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): The plot renders on a warm off-white background (~#FAF8F1, not pure white). Six circular markers and symmetric error bars with caps are shown in #009E73 (Okabe-Ito brand green) across categorical samples A–F. The title "errorbar-basic · plotnine · anyplot.ai" appears in dark text. Y-axis label "Measurement Value (mg/L)" and x-axis label "Experimental Sample" are descriptive and readable. Subtle horizontal grid lines at 30, 40, 50 mg/L. Only the bottom spine is visible (top/right removed). All text is clearly readable against the light background — PASS.

Dark render (plot-dark.png): The plot renders on a near-black background (~#1A1A17). Data colors are identical to the light render — same #009E73 green markers and error bars, confirming Okabe-Ito positions 1–7 are theme-independent. Title, axis labels, and tick labels all appear in light-colored text against the dark surface. Grid lines remain subtle. No dark-on-dark failures detected. All text is readable against the dark background — PASS.

⚠️ Code–Image Discrepancy: The committed code uses color="#306998" (Python Blue) and title "pyplots.ai", yet the images show #009E73 and "anyplot.ai". The images were likely generated from a different (correct) version of the code. The code as committed would not reproduce these images if re-run.

Score: 76/100

Category	Score	Max
Visual Quality	25	30
Design Excellence	9	20
Spec Compliance	15	15
Data Quality	13	15
Code Quality	8	10
Library Mastery	6	10
Total	76	100

Visual Quality (25/30)

• VQ-01: Text Legibility (6/8) — Sizes explicitly set (title 24pt, labels 20pt, ticks 16pt); both renders readable. Deducted 2: code has no INK/INK_SOFT tokens, dark-theme text readability relies on plotnine defaults rather than explicit theming.
• VQ-02: No Overlap (6/6) — All six category labels and values well-spaced, no collisions.
• VQ-03: Element Visibility (5/6) — Markers and error bars clearly visible; appropriate for 6 data points.
• VQ-04: Color Accessibility (2/2) — Single series #009E73 in images; CVD-safe, good luminance contrast in both themes.
• VQ-05: Layout & Canvas (3/4) — Good use of canvas; slight excess whitespace on the right side.
• VQ-06: Axis Labels & Title (2/2) — Y-axis has units (mg/L); x-axis label descriptive for categorical.
• VQ-07: Palette Compliance (1/2) — Images display correct #009E73 and theme-adaptive backgrounds. However, committed code has color="#306998" (Python Blue — explicit 0 per criteria) with no ANYPLOT_THEME handling. Partial credit acknowledges image correctness; code must be fixed.

Design Excellence (9/20)

• DE-01: Aesthetic Sophistication (4/8) — theme_minimal with explicit sizing; clean but a well-configured library default. No intentional visual hierarchy or palette customization.
• DE-02: Visual Refinement (3/6) — Images show minimal spines and subtle grid; moderate refinement. Code doesn't explicitly remove spines or configure grid opacity.
• DE-03: Data Storytelling (2/6) — Six samples with varied measurements and uncertainties are displayed correctly, but no focal point, emphasis, or narrative guides the viewer.

Spec Compliance (15/15)

• SC-01: Plot Type (5/5) — Correct: geom_point + geom_errorbar with visible caps.
• SC-02: Required Features (4/4) — Visible caps, consistent widths, symmetric error bars, 3–20 groups (6 used) all present.
• SC-03: Data Mapping (3/3) — Categorical samples on x, measurement values on y; all data visible.
• SC-04: Title & Legend (3/3) — Images show correct "errorbar-basic · plotnine · anyplot.ai"; no legend for single series (correct). ⚠️ Code has "pyplots.ai" — would fail if re-run.

Data Quality (13/15)

• DQ-01: Feature Coverage (5/6) — Six groups with varied means and error sizes demonstrate range variation. Missed opportunity to show asymmetric error bars mentioned in spec notes.
• DQ-02: Realistic Context (4/5) — Lab chemical measurements (mg/L) is a plausible scientific scenario. "Sample A–F" naming is slightly generic.
• DQ-03: Appropriate Scale (4/4) — Values 33–55 mg/L with errors 3–5.5 are realistic for chemical measurement contexts.

