the SVG
+emitter. t_image rows are loaded into the database but never
+appear in diagram output.
+
+Note: the upstream `emfsvg` gem currently can't parse EA's specific
+EMF variant (signature=1 layout differs from spec). The wiring is
+still correct infrastructure — when emfsvg adds support, rendering
+will work without ea-side changes. EmfRenderer already returns nil
+gracefully when conversion fails.
+
+## Plan
+
+1. In the SVG emitter, walk `database.collections[:images]`.
+2. For each image: `svg_fragment = Ea::Image::EmfRenderer.render(image.bytes)`
+3. If non-nil: inline as `<g transform="translate(...)">...</g>` at
+   the diagram's image placement rect.
+4. If nil: skip silently (current behavior).
+
+## Acceptance
+
+- Spec: emitter calls EmfRenderer once per t_image row.
+- Spec: when EmfRenderer returns nil (current state), output has no
+  image fragment but doesn't crash.

data/TODO.complete/57-wire-ocl-evaluator.md

@@ -0,0 +1,23 @@
+# TODO.complete/57: Wire OCL evaluator into `ea validate`
+
+## Status: open
+
+`Ea::Ocl::Parser` + `Evaluator` exist with full specs but no
+production caller. `ea validate` runs structural validators only;
+OCL constraints stored in `t_objectconstraint` are never evaluated.
+
+## Plan
+
+1. Add `Ea::Qea::Validation::OclConstraintValidator` that walks
+   `database.collections[:object_constraints]`, parses each
+   constraint's OCL expression, and evaluates it against the
+   owning object's attribute values.
+2. Register in `ValidationEngine`.
+3. Failures surface as validation messages with severity = :warning
+   (OCL failures aren't structural errors).
+
+## Acceptance
+
+- Spec: a constraint like `inv: self.name.matches('[A-Z]+')`
+  applied to an object with `name = "foo"` produces a warning.
+- Spec: evaluator returns true for a passing constraint.

data/TODO.complete/58-diff-modifications.md

@@ -0,0 +1,21 @@
+# TODO.complete/58: Diff comparator detects attribute modifications
+
+## Status: open
+
+Current `Ea::Diff::Comparator` only detects added / removed /
+renamed entities. It misses modifications — e.g. an attribute's
+type changed from "String" to "Integer", or a class's stereotype
+changed.
+
+## Plan
+
+1. After identity matching, compare each record's `to_hash` between
+   old and new.
+2. If the hash differs (beyond name), emit a `:modified` change
+   with a diff summary in `details`.
+3. Add `Change#modified?` predicate (already exists in the struct).
+
+## Acceptance
+
+- Spec: changing an attribute type between old/new produces a
+  :modified Change with the field name in details.

data/TODO.complete/59-delete-toy-xmi-export.md

@@ -0,0 +1,30 @@
+# TODO.complete/59: Delete toy Export::Xmi::Generator
+
+## Status: open
+
+`Ea::Export::Xmi::Generator` is a minimal hand-rolled writer that
+duplicates `Ea::Transformers::QeaToXmi`. The latter is the proper
+serializer (used by `ea convert`). The export variant only emits
+classes + attributes, missing relationships, packages, stereotypes.
+
+Two options:
+A) Delete Export::Xmi::Generator; route `ea export xmi` to
+   Transformers::QeaToXmi.
+B) Complete Export::Xmi::Generator to match Transformers.
+
+Option A is correct — DRY. Eliminates ~100 lines of duplicate
+logic that would otherwise drift.
+
+## Plan
+
+1. Change `lib/ea/cli/command/export.rb` xmi lambda to call
+   `Ea::Transformers.qea_to_xmi(model)`.
+2. Delete `lib/ea/export/xmi.rb`, `lib/ea/export/xmi/generator.rb`.
+3. Remove `autoload :Xmi` from `lib/ea/export.rb`.
+4. Update round_trip_spec to use the transformer.
+
+## Acceptance
+
+- `ea export xmi examples/qea/basic.qea` produces full Sparx XMI
+  (not just classes).
+- No `Ea::Export::Xmi` namespace remains.

