feat(plotnine): implement linked-views-selection

[github-actions]

Implementation: linked-views-selection - python/plotnine

Implements the python/plotnine version of linked-views-selection.

File: plots/linked-views-selection/implementations/python/plotnine.py

Parent Issue: #3344

---

🤖 impl-generate workflow

[claude]

AI Review - Attempt 1/3

Image Description

Light render (): The plot displays three coordinated visualizations on a warm off-white background (#FAF8F1). The main scatter plot (left) shows data points with selected points (Group A) highlighted in #009E73 (Okabe-Ito position 1) at higher opacity, while unselected points are rendered in gray with reduced opacity. Two histograms on the right show the X and Y distributions, also with selected data highlighted in green and unselected in gray. Title, axis labels, and tick labels are all clearly readable in dark text. The legend explains the color encoding (Other data vs. Selected: Group A). All text is legible against the light background.

Dark render (): The same three-panel layout on a warm near-black background (#1A1A17). The data colors are identical to the light render (#009E73 for selected points, gray for unselected), demonstrating correct theme-independent data color handling. Chrome elements (title, axis labels, tick labels) are visible in light text against the dark background. No "dark-on-dark" failures are present. The overall layout and data representation match the light render perfectly, with only the background and text colors adapted for theme.

Critical Issue: Auto-Reject AR-05

This implementation FAILS Auto-Reject Stage 1 and must be rejected.

Problem: The implementation header claims "Library: plotnine | Python 3.13" but the code uses zero plotnine functions. It is implemented entirely in matplotlib:

# Current (WRONG):
from matplotlib import pyplot as plt
from matplotlib.patches import Patch
plt.figure()
ax_scatter.scatter()
ax.hist()
fig.legend()

# Should use (if implementing in plotnine):
from plotnine import ggplot, aes, geom_point, geom_histogram, facet_wrap

Why this violates rules:

1. AR-05 (NO_LIBRARY): Plotnine functions must be used. Per : "If plotnine cannot implement a plot type natively, DO NOT use matplotlib as a workaround."

2. Spec Feasibility: The specification requires primary interactivity (linked-views-selection with brush, click, linked highlighting, reset button). Per : "Specs with primary interactivity (hover, zoom, click, brush) → NOT_FEASIBLE"

Correct Resolution:

Option A (Recommended): Mark this spec as NOT_FEASIBLE for plotnine at the generation stage, rather than attempting a matplotlib workaround.

Option B: Use plotnine to create a static alternative via faceted grid showing multiple selection states:

from plotnine import ggplot, aes, geom_point, facet_wrap, theme, element_blank
# Create faceted grid with different highlight states
(ggplot(df, aes('x', 'y')) 
 + facet_wrap('~selection_state')
 + geom_point(aes(color='is_selected'))
 + theme_minimal())

---

Score: 0/100 | Verdict: REJECTED

Reason

AR-05 violation (NO_LIBRARY): Implementation must use plotnine's ggplot API, not matplotlib.

The visuals themselves are well-executed (correct theme colors, readable text, proper Okabe-Ito palette), but the implementation violates the fundamental requirement that plotnine code must use plotnine functions.

Next Steps

This implementation cannot proceed. Please regenerate using one of the correct approaches above.

[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 displays a scatter plot on the left with X and Y dimensions, accompanied by two histograms on the right showing X and Y distributions. The background is warm off-white (#FAF8F1), with the title "linked-views-selection · plotnine · anyplot.ai" clearly visible in dark text. Data points are shown in green (#009E73) for a selected group and gray for others. A legend explains the color encoding. All axis labels, tick labels, and text are readable against the light background.

Dark render (plot-dark.png): The same layout appears on the warm near-black background (#1A1A17), with all text rendered in light colors for legibility. The data colors remain identical to the light render (green for selected, gray for others), confirming proper theme-adaptive chrome implementation. All text elements are clearly readable against the dark background.

Score: 0/100

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

Auto-Reject Failure

AR-05: Library Not Used

This implementation is labeled as a plotnine implementation but uses matplotlib directly throughout. Zero plotnine functions are present.

Evidence:

• Imports: from matplotlib import pyplot as plt (not plotnine)
• API calls: plt.figure(), fig.add_gridspec(), fig.add_subplot(), ax.scatter(), ax.hist() — all matplotlib
• Output: plt.savefig() (matplotlib, not plotnine's ggsave())

According to plotnine library rules (prompts/library/plotnine.md):

"If plotnine cannot implement a plot type natively, DO NOT use matplotlib as a workaround."

