jpmschweitzer/settled-reach
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(simulation): lakes from settled hydrology (D-227, T-1184) #200

Closed
jpmschweitzer wants to merge 0 commits from t1184-hydrology into main
pull from: t1184-hydrology
merge into: :main
Conversation 2 Commits - Files Changed -
jpmschweitzer commented 2026-07-25 02:28:36 +02:00
Owner
Copy Link
Copy Source

Summary

Productionizes the T-1177 equilibrium solver and sources MorphologyZone::Lake from it, per the D-227 lakes amendment and D-255(f) mechanism B. run_layer1 solves hydrology once per body (~24ms) and carries it as TerrainAnalysis.hydrology; run_layer1_with_moisture threads the real body moisture ceiling. At derive time a gridunit is Lake when the bilinear-sampled settled water surface exceeds the bilinear-sampled ground — the continuous comparison, so lake edges refine with rung like coastlines. The old ocean-fraction heuristic survives byte-identically as the no-solve fallback.

Scope notes

  • The keep-always rung-0 tier does not exist until T-1181 — hydrology rides TerrainAnalysis and lands in that tier for free (lead-accepted deviation, recorded on the ticket).
  • No endorheic bit: the drains-vs-closed cue is T-1185 outlet-course presence (honest sequencing per D-227 amendment (4)). Basin spill points are computed and currently discarded — cheap to extend at the same call site.
  • Zero new wire bytes; static classification, distinct from the sim-state flooded plane.
  • Known merge seam with the parallel t1181-step-canvas branch: TerrainAnalysisCache::get_or_derive gained a body_params parameter (second-to-land rebases).

Verification

  • Acceptance: lake_classification_cache_hit_equals_cache_miss (independent double-solve, byte-identical zones, non-vacuous Lake hit) + hydrology determinism tests.
  • Golden fidelity hardened in lead review: the window golden fixture now builds TerrainAnalysis through the production entry point per body, and a dedicated lake_bowl golden body pins the hydrology-sourced Lake path (morphology 1 at ocean_fraction_q 0 — provably not the heuristic). The 108 pre-existing golden rows are byte-identical (pure append). believability.json moved one lake-shaped line (GJ338Bd voxel_relief_m 27->28, traced). river_course + derivation-harness goldens unchanged.
  • Full gate green: fmt, clippy -D warnings, 1904 lib + all integration binaries, deny. Paired-set screenshot eyeball (incl. a lake-targeted window capture) in flight — merge waits on its PASS.

Unblocks T-1185 (basin-outlet -> course-network wiring).

🤖 Generated with Claude Code

## Summary Productionizes the T-1177 equilibrium solver and sources MorphologyZone::Lake from it, per the D-227 lakes amendment and D-255(f) mechanism B. run_layer1 solves hydrology once per body (~24ms) and carries it as TerrainAnalysis.hydrology; run_layer1_with_moisture threads the real body moisture ceiling. At derive time a gridunit is Lake when the bilinear-sampled settled water surface exceeds the bilinear-sampled ground — the continuous comparison, so lake edges refine with rung like coastlines. The old ocean-fraction heuristic survives byte-identically as the no-solve fallback. ## Scope notes - The keep-always rung-0 tier does not exist until T-1181 — hydrology rides TerrainAnalysis and lands in that tier for free (lead-accepted deviation, recorded on the ticket). - No endorheic bit: the drains-vs-closed cue is T-1185 outlet-course presence (honest sequencing per D-227 amendment (4)). Basin spill points are computed and currently discarded — cheap to extend at the same call site. - Zero new wire bytes; static classification, distinct from the sim-state flooded plane. - Known merge seam with the parallel t1181-step-canvas branch: TerrainAnalysisCache::get_or_derive gained a body_params parameter (second-to-land rebases). ## Verification - Acceptance: lake_classification_cache_hit_equals_cache_miss (independent double-solve, byte-identical zones, non-vacuous Lake hit) + hydrology determinism tests. - Golden fidelity hardened in lead review: the window golden fixture now builds TerrainAnalysis through the production entry point per body, and a dedicated lake_bowl golden body pins the hydrology-sourced Lake path (morphology 1 at ocean_fraction_q 0 — provably not the heuristic). The 108 pre-existing golden rows are byte-identical (pure append). believability.json moved one lake-shaped line (GJ338Bd voxel_relief_m 27->28, traced). river_course + derivation-harness goldens unchanged. - Full gate green: fmt, clippy -D warnings, 1904 lib + all integration binaries, deny. Paired-set screenshot eyeball (incl. a lake-targeted window capture) in flight — merge waits on its PASS. Unblocks T-1185 (basin-outlet -> course-network wiring). 🤖 Generated with [Claude Code](https://claude.com/claude-code)
jpmschweitzer added 2 commits 2026-07-25 02:28:37 +02:00
feat(simulation): lakes from settled hydrology (D-227, T-1184) 24fad7090f 
Productionizes the T-1177 equilibrium solver: run_layer1 now solves
hydrology once per body (~24ms, mirrors drainage::analyze) and carries
it as TerrainAnalysis.hydrology; run_layer1_with_moisture threads the
real body moisture ceiling (extracted derive_moisture_ceiling_q), with
the T-1177 population-survey default as fallback. The resident rung-0
global tier does not exist yet (T-1181's scope) — hydrology rides
TerrainAnalysis and lands in that tier for free when it is built
(deviation recorded on the ticket).

