ehm

workshop/steps/13-crush/README.md

@@ -0,0 +1,96 @@
+# Step 13 — The crush (when there is no way out)
+
+You found this one yourself too, chasing the capstone's ledge: the hero gets
+pushed out of the moving ledge, lands **inside the pillar**, and the `NaN` you
+buried in step 12 climbs right back out of its grave. Your autopsy chain from
+last time is unchanged — the only new thing is *how a box you just freed ends
+up inside a wall again in the very same frame*.
+
+> Still a *fourth* thing, separate from the two bugs you're hunting in
+> `engine/system/physics.ts` — no spoilers there.
+
+## Why one pass isn't enough
+
+Step 12's `safeMoveAndSlide` walks the walls **once, in array order**, fixing
+each overlap it meets. For one wall that's airtight. But a depenetration push is
+a *teleport* — and a teleport can land you inside a wall the loop already
+checked and cleared, or one it hasn't reached yet (in which case it works, by
+luck of the ordering). A resolver whose correctness depends on the order of the
+wall array isn't a resolver — it's a coin flip.
+
+Two experiments in a scratch file, **predicting each outcome before running**
+(use your step-12 code as-is):
+
+1. Hero `{x: 9, y: 1, w: 2, h: 2}`, walls `A = {x: 10, y: 0, w: 4, h: 4}` and
+   `B = {x: 4, y: 0, w: 4.5, h: 1.4}`. The hero overlaps only `A`. Run one pass
+   with the array `[A, B]`, then with `[B, A]`. Where does the hero end up in
+   each case, and is it free? Explain the difference before moving on.
+2. Walls `{x: 10, y: -5, w: 4, h: 10}` and `{x: 6.5, y: -5, w: 2, h: 10}` — a
+   gap 1.5 wide. Hero (2 wide) at `{x: 9, y: 0}`. Apply `penetrationVector`
+   pushes in a loop and log `x` each time. Does it converge? What number does
+   `x` bounce between, and *why will it never stop*?
+
+## The negotiation, and when it honestly fails
+
+The fix for experiment 1 is patience: don't do one pass — **repeat whole passes
+until a full pass finds nothing to fix**. That clean pass is your proof of
+freedom. Each pass is cheap, and in sane geometry it settles in one or two.
+
+But experiment 2 shows the negotiation can be *unwinnable*: when the gap is
+narrower than the box, **no overlap-free position exists**. No amount of math
+fixes that, because it isn't a math problem — it's a game-design question, and
+every game answers it differently. Mario between a Thwomp and the floor:
+crushed = death. Some engines let the wall shove you *through* its partner.
+Zelda-flavored games mostly refuse the situation: solid wins, the hero holds
+still until the gap opens. We take that one — it's the smallest honest answer:
+**cap the passes, and if the cap fires, report it** (`settled: false`) instead
+of pretending. You already believe in caps; your step-10 loop carries one for
+exactly the same reason.
+
+Last session you proposed armoring `sweepInterval` against the `Infinity`
+directly. You can — see the optional section — but notice what that answer
+skips over: even with the `NaN` gone, *what should a crushed hero do?* The
+kernel can't know; it only measures. Deciding is the resolver's job. Keeping
+**detection** and **policy** separate is the actual lesson of this step.
+
+## Task
+
+Two functions in `crush.ts`:
+
+1. `resolveOverlaps(box, walls)` — the negotiation: passes until clean or
+   capped, returning `{x, y, settled}`.
+2. `safeMoveAndSlide(box, v, walls)` — step 12's version rebuilt on top of it.
+   Settled → sweep as usual. Crushed → **don't feed the sweep an overlapping
+   box** (you know its opinion of those); the box stays where the resolver left
+   it and waits.
+
+One warning on the `EPSILON` slack: apply it **only along the axis you actually
+pushed**. Before you port your step-12 slack code verbatim, play computer with
+`pv = { x: -3, y: 0 }` and watch what your two lines do to `y`. (That was my
+"bonus" question last session — it's still open, and one of the tests refuses
+to look away.)
+
+```sh
+bun test workshop/steps/13-crush
+```
+
+## Optional 1 — the airbag
+
+Defense in depth: even if some future caller hands `moveAndSlide` an
+overlapping box directly, it should return finite numbers — wrong-ish, maybe,
+but *finite*. You traced in step 12 exactly which value poisons the well. The
+loop already clamps it once (`Math.max(0, …)`) — find the **other** line that
+trusts `nearest.time` to be non-negative. The fix is almost nothing. Then write
+the test step 12 should have had: `moveAndSlide` (not `safe…`) with an
+overlapping start returns finite coordinates.
+
+## Optional 2 — the capstone payoff
+
+Port `resolveOverlaps` into `11-capstone/game.js` and rebuild its
+`safeMoveAndSlide` on it. Now let the ledge squeeze the hero against the pillar,
+and against the border. Watch closely: instead of vanishing, the hero should
+squirt around the ledge like a watermelon seed pinched between two fingers.
+
+Then earn the effect: the ledge is 16 tall and the hero is 12. As the ledge digs
+deeper, which of `penetrationVector`'s four escapes wins, and at what depth does
+the winner change? That flip *is* the watermelon seed.

