feat(matplotlib): implement bar-3d-categorical

[github-actions]

Implementation: bar-3d-categorical - python/matplotlib

Implements the python/matplotlib version of bar-3d-categorical.

File: plots/bar-3d-categorical/implementations/python/matplotlib.py

Parent Issue: #5248

---

🤖 impl-generate workflow

[claude]

AI Review - Attempt 1/3

Image Description

Light render (plot-light.png): A 3D bar chart on a warm off-white (#FAF8F1) background showing retail sales ($ thousands) across 4 product categories (Electronics, Clothing, Home & Garden, Sports) and 4 regions (North, South, East, West). Bars are colored using the viridis colormap — purple/dark blue for low values (~49k) through teal and green to yellow for high values (~142k). A viridis colorbar sits on the right side. Value labels appear on top of each bar in dark ink. Axis labels "Product Category", "Region", and "$ thousands" are visible. All text is readable against the light background. The viewing angle (elev=28°, azim=-48°) provides reasonable depth perspective. LEGIBILITY: PASS — however, font sizes (title 20pt, labels 15pt, ticks 13pt) are below the recommended minimums of 24/20/16pt.

Dark render (plot-dark.png): Same 3D bar chart on a warm near-black (#1A1A17) background. The viridis colormap data colors are identical to the light render — confirming only the chrome flips between themes. Title text is light cream (#F0EFE8), axis labels and tick labels are light-colored (B8B7B0), and value labels on bars appear light/white. The colorbar background matches the dark page background. No dark-on-dark failures detected. LEGIBILITY: PASS — all text is readable in dark mode, though font sizes remain below the recommended minimums. Some value labels on back-row bars (e.g., 115, 87, 118 positions) are partially occluded by foreground bars — an inherent limitation of the 3D perspective from this viewing angle.

Both paragraphs are required. A review that only describes one render is invalid.

Score: 80/100

Category	Score	Max
Visual Quality	23	30
Design Excellence	10	20
Spec Compliance	15	15
Data Quality	15	15
Code Quality	10	10
Library Mastery	7	10
Total	80	100

Visual Quality (23/30)

• VQ-01: Text Legibility (5/8) — All sizes explicitly set but below guidelines: title 20pt (needs ≥24), axis labels 15pt (needs ≥20), tick labels 13pt (needs ≥16), value labels 10pt (quite small at 4800px)
• VQ-02: No Overlap (4/6) — Some value labels on back-row bars partially occluded by closer bars due to 3D perspective (e.g. 115, 87, 49 positions)
• VQ-03: Element Visibility (5/6) — Bars are well-sized with appropriate depth/width (0.55), shade=True adds realism, alpha=0.88 good
• VQ-04: Color Accessibility (2/2) — Viridis is perceptually-uniform and CVD-safe; value labels provide per-bar readout
• VQ-05: Layout & Canvas (3/4) — Good layout; colorbar well-placed; some empty foreground space due to 3D perspective
• VQ-06: Axis Labels & Title (2/2) — "Product Category", "Region", "$ thousands", colorbar "Sales ($ thousands)" — all descriptive with context/units
• VQ-07: Palette Compliance (2/2) — Viridis is the correct continuous sequential colormap; backgrounds are #FAF8F1 (light) / #1A1A17 (dark); all chrome is theme-adaptive in both renders

Design Excellence (10/20)

• DE-01: Aesthetic Sophistication (4/8) — Well-configured 3D chart with thoughtful choices (disabled pane fills, proper colorbar styling, shade=True), but doesn't rise above a polished library default
• DE-02: Visual Refinement (4/6) — Pane fill disabled, grid at alpha=0.10 (very subtle), deliberate colorbar axis placement via add_axes; good refinement for 3D context
• DE-03: Data Storytelling (2/6) — Height and color both encode value (dual encoding), making extremes visually obvious, but no focal point, no narrative, no visual hierarchy guiding the viewer to an insight

Spec Compliance (15/15)

