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

feat(ui): river source tapering + width-grammar retune — map fluency pass (T-1175) #207

Closed
jpmschweitzer wants to merge 0 commits from atlas-map-polish into main
pull from: atlas-map-polish
merge into: :main
Conversation 2 Commits - Files Changed -
jpmschweitzer commented 2026-07-25 15:28:26 +02:00
Owner
Copy Link
Copy Source

Summary

The single large map-reading polish pass Jeroen requested off the T-1170 captures, benchmarked against RimWorld's world-map fluency (docs/design/references/rimworld-world-map-fluency.jpg).

Source tapering — courses draw as per-vertex-width ribbons instead of uniform draw_polyline: width ramps from a hairline at the upstream source to full class width over 15% of the course's arc length (arc-length parameterized so point density doesn't change the read; mitred joins; degenerate zero-length and two-point courses handled), then holds to the mouth. points[0]-is-source confirmed against the server contract and the existing mouth-ring draw.

Width grammar — class table retuned 0.9/1.4/2.2 → 0.6/1.2/2.4: a clean ~2× per-class ladder so a tributary-joins-trunk confluence reads as a join, not a color change on a uniform line. Trunk held at its visually-proven weight; stream thinner per the benchmark's thin/consistent/restrained grammar. Opacity table untouched (Araminta's T-1170 ruling stands); hue left alone (her prior ruling, reasoning documented inline).

Coast/shore gradient — resolved as already-correct, no change: the coastal transition morphology zones + elevation lightness render the shoreline band; capture-verified.

Biome/relief stipple — assess-only per scope: the design write-up (D-255(e)-legal texture-space dithering of already-served vegetation/elev planes, with the three open design questions) is preserved as T-1194.

Verification

Annotation-layer suite 24/24 (8 new taper-geometry tests: source hairline, full-width hold, monotone ramp, degenerate/two-point courses, vertex parity, normal perpendicularity); gdlint clean; full gate green on push. Author self-capture confirms the taper reads at District rung on live data. The formal eyeball on the final tip includes a confluence spot-check (the one frame the author's search budget didn't land) — merge waits on it.

🤖 Generated with Claude Code

## Summary The single large map-reading polish pass Jeroen requested off the T-1170 captures, benchmarked against RimWorld's world-map fluency (docs/design/references/rimworld-world-map-fluency.jpg). **Source tapering** — courses draw as per-vertex-width ribbons instead of uniform draw_polyline: width ramps from a hairline at the upstream source to full class width over 15% of the course's arc length (arc-length parameterized so point density doesn't change the read; mitred joins; degenerate zero-length and two-point courses handled), then holds to the mouth. `points[0]`-is-source confirmed against the server contract and the existing mouth-ring draw. **Width grammar** — class table retuned 0.9/1.4/2.2 → 0.6/1.2/2.4: a clean ~2× per-class ladder so a tributary-joins-trunk confluence reads as a join, not a color change on a uniform line. Trunk held at its visually-proven weight; stream thinner per the benchmark's thin/consistent/restrained grammar. Opacity table untouched (Araminta's T-1170 ruling stands); hue left alone (her prior ruling, reasoning documented inline). **Coast/shore gradient** — resolved as already-correct, no change: the coastal transition morphology zones + elevation lightness render the shoreline band; capture-verified. **Biome/relief stipple** — assess-only per scope: the design write-up (D-255(e)-legal texture-space dithering of already-served vegetation/elev planes, with the three open design questions) is preserved as **T-1194**. ## Verification Annotation-layer suite 24/24 (8 new taper-geometry tests: source hairline, full-width hold, monotone ramp, degenerate/two-point courses, vertex parity, normal perpendicularity); gdlint clean; full gate green on push. Author self-capture confirms the taper reads at District rung on live data. The formal eyeball on the final tip includes a confluence spot-check (the one frame the author's search budget didn't land) — merge waits on it. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
jpmschweitzer added 2 commits 2026-07-25 15:28:26 +02:00
feat(ui): river source tapering + width-grammar retune — map fluency pass (T-1175) 40cf89c3cb 
The single large polish pass Jeroen requested off the T-1170 captures,
benchmarked against RimWorld's world-map fluency. Courses now draw as
source-tapered ribbons: per-vertex width ramps from a hairline at the
upstream source to full class width over 15% of the course's arc length
(arc-length parameterized, not vertex-indexed, so point density doesn't
change the read), built as one draw_polygon ribbon with mitred joins.
Class width table retuned 0.9/1.4/2.2 -> 0.6/1.2/2.4: a clean ~2x
per-class ladder so a tributary-joins-trunk confluence reads as a join,
trunk held at its visually-proven weight, stream thinner per the
benchmark's thin/consistent/restrained grammar. Opacity untouched
(Araminta's T-1170 ruling stands). Coast-gradient item resolved as
already-correct (the coastal transition zones + elevation lightness
render the shoreline band; capture-verified) — no wire change. Stipple
assessment filed as T-1194. Eight new taper-geometry tests.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
chore(meta): update changelog 28d2e6c80b 
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
jpmschweitzer added 1 commit 2026-07-25 16:07:39 +02:00
fix(ui): PR #207 review round — crop-gated taper, true mitres, hybrid AA (T-1175) 686021bee8 
Five findings from the araminta+hoshe round. Tapering now fires only on
a TRUE upstream source: the course's first raw world point is tested
against the canvas's own world bounds (conservative 1m epsilon) —
exact detection because the server crop keeps one point beyond the
window (layer_proxy crop_course_to_window lo = first_in-1, contract
documented), so crop passthroughs draw the old flat full-width cut and
never a false headwater. The averaged-normal joint is replaced by a
real mitre (half_w/cos(theta/2) recovered trig-free via the bisector
normal), clamped by a 2x mitre limit AND 0.45x the shorter adjacent
segment — restoring true perpendicular width at bends (the 29% pinch at
confluences is gone) and preventing the hairpin bowtie; the winding doc
now states the actual bounded guarantee. Antialiasing restored via the
hybrid: only the varying-width taper head draws as a ribbon; the
constant-width ~85% of every course keeps the original antialiased
draw_polyline (byte-identical for untapered courses), split at an
interpolated arc-length point sharing position and width — junction
capture evidence in .cache/screenshots/t1175-fix-round/. Flat-fill
single-element color array. Suite 24 -> 48 tests.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
jpmschweitzer commented 2026-07-25 16:08:33 +02:00
Author
Owner
Copy Link
Copy Source