MorphologyZone::Lake is now sourced from the settled solver at derive
time: a gridunit is Lake when bilinear-sampled filled surface exceeds
bilinear-sampled original elevation at the sample's own (px, py) — the
continuous comparison, so lake edges refine with rung like coastlines;
never a discrete basin-cell projection. The gate sits strictly between
OpenOcean (>= 80) and the old ocean_fraction heuristic (>= 60), which
survives as the derive-fresh fallback when no solve is attached —
byte-identical to pre-T-1184 output in that case. Static
classification, distinct from the sim-state flooded plane; no
endorheic bit (the drains-vs-closed cue is T-1185's outlet-course
presence, per the D-227 amendment (4) sequencing). Zero new wire
bytes.

Acceptance: lake_classification_cache_hit_equals_cache_miss (solve
twice independently, byte-identical zones, non-vacuous Lake hit) plus
hydrology determinism tests. Golden fidelity: the window golden
fixture now builds TerrainAnalysis through the production entry point
(run_layer1_with_moisture, per-body), and a dedicated lake_bowl golden
body pins the hydrology-sourced Lake path (morphology 1 at
ocean_fraction_q 0 — provably not the heuristic); the 108 pre-existing
golden rows are byte-identical (pure append). believability.json moved
by one lake-shaped line (GJ338Bd voxel_relief_m 27->28, a correctly
reclassified lake district leaving the dry-relief sample set).
river_course and derivation-harness goldens unchanged. Full suite:
2114 passed.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
chore(meta): update changelog 3f4c868daa 
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
jpmschweitzer added 2 commits 2026-07-25 02:44:04 +02:00
fix(simulation): PR #200 review round — cache sizing truth + endorheic assertion 4ef79b6b45 
Hoshe finding 1: dropping HydrologySample.elevation was PROVEN unsafe
(TerrainAnalysis.elev_pct is a rank percentile, not raw elevation;
HydrologyResult carries no elevation) — the copy stays, and the truth
moves into the docs instead: gen_queue's TerrainAnalysisCache sizing
comment corrected to real 512x256 working-grid numbers (~1.57 ->
~2.62 MB/entry, capacity-8 worst case ~21 MB), clone-on-hit cost
documented, byte-safety proof recorded on HydrologySample itself.
Arc<TerrainAnalysis> follow-up filed as T-1187.

Hoshe finding 2: the endorheic-split test now asserts the bowl basin's
BasinOutcome actually diverges (Endorheic at moisture 0, Overflow at
100, straddling ENDORHEIC_MOISTURE_CEILING=60) plus basin-count sanity
— mutation-verified by stubbing is_endorheic and watching it fail.

Doc/test-only round: goldens byte-unchanged, full suite 2114 green.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
style(simulation): cargo fmt on the review-round test (gate bounce) b222cec9be 
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
jpmschweitzer commented 2026-07-25 02:44:35 +02:00
Author
Owner
Copy Link
Copy Source

Review: t1184-hydrology -> main (type: code)

Tyre (Architecture): APPROVE

Full D-record conformance verified directly: the continuous filled-surface comparison with no discrete basin-cell lookup anywhere, D-239 §6 vocabulary not widened, no endorheic bit or wire change leaked from the superseded workshop drafts (grepped), static classification with no TTL coupling, D-255(f) mechanism B shape exact, and the one-derive-core (D-256) respected — batch inherits the lake gate through the shared core, no second sampling path. The lead-accepted scope deviation (solve inside run_layer1, resident tier deferred to T-1181) ruled architecturally sound and correctly recorded. The golden-fixture change is a pure strengthening of the T-1162 charter. Independently re-ran the new tests and both golden regressions. No comments.

