feat(plotnine): implement errorbar-basic

[github-actions]

Implementation: errorbar-basic - python/plotnine

Implements the python/plotnine version of errorbar-basic.

File: plots/errorbar-basic/implementations/python/plotnine.py

Parent Issue: #973

---

🤖 impl-generate workflow

[claude]

AI Review - Attempt 1/3

Image Description

Light render (plot-light.png): Warm off-white background (#FAF8F1) — correct. Title "errorbar-basic · python · plotnine · anyplot.ai" is bold, dark (#1A1A17), spans ~70% of width (expected for the long mandated format). Y-axis label "Concentration (mg/L)" and X-axis label "Sampling Site" are clearly legible at 10pt in dark ink. Tick labels at 8pt in INK_SOFT (#4A4A44) are readable. Three series rendered with circular markers (size=5) and error bars with horizontal caps: Automated (#009E73 brand green), Enhanced (#C475FD lavender), Standard (#4467A3 blue). Y-grid only, very subtle (alpha=0.10). Bottom x-axis spine kept; top/right spines removed; tick marks removed. Legend on right with "Method" header. All text clearly readable against the light background. Legibility verdict: PASS.

Dark render (plot-dark.png): Warm near-black background (#1A1A17) — correct. Title is rendered in near-white (#F0EFE8), clearly legible. Axis labels and tick labels switch to light INK_SOFT (#B8B7B0), fully readable. Legend entries visible in light text. Data colors are identical to the light render — Automated (#009E73), Enhanced (#C475FD), Standard (#4467A3) unchanged across themes as required. Error bars and point markers clearly visible against the dark surface. Grid lines are subtle but present. No dark-on-dark failures observed anywhere. Legibility verdict: PASS.

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

Score: 84/100

Category	Score	Max
Visual Quality	30	30
Design Excellence	10	20
Spec Compliance	15	15
Data Quality	13	15
Code Quality	10	10
Library Mastery	6	10
Total	84	100

Visual Quality (30/30)

• VQ-01: Text Legibility (8/8) — All font sizes explicitly set (title=12pt, axis_title=10pt, axis_text=8pt, legend=8pt). Readable in both themes. Title ~70% of width as expected for the long mandated format.
• VQ-02: No Overlap (6/6) — No overlap between any elements. Dodged positions well-separated across sites.
• VQ-03: Element Visibility (6/6) — 18 data points (sparse). size=5 markers are prominent; error bar size=1.4 with visible caps. Density-appropriate sizing.
• VQ-04: Color Accessibility (2/2) — Imprint palette is colorblind-safe. Green/lavender/blue have good luminance separation.
• VQ-05: Layout & Canvas (4/4) — Plot fills ~80% of canvas width. Balanced margins. Legend well-positioned on right.
• VQ-06: Axis Labels & Title (2/2) — Y: "Concentration (mg/L)" with units. X: "Sampling Site" descriptive.
• VQ-07: Palette Compliance (2/2) — First series (Automated) = #009E73 ✓. Next two from Imprint canonical order. Backgrounds #FAF8F1 light / #1A1A17 dark. Theme-adaptive chrome correct in both renders.

Design Excellence (10/20)

• DE-01: Aesthetic Sophistication (4/8) — Well-configured library default. Imprint palette, minimal spines, clean look — but no exceptional design thinking. Polished standard chart, not publication-level design.
• DE-02: Visual Refinement (4/6) — Good refinement: x-grid removed, y-grid only at alpha=0.10, tick marks removed, spines minimized to bottom spine only, elevated legend background. Missing: legend border rule, further whitespace optimization.
• DE-03: Data Storytelling (2/6) — Data displayed without editorial emphasis. The key insight (Automated method has 3–4× more uncertainty than Standard, all methods trend together across sites) is visible but not highlighted. No visual focal point guides the viewer.

Spec Compliance (15/15)

• SC-01: Plot Type (5/5) — Correct error bar plot with centered points and visible horizontal caps.
• SC-02: Required Features (4/4) — Caps at bar ends (width=0.35), consistent widths, different colors per group, symmetric SE-based error bars.
• SC-03: Data Mapping (3/3) — X=categorical sites, Y=concentration values, ymin/ymax for error bars. All data visible.
• SC-04: Title & Legend (3/3) — Title: "errorbar-basic · python · plotnine · anyplot.ai" — correct format. Legend: "Method" with Automated/Enhanced/Standard labels.