Code Quality (8/10)

• CQ-01: KISS Structure (3/3) — Linear: imports → DataFrame → ymin/ymax → ggplot chain → save.
• CQ-02: Reproducibility (2/2) — Fully deterministic hardcoded data.
• CQ-03: Clean Imports (2/2) — All plotnine imports are used; no dead imports.
• CQ-04: Code Elegance (1/2) — Code is readable but has fundamental correctness issues: wrong color, no theme adaptation, wrong title string.
• CQ-05: Output & API (0/1) — FAIL: Saves as plot.png not plot-light.png / plot-dark.png. Must use ggsave(plot, filename=f'plot-{THEME}.png', dpi=300, width=16, height=9).

Library Mastery (6/10)

• LM-01: Idiomatic Usage (4/5) — Grammar-of-graphics idiom correct: ggplot + aes + geom chain; geom_errorbar with ymin/ymax aesthetic is the proper plotnine pattern.
• LM-02: Distinctive Features (2/5) — Layer composition (point + errorbar) shows plotnine's strength; nothing beyond basic usage otherwise.

Score Caps Applied

• None

Strengths

• Correct plot type: geom_errorbar + geom_point layer composition is idiomatic plotnine
• Error bar caps visible and consistent widths across all points
• Realistic lab measurement data context with units (mg/L)
• 6 sample groups with varied means and error sizes showcase the plot type well
• KISS code structure; fully deterministic data
• Both rendered images show readable text in light and dark themes

Weaknesses

• CRITICAL: Code uses color="#306998" (Python Blue) — must be color="#009E73" (Okabe-Ito position 1)
• CRITICAL: Code has no ANYPLOT_THEME environment variable reading — no theme-adaptive chrome (PAGE_BG, INK, INK_SOFT tokens all missing)
• CRITICAL: Code saves as plot.png — must save as plot-{THEME}.png via ggsave() called twice
• CRITICAL: Title string contains pyplots.ai — must be anyplot.ai
• Code–image discrepancy: images appear generated from a correct version of code that is not currently committed
• Design Excellence weak: no explicit spine removal, no visual hierarchy, no storytelling

Issues Found

1. CQ-05 FAIL + VQ-07 FAIL: No theme adaptation and wrong output filename
   • Fix: Add import os; THEME = os.getenv("ANYPLOT_THEME", "light"); PAGE_BG = "#FAF8F1" if THEME == "light" else "#1A1A17"; INK = "#1A1A17" if THEME == "light" else "#F0EFE8"; INK_SOFT = "#4A4A44" if THEME == "light" else "#B8B7B0". Apply via theme() tokens. Save with ggsave(plot, filename=f'plot-{THEME}.png', dpi=300, width=16, height=9).
2. VQ-07 FAIL: Wrong color #306998 (Python Blue)
   • Fix: Replace with color="#009E73" for both geom_errorbar and geom_point.
3. SC-04 FAIL (in code): Title string has pyplots.ai
   • Fix: Change to f"errorbar-basic · plotnine · anyplot.ai".
4. DE-01/DE-02 LOW: Generic theme_minimal with no refinements
   • Fix: Remove top/right spines via panel_border, use subtle y-axis-only grid with panel_grid_minor=element_blank(), set explicit background via panel_background.

AI Feedback for Next Attempt

The core logic is correct — geom_point + geom_errorbar is the right plotnine approach. Fix the four critical issues: (1) replace #306998 with #009E73; (2) add ANYPLOT_THEME env var reading with PAGE_BG/INK/INK_SOFT tokens applied via theme(); (3) save as plot-{THEME}.png using ggsave(); (4) fix title to "anyplot.ai". For design, add explicit panel_background=element_rect(fill=PAGE_BG), remove panel_border, set plot_background=element_rect(fill=PAGE_BG, color=PAGE_BG), and apply INK/INK_SOFT to axis_title, axis_text, and plot_title elements.

