@bycrux/editor 0.12.0 → 1.0.2

package/src/video/timeline/canvas/__tests__/TimelineCanvas.pendingDrops.test.tsx

@@ -0,0 +1,316 @@
+/// <reference types="vitest/globals" />
+/**
+ * Pending-import ghosts, at the COMPONENT level — the seam between the host's
+ * `PendingDrop[]` and the pixels.
+ *
+ * `pending-drop.test.ts` proves the painter given an already-resolved
+ * `PendingDropBand`. What it cannot reach is the half that lives in
+ * `TimelineCanvas`: turning a host's `trackIndex` into a row rectangle
+ * (`pendingDropBands`), and repainting the overlay when the prop changes. Both
+ * are exactly the join a unit test of either side alone would miss — a ghost
+ * resolved against the wrong row, or a correct ghost that never gets painted
+ * because nothing marked the layer dirty.
+ *
+ * The redraw case is the one worth being careful about: the overlay layer is
+ * repainted only by the clock, the viewport or a gesture, none of which fire
+ * here. So the assertion is deliberately before/after — nothing painted on the
+ * first pass, a ghost painted after the prop change — because a test that only
+ * looked at the first render would pass with the redraw effect deleted.
+ *
+ * jsdom has no 2D canvas, so `getContext` hands back a RECORDING proxy, one per
+ * canvas element: this surface stacks two (content below, overlay above), and a
+ * single shared recorder could not tell which layer a ghost landed on.
+ */
+import { describe, it, expect, beforeEach, afterEach, vi } from 'vitest'
+import { render, act, cleanup } from '@testing-library/react'
+import { createPlaybackClock } from '../../../playback-clock'
+import type { Project, PendingDrop } from '../../../../types'
+import TimelineCanvas from '../TimelineCanvas'
+import { createViewportStore } from '../viewport'
+import { timeToX } from '../viewport'
+import {
+  PENDING_DROP_INSET_PX,
+  PENDING_DROP_RADIUS_PX,
+  RULER_HEIGHT_PX,
+  computeTimelineLayout,
+} from '../draw'
+
+// ── Recording context, one per canvas element ────────────────────────────
+
+interface RecordedCall { method: string; args: unknown[] }
+
+interface Recorder {
+  calls: RecordedCall[]
+  of: (method: string) => RecordedCall[]
+  reset: () => void
+}
+
+const recorders = new Map<HTMLCanvasElement, Recorder>()
+
+function recorderFor(canvas: HTMLCanvasElement) {
+  let rec = recorders.get(canvas)
+  if (!rec) {
+    const calls: RecordedCall[] = []
+    rec = {
+      calls,
+      of: (method: string) => calls.filter(c => c.method === method),
+      reset: () => { calls.length = 0 },
+    }
+    recorders.set(canvas, rec)
+  }
+  return rec
+}
+
+function contextFor(canvas: HTMLCanvasElement) {
+  const rec = recorderFor(canvas)
+  const props: Record<string, unknown> = {}
+  return new Proxy({}, {
+    get(_t, prop: string) {
+      if (prop in props) return props[prop]
+      if (prop === 'createLinearGradient') return () => ({ addColorStop: () => {} })
+      return (...args: unknown[]) => { rec.calls.push({ method: prop, args }) }
+    },
+    set(_t, prop: string, value: unknown) {
+      props[prop] = value
+      rec.calls.push({ method: `set:${prop}`, args: [value] })
+      return true
+    },
+  })
+}
+
+let realGetContext: typeof HTMLCanvasElement.prototype.getContext
+let realGetRect: typeof Element.prototype.getBoundingClientRect
+
+beforeEach(() => {
+  recorders.clear()
+  realGetContext = HTMLCanvasElement.prototype.getContext
+  realGetRect = Element.prototype.getBoundingClientRect
+  HTMLCanvasElement.prototype.getContext = function (this: HTMLCanvasElement) {
+    return contextFor(this)
+  } as unknown as typeof HTMLCanvasElement.prototype.getContext
+  Element.prototype.getBoundingClientRect = function (this: Element) {
+    return { x: 0, y: 0, top: 0, left: 0, right: 1000, bottom: 100, width: 1000, height: 100, toJSON: () => ({}) } as DOMRect
+  }
+  vi.useFakeTimers()
+})
+
+afterEach(() => {
+  cleanup()
+  vi.useRealTimers()
+  HTMLCanvasElement.prototype.getContext = realGetContext
+  Element.prototype.getBoundingClientRect = realGetRect
+  recorders.clear()
+})
+
+// Two video tracks, so a `trackIndex` actually chooses between rows that sit at
+// DIFFERENT y — with one row every assertion below would pass by accident.
