feat(altair): implement scatter-matrix

[github-actions]

Implementation: scatter-matrix - altair

Implements the altair version of scatter-matrix.

File: plots/scatter-matrix/implementations/altair.py

---

🤖 impl-generate workflow

[claude]

AI Review - Attempt 1/3

Image Description

The plot shows a 4x4 scatter matrix (SPLOM) for the Iris dataset with four variables: Sepal Length, Sepal Width, Petal Length, and Petal Width (all in cm). Each cell displays a scatter plot showing pairwise relationships between variables. Three species are color-coded: Setosa (yellow/gold), Versicolor (light blue), and Virginica (darker blue). The diagonal cells show scatter plots of each variable against itself (resulting in diagonal lines). The title "Iris Dataset · scatter-matrix · altair · pyplots.ai" appears at the top. A legend for Species is positioned on the right side. The overall layout is clean with subtle grid lines and properly labeled axes.

Quality Score: 80/100

Criteria Checklist

Visual Quality (33/40 pts)

• VQ-01: Text Legibility (8/10) - All text is readable but axis labels and tick marks are slightly small for the canvas size
• VQ-02: No Overlap (8/8) - No overlapping text elements
• VQ-03: Element Visibility (6/8) - Points are visible with good alpha, but marker size could be slightly larger for better visibility
• VQ-04: Color Accessibility (5/5) - Blue-Yellow palette is colorblind-safe
• VQ-05: Layout Balance (4/5) - Good layout but legend is somewhat far from the plots
• VQ-06: Axis Labels (2/2) - All axes have descriptive labels with units (cm)
• VQ-07: Grid & Legend (0/2) - Grid is subtle which is good, but legend labels are quite small

Spec Compliance (20/25 pts)

• SC-01: Plot Type (8/8) - Correct scatter matrix implementation
• SC-02: Data Mapping (5/5) - All variable pairs correctly mapped
• SC-03: Required Features (2/5) - Missing diagonal histograms/KDE as specified - diagonal shows scatter instead of distributions
• SC-04: Data Range (3/3) - All data visible within axes
• SC-05: Legend Accuracy (0/2) - Legend labels are correct but extremely small
• SC-06: Title Format (2/2) - Correct format: Iris Dataset · scatter-matrix · altair · pyplots.ai

Data Quality (18/20 pts)

• DQ-01: Feature Coverage (7/8) - Shows correlations between all variable pairs and species clustering well; slight deduction for not showing univariate distributions
• DQ-02: Realistic Context (7/7) - Iris dataset is a classic, realistic multivariate dataset
• DQ-03: Appropriate Scale (4/5) - Values are realistic for iris flowers, slight overlap in scales between species is realistic

Code Quality (9/10 pts)

• CQ-01: KISS Structure (3/3) - Simple imports → data → plot → save structure
• CQ-02: Reproducibility (3/3) - np.random.seed(42) is set
• CQ-03: Clean Imports (2/2) - Only necessary imports
• CQ-04: No Deprecated API (1/1) - Using current Altair API
• CQ-05: Output Correct (0/1) - Saves both plot.png and plot.html which is correct for Altair

Library Features (0/5 pts)

• LF-01: Uses distinctive library features (0/5) - Does not use Altair's built-in repeat/facet capabilities for diagonal histograms. Altair can layer multiple mark types but this was not utilized.

Strengths

• Clean colorblind-safe color palette using Python blue-based colors
• Proper use of Altair's repeat functionality for creating the matrix grid
• Good data generation with realistic iris-like measurements for three species
• Clear species separation visible in petal measurements showing natural clustering
• Appropriate use of tooltips for interactivity
• Correct title format following pyplots.ai conventions

Weaknesses

• Missing diagonal histograms or KDE plots as explicitly required by the specification - the diagonal currently shows scatter plots of variables against themselves instead of univariate distributions
• Legend text is too small compared to the canvas size - should be larger for better readability
• Could utilize Altair's layering capabilities to show different marks on diagonal vs off-diagonal cells

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 image displays a 4×4 scatter plot matrix of the Iris dataset with four variables: Sepal Length (cm), Sepal Width (cm), Petal Length (cm), and Petal Width (cm). Three species are color-coded: Setosa (dark blue #306998), Versicolor (yellow #FFD43B), and Virginica (light blue #4B8BBE). The title reads "Iris Dataset · scatter-matrix · altair · pyplots.ai" and is positioned at the top. A legend labeled "Species" appears in the top-right corner. Critical observation: The diagonal cells display scatter plots of each variable against itself (showing diagonal lines of points) rather than the required histograms or KDE distributions. The off-diagonal cells correctly show pairwise scatter plots with clear species clustering patterns. Font sizes are readable but relatively small.

