feat(letsplot): implement spectrum-nmr

[github-actions]

Implementation: spectrum-nmr - letsplot

Implements the letsplot version of spectrum-nmr.

File: plots/spectrum-nmr/implementations/letsplot.py

Parent Issue: #4676

---

🤖 impl-generate workflow

[claude]

AI Review - Attempt 1/3

Image Description

The plot shows a synthetic ¹H NMR spectrum of ethanol rendered as a blue line (#306998) on a white background with a minimal theme. The x-axis is labeled "δ Chemical Shift (ppm)" and runs from 12 on the left to -1 on the right (correctly reversed per NMR convention). The y-axis is labeled "Intensity (a.u.)" and runs from 0 to ~1.5. The title reads "Ethanol ¹H NMR · spectrum-nmr · letsplot · pyplots.ai" in bold at the top. Four peak groups are visible: a tall triplet (three closely spaced peaks) at 1.18 ppm labeled "CH₃ (triplet) 1.18 ppm", a quartet (four peaks) at 3.69 ppm labeled "CH₂ (quartet) 3.69 ppm", a single broader peak at 2.61 ppm labeled "OH (singlet) 2.61 ppm", and a reference peak at 0.00 ppm labeled "TMS 0.00 ppm". The baseline is clean with subtle noise. Horizontal grid lines are faint gray; vertical grid lines are removed. The layout uses good proportions with generous margins.

Score: 87/100

Category	Score	Max
Visual Quality	28	30
Design Excellence	13	20
Spec Compliance	15	15
Data Quality	15	15
Code Quality	10	10
Library Mastery	6	10
Total	87	100

Visual Quality (28/30)

• VQ-01: Text Legibility (8/8) — All font sizes explicitly set: title 24pt, axis titles 20pt, axis text 16pt. All perfectly readable.
• VQ-02: No Overlap (5/6) — The CH₃ label at 1.18 ppm slightly overlaps with the top of its peak line. All other labels are well-positioned.
• VQ-03: Element Visibility (6/6) — Peaks are sharp, well-resolved, and clearly visible. Line width of 1.0 is appropriate for this spectrum type.
• VQ-04: Color Accessibility (4/4) — Single blue color on white background, fully accessible.
• VQ-05: Layout & Canvas (3/4) — Good proportions overall, but the left half of the plot (4–12 ppm) is entirely empty baseline. Scientifically accurate but wastes canvas space.
• VQ-06: Axis Labels & Title (2/2) — "δ Chemical Shift (ppm)" and "Intensity (a.u.)" — descriptive with units.

Design Excellence (13/20)

• DE-01: Aesthetic Sophistication (5/8) — Python Blue line, custom dark text colors (#333333, #555555, #222222), minimal theme. Above defaults but not publication-exceptional.
• DE-02: Visual Refinement (4/6) — X-grid removed, subtle y-grid (light gray, 0.3 width), ticks removed, custom margins. Good refinement but spines still present via theme_minimal.
• DE-03: Data Storytelling (4/6) — Peak labels with functional group names and splitting patterns guide the viewer through ethanol's molecular structure. Visual hierarchy from peak heights naturally tells the story.

Spec Compliance (15/15)

• SC-01: Plot Type (5/5) — Correct NMR spectrum as line plot.
• SC-02: Required Features (4/4) — All spec requirements met: reversed x-axis, sharp peaks with splitting patterns (singlet/triplet/quartet), labeled key peaks with chemical shift values, TMS reference at 0 ppm, clean baseline with minimal noise.
• SC-03: Data Mapping (3/3) — Chemical shift on x-axis, intensity on y-axis. Correct mapping.
• SC-04: Title & Legend (3/3) — Title follows format "spectrum-nmr · letsplot · pyplots.ai". No legend needed for single-series plot.

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) — Demonstrates singlet, triplet, quartet, and reference peak. Correct intensity ratios (1:2:1 triplet, 1:3:3:1 quartet). Broader OH peak shows exchange effects.
• DQ-02: Realistic Context (5/5) — Ethanol ¹H NMR is a classic, neutral chemistry example. Scientifically accurate peak positions and splitting patterns.
• DQ-03: Appropriate Scale (4/4) — Chemical shifts (0–4 ppm range for ethanol) and relative intensities are realistic. Lorentzian line shapes with appropriate half-widths.

Code Quality (10/10)

• CQ-01: KISS Structure (3/3) — Clean Imports → Data → Plot → Save structure with no functions or classes.
• CQ-02: Reproducibility (2/2) — np.random.seed(42) set for baseline noise.
• CQ-03: Clean Imports (2/2) — All imports used (numpy, pandas, lets_plot, export_ggsave).
• CQ-04: Code Elegance (2/2) — Lorentzian peak generation using loop with center/amplitude tuples is clean and Pythonic. No over-engineering.
• CQ-05: Output & API (1/1) — Saves as plot.png with current API (export_ggsave with scale=3).