+const project = {
+  id: 'p',
+  tracks: [
+    { id: 'trk-0', items: [{ id: 'a0', type: 'video', src: 'a.mp4', start: 0, end: 10, inPoint: 0, outPoint: 10, sourceDuration: 20 }] },
+    { id: 'trk-1', items: [{ id: 'b0', type: 'video', src: 'b.mp4', start: 0, end: 2, inPoint: 0, outPoint: 2, sourceDuration: 20 }] },
+  ],
+  audio: { tracks: [] },
+} as unknown as Project
+
+const TOTAL_DURATION = 20
+
+function drop(over: Partial<PendingDrop> = {}): PendingDrop {
+  return { id: 'pd-1', atTime: 3, durationSec: 2, trackIndex: 0, label: 'C0042.MP4', ...over }
+}
+
+function mount(pendingDrops?: readonly PendingDrop[]) {
+  const store = createViewportStore()
+  const clock = createPlaybackClock()
+
+  const props = {
+    project,
+    clock,
+    store,
+    totalDuration: TOTAL_DURATION,
+    fps: 30,
+    selectedIds: [] as string[],
+  }
+
+  const utils = render(<TimelineCanvas {...props} pendingDrops={pendingDrops} />)
+  act(() => { vi.advanceTimersByTime(32) })
+
+  const canvases = utils.container.querySelectorAll('canvas')
+  const overlay = recorderFor(canvases[1] as HTMLCanvasElement)
+  const content = recorderFor(canvases[0] as HTMLCanvasElement)
+
+  // Pin the scale so x is readable: x = t × 100. Until this lands the surface
+  // sits at its own fit scale (1000px / 20s = 50px/s), and the paint that
+  // already happened at THAT scale is still in the recorders — so they are
+  // cleared between marking the layers dirty and flushing the frame, leaving
+  // exactly one settled paint to assert on.
+  act(() => { store.set({ pxPerSecond: 100, scrollSeconds: 0, widthPx: 1000 }) })
+  overlay.reset()
+  content.reset()
+  act(() => { vi.advanceTimersByTime(32) })
+
+  return {
+    ...utils,
+    store,
+    // The surface stacks content first, overlay second (see TimelineCanvas's
+    // JSX) — ghosts belong to the overlay.
+    overlay,
+    content,
+    setPendingDrops(next?: readonly PendingDrop[]) {
+      utils.rerender(<TimelineCanvas {...props} pendingDrops={next} />)
+      act(() => { vi.advanceTimersByTime(32) })
+    },
+  }
+}
+
+/** Where the painter should start the ghost's rounded path, derived from the
+ *  layout the canvas itself draws from — never a hard-coded pixel. The fill
+ *  path opens at `(x + radius, y)`; see `roundRectPath`. */
+function expectedGhostOrigin(trackIdx: number, atTime: number, store: ReturnType<typeof createViewportStore>) {
+  const row = computeTimelineLayout(project).rows.find(r => r.trackIdx === trackIdx)
+  if (!row) throw new Error(`no row for trackIdx ${trackIdx}`)
+  return [
+    timeToX(atTime, store.get()) + PENDING_DROP_RADIUS_PX,
+    row.y + PENDING_DROP_INSET_PX,
+  ]
+}
+
+/** The first path the ghost lays down. `setLineDash` is unique to the ghost on
+ *  this layer, so its presence is what proves one was drawn at all. */
+function ghostOrigin(overlay: Recorder): unknown[] | null {
+  if (overlay.of('setLineDash').length === 0) return null
+  return overlay.of('moveTo')[0]?.args ?? null
+}
+
+describe('TimelineCanvas — pending-drop ghosts', () => {
+  it('resolves a trackIndex naming a real row to THAT row\'s rectangle', () => {
+    // trackIdx 1 is drawn ABOVE trackIdx 0 (rows descend from the top), so
+    // this y is nothing like the base row's — the two cases can't be confused.
+    const { overlay, store } = mount([drop({ trackIndex: 1 })])
+    expect(ghostOrigin(overlay)).toEqual(expectedGhostOrigin(1, 3, store))
+  })
+
+  it('resolves the other row to ITS rectangle, at a different y', () => {
+    const { overlay, store } = mount([drop({ trackIndex: 0 })])
+    expect(ghostOrigin(overlay)).toEqual(expectedGhostOrigin(0, 3, store))
+    // Guard against the two expectations having collapsed into one value.
+    expect(expectedGhostOrigin(0, 3, store)).not.toEqual(expectedGhostOrigin(1, 3, store))
+  })
+
+  it('falls back to the base video row for trackIndex -1', () => {
+    // -1 is "no preference" — the base video row is the LOWEST trackIdx that
+    // holds video, which here is trkIdx 0, drawn at the bottom.
+    const { overlay, store } = mount([drop({ trackIndex: -1 })])
+    expect(ghostOrigin(overlay)).toEqual(expectedGhostOrigin(0, 3, store))
+  })
+
+  it('falls back to the base video row for a trackIndex past the end, without throwing', () => {
+    // The host's list is asynchronous: the project can lose a track between the
+    // drop and the ghost being drawn, leaving a trackIndex that names nothing.
+    expect(() => {
+      const { overlay, store } = mount([drop({ trackIndex: 99 })])
+      expect(ghostOrigin(overlay)).toEqual(expectedGhostOrigin(0, 3, store))
+    }).not.toThrow()
+  })
+
+  it('draws a newTrack ghost on a FRESH row — never the base row, never into the ruler', () => {
+    // `trackIndex` here is deliberately a value that names no real row (the
+    // sentinel `resolveDropTrackIndex` returns for "mint a new track") — the
+    // point of `newTrack` is that the renderer must not need `trackIndex` to
+    // resolve this at all.