Quality Score: 65/100

Criteria Checklist

Visual Quality (28/40 pts)

• VQ-01: Text Legibility (7/10) - Text is readable but on the smaller side, especially tick labels
• VQ-02: No Overlap (8/8) - No overlapping text elements
• VQ-03: Element Visibility (6/8) - Markers visible with good alpha, but size=80 is slightly small for 150 points
• VQ-04: Color Accessibility (3/5) - Two blues (dark and light) may be hard to distinguish for some viewers
• VQ-05: Layout Balance (2/5) - Square format appropriate for matrix, but cells feel cramped
• VQ-06: Axis Labels (2/2) - Descriptive labels with units (cm)
• VQ-07: Grid & Legend (0/2) - Grid opacity good at 0.3, but legend is small and distant from plot

Spec Compliance (13/25 pts)

• SC-01: Plot Type (4/8) - MAJOR ISSUE: Diagonal shows scatter plots instead of histograms/KDE as required by spec
• SC-02: Data Mapping (5/5) - X/Y correctly assigned in scatter plots
• SC-03: Required Features (0/5) - Missing diagonal histograms/KDE distributions (spec requirement)
• SC-04: Data Range (3/3) - All data visible
• SC-05: Legend Accuracy (1/2) - Legend present but font size could be larger

Data Quality (17/20 pts)

• DQ-01: Feature Coverage (7/8) - Shows correlations, clusters, and species separation well
• DQ-02: Realistic Context (7/7) - Classic Iris dataset with realistic parameter distributions
• DQ-03: Appropriate Scale (3/5) - Values realistic for iris measurements, though some outliers appear

Code Quality (9/10 pts)

• CQ-01: KISS Structure (3/3) - Linear script structure, no functions/classes
• CQ-02: Reproducibility (3/3) - np.random.seed(42) set
• CQ-03: Clean Imports (2/2) - Only necessary imports
• CQ-04: No Deprecated API (1/1) - Current Altair API used
• CQ-05: Output Correct (0/1) - Title format doesn't match required "{spec-id} · {library} · pyplots.ai"

Library Features (3/5 pts)

• LF-01: Uses Altair's declarative grammar with hconcat/vconcat composition

Strengths

• Clean declarative composition using Altair's hconcat/vconcat for building the matrix
• Good use of colorblind-friendly base color (Python blue palette)
• Proper encoding types (:Q, :N) and tooltips for interactivity
• Reproducible with fixed random seed
• Appropriate transparency (alpha=0.7) to handle overplotting
• Both PNG and HTML outputs generated

Weaknesses

• CRITICAL: Diagonal cells show scatter plots instead of required histograms/KDE (code attempts histogram but rendering shows scatter)
• Title uses "Iris Dataset" instead of required format "{spec-id} · {library} · pyplots.ai" (should be "scatter-matrix · altair · pyplots.ai")
• Two shades of blue (Setosa and Virginica) reduce distinguishability - consider more distinct colors
• Legend font sizes (24/22pt) may be too small relative to the large output image

Verdict: REJECTED

Primary rejection reason: The specification explicitly requires "Diagonal cells should show univariate distributions (histograms or KDE) for each variable" but the rendered output shows scatter plots on the diagonal. The code appears to attempt histograms (lines 74-93) but the visual result shows each variable plotted against itself as a scatter, creating diagonal lines of points rather than distribution bars.

