feat(worldgen): micro-habitat mosaic + native biospheres (T-1084/T-1085) #172
Assignees
jpmschweitzer (Jeroen Schweitzer)
Clear assignees
jpmschweitzer (Jeroen Schweitzer)
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: jpmschweitzer/settled-reach#172
Reference in New Issue
Block a user
Intra-class micro-habitat mosaic (T-1084) + native-biosphere register (T-1085), grounded in D-246, D-247, and a D-228 amendment. Mosaic mechanism (vegetation/water, section-8-safe, hydrology-gated) + grassland biome-awareness; cold-barren worlds now render structured-sparse. Biosphere: authored chirality bit (edible Compatible vs inedible-inert Mirror), Cadwal-to-Linnaeus rename (IP fix), biosphere_class DB pipeline via the stamped regen-db path (111 Compatible : 108 Mirror among inhabited alive), alien-ecology vocabulary preserved as canon. Full test suite green, clippy/ruff clean, systems.db regenerated + stamped. Non-blocking follow-ups: proper_name DB propagation, WS4 Forest biomes, Burnelli food-economy consumer.
D-246 (architecture): intra-class micro-habitat mosaic — a voxel_mosaic() sub-chunk noise band modulates the D-228 axes into spatially-coherent, relief-conditioned micro-habitats; palette keyed on surface-class × climate-biome × biosphere-register; five new D-228 axis values front-loaded for determinism; within-patch believability metric in scope. Governs T-1084. D-247 (content): habitable Reach worlds were alive before humans — the settlement timeline forbids terraforming from dead rock. An authored per-world chirality bit gates whether native life is edible (Compatible) or inedible-but-inert (Mirror); "safe" and "edible" are decoupled; colonization is ecological war that grows an Earth-crop footprint outward from settlements. Grounds D-246's native-vs-introduced register. D-228 amended: +Vegetation{Meadow,Deadfall,Lichen}, +TerrainMaterial{Hardpan,Scree}. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>Sub-chunk mosaic that modulates the D-228 axes into a spatially-coherent, relief/moisture-conditioned mosaic of micro-habitats within a terrain class — the fix for the aliveness probe's uniform "Wetland 100%". - seed.rs: SeedDomain::VoxelMosaic = 12 (append-only; pinned-discriminant guard) - voxel.rs: five new D-228 axis values (Vegetation::{Meadow,Deadfall,Lichen}, TerrainMaterial::{Hardpan,Scree}); the per-family-gated mosaic pass - detail_scatter.rs: voxel_mosaic() selection field (~8-64 m band) - atlas/mosaic.rs: surface-class palette + integer-basis-point weighted selection with moisture-affinity conditioning; temperate/native baseline palettes - data/mosaic_constants.toml: source-canonical palette mirror - believability.rs: micro_habitat_distinct metric + D-245 intra-class criterion v1 scope (D-246-faithful): modulates vegetation + water only, in the vegetated classes (Wetland/Forest/Grassland), so the D-239 §8 lithology law is untouched; water is gated on the T-1040 has_active_channel hydrology rule. Material variants (hardpan/scree) and the extreme classes are a §8-aware follow-up; the biome-aware palette variants (T-1085 vocab) extend the baseline before generation. Goldens regenerated (believability + derivation). Full server suite green. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>Review: micro-habitat-mosaic → main (type: code)
Reviewers: Hoshe (QA / code quality) + Tyre (architecture). Both ran against the branch source and the governing D-records; both confirmed findings against the branch's regenerated
systems.dband the test suites.Pre-push gates were green (fmt, clippy
-D warnings, fullcargo test, tooling tests, systems.db stamp). Process note: the branch author ran no in-game runtime smoke — not applicable here (server/tooling/data branch, no client code path).Hoshe (QA / code quality): REQUEST_CHANGES
The core Rust mosaic mechanism (
mosaic.rsselection math,voxel_mosaic()noise field, thevoxel.rsintegration guard) is correct, deterministic, and honors D-239 §8 (terrain material untouched) + the T-1040 hydrology gate — verified by reading and by running the suites (mosaic/voxel/detail_scatter units,derivation_harness,believability_harnessall pass; golden diffs are legitimate, minimal, correctly-scoped). One confirmed Python bug + a doc/code gap.tooling/economy-db/economy_import/bodies.py:310_default()gate 0 returnsAirlessfornot inhabited, conflating settlement with biosphere and contradicting D-247 ("worlds were alive before humans"). Data: 2996/3240 (92.5%) get Airless; ≥24 uninhabited bodies have a real atmosphere + liquid water (e.g.GJ0g-1: dense atm, rivers) yet are force-Airless; only 220/3240 reach the Compatible/Mirror