Hoshe (Code Quality): REQUEST_CHANGES (2 findings — both fixed, see author response)

Core work verified solid under direct inspection: gate ordering (OpenOcean >= 80 untouched, lake strictly ahead of the >= 60 heuristic), None-fallback byte-identity, moisture threading byte-identical, D-010 determinism (bilinear comparison mirrors the solver's own strict test), and the lake_bowl fixture "genuinely rigorous" (pure-append golden, non-vacuous Lake/rim boundary).

# File Issue
1 server/src/atlas/features.rs (HydrologySample.elevation) + gen_queue.rs:205-212, :716 Unbudgeted footprint: full elevation copy stored per TerrainAnalysis, deep-cloned on every cache HIT; the LRU's load-bearing sizing comment left materially wrong.
2 server/src/atlas/layer1.rs (endorheic-split test) Test name promises "changes endorheic split" but only asserts extent equality — the split itself never checked; a regression making is_endorheic dead code would pass silently.

Author response — both fixed in 4ef79b6b4 + b222cec9b (fmt)

Finding 1: the preferred fix (drop the redundant copy) was investigated and PROVEN unsafe — TerrainAnalysis.elev_pct is a rank-percentile transform, not raw elevation, and HydrologyResult carries no elevation field; comparing filled against any existing TA field would be a silent unit mismatch. The copy therefore stays, with the truth moved into the docs: the gen_queue sizing comment corrected to real 512x256 working-grid numbers (~1.57 -> ~2.62 MB/entry, x1.67; capacity-8 worst case ~21 MB), clone-on-hit documented, and the byte-safety proof recorded on HydrologySample itself. The Arc refactor Hoshe's finding points toward is filed as T-1187.

Finding 2: the test now re-solves and asserts the bowl basin's BasinOutcome diverges — specifically Endorheic at moisture 0 and Overflow at 100 (straddling ENDORHEIC_MOISTURE_CEILING=60), plus basin-count sanity. Mutation-verified: stubbing is_endorheic makes the test fail loudly; the stub was cleanly reverted (zero residual diff).

Round was doc/test-only: all three goldens byte-unchanged, full suite 2114 green, gate green on re-push.

Verdict: CHANGES REQUESTED -> resolved; merge pending the lake-targeted screenshot eyeball (in flight)

Process context: full pre-push gate green twice (initial + fix round); paired-set screenshot protocol running with a dedicated lake-window capture on a known filled basin.

## Review: t1184-hydrology -> main (type: code) ### Tyre (Architecture): APPROVE Full D-record conformance verified directly: the continuous filled-surface comparison with no discrete basin-cell lookup anywhere, D-239 §6 vocabulary not widened, no endorheic bit or wire change leaked from the superseded workshop drafts (grepped), static classification with no TTL coupling, D-255(f) mechanism B shape exact, and the one-derive-core (D-256) respected — batch inherits the lake gate through the shared core, no second sampling path. The lead-accepted scope deviation (solve inside run_layer1, resident tier deferred to T-1181) ruled architecturally sound and correctly recorded. The golden-fixture change is a pure strengthening of the T-1162 charter. Independently re-ran the new tests and both golden regressions. No comments. ### Hoshe (Code Quality): REQUEST_CHANGES (2 findings — both fixed, see author response) Core work verified solid under direct inspection: gate ordering (OpenOcean >= 80 untouched, lake strictly ahead of the >= 60 heuristic), None-fallback byte-identity, moisture threading byte-identical, D-010 determinism (bilinear comparison mirrors the solver's own strict test), and the lake_bowl fixture "genuinely rigorous" (pure-append golden, non-vacuous Lake/rim boundary). | # | File | Issue | |---|------|-------| | 1 | server/src/atlas/features.rs (HydrologySample.elevation) + gen_queue.rs:205-212, :716 | Unbudgeted footprint: full elevation copy stored per TerrainAnalysis, deep-cloned on every cache HIT; the LRU's load-bearing sizing comment left materially wrong. | | 2 | server/src/atlas/layer1.rs (endorheic-split test) | Test name promises "changes endorheic split" but only asserts extent equality — the split itself never checked; a regression making is_endorheic dead code would pass silently. | ## Author response — both fixed in 4ef79b6b4 + b222cec9b (fmt) **Finding 1**: the preferred fix (drop the redundant copy) was investigated and PROVEN unsafe — TerrainAnalysis.elev_pct is a rank-percentile transform, not raw elevation, and HydrologyResult carries no elevation field; comparing filled against any existing TA field would be a silent unit mismatch. The copy therefore stays, with the truth moved into the docs: the gen_queue sizing comment corrected to real 512x256 working-grid numbers (~1.57 -> ~2.62 MB/entry, x1.67; capacity-8 worst case ~21 MB), clone-on-hit documented, and the byte-safety proof recorded on HydrologySample itself. The Arc<TerrainAnalysis> refactor Hoshe's finding points toward is filed as **T-1187**. **Finding 2**: the test now re-solves and asserts the bowl basin's BasinOutcome diverges — specifically Endorheic at moisture 0 and Overflow at 100 (straddling ENDORHEIC_MOISTURE_CEILING=60), plus basin-count sanity. Mutation-verified: stubbing is_endorheic makes the test fail loudly; the stub was cleanly reverted (zero residual diff). Round was doc/test-only: all three goldens byte-unchanged, full suite 2114 green, gate green on re-push. ### Verdict: CHANGES REQUESTED -> resolved; merge pending the lake-targeted screenshot eyeball (in flight) *Process context: full pre-push gate green twice (initial + fix round); paired-set screenshot protocol running with a dedicated lake-window capture on a known filled basin.*
jpmschweitzer commented 2026-07-25 03:09:59 +02:00
Author
Owner
Copy Link
Copy Source

