rev 3 of the curve redesign. The developer's verdict on the 01 baseline was that the LOWLANDS ARE GOOD; the fault is the terracing above them. So this adds a second curve mode that preserves the low plain bit-for-bit and replaces everything above the flood line with one smooth monotone climb. Core/ContinuousCurve — piecewise, and the pieces have different loyalties: - at/below sea: identity, as ever. - above sea to K2: DELEGATES to HeightCurve's own toe+red branches. Not "equivalent" — the same code path, so the same floats. Oracle (d) holds it to that. - above the ceiling: a Fritsch-Carlson (PCHIP) monotone spline to the 420 m cap, C1-joined to the red band's exit slope. Monotone by construction for any ordered control points, which retires the 24-corner sweep; a 10k strict-increase sample runs per seed anyway, because "cannot fail" is worth a millisecond. - Build() REFUSES rather than degrades: a ceiling near the old bench, a drama that folds the summit under its own onset, control-point secants that are not strictly increasing (the no-magnet rule, enforced rather than hoped for). Only BENCH_*/PLATEAU_* are dropped. SEA/ORANGE_CEIL/RED_CEIL survive because they are the storm-ladder FLOOD TIERS and they live inside the preserved lowland; PEAK_CAP and the per-seed spikeMax normalization survive as the summit. Shelf detail is forced off in continuous mode: the flat benches it de-slabbed no longer exist, and painting noise on the climb now would pre-judge what erosion should carve. Oracle, all hard checks passing: - (a1) curve off is bit-identical to Phase 1's dump. - (a2) staircase mode is bit-identical to TASK 01's dump — the control is provably the control, not a re-derivation. (CurveBaselineTool is pinned to Staircase so the config default moving to Continuous cannot drift it.) - (d) lowlands bit-identical to the staircase over 3.6M cells, every continuous variant, both seeds. The lifted_WRONG bookend fails it on 1.6M cells, as intended. - (f) sea identity per CELL, not per count, including 67M cells at 8192. The finding, measured and recorded in the batch scratch: the massif SHRANK. Land above 100 m goes 14.9% -> 4.8%, above 220 m 4.5% -> 0.6%. A feather sweep to the practical floor recovers ~1.3 points, so this is structural, not a tuning miss: the staircase's highland area was an artifact of the bench and plateau acting as magnets, and a curve with no magnets preserves the raw distribution's bottom-heavy shape. "No terraces" and "the same land up high" are not both available from curve work alone. Exploration batch, not convergence. A tuning pass follows once a direction is picked. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DCWNaDZPfTiAy3meGNGgqt
159 lines
6.7 KiB
C#
159 lines
6.7 KiB
C#
using System.Collections.Generic;
|
|
using IslaApocalypse.Core;
|
|
|
|
namespace IslaApocalypse.Tools
|
|
{
|
|
/// <summary>
|
|
/// Everything pass 2 produces: THE TWO HEIGHT FIELDS. → <see cref="Shaping"/>.
|
|
///
|
|
/// ═══ ⭐⭐ THE TWO-FIELD SPLIT (D-046) — THE DISCIPLINE THIS CLASS EXISTS TO HOLD ═══
|
|
///
|
|
/// <see cref="HeightClassify"/> RAW. Uncurved, un-detailed. The ORACLE.
|
|
/// <see cref="Height"/> RENDER. Curved, detailed, and later eroded and carved.
|
|
///
|
|
/// Everything that CLASSIFIES the world — biomes, water bodies, the ocean flood fill — reads the
|
|
/// classify field. Everything that DRAWS or MESHES it reads the render field. The reference's
|
|
/// hardest-won lesson is that this split is what made five rounds of taste-iteration safe: the
|
|
/// biome and water maps stayed md5-identical across every shaping change, so correctness was
|
|
/// never being judged by eye. → `Design - Tooling - Iteration and Batching.md`,
|
|
/// "build the oracle before the taste-iteration".
|
|
///
|
|
/// ⚠ NOTHING CONSUMES THE CLASSIFY FIELD YET. No water, no biomes exist in this phase. The split
|
|
/// is established HERE, at the curve, because the curve is where the second field is BORN — and
|
|
/// retrofitting a classify path after three passes already ran on one array is how the two
|
|
/// silently diverge. The field is produced and asserted now so that when water lands it has
|
|
/// something correct to read.
|
|
///
|
|
/// ═══ ⚠ WHEN THE TWO FIELDS ARE THE SAME ARRAY ═══
|
|
///
|
|
/// With the curve OFF there is nothing to separate, so both properties reference ONE array —
|
|
/// exactly as the reference did (<c>_heightMapClassify = (_curveOn || _erosionOn) ? new float[…]
|
|
/// : _heightMap</c>). <see cref="FieldsAreAliased"/> says so out loud, because a later pass that
|
|
/// writes through one reference while reading the other MUST know: the reference's crater carve
|
|
/// reads both into locals before writing either for precisely this reason, and that is the trap
|
|
/// this flag is here to keep visible until the carve lands.
|
|
/// </summary>
|
|
public sealed class Pass2Result
|
|
{
|
|
/// <summary>Map side in columns.</summary>
|
|
public readonly int MapSize;
|
|
|
|
/// <summary>The resolved seed. Same seed, same two fields.</summary>
|
|
public readonly int Seed;
|
|
|
|
/// <summary>
|
|
/// ⭐ THE RENDER FIELD, <c>[x, y]</c> — curved and detailed. What gets drawn, dumped and
|
|
/// (later) eroded, carved and meshed.
