feat(d3): implement ternary-basic

[github-actions]

Implementation: ternary-basic - javascript/d3

Implements the javascript/d3 version of ternary-basic.

File: plots/ternary-basic/implementations/javascript/d3.js

Parent Issue: #1001

---

🤖 impl-generate workflow

[claude]

AI Review - Attempt 1/3

Image Description

Light render (plot-light.png): Warm off-white background (#FAF8F1-consistent). Bold dark title "ternary-basic · javascript · d3 · anyplot.ai" centered at top, clearly readable. Equilateral triangle outlined in ink-soft stroke, bold vertex labels "Clay" (top), "Sand" (bottom-left), "Silt" (bottom-right). Tick labels (0/20/40/60/80/100) along all three edges are readable, except at the three vertices where two tick labels from different axes land on the exact same pixel position, producing a visibly doubled/bolder "100" with faint chromatic-fringe ghosting under zoom. Grid lines at 20% intervals in three directions, subtle and non-competing. 28 green (#009E73) circles with a light cream stroke, fill-opacity 0.85, clustered mid-lower in the triangle. All text is readable against the light background.

Dark render (plot-dark.png): Warm near-black background (#1A1A17-consistent). Title, vertex labels, and tick labels all render in light ink tokens — fully legible against the dark background, no dark-on-dark failures anywhere. The same vertex tick-label doubling artifact reproduces identically (it's a geometry bug in the tick computation, not a theme problem). Data colors are identical to the light render (#009E73); only the marker stroke flips from cream to a dark tone to match the page background.

Both renders pass the readability check; the vertex tick-label overlap is a real but minor defect, not a legibility failure.

Score: 80/100

Category	Score	Max
Visual Quality	26	30
Design Excellence	10	20
Spec Compliance	15	15
Data Quality	12	15
Code Quality	10	10
Library Mastery	7	10
Total	80	100

Visual Quality (26/30)

• VQ-01: Text Legibility (7/8)
• VQ-02: No Overlap (4/6)
• VQ-03: Element Visibility (6/6)
• VQ-04: Color Accessibility (2/2)
• VQ-05: Layout & Canvas (4/4)
• VQ-06: Axis Labels & Title (1/2)
• VQ-07: Palette Compliance (2/2)

Design Excellence (10/20)

• DE-01: Aesthetic Sophistication (4/8) - Well-configured single-color scatter, not exceptional
• DE-02: Visual Refinement (4/6) - Subtle grid, generous whitespace, clean typography
• DE-03: Data Storytelling (2/6) - Plain uniform scatter, no focal point or emphasis

Spec Compliance (15/15)

• SC-01: Plot Type (5/5)
• SC-02: Required Features (4/4)
• SC-03: Data Mapping (3/3)
• SC-04: Title & Legend (3/3)

Data Quality (12/15)

• DQ-01: Feature Coverage (4/6)
• DQ-02: Realistic Context (5/5)
• DQ-03: Appropriate Scale (3/4)

Code Quality (10/10)

• CQ-01: KISS Structure (3/3)
• CQ-02: Reproducibility (2/2)
• CQ-03: Clean Imports (2/2)
• CQ-04: Code Elegance (2/2)
• CQ-05: Output & API (1/1)

Library Mastery (7/10)

• LM-01: Idiomatic Usage (4/5)
• LM-02: Distinctive Features (3/5) - Custom barycentric coordinate system via D3's low-level SVG binding

Score Caps Applied

• None

Strengths

• Correct barycentric-coordinate math (toXY) drives grid lines, tick marks, and data points from a single consistent transform — all three grid-line families and tick axes line up perfectly with the triangle edges
• Idiomatic D3 data-join pattern (selection.data().join()) used for grid lines, tick labels, vertex labels, and sample points
• Deterministic LCG-seeded data generation, fully self-contained (no network calls), matching the D3 reproducibility rule
• Marker sizing (r=9, fill-opacity=0.85, theme-adaptive stroke) well matched to the sparse 28-point dataset
• Correct Imprint palette usage: first series is #009E73, identical between light and dark renders, with theme-correct #FAF8F1/#1A1A17 backgrounds and adaptive ink tokens for all chrome

Weaknesses

• At all three triangle vertices, two tick labels are rendered at the exact same pixel position and overlap: e.g. the clay-axis lv=0 tick and the sand-axis lv=1 tick both map to the "Sand" vertex via toXY, so their labels ("0" and "100") stack exactly on top of each other — visible as a doubled/bolder "0" fused into the second "0" of "100" with color-fringe ghosting on close zoom. Dedupe the tickLabels array so vertex-coincident ticks (three pairs total) aren't drawn twice, or nudge one of each coincident pair tangentially.
• Vertex labels "Clay"/"Sand"/"Silt" have no unit shown, even though the 0-100 ticks are percentages — add "(%)" to each, matching the sibling matplotlib implementation's "Sand (%)" style.
• The data generator's raw ranges (0.15 + rand()*0.85 per component before normalization) mathematically confine every component to roughly 7-77%, so no sample ever shows a near-vertex/high-purity composition. Widen the floor toward 0 so a few points land close to the vertices/edges for better feature coverage.
• No visual hierarchy or storytelling element — consider highlighting a sub-region or a couple of representative points to give the viewer a focal point.