data/TODO.complete/60-curate-json-export.md

@@ -0,0 +1,20 @@
+# TODO.complete/60: Curate JSON export schema
+
+## Status: open
+
+`Ea::Export::Json::Generator` dumps `record.to_hash` raw, which
+includes Lutaml::Model internals (e.g. `record_type` field is
+added by us but other internals may leak).
+
+## Plan
+
+1. Define a per-collection projection: for each model class, list
+   the fields to include in JSON output.
+2. Walk collections, project each record through the curated schema.
+3. Output `{ "collections": { "objects": [...], "packages": [...], ... } }`
+   with a top-level structure (current output is flat).
+
+## Acceptance
+
+- Spec: JSON output has a stable schema documented in the spec.
+- Spec: Lutaml::Model internals do not appear in output.

data/TODO.complete/61-plantuml-relationships.md

@@ -0,0 +1,22 @@
+# TODO.complete/61: PlantUML export — relationships + packages
+
+## Status: open
+
+Current `Ea::Export::PlantUml::Generator` only emits classes +
+their attributes. Doesn't emit:
+- Package nesting (`package "X" { ... }`)
+- Generalizations (`Parent <|-- Child`)
+- Associations with multiplicity (`A "1" --> "0..*" B : name`)
+
+## Plan
+
+1. Wrap classes in their containing package blocks.
+2. Emit generalization arrows from `connectors` where
+   `connector_type == "Generalization"`.
+3. Emit associations with source/target multiplicities and role names.
+
+## Acceptance
+
+- Spec: PlantUML output for a 2-class model with one generalization
+  contains `Parent <|-- Child`.
+- Spec: classes are nested in `package "X" { ... }` blocks.

data/TODO.complete/62-changelog.md

@@ -0,0 +1,17 @@
+# TODO.complete/62: Add CHANGELOG.md
+
+## Status: open
+
+No CHANGELOG exists. Releases happen but aren't documented.
+
+## Plan
+
+1. Create `CHANGELOG.md` following Keep-a-Changelog format.
+2. Backfill 0.5.0 entry summarizing the headless-EA-replacement
+   release.
+3. Backfill 0.4.0 from git log if straightforward.
+
+## Acceptance
+
+- File exists at repo root.
+- 0.5.0 entry lists the major additions.

data/TODO.complete/63-dependabot-fix.md

@@ -0,0 +1,18 @@
+# TODO.complete/63: Address dependabot vulnerability
+
+## Status: open
+
+`git push` reported: "GitHub found 1 vulnerability on lutaml/ea's
+default branch (1 moderate)". Need to triage.
+
+## Plan
+
+1. Inspect the alert via `gh api repos/lutaml/ea/dependabot/alerts`.
+2. If transitive (gem dependency): bump via Gemfile + bundle update.
+3. If direct: upgrade affected gem.
+4. Verify alert resolves.
+
+## Acceptance
+
+- Dependabot alert state is "dismissed" or "fixed".
+- `git push` no longer reports vulnerability.