workshop/steps/13-crush/crush.test.ts

@@ -0,0 +1,98 @@
+import { expect, test } from "bun:test";
+import { resolveOverlaps, safeMoveAndSlide } from "./crush.ts";
+import { type AABB, aabbOverlap, vec } from "./given.ts";
+
+// --- resolveOverlaps ---------------------------------------------------------
+
+test("free box -> untouched, settled", () => {
+	const r = resolveOverlaps({ x: 0, y: 0, w: 2, h: 2 }, [
+		{ x: 5, y: 5, w: 2, h: 2 },
+	]);
+	expect(r.settled).toBe(true);
+	expect(r.x).toBe(0);
+	expect(r.y).toBe(0);
+});
+
+test("single overlap -> pushed out, settled", () => {
+	const r = resolveOverlaps({ x: 9, y: 0, w: 2, h: 2 }, [
+		{ x: 10, y: -5, w: 4, h: 10 },
+	]);
+	expect(r.settled).toBe(true);
+	expect(r.x).toBeCloseTo(8);
+	// the push was x-only: the axis that needed no fixing must come back
+	// EXACTLY untouched — not "close to" untouched
+	expect(r.y).toBe(0);
+});
+
+// pushed out of A (left) -> lands inside B -> a second push (downward, B is
+// shallow) frees it. One pass can't see this coming; the negotiation can.
+const chainA: AABB = { x: 10, y: 0, w: 4, h: 4 };
+const chainB: AABB = { x: 4, y: 0, w: 4.5, h: 1.4 };
+
+test("a push that lands you in the NEXT wall still resolves", () => {
+	const hero: AABB = { x: 9, y: 1, w: 2, h: 2 };
+	const r = resolveOverlaps(hero, [chainA, chainB]);
+	expect(r.settled).toBe(true);
+	const moved = { ...hero, x: r.x, y: r.y };
+	expect(aabbOverlap(moved, chainA)).toBe(false);
+	expect(aabbOverlap(moved, chainB)).toBe(false);
+});
+
+test("...and the answer must not depend on wall order", () => {
+	const hero: AABB = { x: 9, y: 1, w: 2, h: 2 };
+	const r = resolveOverlaps(hero, [chainB, chainA]);
+	expect(r.settled).toBe(true);
+	const moved = { ...hero, x: r.x, y: r.y };
+	expect(aabbOverlap(moved, chainA)).toBe(false);
+	expect(aabbOverlap(moved, chainB)).toBe(false);
+});
+
+// the gap between these two walls is 1.5 wide; the hero is 2 wide.
+// there is NO overlap-free position — the pushes ping-pong forever.
+const vice: AABB[] = [
+	{ x: 10, y: -5, w: 4, h: 10 },
+	{ x: 6.5, y: -5, w: 2, h: 10 },
+];
+
+test("gap narrower than the box -> reports the crush, stays finite", () => {
+	const r = resolveOverlaps({ x: 9, y: 0, w: 2, h: 2 }, vice);
+	expect(r.settled).toBe(false);
+	expect(Number.isFinite(r.x)).toBe(true);
+	expect(Number.isFinite(r.y)).toBe(true);
+});
+
+// --- safeMoveAndSlide --------------------------------------------------------
+
+test("no overlap at start -> behaves exactly like moveAndSlide", () => {
+	// step 10's diagonal test: x blocked at 3, y keeps going
+	const walls: AABB[] = [{ x: 5, y: -10, w: 2, h: 40 }];
+	const p = safeMoveAndSlide({ x: 0, y: 0, w: 2, h: 2 }, vec(10, 10), walls);
+	expect(p.x).toBeCloseTo(3);
+	expect(p.y).toBeCloseTo(10);
+});
+
+test("resolvable overlap, then walking away -> freed and gone", () => {
+	const p = safeMoveAndSlide({ x: 9, y: 0, w: 2, h: 2 }, vec(-5, 0), [
+		{ x: 10, y: -5, w: 4, h: 10 },
+	]);
+	expect(p.x).toBeCloseTo(3);
+	expect(p.y).toBeCloseTo(0);
+});
+
+test("crushed with zero velocity -> a number, not a ghost", () => {
+	// the capstone bug: ledge pushes hero into the pillar, hero holds still
+	const p = safeMoveAndSlide({ x: 9, y: 0, w: 2, h: 2 }, vec(0, 0), vice);
+	expect(Number.isFinite(p.x)).toBe(true);
+	expect(Number.isFinite(p.y)).toBe(true);
+	// still pinched between the walls — not flung across the room
+	expect(p.x).toBeGreaterThan(7);
+	expect(p.x).toBeLessThan(9.5);
+});
+
+test("crushed and still pushing -> the walls win", () => {
+	const p = safeMoveAndSlide({ x: 9, y: 0, w: 2, h: 2 }, vec(5, 0), vice);
+	expect(Number.isFinite(p.x)).toBe(true);
+	expect(Number.isFinite(p.y)).toBe(true);
+	expect(p.x).toBeGreaterThan(7);
+	expect(p.x).toBeLessThan(9.5);
+});