[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 4×4 scatter plot matrix (SPLOM) visualizing the Iris dataset with four variables: Sepal Length, Sepal Width, Petal Length, and Petal Width (all in cm). The diagonal cells contain layered histograms showing the distribution of each variable, colored by species. Off-diagonal cells show scatter plots for each pairwise variable combination. Three species are color-coded: Setosa (blue), Versicolor (orange/yellow), and Virginica (teal/green). The title "scatter-matrix · altair · pyplots.ai" appears at the top. A species legend is positioned on the right side. The overall layout is clean with a grid structure, subtle gridlines, and good use of transparency in both histograms and scatter points.

Quality Score: 82/100

Criteria Checklist

Visual Quality (32/40 pts)

• VQ-01: Text Legibility (8/10) - Text is readable but axis labels are somewhat small on the high-resolution output
• VQ-02: No Overlap (8/8) - No overlapping text elements
• VQ-03: Element Visibility (6/8) - Markers visible with good transparency (0.7), but could be slightly larger for this data density (150 points)
• VQ-04: Color Accessibility (5/5) - Colorblind-safe palette (blue, orange, green) with good differentiation
• VQ-05: Layout Balance (3/5) - Matrix fills canvas reasonably but the legend appears somewhat disconnected
• VQ-06: Axis Labels (2/2) - Descriptive labels with units: "Sepal Length (cm)", etc.
• VQ-07: Grid & Legend (0/2) - Legend is small and isolated; positioned far from matrix

Spec Compliance (23/25 pts)

• SC-01: Plot Type (8/8) - Correct scatter matrix with histograms on diagonal
• SC-02: Data Mapping (5/5) - All pairwise relationships correctly shown
• SC-03: Required Features (5/5) - Histograms on diagonal, scatter off-diagonal, color encoding by category
• SC-04: Data Range (3/3) - All data visible within axes
• SC-05: Legend Accuracy (0/2) - Legend shows species but is poorly integrated and small
• SC-06: Title Format (2/2) - Correctly uses "scatter-matrix · altair · pyplots.ai"

Data Quality (18/20 pts)

• DQ-01: Feature Coverage (7/8) - Shows correlations, distributions, and species separation well; Setosa clearly separable
• DQ-02: Realistic Context (7/7) - Classic Iris dataset with realistic values
• DQ-03: Appropriate Scale (4/5) - Values realistic for iris flowers, though some generated values slightly out of typical range

Code Quality (9/10 pts)

• CQ-01: KISS Structure (2/3) - Generally simple but uses loops and manual layer building which adds complexity
• CQ-02: Reproducibility (3/3) - np.random.seed(42) is set
• CQ-03: Clean Imports (2/2) - Only necessary imports
• CQ-04: No Deprecated API (1/1) - Uses current Altair API
• CQ-05: Output Correct (1/1) - Saves as plot.png and plot.html

Library Features (0/5 pts)

• LF-01: Uses distinctive library features (0/5) - Does NOT use Altair's built-in repeat/facet for scatter matrices; manual loop construction misses Altair's declarative strengths

Strengths

• Correct implementation of scatter matrix with histograms on diagonal and scatter plots off-diagonal
• Excellent colorblind-safe color palette (blue, orange, green)
• Good use of transparency (opacity=0.6 for histograms, 0.7 for scatter) to handle overlapping
• Appropriate use of the classic Iris dataset as realistic multivariate data
• Proper title formatting following the spec-id · library · pyplots.ai convention
• Interactive HTML output alongside PNG
• Tooltips enabled for interactivity

Weaknesses

• Legend is too small and disconnected from the main matrix - should be larger and better integrated
• Does not leverage Altair's native repeat() or facet() functionality for creating scatter matrices, which would be more idiomatic
• Font sizes could be slightly larger for better readability at full resolution
• The manual loop-based construction is less elegant than Altair's declarative approach

Verdict: APPROVED

This is attempt 3/3. The implementation correctly creates a scatter plot matrix with all required features: histograms on diagonal, scatter plots off-diagonal, color encoding by species, and proper labeling. While it doesn't fully exploit Altair's declarative repeat/facet capabilities, the result is visually correct and functional. The score of 82 exceeds the 50-point threshold for final attempt acceptance.