• SC-01: Plot Type (5/5) — Correct 3D bar chart with bars rising from 2D categorical grid
• SC-02: Required Features (4/4) — Bar spacing ✓, color encoding value magnitude via viridis ✓, viewing angle ~28°/−48° ✓, base plane grid ✓, value labels on all 16 bars ✓, colorbar ✓
• SC-03: Data Mapping (3/3) — Product on X, Region on Y, Sales value on Z; all 16 (4×4) bars visible
• SC-04: Title & Legend (3/3) — "Retail Sales by Product & Region · bar-3d-categorical · matplotlib · anyplot.ai" — spec-id, library, anyplot.ai all present; colorbar serves as legend with clear label

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) — Shows full 4×4 grid, meaningful value variation (49–142k), dual encoding demonstrates color + height relationship
• DQ-02: Realistic Context (5/5) — Retail sales by product and region is a realistic, neutral business scenario; no controversial topics
• DQ-03: Appropriate Scale (4/4) — Values 49–142k USD per product-region are plausible retail figures; Electronics-North highest is realistic

Code Quality (10/10)

• CQ-01: KISS Structure (3/3) — Linear structure: imports → data → plot → save; no functions or classes
• CQ-02: Reproducibility (2/2) — Hardcoded sales array; fully deterministic
• CQ-03: Clean Imports (2/2) — All imports used; Axes3D imported as side-effect with noqa comment (standard pattern for 3D matplotlib)
• CQ-04: Code Elegance (2/2) — Clean nested loop, proper norm/cmap pattern, no fake UI, no over-engineering
• CQ-05: Output & API (1/1) — Saves as plot-{THEME}.png, API usage is current

Library Mastery (7/10)

• LM-01: Idiomatic Usage (4/5) — Correct use of projection="3d", computed_zorder=False, ax.bar3d(), ax.view_init(), fig.add_axes() for colorbar; well-structured 3D matplotlib code
• LM-02: Distinctive Features (3/5) — Uses computed_zorder=False (matplotlib-specific depth ordering control), xaxis.pane styling (3D-only API), ax.view_init(), shade=True on bar3d — distinctive to matplotlib's mpl_toolkits.mplot3d

Score Caps Applied

• None — DE-01=4 (not ≤2), CQ-04=2 (not 0), all other caps not triggered

Strengths

• Perfect spec compliance: all spec features present (spacing, viewing angle, value labels, colorbar, grid, color encoding)
• Perfect data quality: realistic retail business scenario with meaningful variation across product/region combinations
• Perfect code quality: clean KISS structure, deterministic data, proper imports
• Correct viridis colormap for continuous value encoding (both spec-appropriate and guideline-compliant)
• Thorough theme-adaptive chrome: pane edges, colorbar label/ticks/outline, value labels, and axis labels all use theme tokens correctly in both renders
• computed_zorder=False and shade=True show understanding of matplotlib 3D rendering subtleties

Weaknesses

• Font sizes are explicitly set but below guidelines: title should be ≥24pt (currently 20), axis labels ≥20pt (currently 15), tick labels ≥16pt (currently 13), value labels 10pt is too small — increase all to meet the style guide minimums
• Some value labels on back-row bars are partially occluded by foreground bars (positions 115, 87, 49) — consider adjusting azimuth or elevation to improve readability, or check if zorder on text elements can help
• Design Excellence is limited: the dual height+color encoding doesn't add new information (both encode the same sales value); consider using color to encode a different dimension (e.g., a third categorical variable, or deviation from mean) to add analytical depth
• No visual storytelling element: no focal point, no emphasis on the highest bar (Electronics-North at 142k) or lowest (Home & Garden-West at 49k) — a subtle annotation or distinct color accent would help guide the viewer

Issues Found

1. VQ-01 LOW: Font sizes below required minimums for 4800×2700px output
   • Fix: Increase title to 24pt, xlabel/ylabel to 20pt, tick labels to 16pt, z-label to 16pt, value labels to at least 12pt
2. VQ-02 PARTIAL: Back-row value labels partially occluded by foreground bars in 3D perspective
   • Fix: Try azimuth around −55° to −65° to reduce bar overlap at front, or move value labels slightly (offset z by more headroom)
3. DE-01/DE-03 LOW: No storytelling or visual hierarchy guiding the viewer
   • Fix: Add a subtle annotation or emphasize the max bar (e.g., slightly brighter color, text callout "Peak: Electronics North $142k"); or use color to encode deviation from category average rather than absolute value for a more insightful visualization

AI Feedback for Next Attempt

Increase all font sizes to meet style guide minimums: title≥24, axis labels≥20, tick/z labels≥16, value labels≥12. To lift Design Excellence: either annotate the peak bar (Electronics-North, 142k) with a callout to create a focal point, or re-encode color as deviation from regional mean (so blue = below average, yellow = above average) to make the chart tell a different story than height alone. Also try azimuth ~−60° to reduce front-bar occlusion of back-bar value labels.