Issues Found

1. VQ-02 MEDIUM: Overlapping tick labels at all 3 triangle vertices
   • Fix: In the tickLabels construction, detect and skip/offset the duplicate label at each vertex (three vertex-coincident pairs: apex, left, right) instead of drawing both axes' extreme labels at the identical toXY position.
2. VQ-06 LOW: Vertex labels missing "%" units
   • Fix: Change label text to "Clay (%)", "Sand (%)", "Silt (%)".
3. DQ-01 LOW: No near-vertex/high-purity data samples
   • Fix: Lower the raw per-component floor (e.g. from 0.15 toward ~0.02) so the normalized dataset spans closer to 0% and 100% for at least a few points.
4. DE-03 LOW: No visual hierarchy or focal point
   • Fix: Highlight a compositional sub-region (analogous to the matplotlib "Loam" zone) or emphasize a couple of representative samples with size/label to create a focal point.

AI Feedback for Next Attempt

Fix the vertex tick-label overlap first (VQ-02) — it's a straightforward dedup in the tickLabels construction where three vertex-coincident (level, axis) pairs currently produce two labels at the same pixel position. Then add "(%)" units to the vertex labels, widen the data generator's range so a few samples approach the vertices/edges, and consider a small storytelling touch (a highlighted zone or emphasized samples) to lift Design Excellence.

Verdict: REJECTED

[github-actions]

🔧 Repair Attempt 1/4

Applied fixes based on AI review feedback.

Status: Repair completed, re-triggering review...

---

🤖 impl-repair

[claude]

AI Review - Attempt 2/3

Image Description

Light render (plot-light.png): Warm off-white background (#FAF8F1-matching), bold dark title "ternary-basic · javascript · d3 · anyplot.ai" fully visible at top. Triangle border and 20%-interval grid lines render as subtle gray on a properly barycentric-mapped triangle. Tick labels (10-100) along all three edges are dark gray and readable. "Clay (%)" (top) and "Silt (%)" (bottom-right) vertex labels are bold, dark, fully visible with margin to spare. However, the "Sand (%)" vertex label (bottom-left) is clipped at the left canvas border — a 4x pixel-zoomed crop of the x=0-120, y=2140-2220 region shows the left curve of the "S" cut off flat at x=0, i.e. missing pixels, not just a tight margin. 28 green (#009E73) circles, radius scaled by compositional purity, are clearly distinguishable from the background.

Dark render (plot-dark.png): Warm near-black background (#1A1A17-matching). Title, grid, triangle border, and tick labels correctly flip to light colors and stay fully readable — no dark-on-dark failures anywhere. Data colors are identical to the light render (same #009E73 green, same positions/sizes). The same "Sand (%)" clipping defect reproduces identically — the "S" is cut off at the left canvas edge in this render too, confirming it's a layout-math issue (not a per-theme rendering glitch).

Both paragraphs above are based on directly viewing both PNGs plus pixel-level crops to verify the clipping claim.

Score: 0/100

AR-09 EDGE CLIPPING triggered — auto-reject overrides the category math below. The category breakdown is provided for repair-loop diagnostic value; the posted score is forced to 0 per the AR-09 rule.

Category	Score (pre-override)	Max
Visual Quality	21	30
Design Excellence	13	20
Spec Compliance	15	15
Data Quality	14	15
Code Quality	10	10
Library Mastery	6	10
Total (pre-override)	79	100
Posted Score (AR-09 override)	0	100

Visual Quality (21/30)

• VQ-01: Text Legibility (4/8) - Explicit font sizes, readable in both themes, but the "Sand (%)" label's leading "S" is physically clipped in both renders
• VQ-02: No Overlap (6/6)
• VQ-03: Element Visibility (5/6)
• VQ-04: Color Accessibility (2/2)
• VQ-05: Layout & Canvas (0/4) - AR-09 hard rule: clipped vertex label
• VQ-06: Axis Labels & Title (2/2)
• VQ-07: Palette Compliance (2/2)

Design Excellence (13/20)

• DE-01: Aesthetic Sophistication (5/8) - Custom triangle geometry + purity-scaled markers, above a bare default but short of publication-ready polish
• DE-02: Visual Refinement (4/6) - Subtle grid, clean chrome; held back by the clipping and moderate (~30%) canvas fill
• DE-03: Data Storytelling (4/6) - Purity-scaled marker size creates a real visual-hierarchy focal point

Spec Compliance (15/15)

• SC-01: Plot Type (5/5)
• SC-02: Required Features (4/4) - 20% grid, vertex labels, edge tick marks all present
• SC-03: Data Mapping (3/3)
• SC-04: Title & Legend (3/3) - Title format exact; no legend needed for single series

Data Quality (14/15)

• DQ-01: Feature Coverage (5/6)
• DQ-02: Realistic Context (5/5) - Neutral soil-texture (clay/sand/silt) scenario
• DQ-03: Appropriate Scale (4/4) - Every triplet normalized to sum to exactly 100%

Code Quality (10/10)

• CQ-01: KISS Structure (3/3)
• CQ-02: Reproducibility (2/2) - Fixed-seed LCG
• CQ-03: Clean Imports (2/2)
• CQ-04: Code Elegance (2/2)
• CQ-05: Output & API (1/1)

Library Mastery (6/10)

• LM-01: Idiomatic Usage (4/5) - Clean selectAll().data().join() throughout
• LM-02: Distinctive Features (2/5) - Custom barycentric math is D3-friendly but not D3-exclusive