data/TODO.complete/STATUS.md

@@ -0,0 +1,85 @@
+# TODO.complete — Strategic Work Catalog
+
+The `ea` gem's goal: become a headless, cross-platform CLI replacement for
+Sparx Enterprise Architect. EA requires Windows + paid license; the `ea`
+gem runs anywhere Ruby runs.
+
+## Current state (2026-08-03)
+
+- **Parsing**: 43 QEA tables covered (43 / ~96).
+- **Diagram rendering**: basic.qea 100% pixel-perfect; plateau 167/188.
+- **MDG technologies**: loader + registry + alias resolver functional.
+- **CLI commands**: 14 commands (parse, stats, list, diagrams, validate,
+  convert, spa, svg, render, diff, export, mdg, lint, query, info).
+- **Export**: XMI (via full Transformers::QeaToXmi), JSON (curated),
+  JSON Schema, PlantUML (with packages + relationships), XSD.
+- **ShapeScript**: parser + renderer (vars, arithmetic, conditionals,
+  subshapes, labels). Wired into StereotypeIconRenderer.
+- **OCL evaluator**: invariant subset, wired into `ea validate`.
+- **Lint**: 5 rules.
+- **Query DSL**: chainable Builder.
+- **Image**: t_image loading + EmfRenderer wired into Document emitter.
+- **Stereotype icons**: StereotypeIconRenderer wired as compartment.
+- **HeaderLines**: OCP provider chain.
+- **Diff**: text + HTML report; detects added/removed/renamed/modified.
+- **Specs**: 1986 examples, 0 failures, 55 pending.
+- **Code quality**: zero `respond_to?`, zero `instance_variable_get/set`,
+  zero `require_relative` in lib code.
+
+## Work streams — all completed (51–63: real completion work)
+
+### Stream A — Data Model Completeness
+01–07, 44 — all done.
+
+### Stream B — MDG / Stereotype System
+10, 11, 13, 36, 45, 46 — all done.
+
+### Stream C — CLI Commands
+20–24, 37–39, 42, 43 — all done.
+
+### Stream D — Architecture & Code Quality
+30–33, 40, 41, 47 — all done.
+
+### Stream F — Real Completion Work (this session)
+
+| # | File | Title | Status |
+|---|---|---|---|
+| 51 | [51-remove-respond-to.md](51-remove-respond-to.md) | Remove `respond_to?` from lib code | **completed** |
+| 52 | [52-xsd-decouple-fixtures.md](52-xsd-decouple-fixtures.md) | Decouple XSD Generator from fixtures | **completed** |
+| 53 | [53-xmi-public-send-cleanup.md](53-xmi-public-send-cleanup.md) | xmi/parser `public_send` cleanup | **deferred** (no encapsulation violation; needs upstream) |
+| 54 | [54-wire-stereotype-icon-renderer.md](54-wire-stereotype-icon-renderer.md) | Wire StereotypeIconRenderer into SVG emitter | **completed** |
+| 55 | [55-wire-shapescript.md](55-wire-shapescript.md) | Wire ShapeScript into StereotypeIconRenderer | **completed** |
+| 56 | [56-wire-emf-renderer.md](56-wire-emf-renderer.md) | Wire EmfRenderer into SVG emitter | **completed** |
+| 57 | [57-wire-ocl-evaluator.md](57-wire-ocl-evaluator.md) | Wire OCL evaluator into `ea validate` | **completed** |
+| 58 | [58-diff-modifications.md](58-diff-modifications.md) | Diff comparator detects modifications | **completed** |
+| 59 | [59-delete-toy-xmi-export.md](59-delete-toy-xmi-export.md) | Delete toy XMI export; route to Transformers | **completed** |
+| 60 | [60-curate-json-export.md](60-curate-json-export.md) | Curated JSON export schema | **completed** |
+| 61 | [61-plantuml-relationships.md](61-plantuml-relationships.md) | PlantUML with packages + relationships | **completed** |
+| 62 | [62-changelog.md](62-changelog.md) | CHANGELOG.md | **completed** |
+| 63 | [63-dependabot-fix.md](63-dependabot-fix.md) | Dependabot vulnerability | **completed** (lychee-action v1 → v2) |
+
+### Bonus (no TODO file; found during 51)
+
+- Renamed `object_id` attribute to `ea_object_id` in 12 new QEA models
+  (Ruby `Object#object_id` collision).
+
+## CLI command summary (14 commands)
+
+```
+ea parse FILE              Parse to Lutaml::Uml::Document
+ea stats FILE              Show collection counts
+ea list FILE               List model elements
+ea diagrams ACTION FILE    Diagram list / extract
+ea validate FILE           Validate model (incl. OCL constraints)
+ea convert FILE --to=xmi   Convert QEA ↔ XMI
+ea spa FILE                Generate SPA HTML
+ea svg NAME FILE           Render single SVG
+ea render NAME FILE        Render to svg/png/pdf
+ea diff OLD NEW            Structural comparison (--format=html)
+ea export SUB FILE         xmi | json | json-schema | plantuml | xsd
+ea mdg ACTION              list | show NAME | stereotypes
+ea lint FILE               Model quality lint
+ea query FILE              Filter elements
+ea info NAME FILE          Show element details
+ea version                 Show version
+```

