feat(plotnine): implement scatter-matrix-interactive

[github-actions]

Implementation: scatter-matrix-interactive - plotnine

Implements the plotnine version of scatter-matrix-interactive.

File: plots/scatter-matrix-interactive/implementations/plotnine.py

Parent Issue: #3604

---

🤖 impl-generate workflow

[claude]

AI Review - Attempt 1/3

Image Description

The plot displays a 4x4 scatter plot matrix (SPLOM) showing pairwise relationships between four Iris dataset variables: Sepal Length, Sepal Width, Petal Length, and Petal Width. The diagonal cells are empty (no univariate distributions shown). Each off-diagonal cell contains a scatter plot with points colored by species: blue (#306998) for setosa, yellow (#FFD43B) for versicolor, and red/coral (#E74C3C) for virginica. The title reads "scatter-matrix-interactive · plotnine · pyplots.ai" with a subtitle explaining it's static and interactive brushing requires Plotly/Bokeh/Altair. Variable names appear as facet strip labels on top (x-axis variables) and right side (y-axis variables). Points use alpha=0.7 for transparency. Legend is positioned at the bottom with species labels. Background panels are light gray (#f8f8f8) with no grid lines.

Quality Score: 72/100

Criteria Checklist

Visual Quality (30/40)

• VQ-01: Text Legibility (7/10) - Title and labels are readable but tick labels are somewhat small
• VQ-02: No Overlap (8/8) - No overlapping text elements
• VQ-03: Element Visibility (6/8) - Markers visible but size=3 is small for 150 points; could be larger
• VQ-04: Color Accessibility (3/5) - Yellow and red/coral could be difficult for some colorblind users; not ideal palette
• VQ-05: Layout Balance (4/5) - Good use of canvas, though diagonal gaps create some empty space
• VQ-06: Axis Labels (0/2) - No axis labels shown (just facet strip text, no units)
• VQ-07: Grid & Legend (2/2) - Grid is hidden, legend well placed at bottom

Spec Compliance (17/25)

• SC-01: Plot Type (8/8) - Correct scatter plot matrix type
• SC-02: Data Mapping (5/5) - X/Y correctly assigned for each facet
• SC-03: Required Features (0/5) - Missing diagonal univariate distributions (histograms/KDE), no interactive brushing (acknowledged in subtitle)
• SC-04: Data Range (3/3) - All data visible within axes
• SC-05: Legend Accuracy (1/2) - Legend present but could be more prominent
• SC-06: Title Format (0/2) - Uses correct format with spec-id · library · pyplots.ai but subtitle is overly apologetic

Data Quality (17/20)

• DQ-01: Feature Coverage (6/8) - Shows relationships well, but missing diagonal distributions limits feature demonstration
• DQ-02: Realistic Context (7/7) - Iris dataset is classic, appropriate scientific context
• DQ-03: Appropriate Scale (4/5) - Values are realistic for iris measurements

Code Quality (8/10)

• CQ-01: KISS Structure (3/3) - Clean imports → data → plot → save structure
• CQ-02: Reproducibility (3/3) - np.random.seed(42) present
• CQ-03: Clean Imports (1/2) - element_blank imported but could be avoided
• CQ-04: No Deprecated API (1/1) - Modern plotnine API used
• CQ-05: Output Correct (0/1) - Saves as plot.png correctly

Library Features (0/5)

• LF-01: Uses distinctive library features (0/5) - Does not use plotnine's geom_histogram or geom_density for diagonals; facet_grid is standard feature

Strengths

• Clean, professional layout with proper faceting using plotnine's grammar of graphics
• Appropriate use of Iris dataset as a classic multivariate example
• Honest subtitle acknowledging plotnine's limitations for interactive features
• Good color scheme with Python Blue (#306998) as primary color
• Proper categorical handling for facet ordering

Weaknesses

• Missing diagonal univariate distributions (histograms or KDE) - this is a key spec requirement
• No axis labels with units (e.g., "Length (cm)")
• Marker size (3) is too small for 150 data points - should be larger for better visibility
• Yellow/red color combination could be improved for colorblind accessibility

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 displays a 4×4 scatter plot matrix (SPLOM) of the Iris dataset using plotnine. The matrix shows pairwise relationships between Sepal Length, Sepal Width, Petal Length, and Petal Width. Each row and column corresponds to one variable, with column headers at the top and row labels on the right side.