split, so the ~50/50 target applies to a far smaller population than intended. Fix: gate onatmosphere/hydrospherealone. Also: D-247 (content.md:635) says the default derives fromeconomic_base_primary+planet_class— the code never readsplanet_class, and the substitution isn't recorded.governance/decisions/architecture.md:1885vsserver/src/atlas/voxel.rs:537applymatch hard-codes_ => falsefor them. Consequence:TerrainMaterial::{Hardpan, Scree}(2 of the 5 new D-228 values this decision adds) are never constructed on any reachable path. Legit v1 scope cut — but D-246 was never amended to say so (unlike D-239's postscripts for T-1080/81/82). Only record is a code comment + ticket history.server/src/atlas/voxel.rs:529-576applygate, thewater == Dry && has_active_channelguard, wetness conditioning). Passes only because an unmodified pre-existing test happens to exercise a Forest district. Add a purpose-built regression test.server/src/atlas/voxel.rs:512relief_signal ~[-0.5, +0.5]; real bound is ~[-1,+1](envelopesignalclamps to[0,1], can reach 1.0). Not functional (wetness is clamped) — misleading comment.tooling/economy-db/economy_import/bodies.py:261-277{NativeCompatible, NativeMirror, TerraformedSterile, Airless}enum — a quoted value or typo writes to the DB silently. Low real-world risk; worth a guard.server/data/systems.dbbodies.proper_nameforGJ166Acstill reads"Cadwal"despite the rename (already tracked as remaining work in T-1085 — flagged so it isn't lost before merge).Tyre (architecture): REQUEST_CHANGES
Core architecture is sound and faithful:
SeedDomain::VoxelMosaic=12is a clean append-only body-global domain,voxel_mosaic()reuses the T-1081 enveloped-fBm machinery correctly, selection is integer-basis-points with the sanctioned f64-truncate discipline (D-010/D-239 §4), the five new D-228 values are correctly append-only + discriminant-pinned, and the biosphere_class pipeline routes through the correct stamped regen-db path (schema+migration+consumer ship together, D-247). One blocking defect.server/src/atlas/mosaic.rs:106-119+voxel.rs:542-548ClimateBiome::from_climateclassifies every sub-zero temp asTundra(no frozen floor), and the Grassland apply-gate fires on any Barren-dominant Tundra/Boreal district with no moisture check — overridingderive_vegetationexactly where it returned Barren fortemp < -50(surface ice = geology, D-239 §2) or hyper-arid (moisture_q < 5). Data: GJ244Ad (Edict) — a frozen ice world, hydrosphere=ice, moisture 0–20 — goesvegetated_districts: 0 → 39in the committed golden (a frozen world reading 61% vegetated tundra). Violates D-245 §5, D-239 §2/§8. Fix: give the mosaic the same frozen/moisture floors asderive_vegetation(don't let it re-grant withheld cover), and regeneratebelievability.json.tooling/economy-db/economy_import/bodies.py(_default)planet_class, which D-247 names as a default input (same root as H1). Reconcile: use it, or amend D-247's "+ planet_class" phrasing to the atmosphere/hydrosphere signals actually used.server/src/atlas/believability.rs:376-380intra-class varietycriterion (micro_habitat_distinct >= 3) fails on both golden bodies (2, 1). Advisory (K deferred to Q-123), so not independently blocking — but as a raw min over patches against palettes with a dominant entry (dense stand 40/90), min ≈ 1–2 regardless of overall quality: the mechanic's own acceptance metric is structurally near-unreachable as specified. Also pools cross-class triples within a chunk (D-246 framed it same-class). Reconsider the estimator when Q-123 calibrates K.server/src/atlas/voxel.rs:523-524Verdict: CHANGES REQUESTED
Load-bearing fixes (both real, data-confirmed correctness bugs):
inhabitedpre-gate mis-classifies uninhabited-but-alive worlds asAirless(systemic); gate on atmosphere/hydrosphere alone. C2/H1 also flag the missingplanet_classinput — reconcile code ↔ D-247.Follow-through (all actionable): C3 (metric estimator), C4 + H2 (D-246 scope/§8-vegetation fidelity — amend the record), H3 (integration test), H4 (comment range), H5 (frontmatter validation). Plus the already-tracked
proper_nameDB propagation.🤖 Combined review by Hoshe + Tyre via Claude Code
Review findings addressed —
48fb749eeAll 10 items (2 blocking, 8 follow-through) fixed in
c17b4f544; CHANGELOG in0a96d6eb8. Thanks Hoshe + Tyre — both blocking bugs were real and data-confirmed.Blocking