[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 showing pairwise relationships between four Iris flower measurements: Sepal Length (cm), Sepal Width (cm), Petal Length (cm), and Petal Width (cm). Three species are color-coded using a colorblind-safe palette: Setosa (blue #306998), Versicolor (orange #E69F00), and Virginica (green #009E73). The diagonal cells show the same variable on both axes, resulting in linear patterns (x=y). Off-diagonal cells display scatter plots revealing correlations between variable pairs, with clear species clustering visible. The title "scatter-matrix · altair · pyplots.ai" appears at the top center. A legend on the right identifies the three species with appropriately sized symbols.

Quality Score: 85/100

Criteria Checklist

Visual Quality (34/40 pts)

• VQ-01: Text Legibility (9/10) - Title and axis labels are clear and readable; axis tick labels slightly small but acceptable
• VQ-02: No Overlap (8/8) - No overlapping text; all labels and titles well-spaced
• VQ-03: Element Visibility (7/8) - Markers well-sized with good opacity (0.7); slight overlap in dense areas but species distinguishable
• VQ-04: Color Accessibility (5/5) - Excellent colorblind-safe palette (blue/orange/green)
• VQ-05: Layout Balance (3/5) - Good use of canvas but legend placement creates slight imbalance; matrix fills space adequately
• VQ-06: Axis Labels (2/2) - Descriptive labels with units: "Sepal Length (cm)", etc.
• VQ-07: Grid & Legend (0/2) - Grid opacity is good (0.3), but diagonal cells show meaningless x=y lines instead of histograms/KDE as spec requires

Spec Compliance (19/25 pts)

• SC-01: Plot Type (8/8) - Correct scatter plot matrix type
• SC-02: Data Mapping (5/5) - All four variables correctly mapped to rows/columns
• SC-03: Required Features (2/5) - Missing histograms/KDE on diagonal as explicitly required by spec; shows scatter instead
• SC-04: Data Range (3/3) - All data visible within axis ranges
• SC-05: Legend Accuracy (1/2) - Species labels correct but could be more prominent
• SC-06: Title Format (2/2) - Correct format: "scatter-matrix · altair · pyplots.ai"

Data Quality (19/20 pts)

• DQ-01: Feature Coverage (7/8) - Shows clear species separation, correlations visible; could benefit from showing outliers
• DQ-02: Realistic Context (7/7) - Classic Iris dataset with realistic measurements and distributions
• DQ-03: Appropriate Scale (5/5) - Measurement values realistic for iris flowers

Code Quality (10/10 pts)

• CQ-01: KISS Structure (3/3) - Clean script structure without functions/classes
• CQ-02: Reproducibility (3/3) - np.random.seed(42) set correctly
• CQ-03: Clean Imports (2/2) - Only necessary imports (altair, numpy, pandas)
• CQ-04: No Deprecated API (1/1) - Uses current Altair 6.0 API
• CQ-05: Output Correct (1/1) - Saves plot.png and plot.html

Library Features (3/5 pts)

• LF-01: Uses distinctive library features (3/5) - Uses Altair's native repeat() for declarative SPLOM construction, tooltips enabled; could use layering for diagonal histograms

Strengths

• Excellent use of Altair's declarative repeat() pattern for idiomatic SPLOM construction
• Colorblind-safe palette (blue/orange/green) provides clear species distinction
• Clean data generation with realistic Iris-like measurements and proper random seed
• All axis labels include units (cm) for clarity
• Interactive tooltips enhance usability in HTML version

Weaknesses

• Diagonal cells show scatter (x=y lines) instead of required histograms/KDE for univariate distributions as specified
• Legend could be larger or more prominently positioned for better visual balance

Verdict: APPROVED

This is attempt 3/3. While the implementation misses the diagonal histograms/KDE requirement from the specification, the overall quality score of 85/100 exceeds the 50-point threshold for final merge. The scatter matrix effectively demonstrates Altair's declarative approach and provides a visually clear, colorblind-accessible visualization of the Iris dataset relationships.

[github-actions]

AI Review - Final Status

Score: 85/100 (Acceptable)

After 3 repair attempts, altair reached 85/100.

Score ≥ 50 is acceptable for the repository. Merging.

---

🤖 impl-review