+    const { overlay } = mount([drop({ trackIndex: 2, newTrack: true })])
+    const origin = ghostOrigin(overlay)
+    expect(origin).not.toBeNull()
+    const [, y] = origin as [number, number]
+
+    // Not the base row (trackIdx 0, at the bottom of the video block) — the
+    // exact bug this feature fixes: a new-track ghost sitting on top of
+    // whatever footage already occupies the base row.
+    const baseRowY = computeTimelineLayout(project).rows.find(r => r.trackIdx === 0)!.y + PENDING_DROP_INSET_PX
+    expect(y).not.toEqual(baseRowY)
+    // Not even the CURRENT top-of-video-block row (trackIdx 1) — this is a
+    // genuinely fresh row, one slot further toward the ruler than any row
+    // that exists today.
+    const topVideoRowY = computeTimelineLayout(project).rows.find(r => r.trackIdx === 1)!.y + PENDING_DROP_INSET_PX
+    expect(y).not.toEqual(topVideoRowY)
+    // Never intrudes into the ruler strip.
+    expect(y).toBeGreaterThanOrEqual(RULER_HEIGHT_PX)
+    // This fixture has no overlay/caption rows above the video block, so
+    // there is no room for a full extra row without the surface actually
+    // growing by one — the phantom clamps flush against the ruler's bottom
+    // edge. Still a row of its own, and still never past the ruler.
+    expect(y).toEqual(RULER_HEIGHT_PX + PENDING_DROP_INSET_PX)
+  })
+
+  it('ignores newTrack on a drop that already names a real row', () => {
+    // `newTrack` absent (or false) is the ordinary case — unchanged from
+    // before this field existed.
+    const { overlay, store } = mount([drop({ trackIndex: 1 })])
+    expect(ghostOrigin(overlay)).toEqual(expectedGhostOrigin(1, 3, store))
+  })
+
+  it('resolves each drop in a list independently', () => {
+    const { overlay, store } = mount([
+      drop({ id: 'pd-1', trackIndex: 1, atTime: 3 }),
+      drop({ id: 'pd-2', trackIndex: 0, atTime: 6 }),
+    ])
+    // Each band lays two paths (fill, then the dashed outline inset by half the
+    // stroke), so the fill origins are the even entries.
+    const origins = overlay.of('moveTo').filter((_, i) => i % 2 === 0).map(c => c.args)
+    expect(origins).toEqual([
+      expectedGhostOrigin(1, 3, store),
+      expectedGhostOrigin(0, 6, store),
+    ])
+  })
+
+  it('paints no ghost when pendingDrops is absent or empty', () => {
+    for (const value of [undefined, [] as PendingDrop[]]) {
+      const { overlay } = mount(value)
+      expect(overlay.of('setLineDash')).toHaveLength(0)
+      cleanup()
+    }
+  })
+
+  // ── The redraw wiring ──
+  //
+  // Nothing else repaints this layer here: the clock never ticks, the viewport
+  // is already settled, no gesture runs. So if the prop change did not mark the
+  // overlay dirty, the "after" assertion below would still see an empty
+  // recorder — which is exactly what makes this a test of the wiring rather
+  // than of the first render.
+
+  it('repaints the overlay when pendingDrops APPEARS on a mounted surface', () => {
+    const { overlay, store, setPendingDrops } = mount()
+    expect(ghostOrigin(overlay)).toBeNull()
+
+    overlay.reset()
+    setPendingDrops([drop({ trackIndex: 1 })])
+    expect(ghostOrigin(overlay)).toEqual(expectedGhostOrigin(1, 3, store))
+  })
+
+  it('repaints when a ghost MOVES, and the new band is at the new row', () => {
+    const { overlay, store, setPendingDrops } = mount([drop({ trackIndex: 1 })])
+    expect(ghostOrigin(overlay)).toEqual(expectedGhostOrigin(1, 3, store))
+
+    overlay.reset()
+    setPendingDrops([drop({ trackIndex: 0, atTime: 6 })])
+    expect(ghostOrigin(overlay)).toEqual(expectedGhostOrigin(0, 6, store))
+  })
+
+  it('repaints with the ghost GONE when the host retracts it', () => {
+    // The retraction is the half that actually matters in production: a ghost
+    // that never comes down outlives the import it stood for.
+    const { overlay, setPendingDrops } = mount([drop()])
+    expect(ghostOrigin(overlay)).not.toBeNull()
+
+    overlay.reset()
+    setPendingDrops([])
+    expect(overlay.of('clearRect').length).toBeGreaterThan(0)
+    expect(overlay.of('setLineDash')).toHaveLength(0)
+  })
+
+  it('keeps ghosts off the CONTENT layer, which clips and filmstrips own', () => {
+    const { content } = mount([drop()])
+    expect(content.of('setLineDash')).toHaveLength(0)
+  })
+})