Fantasy-Map-Generator/_bmad-output/implementation-artifacts/1-2-framework-core-init-canvas-and-dom-setup.md

Story 1.2: Framework Core — Init, Canvas, and DOM Setup

Status: review Epic: 1 — WebGL Layer Framework Module Story Key: 1-2-framework-core-init-canvas-and-dom-setup Created: (SM workflow) Developer: Amelia (Dev Agent)

---

Story

As a developer, I want init() to set up the WebGL2 canvas, wrap svg#map in div#map-container, create the Three.js renderer/scene/camera, attach the ResizeObserver, and subscribe to D3 zoom events, So that the framework owns the single shared WebGL context and the canvas is correctly positioned in the DOM alongside the SVG map.

---

Context

Prior Art (Story 1.1 — Complete)

Story 1.1 delivered the scaffold in src/modules/webgl-layer-framework.ts:

• Pure exports: buildCameraBounds, detectWebGL2, getLayerZIndex
• Interfaces: WebGLLayerConfig (exported), RegisteredLayer (internal)
• Class: WebGL2LayerFrameworkClass with all 9 private fields (stubs)
• All Seven public API methods: init(), register(), unregister(), setVisible(), clearLayer(), requestRender(), syncTransform() — stubs only
• _fallback backing field + get hasFallback() getter
• register() currently pushes to pendingConfigs[]
• Global: globalThis.WebGL2LayerFramework = new WebGL2LayerFrameworkClass() (last line)
• 16 tests in src/modules/webgl-layer-framework.test.ts — all passing

Files to Modify

• src/modules/webgl-layer-framework.ts — implement init(), observeResize(), subscribeD3Zoom(), syncTransform() (partial), change import type → value imports for WebGLRenderer/Scene/OrthographicCamera/Group
• src/modules/webgl-layer-framework.test.ts — add Story 1.2 tests for init() paths

---

Acceptance Criteria

AC1: init() called + WebGL2 available → div#map-container wraps svg#map (position:relative, z-index:1 for svg), canvas#terrainCanvas is sibling to #map inside container (position:absolute; inset:0; pointer-events:none; aria-hidden:true; z-index:2)

AC2: detectWebGL2() returns false → init() returns false, hasFallback === true, all subsequent API calls are no-ops (guard on _fallback)

AC3: hasFallback uses backing field _fallback (NOT readonly) — already implemented in Story 1.1; verify pattern remains correct

AC4: After successful init() → exactly one WebGLRenderer, Scene, OrthographicCamera exist as instance fields (non-null)

AC5: ResizeObserver on #map-container → calls renderer.setSize(width, height) and requestRender() on resize events

AC6: D3 zoom subscription → viewbox.on("zoom.webgl", () => this.requestRender()) called in init(); guarded with typeof globalThis.viewbox !== "undefined" for Node test env

AC7: Constructor has no side effects → all of canvas/renderer/scene/camera/container are null after construction; only _fallback=false, layers=new Map(), pendingConfigs=[] are initialized

AC8: init() completes in <200ms (NFR-P5) — no explicit test; implementation must avoid blocking operations

AC9: Global pattern unchanged — globalThis.WebGL2LayerFramework = new WebGL2LayerFrameworkClass() remains as last line

---

Technical Notes

init() Sequence (step-by-step)

1. this._fallback = !detectWebGL2() — use probe-less call; document.createElement("canvas") is fine at init time (only called when browser runs init())
2. If _fallback: return false immediately (no DOM mutation)
3. Find #map via document.getElementById("map") — if not found, log WARN, return false
4. Create div#map-container: style.position = "relative"; id = "map-container" — insert before #map in parent, then move #map inside
5. Build canvas#terrainCanvas: set styles (position:absolute; inset:0; pointer-events:none; aria-hidden:true; z-index:2)
6. Size canvas: canvas.width = container.clientWidth || 960; canvas.height = container.clientHeight || 540
7. Create new WebGLRenderer({ canvas, antialias: false, alpha: true })
8. Create new Scene()
9. Create new OrthographicCamera(0, canvas.width, 0, canvas.height, -1, 1) — initial ortho bounds; will be updated on first syncTransform()
10. Store all in instance fields
11. Call subscribeD3Zoom()
12. Process pendingConfigs[] → for each, create new Group(), set group.renderOrder = config.renderOrder, call config.setup(group), scene.add(group), store in layers Map
13. Clear pendingConfigs = []
14. Call observeResize()
15. Return true