Colors: Three species are color-coded using a colorblind-safe palette: setosa (blue/teal), versicolor (yellow/gold), and virginica (red/coral). The legend is positioned at the bottom with clear labels.

Layout: The diagonal cells appear to be empty (gray background only) - no histograms or distributions are shown. Off-diagonal cells contain scatter plots with appropriately sized markers (alpha=0.7). The grid has light gray panel backgrounds.

Missing Features: The specification requires diagonal cells to show univariate distributions (histograms or KDE), but all diagonal cells are blank. The subtitle text mentions "Iris Dataset Pairwise Scatter Plots with Univariate Distributions (Static)" but no distributions are actually visible. The subtitle also correctly notes that interactive brushing requires Plotly/Bokeh/Altair.

Quality Score: 68/100

Criteria Checklist

Visual Quality (30/40)

• VQ-01: Text Legibility (8/10) - Title and labels readable, but axis text slightly small at full size
• VQ-02: No Overlap (8/8) - No overlapping text elements
• VQ-03: Element Visibility (6/8) - Markers visible with good alpha, but sizing could be optimized
• VQ-04: Color Accessibility (5/5) - Excellent colorblind-safe palette (Dark2-inspired)
• VQ-05: Layout Balance (3/5) - Good use of canvas, but diagonal empty cells waste space
• VQ-06: Axis Labels (0/2) - Column headers lack units (show "Sepal Length" not "Sepal Length (cm)")
• VQ-07: Grid & Legend (0/2) - Grid is completely removed (no panel_grid), legend well placed

Spec Compliance (15/25)

• SC-01: Plot Type (8/8) - Correct scatter matrix structure
• SC-02: Data Mapping (5/5) - X/Y correctly assigned in pairwise combinations
• SC-03: Required Features (0/5) - CRITICAL: Diagonal must show histograms/KDE per spec - they are blank
• SC-04: Data Range (0/3) - Data shown correctly, but diagonal ranges not applicable
• SC-05: Legend Accuracy (2/2) - Legend labels correct with species names
• SC-06: Title Format (0/2) - Subtitle changed from spec-required format

Data Quality (16/20)

• DQ-01: Feature Coverage (5/8) - Shows species clustering but diagonal distributions missing
• DQ-02: Realistic Context (7/7) - Iris dataset is a classic, appropriate choice
• DQ-03: Appropriate Scale (4/5) - Real data with appropriate cm scale values

Code Quality (7/10)

• CQ-01: KISS Structure (2/3) - Good structure but uses warnings suppression
• CQ-02: Reproducibility (3/3) - np.random.seed(42) set
• CQ-03: Clean Imports (1/2) - Some unused imports (element_blank imported but grid is just not shown)
• CQ-04: No Deprecated API (1/1) - Current plotnine API
• CQ-05: Output Correct (0/1) - Saves as plot.png correctly

Library Features (0/5)

• LF-01: Uses distinctive library features (0/5) - Faceting is basic; diagonal histograms would showcase layer composition but they do not render

Strengths

• Excellent colorblind-safe color palette (Dark2-inspired teal, orange, purple)
• Clean scatter plot matrix structure with proper facet_grid implementation
• Good marker sizing and alpha transparency for data visibility
• Appropriate subtitle noting plotnine's static nature (interactive brushing not supported)
• Uses real Iris dataset providing meaningful, interpretable results

Weaknesses

• CRITICAL: Diagonal cells are completely empty - spec requires histograms or KDE distributions
• The code includes histogram data preparation (hist_df) and geom_histogram layer, but diagonals render blank
• Panel grid completely removed instead of being subtle (alpha 0.2-0.4)
• Axis labels in facet strips lack units (should show "Sepal Length (cm)" not just variable name)
• Subtitle format differs from standard spec-id format requirement

Verdict: REJECTED

Primary Issue: The diagonal cells must show univariate distributions (histograms or KDE) as explicitly required by the specification. Currently they are blank, which is a critical spec compliance failure. The implementation appears to attempt this with separate histogram data and geom_histogram, but the layer composition with facet_grid is not rendering the histograms on the diagonal panels.