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): 3D bar chart on a warm off-white background (#FAF8F1). Title "Retail Sales by Product & Region · bar-3d-categorical · matplotlib · anyplot.ai" rendered in dark ink at the top. X-axis shows four product categories (Electronics, Clothing, Home & Garden, Sports) in dark tick labels; Y-axis shows four regions (North, South, East, West). Z-axis labeled "$ thousands" in INK_SOFT. All 16 bars (4×4 grid) rendered with viridis colormap — dark purple/blue for lowest values (~49k) through teal/green to yellow for highest (~142k). Value labels (e.g. 142, 98, 115, 87, 76, 103, 91, 118, 55, 71, 63, 48, 88, 95, 72, 110) appear on top of each bar in dark ink. Colorbar on the right side labeled "Sales ($ thousands)" in dark ink with viridis scale visible. All text is readable against the light background; some back-row value labels (115, 87, 49) are partially occluded by foreground bars inherent to the 3D perspective. Legibility verdict: PASS.

Dark render (plot-dark.png): Same chart on a warm near-black background (#1A1A17). Title text renders in light cream (#F0EFE8), clearly readable against the dark surface. Axis labels ("Product Category", "Region", "$ thousands") appear in light INK color. Tick labels for product categories and regions render in light INK_SOFT (#B8B7B0). Value labels on bars switch to light cream, readable against the bar surfaces. Colorbar label and tick values also appear in light colors. Viridis bar colors are identical to the light render — purple for low values, yellow-green for the tallest bar at 142. No dark-on-dark failures detected anywhere: all chrome tokens correctly flip for dark mode. Legibility verdict: PASS.

Both paragraphs are required. A review that only describes one render is invalid.

Score: 80/100

Category	Score	Max
Visual Quality	23	30
Design Excellence	10	20
Spec Compliance	15	15
Data Quality	15	15
Code Quality	10	10
Library Mastery	7	10
Total	80	100

Visual Quality (23/30)

• VQ-01: Text Legibility (5/8) — All sizes explicitly set but below guidelines: title 20pt (needs ≥24), axis labels 15pt (needs ≥20), tick labels 13pt (needs ≥16), value labels 10pt too small for 4800px canvas
• VQ-02: No Overlap (4/6) — Back-row value labels (115, 87, 49) partially occluded by foreground bars; inherent 3D depth perspective issue
• VQ-03: Element Visibility (5/6) — Bars well-sized (width/depth 0.55), shade=True adds depth realism, alpha=0.88 appropriate
• VQ-04: Color Accessibility (2/2) — Viridis is perceptually-uniform and CVD-safe; value labels provide per-bar precision
• VQ-05: Layout & Canvas (3/4) — Colorbar well-placed via add_axes; some empty foreground space due to 3D perspective unavoidable
• VQ-06: Axis Labels & Title (2/2) — Product Category, Region, $ thousands, colorbar Sales ($ thousands) — all descriptive
• VQ-07: Palette Compliance (2/2) — Viridis correct for continuous sequential data; backgrounds #FAF8F1/#1A1A17; all chrome theme-adaptive in both renders

Design Excellence (10/20)

• DE-01: Aesthetic Sophistication (4/8) — Well-configured 3D chart with thoughtful pane styling and colorbar integration; transparent panes add sophistication, but design stays at a polished library default without intentional hierarchy
• DE-02: Visual Refinement (4/6) — Pane fill disabled, grid at alpha=0.10, deliberate colorbar placement via add_axes — good 3D-appropriate refinement
• DE-03: Data Storytelling (2/6) — Dual encoding (height + color) both encode the same variable with no focal point; no visual emphasis on extremes (Electronics-North 142k peak or Home & Garden-West 49k trough)