Score Caps Applied

• AR-09 (Edge Clipping): Score forced to 0/100, verdict REJECTED — see Issues Found below

Strengths

• Correct barycentric-coordinate triangle geometry with a proper three-family 20% grid, tick marks, and tick labels along all three edges
• Marker radius scales with each sample's dominant-component purity, giving the plot a genuine visual-hierarchy focal point rather than uniform dots
• Realistic, neutral soil-science scenario (clay/sand/silt) with values correctly normalized so every triplet sums to 100%
• Clean idiomatic D3 data-join pattern for grid lines, ticks, labels, and points; deterministic seeded LCG for reproducibility
• Both themes correctly use the Imprint palette (#009E73 data points, correct #FAF8F1/#1A1A17 backgrounds, theme-adaptive ink/grid tokens)

Weaknesses

• AR-09 EDGE CLIPPING: the "Sand (%)" vertex label is clipped at the left canvas border in BOTH plot-light.png and plot-dark.png — the left stroke/curve of the "S" is chopped off, missing pixels for good.
• Considerable unused whitespace on the left/right sides outside the vertex labels; the triangle itself covers only ~30% of the canvas area — worth using more of the available square canvas once the clipping is fixed.

Issues Found

1. AR-09 CRITICAL (Edge Clipping): "Sand (%)" vertex label clipped at the left canvas edge in both renders
   • Root cause: the label is positioned with outward(left, 78) and text-anchor="middle". Because "Sand (%)" renders wider than the other vertex labels (wide letters a/n/d vs the narrower i/l/t in "Silt (%)"), its horizontally-centered left edge lands at a negative x coordinate (~-5.5 CSS px before the 2x screenshot scale), pushing part of the "S" off the 1200px mount.
   • Fix (pick one): increase margin.left/margin.right enough to clear the widest vertex label at its outward-push distance; reduce the outward-push distance specifically for the left/right vertex labels; or switch the left vertex label to text-anchor="start" (and the right vertex label to text-anchor="end") so labels grow inward from their vertex instead of extending symmetrically past the canvas edge.
2. VQ-05 / DE-02 LOW: Triangle occupies only ~30% of the 2400×2400 canvas area with sizeable unused margins left/right
   • Fix: after resolving the clipping (which likely requires slightly larger margins anyway), consider increasing the triangle's side length to better fill the available canvas.

AI Feedback for Next Attempt

Priority fix: the "Sand (%)" vertex label is clipped at the left canvas edge in both light and dark renders (AR-09, auto-reject). Switch the left/right vertex labels to text-anchor="start"/"end" (instead of centering them past the canvas edge with text-anchor="middle"), or increase the left/right margins so the widest vertex label clears the canvas bounds at its current outward-push distance. Everything else in this implementation is solid — correct ternary geometry, proper grid/ticks, purity-scaled markers, and correct Imprint theming — so once the clipping is fixed this should score well above the "good" tier.

Verdict: REJECTED

[github-actions]

🔧 AI Review Produced No Score — Auto-Retrying

The Claude Code Action ran but didn't write quality_score.txt. Auto-retrying review once...

---

🤖 impl-review

[claude]

AI Review - Attempt 2/3

Image Description

Light render (plot-light.png): Warm off-white background (#FAF8F1-consistent). Bold dark title "ternary-basic · javascript · d3 · anyplot.ai" centered at top, fully visible. Triangle border and 20%-interval grid render as subtle gray on a correctly barycentric-mapped triangle; the vertex-coincident tick-label doubling from attempt 1 is fixed. "Clay (%)" (top) and "Silt (%)" (bottom-right) vertex labels are bold, dark, and fully clear of the canvas edges. However, the "Sand (%)" vertex label (bottom-left) is clipped at the left canvas border — a 6x pixel-zoomed crop of x=0-150, y=2150-2230 shows the left curve of the "S" cut off flat at x=0, and direct pixel sampling confirms non-background glyph pixels sitting on column x=0 (y=2198-2201). This is missing pixels, not a tight margin. 28 green (#009E73) circles, radius scaled by compositional purity, are clearly distinguishable from the background.

Dark render (plot-dark.png): Warm near-black background (#1A1A17-consistent). Title, grid, triangle border, and tick labels correctly flip to light tokens and stay fully readable — no dark-on-dark failures anywhere. Data colors are identical to the light render (same #009E73 green, same positions/sizes). The same "Sand (%)" clipping defect reproduces identically — the "S" is cut off at the left canvas edge in this render too (verified with the same pixel-level crop), confirming it's a layout-math issue, not a per-theme rendering glitch.

Both paragraphs are based on directly viewing both PNGs plus pixel-level crops/sampling to verify the clipping claim.

Score: 0/100

AR-09 EDGE CLIPPING triggered — auto-reject overrides the category math below. The category breakdown is provided for repair-loop diagnostic value; the posted score is forced to 0 per the AR-09 rule.

