feat(seaborn): implement bar-tornado-sensitivity

[github-actions]

Implementation: bar-tornado-sensitivity - python/seaborn

Implements the python/seaborn version of bar-tornado-sensitivity.

File: plots/bar-tornado-sensitivity/implementations/python/seaborn.py

Parent Issue: #4566

---

🤖 impl-generate workflow

[claude]

AI Review - Attempt 1/3

Image Description

Light render (plot-light.png): The plot renders on a warm off-white #FAF8F1 background. The title "bar-tornado-sensitivity · python · seaborn · anyplot.ai" is clearly readable at the top. Two Imprint palette colors are used: brand green #009E73 for Low Scenario and slate blue #4467A3 for High Scenario. The tornado shape is clear — widest bars (Discount Rate, Revenue Growth) at the bottom, narrowest (Inflation Rate) at the top. A vertical dashed reference line marks the $120M base case. Bar-end annotations show absolute NPV values. The x-axis tick labels are formatted as "$70", "$80", etc. and are readable. Critical issue: Four Y-axis tick labels are clipped at the left canvas edge — "Discount Rate" displays as "iscount Rate", "Revenue Growth" as "evenue Growth", "Operating Margin" as "perating Margin", and "Initial Investment" as "nitial Investment". The first character of each is missing. This is a hard AR-09 violation.

Dark render (plot-dark.png): Background is warm near-black #1A1A17. Title, x-axis labels, tick labels, and bar-end annotations are rendered in light text and are clearly readable against the dark background. No dark-on-dark failures in the text chrome. Data colors (brand green and slate blue) are identical to the light render — only background, grid, and legend frame colors flip. The same four Y-axis label clipping failures are present: "Discount Rate", "Revenue Growth", "Operating Margin", and "Initial Investment" all have their first character chopped at the left canvas edge. The dark render does not introduce any new legibility failures beyond the AR-09 clipping issue.

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

Score: 0/100

(AR-09 auto-reject — see below)

Category	Score	Max
Visual Quality	23	30
Design Excellence	14	20
Spec Compliance	15	15
Data Quality	15	15
Code Quality	10	10
Library Mastery	7	10
Total (raw)	84	100
Total (after AR-09)	0	100

Visual Quality (23/30)

• VQ-01: Text Legibility (4/8) — Title and x-axis labels are clear; four Y-axis labels clipped at left edge (AR-09)
• VQ-02: No Overlap (5/6) — Minor crowding of bar-end annotations near base line but no severe collisions
• VQ-03: Element Visibility (6/6) — Bars fully visible and distinguishable in both themes
• VQ-04: Color Accessibility (2/2) — Green/blue pair is CVD-safe; no red-green sole signal
• VQ-05: Layout & Canvas (2/4) — Canvas gate passed; left margin insufficient for long Y-axis labels causing clipping
• VQ-06: Axis Labels & Title (2/2) — "Net Present Value ($M)" and "Input Parameter" descriptive with units
• VQ-07: Palette Compliance (2/2) — First series #009E73, second #4467A3 (Imprint position 3); backgrounds #FAF8F1 / #1A1A17 correct

Design Excellence (14/20)

• DE-01: Aesthetic Sophistication (5/8) — Clear visual hierarchy via bold top-3 labels; bar-end annotations add professional context; Imprint palette correctly applied; above generic defaults
• DE-02: Visual Refinement (4/6) — Left spine removed, subtle x-axis grid at 15% opacity, edgecolor="none" on bars for clean fill
• DE-03: Data Storytelling (5/6) — Tornado shape immediately ranks parameters by influence; bold labeling draws attention to key drivers; base case reference line clearly explained with annotation

Spec Compliance (15/15)

• SC-01: Plot Type (5/5) — Correct horizontal bar tornado chart
• SC-02: Required Features (4/4) — Reference line at base case, bars sorted by range descending, dual colors for low/high, bar-end value annotations
• SC-03: Data Mapping (3/3) — X-axis correctly shows absolute NPV output, Y-axis shows all 10 parameters sorted by impact
• SC-04: Title & Legend (3/3) — Title matches required format; legend labels "Low Scenario" / "High Scenario"

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) — Demonstrates all tornado chart aspects: inverted relationships, dual-sided bars, sorted ranking, reference line, annotations
• DQ-02: Realistic Context (5/5) — NPV sensitivity for capital investment is realistic, neutral, and professionally relevant
• DQ-03: Appropriate Scale (4/4) — $120M base NPV with ±$38M worst-case swing; financially plausible values

Code Quality (10/10)