Review: atlas-map-polish -> main (type: visual)

Reviewers: Hoshe (QA) + Araminta (art direction). Both REQUEST_CHANGES; all five findings addressed in 686021bee.

Araminta (art direction): REQUEST_CHANGES

# Issue Resolution
1 False headwaters at window-crop edges — taper fired unconditionally on points[0], but the server crops the upstream end silently (no source-side CourseTerminus analog); a mid-river fragment entering the window would draw a fake spring at the crop line Fixed, no wire change needed — detection is exact client-side: the crop keeps one point beyond the window (lo = first_in - 1), so a cropped course's first raw world point lies outside the canvas bounds while a true source lies inside. Taper gates on that test (canvas-space, conservative 1 m epsilon); crop passthroughs draw the old flat full-width cut — her stated preferred default. The one-point-overhang contract this relies on is documented against the server crop so a future change there knows it breaks the gate
2 AA loss (shared with Hoshe #3) See below

Her confirmations on record: width ladder 0.6/1.2/2.4 correct; 15% taper length right when legitimately firing; opacity/hue rulings stand; declined opacity-along-run ramp seconded.

Hoshe (QA): REQUEST_CHANGES

# Issue Resolution
1 "Always simple polygon" claim false — tight hairpin (turn radius < half-width) self-intersects the ribbon, verified numerically Fixed — offset now clamped by 0.45× the shorter adjacent segment (the clamp that actually prevents the bowtie) on top of the 2× mitre limit; doc rewritten to the real bounded guarantee
2 Mitre under-widens by cos(θ/2) at bends — 29% pinch at 90°, right at the confluence moment the retune sells; bend case untested Fixed by the same change — true mitre length half_w/cos(θ/2) (recovered trig-free via the bisector normal, near-fold guard), restoring exact perpendicular width; new bend tests assert the projection onto each adjacent segment's normal equals half_w
3 Silent AA regression — draw_polyline(antialiased=true) → draw_polygon (no AA parameter) Fixed via the hybrid — only the varying-width taper head draws as a ribbon; the constant-width ~85% of every course keeps the original antialiased draw_polyline (byte-identical for untapered courses). Split at an interpolated arc-length point (not index-snapped — station spacing can exceed the taper window at deep rungs) sharing exact position and width: butt joint, no gap/overlap. Junction verified in a live capture at 7× zoom — clean join, smooth AA diagonals: .cache/screenshots/t1175-fix-round/district_overview.png (on disk this time)
4 N-element PackedColorArray for a flat fill Fixed — single-element array

Verification

Annotation suite 24 → 48 tests (crop gate, head/tail split, mitred offsets, hairpin clamp), all green; gdlint clean; full gate green on push. The eyeball on this tip carries the two visual proofs still owed: a confluence in frame (width-grammar) and a crop-edge course drawing flat.

Verdict: CHANGES REQUESTED -> addressed in 686021bee; merge follows the eyeball.