Category	Score (pre-override)	Max
Visual Quality	20	30
Design Excellence	13	20
Spec Compliance	15	15
Data Quality	15	15
Code Quality	10	10
Library Mastery	7	10
Total (pre-override)	80	100
Posted Score (AR-09 override)	0	100

Visual Quality (20/30)

• VQ-01: Text Legibility (3/8) - "Sand (%)" vertex label has glyph pixels physically clipped at the left canvas edge in both renders
• VQ-02: No Overlap (6/6) - vertex tick-label dedup from attempt 1 confirmed fixed
• VQ-03: Element Visibility (6/6)
• VQ-04: Color Accessibility (2/2)
• VQ-05: Layout & Canvas (0/4) - AR-09 hard rule: clipped vertex label
• VQ-06: Axis Labels & Title (1/2)
• VQ-07: Palette Compliance (2/2)

Design Excellence (13/20)

• DE-01: Aesthetic Sophistication (5/8) - Custom triangle geometry + purity-scaled markers, above bare defaults, short of publication-ready polish
• DE-02: Visual Refinement (4/6) - Subtle grid, clean chrome; held back by the clipping and the triangle filling only ~30% of the square canvas
• DE-03: Data Storytelling (4/6) - Purity-scaled marker size creates a real visual-hierarchy focal point

Spec Compliance (15/15)

• SC-01: Plot Type (5/5)
• SC-02: Required Features (4/4) - 20% grid, vertex labels, edge tick marks all present
• SC-03: Data Mapping (3/3)
• SC-04: Title & Legend (3/3) - Title format exact; no legend needed for single series

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) - widened generator range now reaches near-vertex/high-purity samples
• DQ-02: Realistic Context (5/5) - Neutral soil-texture (clay/sand/silt) scenario
• DQ-03: Appropriate Scale (4/4) - Every triplet normalized to sum to exactly 100%

Code Quality (10/10)

• CQ-01: KISS Structure (3/3)
• CQ-02: Reproducibility (2/2) - Fixed-seed LCG
• CQ-03: Clean Imports (2/2)
• CQ-04: Code Elegance (2/2)
• CQ-05: Output & API (1/1)

Library Mastery (7/10)

• LM-01: Idiomatic Usage (4/5) - Clean selectAll().data().join() throughout
• LM-02: Distinctive Features (3/5) - Custom barycentric math is D3-friendly and the only way to build a ternary plot in D3, but not an exclusive/advanced-API feature

Score Caps Applied

• AR-09 (Edge Clipping): Score forced to 0/100, verdict REJECTED — see Issues Found below

Strengths

• Correct barycentric-coordinate triangle geometry with a proper three-family 20% grid, tick marks, and now-deduplicated tick labels along all three edges
• Marker radius scales with each sample's dominant-component purity, giving the plot a genuine visual-hierarchy focal point rather than uniform dots
• Realistic, neutral soil-science scenario (clay/sand/silt), correctly normalized so every triplet sums to 100%, and now spans near-vertex high-purity samples too
• Clean idiomatic D3 data-join pattern for grid lines, ticks, labels, and points; deterministic seeded LCG for reproducibility
• Both themes correctly use the Imprint palette (#009E73 data points, correct #FAF8F1/#1A1A17 backgrounds, theme-adaptive ink/grid tokens)

Weaknesses

• AR-09 EDGE CLIPPING: the "Sand (%)" vertex label is clipped at the left canvas border in BOTH plot-light.png and plot-dark.png — the left stroke/curve of the "S" is chopped off, missing pixels for good.
• Considerable unused whitespace on the left/right sides outside the vertex labels; the triangle itself covers only ~30% of the canvas area — worth using more of the available square canvas once the clipping is fixed.

Issues Found

1. AR-09 CRITICAL (Edge Clipping): "Sand (%)" vertex label clipped at the left canvas edge in both renders
   • Root cause: the label is positioned with outward(left, 78) and text-anchor="middle". Because "Sand (%)" renders wider than "Silt (%)" (letters a/n/d are wider than i/l/t), centering it on the outward-pushed left-vertex anchor point pushes the left half of the string past x=0. (Confirmed: the mirrored "Silt (%)" label on the right does NOT clip.)
   • Fix (pick one): increase margin.left/margin.right enough to clear the widest vertex label at its outward-push distance; reduce the outward-push distance specifically for the left/right vertex labels; or switch the left vertex label to text-anchor="start" (and the right vertex label to text-anchor="end") so labels grow inward from their vertex instead of extending symmetrically past the canvas edge.
2. VQ-05 / DE-02 LOW: Triangle occupies only ~30% of the 2400×2400 canvas area with sizeable unused margins left/right
   • Fix: after resolving the clipping (which likely needs slightly larger margins anyway), consider increasing the triangle's side length to better fill the available canvas.

AI Feedback for Next Attempt

Priority fix: the "Sand (%)" vertex label is still clipped at the left canvas edge in both light and dark renders (AR-09, auto-reject). Switch the left/right vertex labels to text-anchor="start"/"end" (instead of centering them past the canvas edge with text-anchor="middle"), or increase the left/right margins so the widest vertex label clears the canvas bounds at its current outward-push distance. Everything else in this implementation is solid — correct ternary geometry, proper grid/ticks, purity-scaled markers, and correct Imprint theming — so once the clipping is fixed this should score well above the "good" tier.

Verdict: REJECTED

[github-actions]

❌ AI Review Failed (auto-retry exhausted)

The AI review action completed but did not produce valid output files. Auto-retry already tried once.