Verdict: REJECTED

[github-actions]

🔧 Repair Attempt 1/3

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): The plot displays six error bar data points (Sample A through Sample F) on a warm off-white background (#FAF8F1). Each group has a filled circle center point and T-capped error bars — all rendered in brand green (#009E73). The title reads "errorbar-basic · plotnine · anyplot.ai" in dark ink. The y-axis is labelled "Measurement Value (mg/L)" and the x-axis "Experimental Sample", both clearly readable. Tick labels are dark and fully legible. A subtle horizontal grid is present; top and right spines are removed (L-frame). The overall layout is clean with generous whitespace. All text is readable against the light background.

Dark render (plot-dark.png): The same chart on a warm near-black background (#1A1A17). The data colors (green error bars and points) are identical to the light render — only chrome has flipped: title, axis labels, and tick labels are rendered in white/light text, the subtle grid is lighter, and the panel background is the correct near-black. No dark-on-dark failures observed; all text is readable against the dark background.

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

Score: 78/100

Category	Score	Max
Visual Quality	29	30
Design Excellence	9	20
Spec Compliance	14	15
Data Quality	13	15
Code Quality	8	10
Library Mastery	5	10
Total	78	100

Visual Quality (29/30)

• VQ-01: Text Legibility (7/8) — Font sizes explicitly set (title 24, axis 20, ticks 16); fully readable in both themes. Minor: title weight not differentiated (bold).
• VQ-02: No Overlap (6/6) — No text or element collisions.
• VQ-03: Element Visibility (6/6) — Error bars and center points clearly sized; caps visible.
• VQ-04: Color Accessibility (2/2) — Single CVD-safe green; no red-green-only encoding.
• VQ-05: Layout & Canvas (4/4) — 16:9 landscape, balanced margins, nothing cut off.
• VQ-06: Axis Labels & Title (2/2) — Y-axis includes units "(mg/L)"; x-axis is descriptive.
• VQ-07: Palette Compliance (2/2) — Rendered images show #009E73 first series and correct #FAF8F1/#1A1A17 backgrounds; however see code issue below.

Design Excellence (9/20)

• DE-01: Aesthetic Sophistication (4/8) — Professional but relies on theme_minimal() defaults. No intentional visual hierarchy or emphasis beyond choosing the brand green.
• DE-02: Visual Refinement (3/6) — theme_minimal() removes top/right spines; subtle horizontal grid; generous whitespace. Some extra refinement (spine removal per Okabe-Ito tokens, custom grid opacity) would lift this.
• DE-03: Data Storytelling (2/6) — Data is displayed faithfully but no visual hierarchy: all bars uniform weight and colour, no focal point, no ordering by magnitude.

Spec Compliance (14/15)

• SC-01: Plot Type (5/5) — Correct: geom_errorbar + geom_point implements error-bar-with-center correctly.
• SC-02: Required Features (4/4) — Caps present (width=0.3); consistent widths; symmetric errors computed from ymin/ymax.
• SC-03: Data Mapping (3/3) — Categorical x, numeric y, errors mapped correctly.
• SC-04: Title & Legend (2/3) — Images show correct title; code hard-codes "pyplots.ai" instead of "anyplot.ai". No legend needed (single series — correct).

Data Quality (13/15)

• DQ-01: Feature Coverage (4/6) — Demonstrates the core pattern (center + symmetric error caps) but shows only a single series. Spec explicitly suggests considering multiple series for group comparison and notes asymmetric error bars as a common variant; neither is present.
• DQ-02: Realistic Context (5/5) — Lab sample measurements in mg/L — neutral, plausible scientific scenario.
• DQ-03: Appropriate Scale (4/4) — Values 33–55, errors 2–5; realistic measurement uncertainty.