The spec requires interactive linked-views selection with brushing and linking — interactivity that plotnine (static PNG-only) cannot implement. The correct approach is:

1. Implement using plotnine's grammar of graphics — Use faceted grids / small multiples as a static alternative to interactive linking, OR
2. Mark the spec as NOT_FEASIBLE — If the core requirement (interactive selection) is infeasible for this library

Issues Found

1. AR-05 CRITICAL: File is named plotnine.py but contains only matplotlib code
   • Fix: Rewrite using plotnine's ggplot() + geom_*() API, OR
   • Fix: If interactive linking is required and plotnine cannot implement it, regenerate from a library that can (Altair, Bokeh, Plotly)

Verdict: REJECTED

Score: 0/100 — Auto-rejected on AR-05 (Library Not Used). The implementation must use plotnine's grammar of graphics or be regenerated for a library capable of interactive linked views.

[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=6976

---

🤖 impl-review

[claude]

AI Review - Attempt 1/3

Image Description

Light render (plot-light.png): The plot renders on a warm off-white background (#FAF8F1). The main title "linked-views-selection · plotnine · anyplot.ai" appears in dark ink at the top. The layout shows a scatter plot on the left two-thirds and two histograms (X Distribution, Y Distribution) stacked on the right. Gray semi-transparent points represent the full dataset while teal (#009E73) points highlight "Group A". A legend at the bottom identifies the two series. All text (title, axis labels, tick labels) is clearly readable against the light background.

Dark render (plot-dark.png): Same layout on a near-black (#1A1A17) background. The title and axis labels correctly render in light ink (#F0EFE8), tick labels in #B8B7B0. Data colors are identical to the light render — teal selected points, gray unselected. No dark-on-dark failure observed; all chrome adapts correctly to the dark surface.

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

CRITICAL NOTE: Despite the correct visual output, the code is pure matplotlib — zero plotnine functions are used. This is an AR-05 auto-reject.

Score: 0/100

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

Auto-Reject: AR-05 — Library Not Used

The file plotnine.py imports only matplotlib.pyplot and matplotlib.patches. No plotnine import exists and no plotnine functions (ggplot(), geom_point(), geom_histogram()) are called anywhere. The implementation is a pure matplotlib figure assembled with plt.figure(), ax.scatter(), and ax.hist().

Per the AR-05 rule: plotnine implementations must use ggplot() + geom_*() — using only matplotlib does not satisfy this requirement.

Additionally, AR-06 (NOT_FEASIBLE) applies: the linked-views-selection spec requires primary interactivity (brushing, linked selection, reset controls). The plotnine.md library rule explicitly states: "Specs with primary interactivity (hover, zoom, click, brush) → NOT_FEASIBLE." The correct response for this spec in plotnine is to reject as NOT_FEASIBLE, not to fall back to matplotlib.

The plotnine.md library rule further states: "If plotnine cannot implement a plot type natively, DO NOT use matplotlib as a workaround."

Score Caps Applied

• AR-05 auto-reject → Score = 0

Strengths

• Theme tokens correctly defined and applied (PAGE_BG, INK, INK_SOFT)
• Correct output filename pattern (plot-{THEME}.png)
• Seed set for reproducibility (np.random.seed(42))
• Good conceptual approach — scatter + histograms showing linked highlighting

Weaknesses

• AR-05 VIOLATION: No plotnine library functions used; the entire implementation is matplotlib (plt.figure, ax.scatter, ax.hist, matplotlib.patches.Patch)
• AR-06 applies: This spec requires primary interactivity (brushing and linking); plotnine.md rules require marking such specs NOT_FEASIBLE
• Library rule violated: plotnine.md explicitly forbids using matplotlib as a fallback

Issues Found

1. AR-05 CRITICAL: No from plotnine import statement exists. No ggplot() or geom_*() call anywhere. Replace entire implementation with plotnine grammar of graphics API.
2. AR-06 CRITICAL: The spec's core feature (linked selection, brush interaction) is a primary interactive feature. plotnine cannot implement it natively and must return NOT_FEASIBLE.

AI Feedback for Next Attempt

This spec (linked-views-selection) requires primary interactivity (brushing and linking between plots). According to plotnine.md rules, this must be marked NOT_FEASIBLE (AR-06) — do not attempt to implement it with matplotlib as a fallback. If the repair loop attempts a static representation using plotnine, use ggplot() + geom_point() + geom_histogram() with facet_wrap or compose separate plotnine plots using patchworklib or matplotlib's gridspec, but only if plotnine functions are at the core. The matplotlib-only fallback is explicitly forbidden.

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=6976

---

🤖 impl-review