feat(simulation): D-243 region-tier finish — region climate stack + cross-district blending (T-1078, T-1042) #167
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#167
Reference in New Issue
Block a user
Batch A from /whats-next — finishes the D-243 region tier on top of the just-landed re-scale (T-1077) and Region→Chunk→Voxel wiring (T-1046).
T-1078 — Region climate stack (D-243 §3/§4)
region_profile.rs: region (~205 km) climate tier above the DistrictProfile carrier —RegionProfile/RegionClock,derive_region_baseline_c(latitude lerp + greenhouse, D-240 class-envelope, no orbit/star inputs), andregion_baseline_at_district— the D-243 §4 edge-fuzz warp-perturbed bilinear blend so the ~205 km grid is invisible.derive_district_temperature_cnow modulates the region baseline with elevation lapse only.build_district_profiletakes anOption<f32>baseline;Nonepreserves the legacy path (existing call sites unchanged).derive_coverconsumes the post-refactor value — no signature change.T-1042 — Cross-district parameter blending (D-239 §4/§7/§8)
ChunkContextgainssecondary: Option<DistrictProfile>+blend_weight;derive_voxel_columninteger-blendselev_q/moisture_qbefore material selection, reusing the warp offset so the seam can't align with a chunk edge.The two are independent mechanisms (climate seam 205 km vs terrain seam 2 km) — they do not share the blend field.
Verification
cargo test(full): 1547 lib + 37 harness + all integration — 0 failures.--all-targets -D warningsclean; fmt clean.🤖 Generated with Claude Code
Phase 2 of the D-243 re-scale. The ~1km carrier is renamed and re-scaled to the 2km district, and every consumer reads scale from the canonical scale.rs instead of a local literal. - region_profile.rs -> district_profile.rs; RegionProfile -> DistrictProfile, RegionPos -> DistrictPos (re-exported from scale::DistrictPos), derive_*_regions -> derive_*_districts, across all 10 consumers + the derivation harness. - chunk_context now references scale:: for the carrier cell: district 1024m -> 2048m, shift 4 -> 5 (CHUNK_DISTRICT_SHIFT). This caught a real latent bug: the seed-district index used a literal '>> 4' while the anchor used the constant, so same-district chunks could derive different anchors — now both use scale::CHUNK_DISTRICT_SHIFT. - derivation_harness scale literals ('>> 4', '* 1024', '0..16') converted to scale:: constants (the same drift the canonical source eliminates). Believability tests pass at 2km; the voxel determinism golden re-pinned (anchors moved with the re-scale — deterministic, intended). Full suite + clippy --all-targets -D warnings green. The district DERIVATION still tiles the heightmap at 8 cells (the old coarse path); replacing it with on-demand heightmap interpolation + detail-scatter is phase 3. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>Client half — the cascade's district tier is now visible in the Atlas map. - protocol.gd: decode the district_grid field from AtlasLayerResponse. - atlas_viewer.gd: a gen_district ('MRPH') toggle in OVERLAY_DEFS + the district-grid generation state (set/get) + wire it from the Ready response. - atlas_marker_overlay.gd: _draw_gen_district() paints the coarse cols×rows grid, each cell coloured by its MorphologyZone discriminant (D-239 §6, 17-zone palette), semi-transparent under the Layer-1 line overlays. Planetary map view (not the 2km on-demand districts — those are Phase 5 in-world). gdUnit4 test: overlay registered + district grid round-trips through the viewer and the protocol decode. Visual tuning of the palette can follow once eyeballed in the running Atlas. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>PR #167 Review —
region-climate-district-blend→ main (code)Reviewers: Hoshe (code quality) + Tyre (architecture), both on the branch via
git show.Verdict: CHANGES REQUESTED (all comments actionable per review policy).
Strengths: integer blend lerp is overflow-safe and symmetric; the Cow interior fast-path preserves far-from-border determinism bit-for-bit; the two-phase temperature split is correct; D-240 class-envelope anchoring is enforced (no orbit/star inputs); morphology family seams stay sharp (D-239 §7); integer discipline (D-010) maintained.
Hoshe (Code Quality) — REQUEST_CHANGES
server/src/atlas/voxel.rs—VoxelCachedoc +get_or_deriveserver/src/atlas/chunk_context.rs—sec_phase(~L302)primary_meander_phase / 2, which varies with the primary district's per-position seed (biases the blend toward primary by ~25% at w=128). Cosmetic (boundary meander phase only), but the comment is actively misleading — fix the comment or drop the term.server/src/atlas/voxel.rstests /derivation_harness.rsVoxelCache.get_or_derivewith a blended boundary chunk (secondary=Some(...),blend_weight=128). The harness T-1042 tests callderive_voxel_columndirectly, bypassing the cache. Add a test that caches a boundary tile twice and asserts identical blended elevation — guards against future cache-key drift and exercises the Cow::Owned path through the cache.Tyre (Architecture) — REQUEST_CHANGES
server/src/atlas/district_profile.rs—derive_district_profile(~L1038) &derive_district(~L1220)region_baseline_c: None, soregion_baseline_at_district(the D-243 §4 edge-fuzz blend) is never consulted by any cascade path (cascade.rshas 0 references) — district temperature is still derived from scratch, not as a modulation of the region baseline (D-243 §3's core deliverable). TheOption<f32>seam andregion_baseline_at_districtexist but are dead in production. Either thread the region cache through now, or file a tracked ticket (status ready/in_progress) and reference its ID in theNonecomments so the wiring isn't deferred indefinitely as T-1078 closes.server/src/atlas/chunk_context.rs—sec_phase(~L302)server/src/atlas/region_profile.rs—derive_region_baseline_c/build_region_profileBodyParams.district_latitude_degis repurposed to mean the region centre latitude — a call-site trap on thepubderive_region_baseline_c(field name says "district"). Rename the field tolatitude_deg, introduce aRegionParamsnewtype, or at minimum add a precondition/# Panicsdoc on the callee.Disposition to follow.
Review findings addressed —
ff5e7ee94All six findings resolved (commits
d974cf42bT-1078,ff5e7ee94T-1042). Full gate re-run green:cargo test1548 lib + 37 harness + all integration (0 failures), clippy--all-targets -D warningsclean, fmt clean.derive_district_profile(cascade) andderive_district(on-demand) now computeregion_baseline_at_district(...)and pass it tobuild_district_profile, so the two-phase modulation (D-243 §3) + edge-fuzz blend (§4) run live. Cascade builds aBTreeMap<RegionPos, RegionProfile>(all district positions + ±1 neighbours) once per body; on-demand derives the 4 surrounding baselines on the fly.Noneonly for genuinely airless bodies.get_or_derive: caller must pass theChunkContextof the chunk that contains the tile.sec_phasemisleading / primary-contaminated+ primary_meander_phase/2term;sec_phaseis now the secondary's own morphology scalars (slope_q + moisture_q). Comment rewritten to name it a bounded cosmetic approximation and where to thread the real secondary seed.cache_boundary_tile_blend_is_stable— boundaryChunkContext(blend_weight=128,secondary=Some), twoget_or_derivecalls, asserts cache hit + identical blended elevation + that theCow::Ownedpath ran.district_latitude_degcall-site trapBodyParams.district_latitude_deg → latitude_deg(dual-scale), propagated to all sites.Note on the golden fixture: no re-pin needed — the harness golden constructs districts directly via
make_region(explicit temperatures), bypassingbuild_district_profile, so the production temperature change doesn't reach it. 37/37 harness green.Pull request closed