# Review: atlas-map-polish -> main (type: visual) Reviewers: Hoshe (QA) + Araminta (art direction). Both REQUEST_CHANGES; all five findings addressed in `686021bee`. ### Araminta (art direction): REQUEST_CHANGES | # | Issue | Resolution | |---|-------|------------| | 1 | **False headwaters at window-crop edges** — taper fired unconditionally on points[0], but the server crops the upstream end silently (no source-side CourseTerminus analog); a mid-river fragment entering the window would draw a fake spring at the crop line | **Fixed, no wire change needed** — detection is exact client-side: the crop keeps one point *beyond* the window (`lo = first_in - 1`), so a cropped course's first raw world point lies outside the canvas bounds while a true source lies inside. Taper gates on that test (canvas-space, conservative 1 m epsilon); crop passthroughs draw the old flat full-width cut — her stated preferred default. The one-point-overhang contract this relies on is documented against the server crop so a future change there knows it breaks the gate | | 2 | AA loss (shared with Hoshe #3) | See below | Her confirmations on record: width ladder 0.6/1.2/2.4 correct; 15% taper length right *when legitimately firing*; opacity/hue rulings stand; declined opacity-along-run ramp seconded. ### Hoshe (QA): REQUEST_CHANGES | # | Issue | Resolution | |---|-------|------------| | 1 | "Always simple polygon" claim false — tight hairpin (turn radius < half-width) self-intersects the ribbon, verified numerically | **Fixed** — offset now clamped by 0.45× the shorter adjacent segment (the clamp that actually prevents the bowtie) on top of the 2× mitre limit; doc rewritten to the real bounded guarantee | | 2 | Mitre under-widens by cos(θ/2) at bends — 29% pinch at 90°, right at the confluence moment the retune sells; bend case untested | **Fixed by the same change** — true mitre length `half_w/cos(θ/2)` (recovered trig-free via the bisector normal, near-fold guard), restoring exact perpendicular width; new bend tests assert the projection onto *each* adjacent segment's normal equals half_w | | 3 | Silent AA regression — draw_polyline(antialiased=true) → draw_polygon (no AA parameter) | **Fixed via the hybrid** — only the varying-width taper head draws as a ribbon; the constant-width ~85% of every course keeps the original antialiased draw_polyline (byte-identical for untapered courses). Split at an *interpolated* arc-length point (not index-snapped — station spacing can exceed the taper window at deep rungs) sharing exact position and width: butt joint, no gap/overlap. Junction verified in a live capture at 7× zoom — clean join, smooth AA diagonals: `.cache/screenshots/t1175-fix-round/district_overview.png` (on disk this time) | | 4 | N-element PackedColorArray for a flat fill | **Fixed** — single-element array | ### Verification Annotation suite 24 → 48 tests (crop gate, head/tail split, mitred offsets, hairpin clamp), all green; gdlint clean; full gate green on push. The eyeball on this tip carries the two visual proofs still owed: a confluence in frame (width-grammar) and a crop-edge course drawing flat. ### Verdict: CHANGES REQUESTED -> addressed in `686021bee`; merge follows the eyeball.
jpmschweitzer commented 2026-07-25 16:40:50 +02:00
Author
Owner
Copy Link
Copy Source

Eyeball — PASS on all three owed judgments (merge gate cleared)

Live runs on 686021bee, GJ251c, targets constructed from the real drainage arrays (a scratch probe walking confluences/river_class/fdir, cross-checked against the server's own inflow computation) rather than blind panning.

  • Confluence (width grammar + mitre): at a genuine class-transition join, column-scan measurement shows a flat 2px stream run jumping to a flat 4px trunk run immediately post-join — the exact 2× ladder — with no pinch at the bend (never dips below pre-join width). confluence.png / -zoom.png.
  • Crop-edge flat cut (araminta's finding): a window constructed to bisect a known course; the cropped course's overhang point confirmed ~306 m outside the window bound (the one-station overhang the gate relies on); pixel-flat 2px width from the crop line — no false headwater. crop-edge.png.
  • True-source taper + AA junction: a genuine zero-inflow headwater (104 found on the body) centered in-window tapers in cleanly (renders as a coverage ramp at this class's 2px width — correct); the ribbon-to-polyline junction is invisible at 1×; diagonal spans show clean AA, no stairstepping. taper-source.png, diagonal-aa-zoom.png.
  • Zero errors across 12 driver runs + server logs; process table confirmed clean afterward; worktree source untouched.

Captures: .cache/screenshots/t1175-eyeball/. Merging.

# Eyeball — PASS on all three owed judgments (merge gate cleared) Live runs on `686021bee`, GJ251c, targets *constructed* from the real drainage arrays (a scratch probe walking `confluences`/`river_class`/`fdir`, cross-checked against the server's own inflow computation) rather than blind panning. - **Confluence (width grammar + mitre):** at a genuine class-transition join, column-scan measurement shows a flat 2px stream run jumping to a flat 4px trunk run immediately post-join — the exact 2× ladder — with **no pinch** at the bend (never dips below pre-join width). `confluence.png` / `-zoom.png`. - **Crop-edge flat cut (araminta's finding):** a window constructed to bisect a known course; the cropped course's overhang point confirmed ~306 m outside the window bound (the one-station overhang the gate relies on); pixel-flat 2px width from the crop line — **no false headwater**. `crop-edge.png`. - **True-source taper + AA junction:** a genuine zero-inflow headwater (104 found on the body) centered in-window tapers in cleanly (renders as a coverage ramp at this class's 2px width — correct); the ribbon-to-polyline junction is invisible at 1×; diagonal spans show clean AA, no stairstepping. `taper-source.png`, `diagonal-aa-zoom.png`. - Zero errors across 12 driver runs + server logs; process table confirmed clean afterward; worktree source untouched. Captures: `.cache/screenshots/t1175-eyeball/`. Merging.
jpmschweitzer closed this pull request 2026-07-25 16:43:55 +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#207