Data Quality (13/15)

• DQ-01: Feature Coverage (6/6) — Three methods with distinct SE magnitudes (Standard ±1.6–2.3, Enhanced ±2.6–3.5, Automated ±4.8–7.2). Multiple sites with varying concentrations.
• DQ-02: Realistic Context (5/5) — Water quality monitoring scenario comparing analytical methods across sampling sites. Neutral scientific domain.
• DQ-03: Appropriate Scale (2/4) — Dissolved oxygen (DO) concentrations of 36–69 mg/L are physically impossible in water. DO saturates at ~14.6 mg/L at 0°C and is typically 6–12 mg/L in field measurements. The data structure is sound but the analyte label is factually wrong at these values.

Code Quality (10/10)

• CQ-01: KISS Structure (3/3) — Clean imports → constants → data → plot → save. No functions or classes.
• CQ-02: Reproducibility (2/2) — Fully deterministic hardcoded data.
• CQ-03: Clean Imports (2/2) — All imports verified as used.
• CQ-04: Code Elegance (2/2) — Pythonic, appropriate complexity. No fake UI elements.
• CQ-05: Output & API (1/1) — Saves as plot-{THEME}.png. Current plotnine API used.

Library Mastery (6/10)

• LM-01: Idiomatic Usage (4/5) — Solid grammar-of-graphics approach: shared aesthetic mapping with position_dodge across geom layers, scale_x_discrete for ordered categorical axis, theme() customization throughout.
• LM-02: Distinctive Features (2/5) — position_dodge with layered geoms is well-executed but replicated easily in ggplot2/letsplot. No distinctively plotnine capability leveraged (e.g., stat_summary, facet_wrap for method comparison, geom_crossbar for a richer error representation).

Score Caps Applied

• None — no caps triggered (DE-01=4 > 2, DE-02=4 > 2; other cap thresholds not hit)

Strengths

• Full theme-adaptive chrome: backgrounds, text, grid, and legend correctly switch between light (#FAF8F1) and dark (#1A1A17) themes — both renders pass readability checks
• Imprint palette applied correctly — Automated=#009E73 (first series), Enhanced=#C475FD, Standard=#4467A3 in canonical order
• Clean idiomatic plotnine grammar: geom_errorbar + geom_point layers sharing position_dodge; scale_x_discrete for custom site ordering
• Error bars with visible caps, appropriate size=5 markers for sparse 18-point dataset
• Sites sorted by Standard-method concentration creating a meaningful left-to-right ascending visual pattern
• Excellent code quality: deterministic data, KISS structure, clean imports, correct output naming

Weaknesses

• DQ-03 — Factually incorrect data: Dissolved oxygen concentrations of 36–69 mg/L are physically impossible in water (saturates at ~14.6 mg/L). Relabel the analyte (e.g., Nitrate, BOD, Total Dissolved Solids) or change the concentration range to 6–12 mg/L for real DO measurements
• DE-03 — No visual storytelling: The key finding (Automated has 3–4× greater uncertainty than Standard; all methods share the same site ranking) is not visually emphasized. Consider annotating the Automated series bars or using a reference line at the Standard SE level to draw the viewer's eye to the precision difference
• DE-01 — Well-configured default, not exceptional: No intentional typographic hierarchy beyond size, no legend frame border, no secondary visual trick to lift design above polished defaults
• DE-02 — Missing legend frame: Adding a thin INK_SOFT border on the legend box (legend_box_color in theme) would complete the refinement detail
• LM-02 — Generic plotnine usage: Could use stat_summary to compute and display SE directly from raw data, or geom_crossbar instead of plain geom_errorbar for a more information-dense representation unique to ggplot grammar

Issues Found

1. DQ-03 MEDIUM: Dissolved oxygen values (36–69 mg/L) are physically impossible in water
   • Fix: Rename analyte to "Nitrate (mg/L)" or "BOD (mg/L)" (with values in realistic range), OR keep "Dissolved Oxygen" but use values 5–12 mg/L
2. DE-03 LOW: No visual hierarchy or data storytelling
   • Fix: Add emphasis — e.g., annotate the Automated series with a label "Highest variability", or draw a horizontal band showing the ±1 SE range of the Standard method as a reference baseline
3. DE-01 LOW: Design is at well-configured-default level (4/8)
   • Fix: Add intentional visual refinement — legend border rule, subtle annotation, or size variation to create hierarchy; aim for DE-01=6
4. LM-02 LOW: Standard plotnine features only
   • Fix: Replace hardcoded ymin/ymax with stat_summary(fun_data='mean_se') to compute SE from raw replicates, or use geom_crossbar for the combined mean+CI representation unique to ggplot grammar

AI Feedback for Next Attempt

Fix the factual data issue first: rename the analyte from dissolved oxygen to something like Nitrate or BOD, adjusting concentration values to the realistic range for that analyte. Then improve design storytelling: add a visual focal point that guides the reader toward the key insight (Automated method carries the most measurement uncertainty). This can be achieved without annotations — e.g., bolder error bars for Automated via custom alpha or a reference band showing Standard method precision. Finally, add a thin legend border (legend_box_color=INK_SOFT) and consider using stat_summary instead of hardcoded ymin/ymax to leverage a distinctively ggplot-grammar feature.