What happened:

• The Claude Code Action ran
• No quality_score.txt file was created

Manual rerun:

gh workflow run impl-review.yml -f pr_number=10003

---

🤖 impl-review

[claude]

AI Review - Attempt 2/3

Image Description

Light render (plot-light.png): Warm off-white background (#FAF8F1-consistent). Bold dark title "ternary-basic · javascript · d3 · anyplot.ai" centered at top, fully visible with clear margin. Equilateral triangle border and 20%-interval grid lines render as subtle gray, correctly barycentric-mapped (the vertex tick-label doubling from attempt 1 is fixed — labels only show at their owning vertex). "Clay (%)" (top) and "Silt (%)" (bottom-right) vertex labels are bold, dark, and fully clear of the canvas edges. However, the "Sand (%)" vertex label (bottom-left) is clipped at the left canvas border: a 3x pixel-zoomed crop of the x=0–300, y=2050–2300 region shows the leading curve of the "S" cut off flat at the canvas edge, and direct pixel sampling at column x=0 finds non-background glyph pixels at y=2198–2200 (RGB (236,213,186), (222,164,91), (243,220,166) against a (250,248,241) background) — i.e. actual glyph pixels sitting on the very last column, confirming missing content rather than a tight margin. 28 green (#009E73) circles, radius scaled by compositional purity, are clearly distinguishable from the background; fill and stroke are correctly themed.

Dark render (plot-dark.png): Warm near-black background (#1A1A17-consistent). Title, grid, triangle border, and tick labels correctly flip to light tokens and stay fully readable — no dark-on-dark failures anywhere. Data colors are identical to the light render (same #009E73 green, same positions/sizes); only chrome and marker stroke flip. The same "Sand (%)" clipping defect reproduces identically — pixel sampling at column x=0 finds glyph pixels at the same y=2198–2200 rows (RGB (46,72,94), (66,119,179), (32,65,114) against a (26,26,23) background), confirming this is a layout-math issue independent of theme.

Both paragraphs are based on directly viewing both PNGs plus pixel-level crops and column-0 pixel sampling to verify the clipping claim quantitatively.

Score: 0/100

AR-09 EDGE CLIPPING triggered — auto-reject overrides the category math below. The "Sand (%)" vertex label is physically clipped at the left canvas edge in both renders (glyph pixels present at x=0). The category breakdown below is the honest pre-override assessment, kept for repair-loop diagnostic value, but the posted score is forced to 0 per the AR-09 rule.

Category	Score (pre-override)	Max
Visual Quality	21	30
Design Excellence	13	20
Spec Compliance	15	15
Data Quality	14	15
Code Quality	10	10
Library Mastery	7	10
Total (pre-override)	80	100
Posted Score (AR-09 override)	0	100

Visual Quality (21/30)

• VQ-01: Text Legibility (3/8) - Explicit font sizes, readable everywhere else in both themes, but the "Sand (%)" vertex label's leading "S" is physically clipped at the left canvas edge in both renders
• VQ-02: No Overlap (6/6) - Vertex tick-label doubling from attempt 1 confirmed fixed; no other overlaps
• VQ-03: Element Visibility (6/6) - Purity-scaled markers clearly visible, appropriate for 28-point density
• VQ-04: Color Accessibility (2/2) - Single-series brand green, no red/green-only signaling
• VQ-05: Layout & Canvas (0/4) - AR-09 hard rule: clipped vertex label overrides this category
• VQ-06: Axis Labels & Title (2/2) - Vertex labels now carry units ("Clay (%)" etc.), title format correct
• VQ-07: Palette Compliance (2/2) - Brand green first/only series, correct theme backgrounds both renders

Design Excellence (13/20)

• DE-01: Aesthetic Sophistication (5/8) - Custom barycentric triangle geometry + purity-scaled markers, above bare defaults, short of full publication polish
• DE-02: Visual Refinement (4/6) - Subtle grid, generous whitespace, clean typography; held back by the clipping defect
• DE-03: Data Storytelling (4/6) - Purity-scaled marker size creates a genuine visual-hierarchy focal point toward high-purity, near-vertex samples

Spec Compliance (15/15)

• SC-01: Plot Type (5/5) - Correct ternary/barycentric plot
• SC-02: Required Features (4/4) - 20%-interval grid, vertex labels, tick marks on all three edges, distinct markers all present
• SC-03: Data Mapping (3/3) - Clay/Sand/Silt correctly mapped to the three vertices, full 0-100 range represented
• SC-04: Title & Legend (3/3) - Title format matches spec exactly; no legend needed for a single series

Data Quality (14/15)

• DQ-01: Feature Coverage (5/6) - Good spread across the compositional space including near-vertex high-purity samples and mid-triangle mixtures
• DQ-02: Realistic Context (5/5) - Soil-texture (clay/sand/silt) domain is realistic, neutral, and directly named in the spec's Applications section
• DQ-03: Appropriate Scale (4/4) - 28 points (within the spec's 20-200 recommendation), all triplets correctly normalized to sum to 100

Code Quality (10/10)

• CQ-01: KISS Structure (3/3) - No classes, small deterministic helper functions only
• CQ-02: Reproducibility (2/2) - Fixed-seed LCG, fully deterministic
• CQ-03: Clean Imports (2/2) - Only the d3 global, no extraneous imports
• CQ-04: Code Elegance (2/2) - Clean, appropriately commented, no fake interactivity
• CQ-05: Output & API (1/1) - Builds SVG into #container sized from ANYPLOT_SIZE, no animation, correct mount-node contract

Library Mastery (7/10)