• CQ-01: KISS Structure (3/3) — No unnecessary classes/functions; linear top-to-bottom flow
• CQ-02: Reproducibility (2/2) — All data is deterministic; no random generation needed
• CQ-03: Clean Imports (2/2) — All five imports (os, sys, matplotlib, numpy, pandas, seaborn) are used
• CQ-04: Code Elegance (2/2) — Appropriate complexity; no fake UI or simulated interactivity
• CQ-05: Output & API (1/1) — Saves as plot-{THEME}.png; no bare plot.png

Library Mastery (7/10)

• LM-01: Idiomatic Usage (4/5) — sns.barplot with hue/dodge=False for tornado overlap, sns.set_theme rc params, sns.despine — idiomatic seaborn; minor deduction as tick relabeling and per-bar annotation loops drop to matplotlib level
• LM-02: Distinctive Features (3/5) — dodge=False with hue for the two-scenario overlap pattern, full sns.set_theme rc customization, sns.despine(left=True) — seaborn-specific features put to good use

Score Caps Applied

• AR-09 (Edge Clipping) — Hard auto-reject: score overridden to 0/100

Strengths

• Correct tornado sort order with widest bars at bottom; characteristic shape is immediately clear
• Imprint palette applied correctly — #009E73 for first series, #4467A3 for second
• Bold emphasis on top-3 most impactful parameters adds useful visual hierarchy
• Bar-end absolute NPV annotations add strong interpretive value
• Full theme-adaptive chrome: background, ink, legend frame all flip cleanly between light and dark
• Deterministic, realistic financial dataset with realistic inverted relationships (lower Material Cost → higher NPV)
• All spec requirements met: reference line, sort order, dual colors, bar-end labels

Weaknesses

• AR-09 violation: Y-axis tick labels for "Discount Rate", "Revenue Growth", "Operating Margin", and "Initial Investment" have their first character(s) clipped at the left canvas edge. Fix: add plt.subplots_adjust(left=0.18) (or fig.subplots_adjust(left=0.18)) after plt.subplots() to widen the left margin so all Y-axis labels are fully visible within the canvas.
• The ax.set_xlim call widens the view to prevent right-side annotation clipping, but no equivalent left-margin adjustment is made for Y-axis labels — the two fixes are symmetric and both needed.

Issues Found

1. AR-09 CRITICAL: Y-axis tick labels clipped at left canvas edge
   • Affected labels: "Discount Rate" → "iscount Rate", "Revenue Growth" → "evenue Growth", "Operating Margin" → "perating Margin", "Initial Investment" → "nitial Investment"
   • Fix: Add fig.subplots_adjust(left=0.18) immediately after fig, ax = plt.subplots(...). This reserves ~18% of the figure width for Y-axis labels, which is sufficient for the longest label ("Initial Investment" at 18 chars).

AI Feedback for Next Attempt

The implementation is otherwise high quality — correct data, palette, theme adaptation, and spec compliance. The single blocking fix is adding fig.subplots_adjust(left=0.18) (or a similar value like 0.16–0.20) after plt.subplots() to prevent Y-axis label clipping at the left canvas edge. Do NOT add bbox_inches='tight' as that would cause canvas dimension drift. Use subplots_adjust to push the plot area rightward so all 10 parameter labels fit within the PNG boundary. Everything else (data, colors, annotations, theme tokens, legend) should be kept as-is.

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 1/3

Image Description

Light render (plot-light.png): The plot renders on a warm off-white #FAF8F1 background. It shows a horizontal tornado diagram with 10 parameter bars sorted by NPV impact range. Two colors distinguish the scenarios: brand green #009E73 for Low Scenario and Imprint blue #4467A3 for High Scenario. A dashed vertical reference line marks the Base Case at $120M. Bar-end value annotations ($82M–$152M) appear at both ends of each bar. The title "bar-tornado-sensitivity · python · seaborn · anyplot.ai" is readable at the top. X-axis tick labels show dollar amounts ($70–$170) and are readable. CRITICAL AR-09 FAILURE: Y-axis tick labels for the four longest parameter names are clipped at the left canvas edge — "Operating Margin" renders as "perating Margin" (leading "O" missing), "Initial Investment" as "nitial Investment" (leading "I" missing), "Revenue Growth" as "evenue Growth" (leading "Re" missing), and "Discount Rate" as "iscount Rate" (leading "Di" missing). All four are chopped at the left canvas boundary.