Three.js Import Change

Converting from import type → value imports:

import type {Group} from "three"; // Group stays type-only until Story 1.3 uses it at runtime
import {WebGLRenderer, Scene, OrthographicCamera} from "three";
// Note: Group is created in init() → must also be a value import in 1.2

→ Final: import { Group, WebGLRenderer, Scene, OrthographicCamera } from "three"; → Remove import type line

subscribeD3Zoom() Implementation

private subscribeD3Zoom(): void {
  if (typeof (globalThis as any).viewbox === "undefined") return;
  (globalThis as any).viewbox.on("zoom.webgl", () => this.requestRender());
}

observeResize() Implementation

private observeResize(): void {
  if (!this.container || !this.renderer) return;
  this.resizeObserver = new ResizeObserver(entries => {
    const { width, height } = entries[0].contentRect;
    if (this.renderer && this.canvas) {
      this.renderer.setSize(width, height);
      this.requestRender();
    }
  });
  this.resizeObserver.observe(this.container);
}

Fallback Guard Pattern

All public methods of the class must guard against _fallback (and against null init state). For Story 1.2, register() already works pre-init; init() has the primary guard. Story 1.3 lifecycle methods will add _fallback guards.

syncTransform() (Partial — Story 1.2)

Story 1.3 implements the full syncTransform(). Story 1.2 may leave stub. Story 1.3 reads globalThis.viewX, globalThis.viewY, globalThis.scale, globalThis.graphWidth, globalThis.graphHeight and calls buildCameraBounds().

requestRender() — Story 1.2 transition

Current stub in Story 1.1 calls this.render() directly. Story 1.2 still leaves requestRender() as-is (direct render call) since render() private impl is Story 1.3. Just remove the direct this.render() call from requestRender() stub or leave it — tests will tell us.

Actually, requestRender() stub currently calls this.render() which is also a stub (no-op). This is fine for Story 1.2. Story 1.3 will replace requestRender() with RAF-coalescing.

---

Tasks

• T1: Implement init() in webgl-layer-framework.ts following the sequence above
  • T1a: Change import type { Group, ... } to value imports import { Group, WebGLRenderer, Scene, OrthographicCamera } from "three"
  • T1b: detectWebGL2() fallback guard
  • T1c: DOM wrap (#map → #map-container > #map + canvas#terrainCanvas)
  • T1d: Renderer/Scene/Camera creation
  • T1e: subscribeD3Zoom() call
  • T1f: pendingConfigs[] queue processing
  • T1g: observeResize() call
• T2: Implement private subscribeD3Zoom() method
• T3: Implement private observeResize() method
• T4: Remove biome-ignore comments for fields now fully used (canvas, renderer, camera, scene, container, resizeObserver)
• T5: Add Story 1.2 tests for init() to webgl-layer-framework.test.ts:
  • T5a: init() with failing WebGL2 probe → hasFallback=true, returns false
  • T5b: init() with missing #map element → returns false, no DOM mutation
  • T5c: init() success: renderer/scene/camera all non-null after init
  • T5d: init() success: pendingConfigs[] processed (setup called, layers Map populated)
  • T5e: observeResize() ResizeObserver callback calls renderer.setSize()
• T6: npm run lint clean
• T7: npx vitest run modules/webgl-layer-framework.test.ts all pass
• T8: Set story status to review

---

Dev Agent Record

To be filled by Dev Agent

Implementation Notes

(pending)

Files Modified

(pending)

Test Results

(pending)