[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

The plot displays a 4x4 scatter plot matrix (SPLOM) of the Iris dataset showing pairwise relationships between Sepal Length, Sepal Width, Petal Length, and Petal Width. Three species are color-coded: setosa (blue), versicolor (yellow/gold), and virginica (coral/red). The title correctly shows "scatter-matrix-interactive · plotnine · pyplots.ai" with a subtitle noting interactive brushing requires Plotly/Bokeh/Altair. The diagonal cells appear empty (histograms are not visible), while off-diagonal cells show scatter plots with good separation between species clusters. The legend is positioned at the bottom. Strip labels appear on top (column) and right (row) sides.

Quality Score: 68/100

Criteria Checklist

Visual Quality (30/40)

• VQ-01: Text Legibility (8/10) - Title and labels readable, but tick labels slightly small
• VQ-02: No Overlap (8/8) - No overlapping text elements
• VQ-03: Element Visibility (6/8) - Markers visible but diagonal histograms missing/not visible
• VQ-04: Color Accessibility (5/5) - Colorblind-safe palette (teal, gold, coral)
• VQ-05: Layout Balance (1/5) - Diagonal cells appear empty, wasting significant space
• VQ-06: Axis Labels (0/2) - No units on axis labels, labels only in strip text
• VQ-07: Grid & Legend (2/2) - Subtle grid, legend well placed at bottom

Spec Compliance (20/25)

• SC-01: Plot Type (8/8) - Correct scatter matrix type
• SC-02: Data Mapping (5/5) - X/Y correctly assigned per cell
• SC-03: Required Features (2/5) - Missing diagonal histograms (spec requires univariate distributions)
• SC-04: Data Range (3/3) - All data visible
• SC-05: Legend Accuracy (2/2) - Species labels correct
• SC-06: Title Format (0/2) - Subtitle deviates from standard (adds parenthetical note)

Data Quality (18/20)

• DQ-01: Feature Coverage (6/8) - Shows species separation well, but no univariate view
• DQ-02: Realistic Context (7/7) - Iris dataset is classic, appropriate
• DQ-03: Appropriate Scale (5/5) - Iris measurements in cm are realistic

Code Quality (7/10)

• CQ-01: KISS Structure (2/3) - Mostly linear but has iteration loops for data prep
• CQ-02: Reproducibility (3/3) - np.random.seed(42) set
• CQ-03: Clean Imports (1/2) - sklearn imported for dataset loading only (acceptable)
• CQ-04: No Deprecated API (1/1) - Using current plotnine API
• CQ-05: Output Correct (0/1) - Saves correctly but histogram data not rendering

Library Features (3/5)

• LF-01: Uses distinctive library features (3/5) - Uses facet_grid, grammar of graphics, but histogram overlay fails

Strengths

• Colorblind-safe palette (Dark2 inspired) with good contrast
• Proper use of facet_grid for scatter matrix layout
• Clear species separation visible in scatter plots
• Appropriate subtitle noting plotnine's static nature limitations
• Good data choice using classic Iris dataset

Weaknesses

• Diagonal histograms not rendering - cells appear empty (critical for spec compliance)
• Axis labels lack units (should be "cm" as noted in var_labels but not displayed on axes)
• Subtitle format deviates from standard title format requirement

Verdict: APPROVED

Note: This is attempt 3/3 with score 68/100 (≥50 threshold). The implementation demonstrates plotnine's grammar of graphics approach well, with the main limitation being the diagonal histogram rendering issue. The static nature of plotnine is appropriately documented in the subtitle.

[github-actions]

🔧 Repair Attempt 3/3

Applied fixes based on AI review feedback.

Status: Repair completed, re-triggering review...