Dark render (plot-dark.png): The same plot on a warm near-black #1A1A17 background. Chrome correctly flips to light text — title, axis labels, tick labels, and legend text are all readable in light tones against the dark background. Data colors are identical to the light render (#009E73 green and #4467A3 blue). The same AR-09 Y-axis label clipping is visible: "Operating Margin", "Initial Investment", "Revenue Growth", and "Discount Rate" have their leading characters chopped at the left canvas edge. No dark-on-dark text failures beyond the clipping issue. Legibility verdict for chrome elements (other than the clipped labels): PASS. Legibility verdict for clipped Y-axis labels: FAIL (AR-09).

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

Score: 0/100

Category	Score	Max
Visual Quality	21	30
Design Excellence	13	20
Spec Compliance	15	15
Data Quality	15	15
Code Quality	10	10
Library Mastery	7	10
Total (pre-AR-09)	81	100
Total (AR-09 applied)	0	100

Visual Quality (21/30)

• VQ-01: Text Legibility (3/8) — Font sizes explicitly set (title 12pt, labels 10pt, ticks 8pt), but Y-axis labels for 4 of 10 parameter names are clipped at the left canvas edge (AR-09). Remaining text proportions are good.
• VQ-02: No Overlap (6/6) — No text overlaps with other text or data elements.
• VQ-03: Element Visibility (6/6) — Bars are clearly visible with appropriate sizing for the 10-parameter dataset.
• VQ-04: Color Accessibility (2/2) — Green vs. blue (not red-green), CVD-safe Imprint palette.
• VQ-05: Layout & Canvas (0/4) — Left margin is too narrow to accommodate the longest Y-axis labels; fig.subplots_adjust(left=...) was not set explicitly, causing label clipping at the canvas boundary.
• VQ-06: Axis Labels & Title (2/2) — "Net Present Value ($M)" and "Input Parameter" — descriptive with units.
• VQ-07: Palette Compliance (2/2) — Low Scenario = #009E73 (brand green, correct first series); High Scenario = #4467A3 (Imprint blue, position 3); backgrounds #FAF8F1 / #1A1A17 correct; chrome theme-adaptive in both renders.

Design Excellence (13/20)

• DE-01: Aesthetic Sophistication (5/8) — Bold top-3 Y-axis labels for visual hierarchy, base case annotation, bar-end value labels, clean despined layout. Clearly above generic defaults, but not quite publication-ready.
• DE-02: Visual Refinement (4/6) — Left spine removed via sns.despine(left=True), X-axis-only grid at subtle alpha, legend styled with ELEVATED_BG. Good refinement overall.
• DE-03: Data Storytelling (4/6) — Tornado shape communicates ranking immediately; bold top-3 labels draw the eye to highest-impact parameters; base case reference line gives clear context. Visual hierarchy guides the viewer.

Spec Compliance (15/15)

• SC-01: Plot Type (5/5) — Correct tornado/sensitivity horizontal bar chart, bars sorted by impact range.
• SC-02: Required Features (4/4) — Vertical reference line at base case, sorted bars (widest at bottom), dual colors for low/high, parameter names on Y-axis, bar-end value annotations.
• SC-03: Data Mapping (3/3) — X-axis shows NPV values, Y-axis shows parameter names, base case at $120M.
• SC-04: Title & Legend (3/3) — Title is exactly "bar-tornado-sensitivity · python · seaborn · anyplot.ai"; legend shows "Low Scenario" / "High Scenario".

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) — Shows all tornado diagram aspects: sorted bars, dual colors, inverted relationships (Material Cost, Tax Rate), base case reference, value annotations.
• DQ-02: Realistic Context (5/5) — NPV sensitivity analysis for a capital investment project — realistic, neutral, and comprehensible.
• DQ-03: Appropriate Scale (4/4) — NPV range $82M–$152M around $120M base case; plausible for a mid-size capital project.

Code Quality (10/10)

• CQ-01: KISS Structure (3/3) — Flat script: imports → theme → data → plot → save. No functions or classes.
• CQ-02: Reproducibility (2/2) — Fully deterministic synthetic data; no randomness.
• CQ-03: Clean Imports (2/2) — All imports (os, sys, matplotlib, numpy, pandas, seaborn) are used.
• CQ-04: Code Elegance (2/2) — Clean, well-organized Pythonic code. sys.path guard is documented and appropriate.
• CQ-05: Output & API (1/1) — Saves as plot-{THEME}.png without bbox_inches='tight'. Current seaborn API.

Library Mastery (7/10)

• LM-01: Idiomatic Usage (4/5) — sns.barplot(hue=...) for dual-color bars, sns.set_theme(rc={...}) for theme configuration, sns.despine() for spine removal. Idiomatic and correct.
• LM-02: Distinctive Features (3/5) — Uses seaborn's hue-based barplot for the dual-scenario bars (seaborn-specific convenience), and sns.set_theme rc overrides for cohesive theme application.