Code Quality (8/10)

• CQ-01: KISS Structure (3/3) — Flat: imports → data → plot → save. No functions or classes.
• CQ-02: Reproducibility (2/2) — Fully deterministic (hardcoded DataFrame).
• CQ-03: Clean Imports (2/2) — Only used symbols imported.
• CQ-04: Code Elegance (0/2) — Critical: color="#306998" (Python logo blue) is used instead of #009E73 (Okabe-Ito brand green). The code does not import os or read ANYPLOT_THEME, so no theme adaptation is implemented in code. The visible images appear to come from a prior run; the current code would produce blue, light-only output.
• CQ-05: Output & API (1/1) — plot.save("plot.png", dpi=300) matches the CQ-05 criterion.

Library Mastery (5/10)

• LM-01: Idiomatic Usage (4/5) — Grammar-of-graphics layer composition (+ operator) used correctly; aes() mapping for ymin/ymax is idiomatic plotnine.
• LM-02: Distinctive Features (1/5) — Uses only two basic geoms (geom_errorbar, geom_point). No plotnine-distinctive features: no stat_summary, no scale_*_manual for Okabe-Ito, no facet_wrap, no geom_crossbar.

Score Caps Applied

• None — DE-01=4 and DE-02=3, so the "generic + no refinement" cap (DE-01≤2 AND DE-02≤2) does not apply.

Strengths

• Correct plot type: geom_errorbar + geom_point is the idiomatic plotnine approach.
• Realistic, neutral scientific context (lab samples in mg/L) with plausible error magnitudes.
• KISS code structure — flat, readable, no over-engineering.
• Font sizes explicitly set at all three levels (title 24, axes 20, ticks 16).

Weaknesses

• Critical code bug: color="#306998" must be replaced with color="#009E73" (Okabe-Ito position 1). The rendered images appear to come from a prior version; the current code would ship blue error bars.
• No ANYPLOT_THEME support: os.getenv("ANYPLOT_THEME", "light") must be added; theme tokens (PAGE_BG, INK, INK_SOFT, ELEVATED_BG) must be threaded into theme() via element_rect/element_text/element_line. Without this the dark render cannot be produced correctly.
• Wrong title domain: "pyplots.ai" → "anyplot.ai".
• Single-series only: Add a second or third group (e.g., different experimental conditions) to demonstrate multi-series error bar comparison with Okabe-Ito colour ordering.
• Design depth: Sort bars by central value to create visual hierarchy; use scale_color_manual(values=OKABE_ITO) when multi-series. Remove panel border or reduce grid to y-only with explicit opacity.

Issues Found

1. CQ-04 CRITICAL — Wrong colour constant: color="#306998" must become the Okabe-Ito brand green color="#009E73". Fix affects both geom_errorbar and geom_point.
2. CQ-04 CRITICAL — No theme adaptation: Add import os, read THEME = os.getenv("ANYPLOT_THEME", "light"), define PAGE_BG/INK/INK_SOFT/ELEVATED_BG tokens, and apply them through theme(plot_background=element_rect(fill=PAGE_BG), panel_background=element_rect(fill=PAGE_BG), axis_text=element_text(color=INK_SOFT), axis_title=element_text(color=INK), plot_title=element_text(color=INK), ...). Save as f"plot-{THEME}.png".
3. SC-04 — Wrong site name in title: Change "pyplots.ai" → "anyplot.ai".
4. DQ-01 — Feature coverage: Add a second condition/group to show multi-series comparison (the spec explicitly mentions it); optionally demonstrate asymmetric error bars.
5. DE-03 — No visual hierarchy: Order categories by measurement value; use colour variation (multi-series Okabe-Ito) or size/alpha to create a focal point.