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 matching #FAF8F1. Bold dark title "errorbar-basic · python · plotnine · anyplot.ai" spans ~60% of canvas width, clearly readable. Muted grey subtitle "Automated sampling shows 3–4× wider uncertainty than Standard across all sites" is legible. Axis labels "Sampling Site" (x) and "Nitrate Concentration (mg/L)" (y) in darker ink with units. Tick labels (site names, y-values) in INK_SOFT tone. Three groups of narrow crossbars per site dodged horizontally: Automated in #009E73 (brand green, wide boxes ~10–13 mg/L tall), Enhanced in #C475FD (lavender, medium ~5–7 mg/L), Standard in #4467A3 (blue, very narrow ~3–4 mg/L). Semi-transparent fills (alpha=0.2). Only bottom x-axis spine, no tick marks, subtle y-axis-only grid. Legend box in ELEVATED_BG with soft border at right. All text is readable against the light background — no light-on-light issues.

Dark render (plot-dark.png): Warm near-black background matching #1A1A17 — not pure black. Title and subtitle render in light off-white/cream, fully readable. Axis labels and tick labels in lighter INK_SOFT (#B8B7B0) — all visible, no dark-on-dark failures. Subtle y-axis grid lines visible but not dominant. Legend box uses ELEVATED_BG (#242420) with soft border. Data colors #009E73 (green), #C475FD (lavender), #4467A3 (blue) are identical to the light render — only chrome flips. All three method groups clearly distinguishable on the dark surface.

Both renders pass theme-readability checks. No AR-08 (fake interactivity) or AR-09 (edge clipping) violations found. Canvas gate file absent — dimensions confirmed correct (3200×1800).

Score: 84/100

Category	Score	Max
Visual Quality	30	30
Design Excellence	10	20
Spec Compliance	13	15
Data Quality	15	15
Code Quality	10	10
Library Mastery	6	10
Total	84	100

Visual Quality (30/30)

• VQ-01: Text Legibility (8/8) — All font sizes explicitly set with proper hierarchy (title 12pt bold, subtitle 9pt, axis_title 10pt, axis_text/legend 8pt). Proportioned correctly in both renders.
• VQ-02: No Overlap (6/6) — No text overlaps. position_dodge cleanly separates crossbars. Legend in right margin with no data collision.
• VQ-03: Element Visibility (6/6) — 18 data points, 6 sites × 3 methods. Crossbars appropriately sized. Narrow Standard boxes reflect tight SE — intentional, not a defect.
• VQ-04: Color Accessibility (2/2) — Imprint palette positions 1–3, CVD-safe. Good luminance contrast in both themes.
• VQ-05: Layout & Canvas (4/4) — Canvas gate passed (3200×1800). Balanced margins. Legend correctly positioned. No clipping.
• VQ-06: Axis Labels & Title (2/2) — "Sampling Site" and "Nitrate Concentration (mg/L)" — descriptive with units.
• VQ-07: Palette Compliance (2/2) — First series #009E73 (Automated). #C475FD (Enhanced). #4467A3 (Standard). Backgrounds #FAF8F1 / #1A1A17. Data colors identical across themes. Chrome flips correctly.

Design Excellence (10/20)

• DE-01: Aesthetic Sophistication (4/8) — Well-configured above defaults (Imprint palette, refined chrome, meaningful subtitle), but no exceptional design element elevates it beyond a polished standard: no reference line, no emphasized focal group, no size-variation technique beyond the natural SE difference.
• DE-02: Visual Refinement (4/6) — Good refinement: all border spines removed except bottom x-axis, tick marks removed, subtle y-axis-only grid (alpha=0.10), elevated legend background with soft border. Conventional legend box treatment holds it back from perfect.
• DE-03: Data Storytelling (2/6) — Story carried primarily by the subtitle text, not by visual encoding. Without reading the subtitle the viewer sees three colored groups but no additional visual emphasis (reference line, annotation, highlighted method) to create hierarchy beyond the natural bar-height difference.

Spec Compliance (13/15)

• SC-01: Plot Type (3/5) — geom_crossbar is a valid ggplot2/plotnine uncertainty geom, but departs significantly from the canonical error bar (point + whisker). The spec says "bars extending above and below each point" implying a visible central point; the crossbar's thick middle bar (fatten=3) approximates this but is not the standard form.
• SC-02: Required Features (4/4) — Visible caps (crossbar top/bottom horizontal lines), consistent widths (width=0.18), multiple colored groups (3 methods), SE-based error range shown.
• SC-03: Data Mapping (3/3) — X = sampling sites (categorical), Y = nitrate concentration (mg/L), color = method. All 6 sites and 3 methods shown.
• SC-04: Title & Legend (3/3) — Title exactly "errorbar-basic · python · plotnine · anyplot.ai". Legend labelled "Method" with Automated / Enhanced / Standard entries.