!Barren). A climatically-barren district (frozen< -50 °C= surface ice/geology D-239 §2, or hyper-aridmoisture < 5— both resolved toVegetationClass::Barrenupstream) is skipped, so the mosaic never re-grants withheld cover. Genuine tundra (cold + moisture ≥ 15) isScrubupstream and still reaches the tundra palette via the same!Barrenpath. believability.json regenerated: Edictvegetated_districts39 → 0 (frozen ice world now reads barren, not 61 % tundra); Arbour 20 → 8 (its real cold-pole/arid districts revert to barren) with intra-class variety preserved (vegetation_classes 3, micro_habitat_distinct 2). Added 3 dedicated regression tests (H3).atmosphere/hydrosphere/planet_class, notinhabited. An uninhabited world with a real atmosphere + liquid water is alive (D-247). 24 uninhabited-alive worlds now classified (was 0);GJ0g-1(dense atm + rivers) →NativeMirror. Split is 122 Compatible : 121 Mirror; all 9 canon exemplars hold.Follow-through
planet_class(barren → Airless) as D-247's "+ planet_class" phrasing names. D-247 amended with a dated postscript recording the actual two-gate signal set (atm/hydro/planet_class habitability;economic_base_primarychirality; settlement-independence).micro_habitat_distinctcriterion relaxed to== 0 || >= 2— the min-over-patches estimator makes K=3 structurally unreachable against a dominant-entry palette. K=2 provisional; the estimator + cross-class-pooling note deferred to Q-123 (doc-commented inbelievability.rs)._ => false, soHardpan/Screeare authored-but-unreachable in v1; no elevation change in v1; and the mosaic owns the §8 climate→vegetation law for the classes it touches (no-skip + climate-appropriateness both upheld; the guaranteed riparian band is softened to a T-1040-gated probabilistic wet-habitat).voxel.rs: C1 barren-respect, §8 material-untouched + forest-textured (≥2 micro-habitats), and T-1040 no-water-without-active-channel.relief_signalcomment corrected to~[-1, +1].bodies.proper_nameforGJ166Acstill reads "Cadwal" — remains the tracked T-1085 follow-up (proper_name is atlas-CLI-owned; a wipe+reimport would drop non-proposal columns). The shippedbiosphere_classis correctlyNativeCompatible.Gates: full
cargo testgreen (incl. regenerated believability + derivation goldens),clippy --all-targets -D warningsclean,ruffclean, systems.db regenerated + stamped.🤖 Fixes by Claude Code
Tyre's re-review found the C1 !Barren guard only covered the Grassland arm — the Wetland arm was an unconditional => true. Since derive_morphology_zone has no temperature gate, a frozen + wet + flat district still yields Wetland material, so the wetland palette would paint copse/bog onto frozen marsh: the exact C1 symptom via the material path. Not a regression (the arm was unchanged) and unreachable by the 2 harness bodies, but the D-246 amendment + the voxel.rs comment already claim the guarantee universally, so the gate must match the claim. Gate is now: apply = !Barren && class in {Wetland, Forest, Grassland}. Behavior change is limited to frozen/hyper-arid Wetland districts (a temperate marsh is never Barren-dominant; Forest is Barren-free by construction). believability.json unchanged (neither harness body reaches the frozen-wet case). New regression test mosaic_pass_does_not_revegetate_frozen_wetland; D-246 amendment reworded to the unified gate. Full cargo test green; clippy -D warnings clean. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>Re-review: both APPROVE — N1 folded in (
895bb83ba)Hoshe + Tyre both APPROVE all prior findings (C1–C4, H1–H5) on re-review — confirmed by reading, DB queries, and running the suites. Hoshe even proved the H3 tests aren't vacuous by temporarily reverting the C1 gate (test failed as it should) and restoring.
Tyre surfaced one new same-class latent bug (N1), non-blocking but flagged for a lead call: the C1
!Barrenguard only covered the Grassland arm, whileWetland => truewas unconditional. Becausederive_morphology_zonehas no temperature input, a frozen + wet + flat district still yields Wetland material — so the wetland palette would paint copse/bog onto a frozen marsh (the C1 symptom via the material path). Not a regression, and unreachable by the two harness bodies.Folded in before merge rather than ticketed — because the D-246 amendment and the
voxel.rscomment already state the guarantee universally ("the mosaic never re-grants cover the climate withheld"), so leaving Wetland unguarded would ship the exact record/code drift C4/H2 just fixed. The gate is now unified:Behavior change is confined to frozen/hyper-arid Wetland districts (a temperate marsh is never Barren-dominant; Forest is Barren-free by construction).
believability.jsonunchanged (neither body reaches the frozen-wet case); new regression testmosaic_pass_does_not_revegetate_frozen_wetland; D-246 amendment reworded to the unified gate. Fullcargo testgreen, clippy-D warningsclean.Merging.
🤖 via Claude Code
Pull request closed