AI Feedback for Next Attempt

Replace #306998 with #009E73. Add full ANYPLOT_THEME support (import os, define PAGE_BG/INK/INK_SOFT tokens, apply via theme() element calls, save as plot-{THEME}.png). Fix title to anyplot.ai. Expand data to 2-3 groups (e.g., Method A / Method B / Method C) to demonstrate multi-series error bar comparison with scale_color_manual(values=OKABE_ITO) colouring. Sort bars by mean value for visual hierarchy (DE-03). Use element_rect(fill=None) to drop the panel border, and constrain the grid to horizontal lines only with panel_grid_minor=element_blank().

Verdict: REJECTED

[github-actions]

🔧 Repair Attempt 2/3

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): The plot renders on a warm off-white background (#FAF8F1). The title "errorbar-basic · plotnine · anyplot.ai" is displayed in bold dark ink at the top. The X-axis is labeled "Experimental Sample" with six categories (Sample A through F in alphabetical order), and the Y-axis is labeled "Measurement Value (mg/L)" with tick marks at 30, 40, and 50. Three methods (Method A, B, C) are shown with position-dodged errorbars and center points using Okabe-Ito colors (#009E73 green, #D55E00 orange, #0072B2 blue), though the three series cluster closely together because the method values are very similar for each sample. Subtle horizontal-only grid lines and no tick marks are present. A small legend labeled "Method" appears on the right side. All text is clearly readable against the light background — no light-on-light issues.

Dark render (plot-dark.png): The same plot renders on a warm near-black background (#1A1A17). All chrome elements (title, axis labels, tick labels) flip to light values — INK (#F0EFE8) for primary text and INK_SOFT (#B8B7B0) for secondary text — and are clearly readable against the dark background. No dark-on-dark failures observed. The data colors (green #009E73 and the Okabe-Ito sequence) are identical to the light render, as required. The horizontal grid lines, subtle axis line, and overall structure are identical. Brand green #009E73 reads well on the dark surface. Both renders pass the theme-readability check.

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

Score: 85/100

Category	Score	Max
Visual Quality	28	30
Design Excellence	11	20
Spec Compliance	15	15
Data Quality	13	15
Code Quality	10	10
Library Mastery	8	10
Total	85	100

Visual Quality (28/30)

• VQ-01: Text Legibility (8/8) — All sizes explicitly set: title 24pt, axis labels 20pt, ticks 16pt; readable in both themes
• VQ-02: No Overlap (6/6) — No text/label collisions; x-axis labels well-spaced
• VQ-03: Element Visibility (5/6) — Errorbars and points clearly visible, but three methods cluster tightly due to similar values
• VQ-04: Color Accessibility (2/2) — Okabe-Ito is CVD-safe; no red-green sole signal
• VQ-05: Layout & Canvas (3/4) — Good 16:9 proportions; canvas well-utilized; legend could be more prominent
• VQ-06: Axis Labels & Title (2/2) — "Measurement Value (mg/L)" includes units; title format correct
• VQ-07: Palette Compliance (2/2) — First series #009E73 ✓; Okabe-Ito order ✓; #FAF8F1 light / #1A1A17 dark backgrounds ✓; chrome tokens correct in both renders

Design Excellence (11/20)

• DE-01: Aesthetic Sophistication (5/8) — Above library defaults: Okabe-Ito custom palette, panel border removed, horizontal-only axis line, no tick marks; not yet FiveThirtyEight-level
• DE-02: Visual Refinement (4/6) — Notable refinements: panel_border blank, axis_ticks blank, panel_grid_major_x blank, horizontal-only subtle grid (alpha=0.10), elevated legend background
• DE-03: Data Storytelling (2/6) — Data is displayed but no clear visual hierarchy or emphasis; the three methods are too similar in value, preventing meaningful method comparison; the intended sample ordering by mean did not render (alphabetical order shown instead)

Spec Compliance (15/15)

• SC-01: Plot Type (5/5) — geom_errorbar + geom_point is correct; error bar plot with center points
• SC-02: Required Features (4/4) — Visible caps (width=0.35) ✓; symmetric errors ✓; multiple series in different colors ✓; consistent widths ✓
• SC-03: Data Mapping (3/3) — X: categorical samples, Y: measurement values, error: ±magnitude; all data visible
• SC-04: Title & Legend (3/3) — Title "errorbar-basic · plotnine · anyplot.ai" ✓; legend "Method" with Method A/B/C labels ✓