|
|
/// </summary>
|
|
public readonly float[,] Height;
|
|
|
|
/// <summary>
|
|
/// ⭐ THE CLASSIFY FIELD, <c>[x, y]</c> — bit-for-bit the raw pre-curve pass-1 height.
|
|
///
|
|
/// ⚠ Nothing may write to this after pass 2 except the crater carve, which is the one pass
|
|
/// that legitimately moves both fields. Erosion, rivers and detail are render-only.
|
|
/// </summary>
|
|
public readonly float[,] HeightClassify;
|
|
|
|
/// <summary>Was the curve applied? The primary A/B gate.</summary>
|
|
public readonly bool CurveOn;
|
|
|
|
/// <summary>
|
|
/// Which curve shaped the render field: "off", "staircase", "continuous" or "lifted_WRONG".
|
|
/// A label, not logic — the oracle and the INDEX read it so a result can never be mistaken
|
|
/// for the wrong mode's.
|
|
/// </summary>
|
|
public readonly string CurveModeLabel;
|
|
|
|
/// <summary>
|
|
/// The batch variant this result belongs to, carried straight from
|
|
/// <see cref="TerrainGenConfig.VariantLabel"/>.
|
|
///
|
|
/// ⚠ CARRIED, NOT INFERRED. The first cut of the chat2/02 tool reconstructed this from the
|
|
/// knob values ("ceiling > 31 ⇒ hold_higher"), which works only for exactly today's
|
|
/// variant set: add a second variant sharing a knob value and two different runs silently
|
|
/// write to one folder. The label is an identity, so it travels with the result.
|
|
/// </summary>
|
|
public readonly string VariantLabel;
|
|
|
|
/// <summary>
|
|
/// The per-seed continuous spline, when <see cref="CurveModeLabel"/> is "continuous" — the
|
|
/// oracle samples its slopes (check e) and the INDEX prints its control points. Null for
|
|
/// every other mode.
|
|
/// </summary>
|
|
public readonly ContinuousCurve Continuous;
|
|
|
|
/// <summary>Was shelf detail applied? Requires <see cref="CurveOn"/> — it warps the curve's knots.</summary>
|
|
public readonly bool DetailOn;
|
|
|
|
/// <summary>The knot set used. Null when the curve is off.</summary>
|
|
public readonly CurveKnots Knots;
|
|
|
|
/// <summary>The output anchors used. Null when the curve is off.</summary>
|
|
public readonly CurveAnchors Anchors;
|
|
|
|
/// <summary>The per-seed spike input, carried through from pass 1.</summary>
|
|
public readonly float HMaxSeed;
|
|
|
|
/// <summary>The edge-warp amplitude actually APPLIED, raw units (post-clamp). Zero when detail is off.</summary>
|
|
public readonly float EdgeAmpRaw;
|
|
|
|
/// <summary>The knot set's safe warp bound, raw units — what <see cref="EdgeAmpRaw"/> was clamped to.</summary>
|
|
public readonly float MaxEdgeShiftRaw;
|
|
|
|
/// <summary>Render-field extremes after shaping. For the ramp and the report.</summary>
|
|
public readonly float HMin, HMax;
|
|
|
|
/// <summary>Wall-clock milliseconds pass 2 took.</summary>
|
|
public readonly ulong ElapsedMs;
|
|
|
|
/// <summary>
|
|
/// Lines worth printing: the monotonicity confirmation, any loud clamp. Collected rather than
|
|
/// printed inside the pass so the shaping code stays a pure function of its inputs and the
|
|
/// tool owns the console.
|
|
/// </summary>
|
|
public readonly List<string> Notes;
|
|
|
|
public Pass2Result(int mapSize, int seed, float[,] height, float[,] heightClassify,
|
|
bool curveOn, bool detailOn, string curveModeLabel, string variantLabel,
|
|
ContinuousCurve continuous, CurveKnots knots, CurveAnchors anchors, float hMaxSeed,
|
|
float edgeAmpRaw, float maxEdgeShiftRaw, float hMin, float hMax, ulong elapsedMs,
|
|
List<string> notes)
|
|
{
|
|
MapSize = mapSize;
|
|
Seed = seed;
|
|
Height = height;
|
|
HeightClassify = heightClassify;
|
|
CurveOn = curveOn;
|
|
DetailOn = detailOn;
|
|
CurveModeLabel = curveModeLabel;
|
|
VariantLabel = variantLabel;
|
|
Continuous = continuous;
|
|
Knots = knots;
|
|
Anchors = anchors;
|
|
HMaxSeed = hMaxSeed;
|
|
EdgeAmpRaw = edgeAmpRaw;
|
|
MaxEdgeShiftRaw = maxEdgeShiftRaw;
|
|
HMin = hMin;
|
|
HMax = hMax;
|
|
ElapsedMs = elapsedMs;
|
|
Notes = notes;
|
|
}
|
|
|
|
/// <summary>
|
|
/// ⚠ True when the two fields ARE the same array (curve off). Any pass that writes one while
|
|
/// reading the other must read both into locals first. See the type header.
|
|
/// </summary>
|
|
public bool FieldsAreAliased => ReferenceEquals(Height, HeightClassify);
|
|
|
|
/// <summary>Fraction of the RENDER field at or above the sea threshold.</summary>
|
|
public float LandFraction(float seaLevel)
|
|
{
|
|
long land = 0;
|
|
for (int x = 0; x < MapSize; x++)
|
|
for (int y = 0; y < MapSize; y++)
|
|
if (Height[x, y] >= seaLevel) land++;
|
|
return land / (float)((long)MapSize * MapSize);
|
|
}
|
|
}
|
|
}
|