Library Mastery (6/10)

• LM-01: Idiomatic Usage (4/5) — Good use of ggplot grammar: aes, geom_line, geom_text, labs, scale_x_reverse, scale_y_continuous with expand, theme_minimal + theme customization, ggsize. Idiomatic but standard.
• LM-02: Distinctive Features (2/5) — Uses standard ggplot-style features available in many libraries. No lets-plot-specific features like tooltips, interactive layers, or geom_area_ridges that would distinguish it from plotnine.

Score Caps Applied

• None — no score caps triggered.

Strengths

• Excellent scientific accuracy: Lorentzian line shapes, correct splitting patterns, proper intensity ratios, realistic chemical shifts for ethanol
• Perfect spec compliance: all required features implemented (reversed x-axis, peak labels, TMS reference, clean baseline)
• Clean, well-structured code with appropriate use of the ggplot grammar
• Good visual design with customized theme, subtle grid, and clear peak annotations

Weaknesses

• The CH₃ peak label overlaps slightly with the peak itself — could use a y-offset or nudge
• Canvas underutilized: the 4–12 ppm region is empty baseline; narrowing the x-range to -0.5 to 5 ppm would better utilize space
• No lets-plot-distinctive features used — could leverage geom_area for filled peaks or interactive tooltips
• Design could be elevated with area fill under peaks for visual impact

Issues Found

1. VQ-02 MINOR: CH₃ label slightly overlaps peak top
   • Fix: Increase the y-offset for the CH₃ label in peak_labels DataFrame
2. VQ-05 MINOR: Large empty region from 4–12 ppm wastes canvas space
   • Fix: Narrow x-range to approximately -0.5 to 5 ppm, or reduce linspace range to focus on the spectral region of interest
3. LM-02 LOW: Generic ggplot usage without lets-plot-distinctive features
   • Fix: Consider using geom_area for filled spectrum, or leverage lets-plot's interactive capabilities with tooltips/hover

AI Feedback for Next Attempt

To reach 90+: (1) Fix CH₃ label overlap by increasing its y-offset. (2) Consider narrowing x-range to ~-0.5–5 ppm to better utilize canvas space since ethanol has no peaks above 4 ppm. (3) Use a lets-plot-distinctive feature such as geom_area with alpha fill under the spectrum curve for visual impact, or leverage interactive tooltips. (4) Consider adding subtle area fill or color variation to enhance design excellence.

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

The plot shows an ethanol ¹H NMR spectrum with chemical shift (ppm) on the x-axis (reversed from 5.0 to -0.5, following NMR convention) and intensity (a.u.) on the y-axis. The spectrum is rendered in Python Blue (#306998) as a line plot with a subtle light blue area fill underneath. Four distinct peak groups are visible and labeled in bold dark text: CH₂ (quartet) at 3.69 ppm showing four peaks with 1:3:3:1 intensity ratio, OH (singlet) at 2.61 ppm as a single broader peak, CH₃ (triplet) at 1.18 ppm showing three peaks with 1:2:1 ratio (the tallest feature), and TMS at 0.00 ppm as a reference singlet. The baseline is clean with minimal noise. The title reads "Ethanol ¹H NMR · spectrum-nmr · letsplot · pyplots.ai" in bold at the top. Axis labels include "δ Chemical Shift (ppm)" and "Intensity (a.u.)". The theme is minimal with subtle horizontal grid lines and no axis ticks.

Score: 90/100

Category	Score	Max
Visual Quality	28	30
Design Excellence	15	20
Spec Compliance	15	15
Data Quality	15	15
Code Quality	10	10
Library Mastery	7	10
Total	90	100

Visual Quality (28/30)

• VQ-01: Text Legibility (7/8) - Font sizes explicitly set (title=24, axis_title=20, axis_text=16, geom_text=11). All text readable, though annotation text could be slightly larger for the canvas size.
• VQ-02: No Overlap (5/6) - Labels are well-positioned above peaks. The CH₃ label renders with minor visual artifacts on the subscript character in bold, but all text remains readable.
• VQ-03: Element Visibility (6/6) - Sharp, well-resolved peaks clearly visible. Line width of 1.2 is appropriate. Area fill with alpha=0.15 adds depth without obscuring detail.
• VQ-04: Color Accessibility (4/4) - Single blue color scheme, no colorblind issues. Good contrast against white background.
• VQ-05: Layout & Canvas (4/4) - Plot fills canvas well with appropriate margins (30, 40, 20, 20). Good proportions at 1600×900 base size.
• VQ-06: Axis Labels & Title (2/2) - "δ Chemical Shift (ppm)" and "Intensity (a.u.)" — descriptive with units.