Data Quality (13/15)

• DQ-01: Feature Coverage (4/6) — Shows symmetric error bars, multi-group comparison, and variation in error sizes; however, the three methods have nearly identical measurement values for each sample (within 3-5 mg/L), making the multi-method comparison visually uninformative
• DQ-02: Realistic Context (5/5) — Lab methods measuring across experimental samples; scientific context is neutral, realistic, and plausible
• DQ-03: Appropriate Scale (4/4) — Values 33–57 mg/L with errors 2.5–5.5 mg/L are realistic; 6 samples × 3 methods = appropriate scope

Code Quality (10/10)

• CQ-01: KISS Structure (3/3) — Flat: imports → theme tokens → data → categorical ordering → plot → save; no functions/classes
• CQ-02: Reproducibility (2/2) — Deterministic hardcoded data; no random calls needed
• CQ-03: Clean Imports (2/2) — All plotnine imports used; no unused imports
• CQ-04: Code Elegance (2/2) — Clean and Pythonic; no fake UI; appropriate complexity; no AR-08 violations
• CQ-05: Output & API (1/1) — Saves as plot-{THEME}.png using plot.save() with correct parameters

Library Mastery (8/10)

• LM-01: Idiomatic Usage (5/5) — Expert ggplot grammar: ggplot() + geom layers, aes() mappings, position_dodge(), scale_color_manual(), theme() with element_* system; fully idiomatic
• LM-02: Distinctive Features (3/5) — Uses plotnine-distinctive features: position_dodge() (grammar of graphics concept), pd.Categorical for factor ordering (ggplot2-style), element_blank() / element_line() theming; could go further with stat_* or annotate layers

Score Caps Applied

• None — no cap conditions triggered (DE-01=5>2, DE-02=4>2, all other caps clear)

Strengths

• Perfect code quality: flat, reproducible, deterministic, clean imports, correct output naming
• Full spec compliance: geom_errorbar with visible caps, position_dodge for multi-series, correct Okabe-Ito colors
• Theme-adaptive implementation is correct in both renders: token values, backgrounds, and chrome all flip properly
• Idiomatic plotnine grammar of graphics usage throughout
• Excellent visual refinement: no tick marks, horizontal-only subtle grid, no panel border, L-shaped axis line only

Weaknesses

• Three methods have nearly identical measurement values (within 3–5 mg/L per sample), causing their errorbars to cluster and overlap visually despite position_dodge — the multi-method comparison loses informational value
• Sample ordering by mean (pd.Categorical) did not render in sorted order — x-axis shows alphabetical order instead of ascending mean measurement
• DE-03 storytelling is weak: no visual emphasis, annotation, or data design that guides the viewer to an insight

Issues Found

1. DQ-01 / DE-03 LOW: Three methods' measurement values are too similar — they nearly overlap even with dodging, defeating the purpose of a grouped comparison. Fix: spread Method B and C values further from Method A (e.g., 8–15 mg/L differences) so the three groups are clearly visually separated and the comparison tells a story.
2. DE-03 LOW: Sample ordering by mean value did not take effect in the rendered output. Fix: verify pd.Categorical ordering works in plotnine or enforce order via scale_x_discrete(limits=sample_order).

AI Feedback for Next Attempt

The grouped error bar design is sound but the data undermines it: three methods with near-identical values collapse into one cluster even after dodging. Redesign the data so Method A, B, C have clearly different mean measurements (e.g., Method B ~10 mg/L higher, Method C ~15 mg/L lower than Method A per sample) to make the three-group comparison visually meaningful. Also enforce sample ordering via scale_x_discrete(limits=sample_order) rather than relying on pd.Categorical alone.

Verdict: APPROVED