Score Caps Applied

• AR-09 fires → Score = 0 (Y-axis tick labels clipped at left canvas edge)

Strengths

• Excellent spec compliance — all required tornado features implemented correctly (sorted bars, reference line, dual colors, value annotations, parameter sorting by impact)
• Strong data quality: realistic NPV sensitivity analysis with plausible inverted relationships (Material Cost, Tax Rate) and correct value ranges
• Good design storytelling: bold top-3 labels create visual hierarchy; base case annotation adds context
• Perfect palette compliance: Imprint colors correctly applied, theme-adaptive chrome in both renders
• Clean, flat code structure with no unnecessary complexity

Weaknesses

• AR-09 (BLOCKER): Y-axis tick labels clipped at left canvas edge — "Operating Margin" → "perating Margin", "Initial Investment" → "nitial Investment", "Revenue Growth" → "evenue Growth", "Discount Rate" → "iscount Rate". Leading characters are missing from the canvas in both light and dark renders. Fix: add fig.subplots_adjust(left=0.22) (or higher) before savefig to reserve enough left margin for the longest parameter names. Do NOT use bbox_inches='tight' — use explicit subplots_adjust.

Issues Found

1. AR-09 BLOCKER: Y-axis tick labels clipped at left canvas edge
   • Cause: Default matplotlib left margin (0.125) is too narrow for long parameter names like "Operating Margin" (15 chars) and "Initial Investment" (18 chars) at 8pt bold
   • Fix: fig.subplots_adjust(left=0.22) before plt.savefig(...). Check all 10 labels are fully visible after adjustment. The bold top-3 labels need slightly more room than regular labels.

AI Feedback for Next Attempt

The implementation is nearly correct — spec compliance, data quality, and code quality are all excellent. The only blocker is insufficient left margin causing Y-axis label clipping at the canvas edge. Add fig.subplots_adjust(left=0.22) (or tune to 0.20–0.24) before plt.savefig(...). Do not use bbox_inches='tight'. Verify that all 10 parameter names — especially the bold "Revenue Growth" and "Discount Rate" — are fully visible with no clipping in both light and dark renders.

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=8257

---

🤖 impl-review

[claude]

AI Review - Attempt 1/3

Image Description

Light render (plot-light.png): The plot renders on a warm off-white #FAF8F1 background. The title "bar-tornado-sensitivity · python · seaborn · anyplot.ai" is fully readable at top. X-axis label "Net Present Value ($M)" and Y-axis label "Input Parameter" are readable. Horizontal bars form a classic tornado shape (widest at bottom = Discount Rate, narrowest at top = Inflation Rate). Green (#009E73) represents Low Scenario and blue (#4467A3) represents High Scenario. A dashed vertical base case reference line sits at $120M. Bar-end annotations show absolute NPV values. Legend in upper-right. Bold labels highlight the top-3 most impactful parameters. HOWEVER: 4 Y-axis tick labels are clipped at the left canvas edge — "Operating Margin" renders as "perating Margin", "Initial Investment" as "nitial Investment", "Revenue Growth" as "evenue Growth", "Discount Rate" as "iscount Rate". The first character of each is chopped off at the canvas left border.

Dark render (plot-dark.png): Background is warm near-black #1A1A17 — correct. Title, axis labels, and tick labels use light-colored text against the dark background with no dark-on-dark failures. Data colors are identical to the light render (#009E73 and #4467A3 unchanged — only chrome flips). Legend box has the correct dark elevated background with light text. The same AR-09 clipping issue is present: the same 4 Y-axis parameter name labels ("Operating Margin", "Initial Investment", "Revenue Growth", "Discount Rate") have their first character cut off at the left canvas edge.

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

Score: 0/100

Category	Score	Max
Visual Quality	0	30
Design Excellence	13	20
Spec Compliance	15	15
Data Quality	15	15
Code Quality	10	10
Library Mastery	7	10
Total	0	100

Score forced to 0 by AR-09 auto-reject. Indicative score without AR-09: ~79/100.