• LM-01: Idiomatic Usage (4/5) - Idiomatic selectAll/data/join pattern, proper token usage
• LM-02: Distinctive Features (3/5) - Custom barycentric-coordinate geometry is a reasonable D3-flexibility showcase, but doesn't leverage D3-specific modules (scales/axes) since a ternary layout doesn't map cleanly onto them

Score Caps Applied

• AR-09 (edge clipping) — score forced to 0, verdict REJECTED, regardless of category math above

Strengths

• Correctly barycentric-mapped triangle geometry with clean grid lines at 20% intervals in all three directions
• Purity-scaled marker radius creates a genuine focal point/visual hierarchy (near-vertex high-purity samples read as more prominent)
• Realistic, spec-aligned soil-texture (clay/sand/silt) dataset, correctly normalized so every triplet sums to 100
• Attempt-1 feedback was properly addressed: vertex tick-label doubling at shared corners is fixed, and vertex labels now carry units

Weaknesses

• AR-09: "Sand (%)" vertex label is clipped at the left canvas edge in both light and dark renders. Pixel sampling at column x=0 shows non-background glyph pixels at y≈2198-2200 in both themes — the leading "S" is missing pixels, not just tight-margined. Root cause: the label is centered (text-anchor="middle") on a point only ~78px outward from the left vertex (at margin.left = 110), and the vertex-outward direction points left-and-down, so half of the ~140px-wide bold "Sand (%)" string lands left of canvas x=0. Fix needs to either: increase margin.left enough to give the outward-projected label room, reduce the outward distance for the left vertex specifically, switch that label's text-anchor to "start" so it grows rightward instead of centering on the (already near-edge) projected point, or clamp the vertex-label positions to stay within the canvas bounds after the outward projection. Apply an equivalent safety margin to the "Silt (%)" label on the opposite side in case symmetric tuning pushes it toward the right edge.

Issues Found

1. AR-09 CRITICAL: "Sand (%)" vertex label clipped at the left canvas edge in both renders
   • Fix: Increase left margin and/or change the label's text-anchor to "start" so the centered-text math doesn't push glyph pixels past x=0; verify with pixel sampling at column x=0 after the fix, in both themes.
2. LM-02 LOW: Ternary geometry is custom-built rather than leveraging D3-specific scale/axis modules
   • Fix: Not required (ternary layouts don't map onto d3-axis cleanly) — optional, low-priority polish only.

AI Feedback for Next Attempt

The only blocking issue is the AR-09 edge clipping on the "Sand (%)" vertex label — it reproduces identically in both themes, so it's a layout-math bug, not a theme problem. Fix the left-vertex label geometry (bigger margin, shorter outward offset for that vertex, text-anchor: "start", or a canvas-bounds clamp) and re-verify with a zoomed crop of the bottom-left corner in both renders before resubmitting. Everything else in this implementation is strong — the purity-scaled markers, the fixed tick-label doubling, and the realistic soil-texture dataset should be kept as-is.

Verdict: REJECTED

[github-actions]

❌ AI Review Failed (auto-retry exhausted)

The AI review action completed but did not produce valid output files. Auto-retry already tried once.

What happened:

• The Claude Code Action ran
• No quality_score.txt file was created

Manual rerun:

gh workflow run impl-review.yml -f pr_number=10003

---

🤖 impl-review

[claude]

AI Review - Attempt 2/3

Image Description

Light render (plot-light.png): Warm off-white background (#FAF8F1-matching). Bold dark title "ternary-basic · javascript · d3 · anyplot.ai" centered and fully visible at top. The equilateral triangle is outlined in a soft ink stroke with a three-family 20%-interval grid rendered as subtle gray lines. Tick labels (10–100) along all three edges are dark gray and clearly readable, with no doubled/overlapping labels at the vertices (the prior tick-dedup fix holds). "Clay (%)" (top) and "Silt (%)" (bottom-right) vertex labels are bold, dark, and fully clear of the canvas edge. However, a 3×-zoomed pixel crop of the bottom-left corner (x=0–300, y=2100–2300) shows the "Sand (%)" vertex label clipped at the left canvas border — the leading curve of the "S" is cut off flat at x=0 with visible color-fringe ghosting, not merely a tight margin. 28 green (#009E73) circles, radius scaled by each sample's compositional purity, are clearly visible and distinguishable from the background; all other text is readable against the light background.

Dark render (plot-dark.png): Warm near-black background (#1A1A17-matching). Title, grid, triangle border, and tick labels all correctly flip to light ink tokens and remain fully legible — no dark-on-dark failures anywhere. Data colors are identical to the light render (same #009E73 green, same positions and purity-scaled sizes; only the marker stroke flips to match the page background). The same "Sand (%)" clipping defect reproduces identically in the equivalent pixel crop — the "S" is cut off at the left canvas edge in this render too, confirming a layout-math issue rather than a per-theme rendering glitch.

Both paragraphs are based on directly viewing both full PNGs plus 3×-zoomed pixel crops of the flagged region to verify the clipping claim.

Score: 0/100

AR-09 EDGE CLIPPING triggered — auto-reject overrides the category math below. The category breakdown is provided for repair-loop diagnostic value; the posted score is forced to 0 per the AR-09 rule.