workshop/steps/13-crush/crush.ts

@@ -0,0 +1,41 @@
+import {
+	type AABB,
+	moveAndSlide,
+	penetrationVector,
+	type Vec,
+} from "./given.ts";
+
+export type Resolution = { x: number; y: number; settled: boolean };
+
+/**
+ * Push `box` out of EVERY wall — even when one push shoves it into another.
+ *
+ * One pass over the walls is not enough: a push is a teleport, and a teleport
+ * can land you inside a wall the loop already cleared. So: run whole passes
+ * over the wall list, applying pushes (with the EPSILON slack — on the pushed
+ * axis only!), until a full pass finds nothing to fix. That pass proves you're
+ * settled.
+ *
+ * But some arrangements have NO free spot (a gap narrower than the box), and
+ * the passes would ping-pong forever. Cap them — 8 is plenty — and if the cap
+ * fires, report `settled: false`. You already believe in caps: your step-10
+ * loop has one for exactly the same reason.
+ *
+ * Returns where the box ended up and whether it truly got free.
+ */
+export function resolveOverlaps(box: AABB, walls: AABB[]): Resolution {
+	throw new Error("not implemented");
+}
+
+/**
+ * Step 12's safeMoveAndSlide, rebuilt on resolveOverlaps.
+ *
+ * Settled -> sweep from the safe spot, business as usual.
+ * Crushed (not settled) -> the sweep would be handed an overlapping box and
+ * you know exactly what it does with one of those. Don't feed it. Our crush
+ * policy: the walls win — the box stays where the resolver left it, finite,
+ * and waits for the gap to open.
+ */
+export function safeMoveAndSlide(box: AABB, v: Vec, walls: AABB[]): Vec {
+	throw new Error("not implemented");
+}