Visual Quality (0/30)

• VQ-01: Text Legibility (0/8) — AR-09 failure: 4 Y-axis labels clipped at left canvas edge
• VQ-02: No Overlap (5/6) — Minor: value annotations slightly close on upper rows
• VQ-03: Element Visibility (6/6) — Bars well-sized and distinguishable
• VQ-04: Color Accessibility (2/2) — Good luminance contrast, CVD-safe
• VQ-05: Layout & Canvas (2/4) — Canvas passes dimension gate but left margin too tight (causes AR-09)
• VQ-06: Axis Labels & Title (2/2) — Descriptive with units
• VQ-07: Palette Compliance (2/2) — First series #009E73, second #4467A3; correct backgrounds both themes

Design Excellence (13/20)

• DE-01: Aesthetic Sophistication (5/8) — Custom Imprint palette, bold top-3 labels for emphasis, base case annotation — above configured defaults
• DE-02: Visual Refinement (4/6) — Spines removed, subtle x-axis-only grid, good whitespace
• DE-03: Data Storytelling (4/6) — Bold top-3 parameter labels create visual hierarchy; tornado shape clearly ranks impact

Spec Compliance (15/15)

• SC-01: Plot Type (5/5) — Correct tornado diagram
• SC-02: Required Features (4/4) — Base case reference line, sorted by range, dual colors, bar-end value labels
• SC-03: Data Mapping (3/3) — 10 parameters, base case $120M, correct axis mapping
• SC-04: Title & Legend (3/3) — Title format correct; legend labels "Low Scenario" / "High Scenario"

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) — Inverted relationships shown, diverse impact ranges, 10 realistic parameters
• DQ-02: Realistic Context (5/5) — Real-world capital investment NPV sensitivity analysis
• DQ-03: Appropriate Scale (4/4) — Realistic $M values, plausible parameter impact ranges

Code Quality (10/10)

• CQ-01: KISS Structure (3/3)
• CQ-02: Reproducibility (2/2) — Deterministic hardcoded data
• CQ-03: Clean Imports (2/2)
• CQ-04: Code Elegance (2/2)
• CQ-05: Output & API (1/1) — plot-{THEME}.png, no bbox_inches='tight'

Library Mastery (7/10)

• LM-01: Idiomatic Usage (4/5) — sns.barplot with hue/palette/dodge=False is idiomatic seaborn for dual-scenario bars
• LM-02: Distinctive Features (3/5) — Seaborn hue+dodge=False pattern for overlapping categorical bars

Score Caps Applied

• AR-09 auto-reject → Score = 0 (overrides all other scoring)

Strengths

• Correct tornado diagram with bars sorted by impact range (widest at bottom)
• Imprint palette correctly applied — #009E73 Low Scenario, #4467A3 High Scenario
• Both themes render with correct backgrounds and readable chrome (no dark-on-dark)
• Bar-end value annotations clearly show absolute NPV outcomes
• Bold top-3 parameter labels add effective visual hierarchy
• All spec requirements met: base case reference line, dual colors, sorted order, bar-end labels
• Realistic NPV sensitivity analysis — neutral, real-world business context
• Clean KISS code structure with no over-engineering
• savefig correctly omits bbox_inches='tight' — canvas stays at exact 3200×1800

Weaknesses

• AR-09 VIOLATION: Y-axis tick labels clipped at the left canvas edge — "Operating Margin" renders as "perating Margin", "Initial Investment" as "nitial Investment", "Revenue Growth" as "evenue Growth", "Discount Rate" as "iscount Rate". The default matplotlib left subplot margin (~12.5%) is too narrow for these long parameter names at this font size. Fix: add fig.subplots_adjust(left=0.22) (or left=0.24 if still tight) before plt.savefig to give the Y-axis labels sufficient space. Do NOT use bbox_inches='tight' as a workaround — it would drift the canvas size.

Issues Found

1. AR-09 CLIPPING: Y-axis parameter name tick labels clipped at left canvas edge
   • Affected labels: "Operating Margin", "Initial Investment", "Revenue Growth", "Discount Rate"
   • Root cause: Default matplotlib left subplot margin (~12.5% ≈ 1.0 inch on 8-inch figure) is insufficient for these 13–18 character labels at 8pt
   • Fix: fig.subplots_adjust(left=0.22) inserted after fig, ax = plt.subplots(...) and before sns.barplot(...)

AI Feedback for Next Attempt

The implementation is near-excellent in all dimensions except a single layout fix. Add fig.subplots_adjust(left=0.22) (or left=0.24 to be safe) right after creating the figure with plt.subplots(). This widens the left margin so all Y-axis parameter name labels ("Operating Margin", "Initial Investment", "Revenue Growth", "Discount Rate") fully display within the canvas bounds. Do NOT add bbox_inches='tight' to savefig — the canvas must stay at exactly 3200×1800 px. All other elements (Imprint palette, theme chrome, tornado shape, spec features, bar-end annotations, bold top-3 labels, base case line) are correct and should be preserved unchanged.

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=8257

---

🤖 impl-review