---

🤖 impl-repair

[claude]

AI Review - Attempt 3/3

Image Description

The plot displays a 4×4 scatter plot matrix (SPLOM) using the Iris dataset with four variables: Sepal Length (cm), Sepal Width (cm), Petal Length (cm), and Petal Width (cm). The matrix uses a colorblind-safe Dark2-inspired palette with teal (#1B9E77) for setosa, orange (#D95F02) for versicolor, and purple (#7570B3) for virginica. Off-diagonal cells show scatter plots with points colored by species at alpha=0.7. Diagonal cells display histogram-based density distributions as filled ribbon areas with alpha=0.5, showing the univariate distribution of each variable per species. The title "scatter-matrix-interactive · plotnine · pyplots.ai" appears at top-left in bold. Strip labels on top show variable names for x-axis, and strip labels on right show variable names for y-axis. A legend at the bottom indicates species color mapping. The grid is subtle with light gray lines. The layout is clean with minimal spacing between panels.

Quality Score: 72/100

Criteria Checklist

Visual Quality (32/40 pts)

• VQ-01: Text Legibility (8/10) - Title and labels readable, strip text slightly small but acceptable
• VQ-02: No Overlap (8/8) - No overlapping text elements
• VQ-03: Element Visibility (6/8) - Points visible with good alpha, density ribbons clear but could use more contrast
• VQ-04: Color Accessibility (5/5) - Colorblind-safe palette (Dark2 teal/orange/purple)
• VQ-05: Layout Balance (3/5) - Good 4×4 grid but some wasted space around edges
• VQ-06: Axis Labels (0/2) - Axis labels are empty (shown in strip text instead)
• VQ-07: Grid & Legend (2/2) - Subtle grid, legend well-placed at bottom

Spec Compliance (17/25 pts)

• SC-01: Plot Type (8/8) - Correct scatter matrix with density diagonals
• SC-02: Data Mapping (5/5) - All 4 variables correctly mapped to pairwise scatter
• SC-03: Required Features (0/5) - Missing interactive features: No brushing/linked selection, no zoom/pan, no box/lasso selection (plotnine is static - spec notes limitations but no workaround provided)
• SC-04: Data Range (3/3) - All data visible within axes
• SC-05: Legend Accuracy (1/2) - Species legend correct but no interactive indicator
• SC-06: Title Format (2/2) - Correct format: scatter-matrix-interactive · plotnine · pyplots.ai

Data Quality (18/20 pts)

• DQ-01: Feature Coverage (7/8) - Shows correlations, clusters separation visible, density distributions on diagonal
• DQ-02: Realistic Context (7/7) - Classic Iris dataset - real botanical data
• DQ-03: Appropriate Scale (5/5) - Real measurements in centimeters

Code Quality (10/10 pts)

• CQ-01: KISS Structure (3/3) - Linear flow: imports → data → plot → save
• CQ-02: Reproducibility (3/3) - np.random.seed(42) set
• CQ-03: Clean Imports (2/2) - All imports used
• CQ-04: No Deprecated API (1/1) - Current plotnine syntax
• CQ-05: Output Correct (1/1) - Saves as plot.png

Library Features (5/5 pts)

• LF-01: Uses distinctive library features (5/5) - Excellent use of ggplot2 grammar: facet_grid, geom_ribbon for density, aes mapping, scale_color_manual, theme customization

Strengths

• Excellent use of plotnine's grammar of graphics with facet_grid for the matrix layout
• Creative solution using geom_ribbon for histogram-based density distributions on diagonal
• Colorblind-safe palette (Dark2-inspired teal, orange, purple)
• Clean data transformation from wide to long format for faceting
• Good visual separation of the three Iris species clusters
• Proper title format following specification

Weaknesses

• Missing interactive features (brushing, linked selection, zoom/pan) - plotnine is a static library and cannot provide these natively; spec notes this limitation but implementation doesn't acknowledge it in comments
• Axis labels are empty strings (x="" and y="") - variable names only shown in strip text
• Diagonal density plots y-axis scale doesn't match the variable's natural scale (normalized density mapped to variable range)

Verdict: APPROVED

Rationale for Attempt 3/3: This is the final attempt. The score of 72/100 exceeds the minimum threshold of 50 required for merge after 3 attempts. While the implementation cannot provide the interactive features specified (as plotnine is fundamentally a static plotting library), it provides an excellent static scatter matrix with creative use of plotnine's grammar of graphics. The spec explicitly notes that "Libraries without native linked selection (matplotlib, seaborn, plotnine) may need alternative approaches or should note limitations" - this implementation demonstrates the best possible static SPLOM that plotnine can produce.