Screenshot eyeball: PASS — merge clear

Paired-set protocol (cold release server, main vs branch, fixed scripted navigation), capture sets at .cache/screenshots/t1184-{main,eyeball}/:

Pair Surface Diff (AE) Verdict
a Cold-start staged frames x3 0 clean start, tiles arrive
b GJ380c continuity: orbital x2 + district/quarter windows at (1695,-1844) 0 no drift where no basin exists
c GJ338Bd lake window at true DistrictPos (13652,4360), district + quarter 1,967,180 (99.997%) the feature: entire 65.5 km / 16.4 km window flips dry plains (main) -> water (branch), with ~2,600 distinct colors of real bathymetric shading (not a flat fill); zoom refinement coherent across the rung boundary
d GJ338Bd whole-body orbital 20,936 (1.06%) new lake dots appear across multiple continents — global lake coverage, not just the probed basin

Lake coordinates were resolved from an aliveness_probe hit at ocean_fraction_q=0 (heuristic provably cannot fire there — hydrology-only classification) via the D-256 survey-to-district bridge. T-1186 (known latitude bug, ticketed) is shared identically by both runs and cannot pollute the pair diff. Both checkouts verified clean post-run; temp drivers deleted; no lingering processes.

Merging.

## Screenshot eyeball: PASS — merge clear Paired-set protocol (cold release server, main vs branch, fixed scripted navigation), capture sets at `.cache/screenshots/t1184-{main,eyeball}/`: | Pair | Surface | Diff (AE) | Verdict | |---|---|---|---| | a | Cold-start staged frames x3 | 0 | clean start, tiles arrive | | b | GJ380c continuity: orbital x2 + district/quarter windows at (1695,-1844) | 0 | no drift where no basin exists | | c | **GJ338Bd lake window at true DistrictPos (13652,4360), district + quarter** | 1,967,180 (99.997%) | **the feature**: entire 65.5 km / 16.4 km window flips dry plains (main) -> water (branch), with ~2,600 distinct colors of real bathymetric shading (not a flat fill); zoom refinement coherent across the rung boundary | | d | GJ338Bd whole-body orbital | 20,936 (1.06%) | new lake dots appear across multiple continents — global lake coverage, not just the probed basin | Lake coordinates were resolved from an `aliveness_probe` hit at `ocean_fraction_q=0` (heuristic provably cannot fire there — hydrology-only classification) via the D-256 survey-to-district bridge. T-1186 (known latitude bug, ticketed) is shared identically by both runs and cannot pollute the pair diff. Both checkouts verified clean post-run; temp drivers deleted; no lingering processes. Merging.
jpmschweitzer closed this pull request 2026-07-25 03:11:56 +02:00

Pull request closed

This pull request cannot be reopened because the branch was deleted.
Sign in to join this conversation.
Reviewers
No Reviewers
Labels
No items
No labels
Milestone
No items
No Milestone
Projects
Clear projects
No projects
Assignees
Clear assignees
jpmschweitzer (Jeroen Schweitzer)
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: jpmschweitzer/settled-reach#200