Category	Score (pre-override)	Max
Visual Quality	22	30
Design Excellence	13	20
Spec Compliance	15	15
Data Quality	14	15
Code Quality	10	10
Library Mastery	6	10
Total (pre-override)	80	100
Posted Score (AR-09 override)	0	100

Visual Quality (22/30)

• VQ-01: Text Legibility (4/8) - Explicit sizes, readable in both themes, but "Sand (%)" label's leading "S" is physically clipped in both renders
• VQ-02: No Overlap (6/6) - Vertex tick-label dedup from the prior fix holds; no doubled labels observed
• VQ-03: Element Visibility (6/6) - Purity-scaled markers clearly visible, well-sized for the sparse 28-point dataset
• VQ-04: Color Accessibility (2/2)
• VQ-05: Layout & Canvas (0/4) - AR-09 hard rule: clipped vertex label
• VQ-06: Axis Labels & Title (2/2) - Vertex labels correctly carry "(%)" units now
• VQ-07: Palette Compliance (2/2)

Design Excellence (13/20)

• DE-01: Aesthetic Sophistication (5/8) - Custom triangle geometry + purity-scaled markers, above a bare default but short of publication-ready polish
• DE-02: Visual Refinement (4/6) - Subtle grid, clean chrome; held back by the clipping and moderate canvas fill
• DE-03: Data Storytelling (4/6) - Purity-scaled marker size creates a real visual-hierarchy focal point

Spec Compliance (15/15)

• SC-01: Plot Type (5/5)
• SC-02: Required Features (4/4) - 20% grid, vertex labels, edge tick marks all present
• SC-03: Data Mapping (3/3)
• SC-04: Title & Legend (3/3) - Title format exact; no legend needed for single series

Data Quality (14/15)

• DQ-01: Feature Coverage (5/6) - Widened data range (0.02 floor) now shows near-vertex/high-purity samples alongside balanced mixtures
• DQ-02: Realistic Context (5/5) - Neutral soil-texture (clay/sand/silt) scenario
• DQ-03: Appropriate Scale (4/4) - Every triplet normalized to sum to exactly 100%

Code Quality (10/10)

• CQ-01: KISS Structure (3/3)
• CQ-02: Reproducibility (2/2) - Fixed-seed LCG
• CQ-03: Clean Imports (2/2)
• CQ-04: Code Elegance (2/2)
• CQ-05: Output & API (1/1)

Library Mastery (6/10)

• LM-01: Idiomatic Usage (4/5) - Clean selectAll().data().join() throughout
• LM-02: Distinctive Features (2/5) - Custom barycentric math is D3-friendly but not D3-exclusive

Score Caps Applied

• AR-09 (Edge Clipping): Score forced to 0/100, verdict REJECTED — see Issues Found below

Strengths

• Correct barycentric-coordinate ternary geometry with a proper three-family 20% grid, tick marks, and tick labels along all three edges, all aligned
• Marker radius scales with each sample's dominant-component purity, giving the plot a genuine visual-hierarchy focal point rather than uniform dots
• Wide compositional data spread (raw floor lowered to 0.02) gives good feature coverage from near-vertex, high-purity samples to balanced mid-triangle mixtures, in a realistic, neutral soil-science (clay/sand/silt) scenario summing to exactly 100%
• Vertex tick-label doubling from the prior review was correctly deduped, and vertex labels now correctly carry "(%)" units
• Clean idiomatic D3 data-join pattern for grid lines, ticks, labels, and points; deterministic seeded LCG; correct Imprint palette/theme-adaptive chrome in both renders

Weaknesses

• AR-09 EDGE CLIPPING (unresolved from prior review): the "Sand (%)" vertex label is still clipped at the left canvas border in BOTH plot-light.png and plot-dark.png — the leading curve of the "S" is chopped off flat at x=0, missing pixels for good. This is the exact same defect flagged in the previous review; the last fix commit addressed the Attempt-1 feedback (tick dedup, "(%)" units, data-range widening, purity-based sizing) but did not touch the vertex-label positioning that causes the clipping.
• Root cause: the label is positioned with outward(left, 78) and text-anchor="middle". "Sand (%)" is the widest of the three vertex labels (wide a/n/d glyphs vs. the narrower glyphs in "Clay (%)"/"Silt (%)"), so its horizontally-centered left edge lands at a negative x coordinate before the 2x screenshot scale, pushing part of the "S" off the 1200px CSS mount.
• Considerable unused whitespace remains on the left/right of the square canvas outside the vertex labels; worth using more of the available area once the clipping fix (which likely needs larger margins anyway) is in place.

Issues Found

1. AR-09 CRITICAL (Edge Clipping): "Sand (%)" vertex label clipped at the left canvas edge in both renders — unresolved from the prior review
   • Fix (pick one): increase margin.left/margin.right enough to clear the widest vertex label's half-width at its current outward-push distance; reduce the outward-push distance specifically for the left/right vertex labels; or switch the left vertex label to text-anchor="start" (and the right vertex label to text-anchor="end") so labels grow inward from their vertex instead of extending symmetrically past the canvas edge.
2. VQ-05 / DE-02 LOW: Triangle + labels still leave sizable unused margins on the square canvas
   • Fix: after resolving the clipping (which likely requires slightly larger margins), consider increasing the triangle's side length to better fill the available canvas.

AI Feedback for Next Attempt

Priority fix (unresolved from the last review): the "Sand (%)" vertex label is clipped at the left canvas edge in both light and dark renders (AR-09, auto-reject). Switch the left/right vertex labels to text-anchor="start"/"end" (instead of centering them past the canvas edge with text-anchor="middle"), or increase the left/right margins so the widest vertex label clears the canvas bounds at its current outward-push distance. Everything else in this implementation is solid — correct ternary geometry, proper grid/ticks, purity-scaled markers, correct units, and correct Imprint theming — so once the clipping is actually fixed this should score well above the "good" tier.