Design Excellence (15/20)

• DE-01: Aesthetic Sophistication (6/8) - Strong design with Python Blue palette, subtle area fill creating depth, intentional gray hierarchy for text ([line-basic] seaborn implementation #222, [area-basic] plotnine implementation #333, feat(plotly): implement donut-labeled #555, #1a1a1a). Clearly above library defaults.
• DE-02: Visual Refinement (5/6) - Minimal theme with removed x-axis grid, subtle y-axis grid (#E0E0E0, size=0.3), removed minor grid, removed axis ticks, generous margins. Well-polished.
• DE-03: Data Storytelling (4/6) - Labeled peaks with molecular group assignments (CH₃, CH₂, OH, TMS) and splitting pattern names guide the viewer through ethanol's molecular structure. Visual hierarchy through peak heights communicates relative proton counts.

Spec Compliance (15/15)

• SC-01: Plot Type (5/5) - Correct NMR spectrum: line plot of intensity vs. chemical shift with sharp peaks.
• SC-02: Required Features (4/4) - All features present: reversed x-axis ✓, sharp/well-resolved peaks ✓, splitting patterns (singlet, triplet, quartet) ✓, peak labels with chemical shift values ✓, TMS reference at 0 ppm ✓, clean baseline with minimal noise ✓.
• SC-03: Data Mapping (3/3) - Chemical shift on x-axis, intensity on y-axis. Axes show full data range (-0.5 to 5.0 ppm).
• SC-04: Title & Legend (3/3) - Title format "spectrum-nmr · letsplot · pyplots.ai" correct. No legend needed for single-series spectrum.

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) - Demonstrates all key NMR features: singlet (OH, TMS), triplet (CH₃), quartet (CH₂). Shows realistic J-coupling splitting patterns with correct intensity ratios (1:2:1 and 1:3:3:1).
• DQ-02: Realistic Context (5/5) - Ethanol ¹H NMR — a classic, universally used chemistry example. Neutral scientific context.
• DQ-03: Appropriate Scale (4/4) - Chemical shifts realistic for ethanol: CH₃ at 1.18, CH₂ at 3.69, OH at 2.61 ppm. All within expected ranges for ¹H NMR.

Code Quality (10/10)

• CQ-01: KISS Structure (3/3) - Clean linear flow: imports → data generation → plot construction → save. No functions or classes.
• CQ-02: Reproducibility (2/2) - np.random.seed(42) set for noise generation.
• CQ-03: Clean Imports (2/2) - All imports used: numpy, pandas, lets_plot, export_ggsave.
• CQ-04: Code Elegance (2/2) - Elegant Lorentzian peak generation with loop over splitting patterns. Appropriate complexity for the domain.
• CQ-05: Output & API (1/1) - Saves as plot.png via export_ggsave with scale=3. Current API usage.

Library Mastery (7/10)

• LM-01: Idiomatic Usage (4/5) - Good grammar-of-graphics usage with proper layer composition (geom_area + geom_line + geom_text), scale_x_reverse, theme customization. Could leverage more advanced lets-plot features.
• LM-02: Distinctive Features (3/5) - Uses geom_area for filled spectrum (a nice lets-plot touch), scale_y_continuous expand parameter, and export_ggsave with scale parameter. These show library awareness beyond generic ggplot usage.

Score Caps Applied

• None

Strengths

• Excellent scientific accuracy with correct Lorentzian peak shapes and realistic J-coupling splitting patterns
• Clean, refined visual design with subtle area fill, minimal chrome, and intentional typography hierarchy
• Complete spec compliance — all required features present including reversed x-axis, labeled peaks, TMS reference, and splitting patterns
• Elegant data generation using parameterized Lorentzian functions

Weaknesses

• Annotation text size (11) could be slightly larger for optimal readability at full resolution
• Minor rendering artifacts on subscript characters in bold annotation labels

Issues Found

1. VQ-01 MINOR: Annotation text at geom_text size=11 is readable but on the smaller side for the 4800×2700 canvas
   • Fix: Increase geom_text size to 13-14 for better readability at full resolution
2. VQ-02 MINOR: CH₃ subscript character rendering in bold text shows slight artifacts
   • Fix: This may be a lets-plot font rendering limitation; consider using "CH3" plain text if subscripts cause issues

AI Feedback for Next Attempt

Strong implementation scoring at the approval threshold. The scientific accuracy and visual design are both excellent. Minor improvements could target annotation text sizing and font rendering of special characters, but overall this is a well-crafted NMR spectrum visualization.

Verdict: APPROVED