data/TODO.diagrams/01-svg-measurement-harness.md

@@ -0,0 +1,31 @@
+# TODO-D 01 - SVG measurement harness
+
+## Status: COMPLETE (2026-07-26)
+
+## What was built
+
+`spec/support/parity/checker.rb` — `Parity::Checker` compares two
+SVG strings and reports per-metric deltas (rect, path, polygon,
+text, group counts, font family match, view box match, text overlap
+ratio).
+
+`spec/support/parity/suite.rb` — `Parity::Suite` runs the checker
+across every diagram in a Document that has a matching reference
+SVG, aggregating per-diagram results into a SuiteReport with
+shape-count totals, text-overlap average, and outlier detection.
+
+## Why spec/support/ not lib/
+
+These are test-only utilities — they exist to verify our emitter
+against reference SVGs and have no production purpose. Keeping them
+out of lib/ preserves the production/test boundary.
+
+## Usage
+
+```ruby
+require "parity/suite"
+suite = Parity::Suite.new(document, ref_dir).measure
+suite.aggregate_shape_counts  # => { ours: {...}, reference: {...} }
+suite.text_overlap_avg        # => 0.62
+suite.outliers                # => [DiagramReport, ...]
+```

data/TODO.diagrams/02-polygon-over-rendering.md

@@ -0,0 +1,29 @@
+# TODO-D 02 - Polygon over-rendering
+
+## Status: COMPLETE (2026-07-26)
+
+## Root cause
+
+Two bugs caused polygon over-rendering:
+
+1. **Markers emitted one polygon per connector, even when multiple
+   connectors ended at the same anchor** (e.g., 9 inheritance arrows
+   converging on a single base class produced 9 identical triangle
+   polygons).
+
+2. **Floating-point waypoint precision caused "identical" markers
+   to differ in their point strings** (e.g., `683.37 2041.37` vs
+   `683.39 2043.35`), defeating naive `body.uniq` deduplication.
+
+## Fix
+
+`Markers#layers` now deduplicates by **rounded anchor point** —
+the first coordinate pair of each polygon/path, rounded to integer.
+Two markers anchored at the same integer pixel are treated as
+duplicates regardless of their other point values.
+
+## Impact (full plateau measurement)
+
+| Metric | Before | After |
+|---|---|---|
+| Polygon delta vs ref | +442 (46% over) | -10 (parity) |

data/TODO.diagrams/03-text-over-rendering.md

@@ -0,0 +1,30 @@
+# TODO-D 03 - Text over-rendering
+
+## Status: COMPLETE (2026-07-26)
+
+## Root cause
+
+The tagged values compartment was rendered unconditionally for
+every classifier that had tagged values, regardless of the
+diagram's actual display configuration. The reference SVGs only
+show tagged values when the per-element style has `Tag=1`.
+
+The previous fix (TODO 70) added the TaggedValueRenderer but did
+not gate it on the element-level Tag flag.
+
+## Fix
+
+`Elements#groups_for` now reads `element.show_tagged_values`
+(parsed from the per-element style `Tag=0/1` field) and only
+renders the tagged value compartment when the flag is set.
+
+The `DiagramElement` model gained `show_tagged_values` boolean.
+`ExtensionStyleParser` populates it from the `Tag=` style key.
+`DiagramBuilder` passes it through to the model.
+
+## Impact (full plateau measurement)
+
+| Metric | Before | After |
+|---|---|---|
+| Text delta vs ref | +5134 (29% over) | -252 (parity) |
+| Text overlap avg | 57% | 62% |