Verdict: REJECTED

[github-actions]

🔧 Repair Attempt 2/4

Applied fixes based on AI review feedback.

Status: Repair completed, re-triggering review...

---

🤖 impl-repair

[claude]

AI Review - Attempt 3/3

Image Description

Light render (plot-light.png): Warm off-white background (#FAF8F1). Title "ternary-basic · javascript · d3 · anyplot.ai" centered at top in bold dark ink. Vertex labels "Clay (%)" (top), "Sand (%)" (bottom-left), "Silt (%)" (bottom-right) in bold dark text, tick labels (0/20/40/60/80/100) along each edge in softer dark gray, triangular border and gridlines in subtle gray. 28 green (#009E73) circular markers with radius scaled to compositional purity and a pale stroke for edge definition. All text is clearly readable against the light background.

Dark render (plot-dark.png): Warm near-black background (#1A1A17). Same title and vertex labels now rendered in light/off-white text, tick labels in a lighter soft gray, gridlines/border in light low-opacity gray. Data markers are the identical #009E73 green as the light render — only chrome flipped. All text is clearly readable against the dark background; no dark-on-dark failures observed.

Both renders pass the theme-readability check.

Score: 92/100

Category	Score	Max
Visual Quality	30	30
Design Excellence	14	20
Spec Compliance	15	15
Data Quality	15	15
Code Quality	10	10
Library Mastery	8	10
Total	92	100

Visual Quality (30/30)

• VQ-01: Text Legibility (8/8) - Explicit font sizes, readable both themes, no overflow
• VQ-02: No Overlap (6/6) - No text overlap; duplicate vertex-label case deliberately skipped
• VQ-03: Element Visibility (6/6) - Purity-scaled 7-14px markers, well suited to 28 sparse points
• VQ-04: Color Accessibility (2/2) - Pale stroke for definition, no CVD risk
• VQ-05: Layout & Canvas (4/4) - Triangle fills ~60-70% of canvas, balanced margins, nothing clipped
• VQ-06: Axis Labels & Title (2/2) - Vertex labels descriptive with units
• VQ-07: Palette Compliance (2/2) - First series #009E73, identical across themes, chrome theme-correct

Design Excellence (14/20)

• DE-01: Aesthetic Sophistication (5/8) - Clean and intentional, but single accent color keeps it below showcase-level
• DE-02: Visual Refinement (5/6) - Subtle gridlines, generous whitespace, custom triangular frame instead of spines
• DE-03: Data Storytelling (4/6) - Purity-scaled marker size creates a real focal point

Spec Compliance (15/15)

• SC-01: Plot Type (5/5) - Correct ternary/barycentric plot
• SC-02: Required Features (4/4) - Regular-interval gridlines, labeled vertices, edge ticks, distinct markers
• SC-03: Data Mapping (3/3) - Barycentric transform correctly maps all three components
• SC-04: Title & Legend (3/3) - Title format exact; single-series legend correctly omitted

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) - Samples span near-vertex to balanced mid-triangle mixtures
• DQ-02: Realistic Context (5/5) - Soil-texture clay/sand/silt scenario, neutral and real-world
• DQ-03: Appropriate Scale (4/4) - Each triplet normalized to sum to 100%, plausible proportions

Code Quality (10/10)

• CQ-01: KISS Structure (3/3) - Flat script, minimal geometry helpers, no classes
• CQ-02: Reproducibility (2/2) - Fixed-seed LCG (seed=42123)
• CQ-03: Clean Imports (2/2) - Only the d3 global
• CQ-04: Code Elegance (2/2) - Appropriate complexity, no fake functionality
• CQ-05: Output & API (1/1) - Correct mount-node contract, single svg, no animation

Library Mastery (8/10)

• LM-01: Idiomatic Usage (5/5) - Consistent .data().join() pattern throughout
• LM-02: Distinctive Features (3/5) - Hand-rolled barycentric transform, triangular grid families, radial outward-label placement

Score Caps Applied

• None

Strengths

• Correct barycentric-to-pixel transform with three grid-line families and edge tick marks at 20% intervals
• Marker radius encodes compositional purity, giving the scatter meaningful visual hierarchy
• Thoughtful start/end text-anchor logic on vertex labels prevents edge clipping
• Fully theme-adaptive chrome; brand green identical and legible on both surfaces
• Deterministic LCG-seeded data, idiomatic D3 data-join usage throughout

Weaknesses

• Single accent color and uniform marker shape keep design at "strong" rather than showcase-level
• No sample lands above ~65% clay — coverage leans toward sand/silt-dominant mixtures

Issues Found

1. DE-01 MODERATE: Single-color, minimalist palette limits aesthetic ceiling
   • Fix (optional): consider a subtle secondary visual cue (alpha or size gradient) tied to purity for extra polish
2. DQ-01 MINOR: No high-clay-purity sample present
   • Fix (optional): add one sample near 70-80% clay for more even vertex coverage

AI Feedback for Next Attempt

This implementation is strong and passes review. If revisited, consider adding a sample or two closer to the clay vertex for more even feature coverage, and explore a secondary visual encoding (subtle alpha/size gradient) to push Design Excellence further.

Verdict: APPROVED