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) — Multiple groups with varied uncertainty magnitudes (Standard tight vs Automated wide), 6 sites for pattern consistency, grouped comparison.
• DQ-02: Realistic Context (5/5) — Environmental monitoring: nitrate concentration at water sampling sites using three analytical methods. Realistic, neutral scientific domain.
• DQ-03: Appropriate Scale (4/4) — Nitrate 34–76 mg/L, SE 1.6–7.2 mg/L. Plausible for water quality monitoring. 3–4× SE difference between methods is realistic.

Code Quality (10/10)

• CQ-01: KISS Structure (3/3) — Clean linear structure: imports → constants → data → plot composition → save.
• CQ-02: Reproducibility (2/2) — Fully hardcoded deterministic data.
• CQ-03: Clean Imports (2/2) — All 14 plotnine imports used. os and pandas used. No unused imports.
• CQ-04: Code Elegance (2/2) — Clean, Pythonic. No over-engineering or fake UI elements.
• CQ-05: Output & API (1/1) — Saves as plot-{THEME}.png with dpi=400, width=8, height=4.5, units='in'.

Library Mastery (6/10)

• LM-01: Idiomatic Usage (5/5) — Exemplary grammar-of-graphics: geom_crossbar + position_dodge + scale_color/fill_manual + scale_x_discrete + labs + theme composed cleanly.
• LM-02: Distinctive Features (1/5) — geom_crossbar and position_dodge are ggplot2-origin features equally available in ggplot2 R or lets-plot. Grammar of graphics layering is distinctive to the ggplot family but not a feature uniquely exploitable in plotnine over similar libraries.

Score Caps Applied

• None — DE-01=4 (>2) and DE-02=4 (>2), so the "generic + no visual refinement" cap does not apply. All other caps not triggered.

Strengths

• Perfect Imprint palette compliance: #009E73 first (Automated), #C475FD second (Enhanced), #4467A3 third (Standard) in canonical order
• Both renders theme-correct with no dark-on-dark failures in the dark render
• All font sizes explicitly set with proper typography hierarchy
• Good visual refinement: minimal spines, tick marks removed, subtle y-axis-only grid
• Smart site ordering by Standard-method concentration creates a coherent left-to-right trend
• Realistic, neutral environmental science scenario with plausible SE values
• Idiomatic plotnine grammar-of-graphics code structure
• Perfect code quality: deterministic, clean imports, KISS structure

Weaknesses

• geom_crossbar renders a box shape rather than the canonical point-plus-whisker error bar. The spec describes "bars extending above and below each point" — consider geom_point + geom_errorbar or geom_pointrange for traditional error bar appearance.
• DE-03 (Data Storytelling) is carried by the subtitle text rather than visual encoding. Adding a reference line, highlighting the highest-uncertainty method, or annotating one site would create visual hierarchy without needing to read the subtitle.
• LM-02 limited: geom_crossbar + position_dodge work identically in ggplot2 R or lets-plot — no uniquely plotnine capability exploited.

Issues Found

1. SC-01 PARTIAL: geom_crossbar is a box-shaped range geom, not the standard point-plus-whisker error bar
   • Fix: Replace with geom_point(size=3, position=dodge) + geom_errorbar(aes(ymin='ymin', ymax='ymax'), width=0.15, position=dodge) for the canonical error bar appearance with visible caps
2. DE-03 LOW: Story depends on reading the subtitle; visual encoding alone doesn't create clear hierarchy
   • Fix: Add geom_hline(yintercept=regulatory_limit, linetype='dashed', color=INK_MUTED) as a reference, OR highlight the Automated group with slightly larger width=0.25 to visually emphasize its wider uncertainty
3. LM-02 LOW: No plotnine-distinctive feature
   • Fix: Could use facet_wrap('~method') as an alternative layout to showcase plotnine's faceting strength, or add stat_summary with a built-in statistic

AI Feedback for Next Attempt

The core issue is plot-type choice: switch from geom_crossbar to geom_point + geom_errorbar (or geom_pointrange) for a canonical error bar with visible central point and whisker caps. The subtitle story ("Automated shows 3–4× wider uncertainty") is correct — reinforce it visually by either adding a horizontal reference line (e.g., regulatory threshold) or using slightly different crossbar widths per method to create visual emphasis. Keep all palette, theme, data, and code quality — those are strong. Focus the repair on (1) chart type canonical form and (2) one visual storytelling element beyond the subtitle.

Verdict: APPROVED