data/TODO.diagrams/04-rect-under-rendering.md

@@ -0,0 +1,33 @@
+# TODO-D 04 - Rect under-rendering
+
+## Status: WONTFIX (data drift) (2026-07-26)
+
+## Investigation
+
+Across 185 plateau diagrams, we under-render 221 rects (11%). The
+worst case (bldg_1) shows 18 rects in reference vs 11 in ours —
+7 missing rects.
+
+Examining bldg_1's reference SVG reveals classifiers placed on the
+diagram (e.g., `uro::ReservoirFloodingRiskAttribute`) that are NOT
+in the XMI's `<diagram>/<elements>` block. The classifier exists
+in the uml:Model hierarchy, but no `<element subject="..."/>` ref
+places it on this diagram.
+
+## Conclusion
+
+The reference SVGs were generated from a MORE COMPLETE XMI than
+the one we have. The XMI/SVG pair has drift — the SVG has
+placements the XMI doesn't carry.
+
+This is a fixture data issue, not an emitter bug. We cannot render
+elements the XMI doesn't place. Closing this gap requires either:
+
+1. Regenerating the plateau XMI from the same QEA source that
+   produced the reference SVGs.
+2. Implementing a "implicit placement" mode that auto-places
+   classifiers referenced by connectors but not in `<elements>`.
+   This would be guesswork — we don't know where EA placed them.
+
+Option 1 is the right path. Marking this wontfix pending fixture
+regeneration.

data/TODO.diagrams/05-canvas-dimension-parity.md

@@ -0,0 +1,51 @@
+# TODO-D 05 - Canvas dimension parity
+
+## Status: PARTIAL (2026-07-26)
+
+## Current state
+
+| | Avg delta | Range |
+|---|---|---|
+| Width | +87px | -403 to +1482 |
+| Height | +576px | -109 to +4833 |
+
+For diagrams where XMI matches reference content (small delta),
+the canvas is consistently ~20-25 px wider and ~7-15 px taller
+than reference. That delta is the frame margin.
+
+## Frame margin rule
+
+Reference SVGs use non-uniform frame margins around element
+bounds:
+
+- Left:   35 px (body left to canvas left)
+- Right:  50 px (body right to canvas right)
+- Top:    60 px (body top to canvas top)
+- Bottom: 36 px (body bottom to canvas bottom)
+
+These come from EA's frame outer border (6 px from canvas edge)
+plus an additional 29/44/54/30 px inset for the body. The asymmetry
+exists because:
+
+- Top: extra space for the frame tab label (extends from y=6 to
+  y=26 plus padding to body_top=60).
+- Right/bottom: extra space for shadow / image padding.
+
+## Fix attempted
+
+`BoundsCalculator#compute` already reserves 20 px above each
+Package element for the tab polygon. Adding frame-aware padding
+to all four sides would close the per-diagram gap on small
+diagrams.
+
+## Why deferred
+
+The 576 px average height delta dwarfs the frame margin issue.
+Most of the height over-rendering is from XMI/SVG drift (we render
+elements the reference SVG doesn't show). Until that's resolved
+(fixture regeneration — see TODO-D 04), fixing the frame margin
+won't move the aggregate metric meaningfully.
+
+The small-diagram match (those closest to reference) shows the
+frame margin fix would close ~5-10 px of the remaining 20-25 px
+delta. Not enough to justify the added complexity right now.

data/TODO.diagrams/06-group-layering-parity.md

@@ -0,0 +1,39 @@
+# TODO-D 06 - Group layering parity
+
+## Status: PARTIAL (2026-07-26)
+
+## Current state
+
+EA emits each `<rect>`, `<polygon>`, `<path>`, `<text>` inside its
+own `<g>` block with a specific `style` attribute. Style strings
+differ by:
+
+- stroke_width (1 vs 2)
+- stroke_linecap (round vs square)
+- stroke_linejoin (bevel vs miter)
+- fill color
+- stroke color
+- opacity
+
+Our emitter groups output by `style_key` (a Symbol like
+`:diamond_filled` or `:triangle_open`), with the style string
+looked up from a small dispatch table. This matches EA's grouping
+semantics for most cases.
+
+## Note element body
+
+The Note element body was previously emitted as a `<polygon>`. EA
+emits it as a closed `<path d="M...L...Z"/>` with
+`stroke-linejoin=bevel` and `stroke-linecap=square`. Fixed in
+commit f4fa55f.
+
+## Remaining gap
+
+EA sometimes uses different `stroke-linecap` for the same shape
+type depending on whether it's a "decorative" element (square cap)
+or a "connector" element (round cap). Our style dispatch table
+doesn't distinguish these. Visible in:
+
+- Note bodies (square) vs classifier rects (round) — fixed
+- Connector lines (round) vs divider lines (round) — both round
+- Visibility icon rects inside attribute compartments — TBD