workshop/steps/13-crush/given.ts

@@ -0,0 +1,132 @@
+// The whole kernel you've built, steps 01-10, finished. Reuse it.
+export type Vec = { x: number; y: number };
+export type AABB = { x: number; y: number; w: number; h: number };
+export type Span = { entry: number; exit: number };
+export type Hit = { time: number; normal: Vec };
+
+/** A hair of slack so we stop just short of a wall instead of inside it. */
+export const EPSILON = 1e-4;
+
+export const vec = (x: number, y: number): Vec => ({ x, y });
+export const add = (a: Vec, b: Vec): Vec => ({ x: a.x + b.x, y: a.y + b.y });
+export const sub = (a: Vec, b: Vec): Vec => ({ x: a.x - b.x, y: a.y - b.y });
+export const scale = (a: Vec, s: number): Vec => ({ x: a.x * s, y: a.y * s });
+export const dot = (a: Vec, b: Vec): number => a.x * b.x + a.y * b.y;
+
+/** Step 04, finished. Strict overlap — merely touching edges is false. */
+export function aabbOverlap(a: AABB, b: AABB): boolean {
+	return (
+		a.x < b.x + b.w && b.x < a.x + a.w && a.y < b.y + b.h && b.y < a.y + a.h
+	);
+}
+
+export function sweepInterval(
+	p: number,
+	v: number,
+	min: number,
+	max: number,
+): Span | null {
+	if (v === 0) {
+		return min <= p && p <= max ? { entry: -Infinity, exit: Infinity } : null;
+	}
+	let t1 = (min - p) / v;
+	let t2 = (max - p) / v;
+	if (t1 > t2) {
+		[t1, t2] = [t2, t1];
+	}
+	return { entry: t1, exit: t2 };
+}
+
+export function rayVsAABB(p: Vec, v: Vec, box: AABB): Hit | null {
+	const spanX = sweepInterval(p.x, v.x, box.x, box.x + box.w);
+	const spanY = sweepInterval(p.y, v.y, box.y, box.y + box.h);
+	if (spanX === null || spanY === null) {
+		return null;
+	}
+	const entry = Math.max(spanX.entry, spanY.entry);
+	const exit = Math.min(spanX.exit, spanY.exit);
+	if (entry > exit || entry >= 1 || exit <= 0) {
+		return null;
+	}
+	const normal =
+		spanX.entry > spanY.entry
+			? { x: v.x > 0 ? -1 : 1, y: 0 }
+			: { x: 0, y: v.y > 0 ? -1 : 1 };
+	return { time: entry, normal };
+}
+
+export function sweptAABB(a: AABB, v: Vec, b: AABB): Hit | null {
+	const inflated: AABB = {
+		x: b.x - a.w,
+		y: b.y - a.h,
+		w: b.w + a.w,
+		h: b.h + a.h,
+	};
+	return rayVsAABB({ x: a.x, y: a.y }, v, inflated);
+}
+
+export function slide(v: Vec, normal: Vec): Vec {
+	return sub(v, scale(normal, dot(v, normal)));
+}
+
+/** Step 10, finished — your loop, exactly as you wrote it. */
+export function moveAndSlide(box: AABB, v: Vec, walls: AABB[]): Vec {
+	let pos: Vec = { x: box.x, y: box.y };
+	let vel: Vec = { x: v.x, y: v.y };
+
+	let timeLeft = 1;
+	let i = 0;
+	while (i++ < 4 && timeLeft > 0) {
+		const move = scale(vel, timeLeft);
+		let nearest: Hit = {
+			normal: { x: 0, y: 0 },
+			time: 2,
+		};
+
+		for (const wall of walls) {
+			const hit = sweptAABB(
+				{
+					x: pos.x,
+					y: pos.y,
+					w: box.w,
+					h: box.h,
+				},
+				move,
+				wall,
+			);
+			if (hit !== null) {
+				nearest = hit.time < nearest.time ? hit : nearest;
+			}
+		}
+
+		if (nearest.time === 2) {
+			pos = add(pos, move);
+			break;
+		} else {
+			pos = add(pos, scale(move, Math.max(0, nearest.time - EPSILON)));
+			vel = slide(vel, nearest.normal);
+			timeLeft = timeLeft * (1 - nearest.time);
+		}
+	}
+
+	return pos;
+}
+
+/** Step 12, finished — your penetrationVector, exactly as you wrote it. */
+export function penetrationVector(a: AABB, b: AABB): Vec | null {
+	const outLeft = a.x + a.w - b.x;
+	const outRight = b.x + b.w - a.x;
+	const outUp = a.y + a.h - b.y;
+	const outDown = b.y + b.h - a.y;
+
+	const minX = Math.min(outLeft, outRight);
+	const minY = Math.min(outUp, outDown);
+
+	if (minX > 0 && minY > 0) {
+		return minX > minY
+			? { x: 0, y: outDown > outUp ? -minY : minY }
+			: { x: outLeft > outRight ? minX : -minX, y: 0 };
+	}
+
+	return null;
+}