Spec Compliance (15/15)

• SC-01: Plot Type (5/5) — Correct 3D bar chart using mpl_toolkits.mplot3d with bars rising from 2D categorical grid
• SC-02: Required Features (4/4) — Bar spacing, viridis color encoding, viewing angle 28°/-48°, base grid, value labels on all 16 bars, colorbar — all present
• SC-03: Data Mapping (3/3) — Product on X, Region on Y, Sales on Z; full 4×4 = 16 bars
• SC-04: Title & Legend (3/3) — Title contains spec-id, library, anyplot.ai; colorbar serves as legend with Sales ($ thousands) label

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) — Full 4×4 grid, meaningful variation (49–142k), dual encoding demonstrates both color and height dimensions
• DQ-02: Realistic Context (5/5) — Retail sales by product and region — realistic, neutral business scenario
• DQ-03: Appropriate Scale (4/4) — Values 49–142k USD plausible for retail; regional/category variation realistic

Code Quality (10/10)

• CQ-01: KISS Structure (3/3) — Linear: imports → data → plot → save; no functions or classes
• CQ-02: Reproducibility (2/2) — Hardcoded sales array; fully deterministic
• CQ-03: Clean Imports (2/2) — All imports used; Axes3D side-effect import with noqa is standard 3D matplotlib pattern
• CQ-04: Code Elegance (2/2) — Clean nested loop, proper norm/cmap pattern, no fake UI, no over-engineering
• CQ-05: Output & API (1/1) — Saves as plot-{THEME}.png; API usage current

Library Mastery (7/10)

• LM-01: Idiomatic Usage (4/5) — Correct use of projection='3d', computed_zorder=False, ax.bar3d(), ax.view_init(), fig.add_axes() for colorbar
• LM-02: Distinctive Features (3/5) — computed_zorder=False (3D depth ordering control), xaxis.pane styling (3D-only API), shade=True on bar3d — distinctive to matplotlib mpl_toolkits

Score Caps Applied

• None

Strengths

• Perfect spec compliance: all spec features present (spacing, viewing angle, value labels, colorbar, grid, color encoding)
• Realistic retail business scenario with meaningful variation across product/region combinations
• Clean KISS structure, deterministic data, proper imports
• Correct viridis colormap for continuous value encoding (spec-appropriate and guideline-compliant)
• Thorough theme-adaptive chrome: pane edges, colorbar label/ticks/outline, value labels, and axis labels all use theme tokens correctly in both renders
• computed_zorder=False and shade=True show understanding of matplotlib 3D rendering subtleties

Weaknesses

• Font sizes below guidelines: title 20pt (needs ≥24), axis labels 15pt (needs ≥20), tick labels 13pt (needs ≥16), value labels 10pt too small for 4800px canvas
• Back-row value labels partially occluded by foreground bars due to 3D perspective (positions 115, 87, 49)
• Dual height+color encoding redundantly encodes the same variable — color could encode a different dimension (e.g. product category as Okabe-Ito categorical) for added analytical depth
• No visual storytelling: no focal point or emphasis on peak/trough bars (Electronics-North 142k, Home & Garden-West 49k)

Issues Found

1. VQ-01 LOW: Font sizes significantly below guidelines for 4800px canvas
   • Fix: title→24pt, axis labels→20pt, tick labels→16pt, value labels→14pt minimum
2. DE-03 LOW: No visual hierarchy or data storytelling
   • Fix: Highlight peak bar (Electronics-North 142k) with distinct edge or annotation; or encode product category with Okabe-Ito instead of repeating value magnitude

AI Feedback for Next Attempt

Increase all font sizes to meet guidelines (title≥24pt, labels≥20pt, ticks≥16pt, value labels≥14pt). Consider encoding product category with Okabe-Ito categorical palette instead of duplicating value magnitude with viridis — this removes the redundant dual encoding and adds a genuine second data dimension. If keeping viridis, add a brief annotation or visual emphasis on the highest-value bar to create a narrative focal point.

Verdict: APPROVED