feat: use calibrated per-rule scores in final percentage calculation

[let-sunny]

Summary

• Replace flat severity weights with calculatedScore from rule engine in density calculation
• Per-rule scores and depthWeight from rule-config.ts now actually influence user-facing scores
• Calibration loop score adjustments flow through to final percentages

What was wrong

calculateScores() used SEVERITY_DENSITY_WEIGHT (blocking=3.0, risk=2.0, missing-info=1.0, suggestion=0.5) for density. This meant:

• All blocking rules contributed 3.0 equally regardless of their calibrated score
• no-auto-layout (score: -10) and missing-size-constraint (score: -3) were treated the same
• depthWeight was computed into calculatedScore but never consumed by scoring
• Adjusting scores in rule-config.ts via calibration loop had zero effect on user-facing scores

What this fixes

Now calculateScores() uses Math.abs(issue.calculatedScore) instead of flat severity weights:

Rule	Before (severity weight)	After (calculatedScore)
no-auto-layout at root (score -10, depthWeight 1.5)	3.0	15
no-auto-layout at leaf (score -10, depthWeight 1.0)	3.0	10
missing-size-constraint (score -3)	3.0	3
unnecessary-node (score -2)	0.5	2

Test plan

• 623 tests pass
• Type check clean
• Re-run calibration fixtures and compare before/after score distributions

Closes #104

Made with Cursor

Summary by CodeRabbit

• Improvements
  • Scoring calculation now weights issues based on individual impact scores rather than severity classification.
  • Issue density assessment is more precise, reflecting actual impact differences across severity categories.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: 143a510b-d995-4b95-9aa9-614532e877ea

📥 Commits

Reviewing files that changed from the base of the PR and between 624a2f5 and 6c071a2.

📒 Files selected for processing (2)

• src/core/engine/scoring.test.ts
• src/core/engine/scoring.ts

---

📝 Walkthrough

Walkthrough

The changes replace severity-based density weighting with per-issue calculatedScore in the scoring pipeline. Previously, SEVERITY_DENSITY_WEIGHT assigned fixed weights by severity category; now density contribution derives from each issue's calculatedScore, enabling depth weights and per-rule calibration to influence final scores.

Changes

Cohort / File(s)	Summary
Scoring Logic Update src/core/engine/scoring.ts	Removed SEVERITY_DENSITY_WEIGHT constant and replaced weightedIssueCount accumulation from severity-based weights to Math.abs(issue.calculatedScore), allowing per-rule calibration and depth weighting to affect density scoring.
Test Suite Updates src/core/engine/scoring.test.ts	Updated density scoring tests to verify that densityScore ordering is driven by individual issue calculatedScore values rather than severity category, with new assertions on weightedIssueCount results.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related issues

• Density scoring needs depth weighting to make node-proportional calculation meaningful #105 — Directly complements this change by requesting depth weighting influence on density, which is enabled by consuming calculatedScore from each issue rather than flat severity weights.

Possibly related PRs

• calibrate: nightly run — 7 score adjustments from 9 fixtures #93 — Updates per-rule scores in RULE_CONFIGS, which directly feed into the calculatedScore now used for density weighting in this PR.
• test: scoring.ts, rule-engine.ts 핵심 모듈 유닛 테스트 추가 #58 — Contains scoring tests that verify the old SEVERITY_DENSITY_WEIGHT behavior, creating a direct conflict with the new per-issue scoring approach.
• fix: invisible-layer severity blocking→suggestion + slot recommendation (#23) #24 — Modifies rule scores/severity definitions that directly affect the calculatedScore values now driving the new density logic.

Poem

🐰 The density now dances with calibrated grace,
No more fixed severities in their rigid place—
Each issue's score tells its own weighted tale,
Depth and calibration make the metrics sail! ✨

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately summarizes the main change: replacing flat severity weights with calibrated per-rule scores in the final percentage calculation.
Linked Issues check	✅ Passed	The PR successfully implements the proposed fix in issue #104: replaces flat SEVERITY_DENSITY_WEIGHT with per-issue calculatedScore from the rule engine, allowing calibrated scores and depthWeight to influence final percentages.
Out of Scope Changes check	✅ Passed	All changes are directly related to the scope of issue #104 and PR objectives; modifications to scoring.ts and scoring.test.ts align with replacing severity weights with calculatedScore-based density weighting.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/use-calculated-score-in-scoring

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[let-sunny]

Before/After Score Comparison

Ran analysis on 4 fixtures comparing old (flat severity weights) vs new (calculatedScore):

Overall

Fixture	OLD	NEW	Delta
material3-52949-27916	C+ (71%)	D (57%)	-14
simple-ds-175-7790	C (69%)	D (52%)	-17
simple-ds-175-9106	C+ (70%)	D (52%)	-18
material3-56615-82356	D (60%)	D (52%)	-8

By Category (selected highlights)

Fixture	Category	OLD	NEW	Delta
material3-52949-27916	token	49%	16%	-33
material3-52949-27916	component	72%	61%	-11
simple-ds-175-7790	component	57%	25%	-32
simple-ds-175-7790	behavior	61%	45%	-16
simple-ds-175-9106	naming	63%	51%	-12
simple-ds-175-9106	behavior	68%	48%	-20

Full category breakdown

material3-52949-27916
  OLD: struct:78 token:49 comp:72 name:74 behav:82
  NEW: struct:63 token:16 comp:61 name:67 behav:77

simple-ds-175-7790
  OLD: struct:73 token:79 comp:57 name:74 behav:61
  NEW: struct:57 token:65 comp:25 name:67 behav:45

simple-ds-175-9106
  OLD: struct:74 token:80 comp:64 name:63 behav:68
  NEW: struct:55 token:64 comp:43 name:51 behav:48

material3-56615-82356
  OLD: struct:5  token:82 comp:74 name:75 behav:65
  NEW: struct:5  token:77 comp:51 name:68 behav:58

Observations

• Scores dropped across the board (expected — high-score rules like no-auto-layout at -10 now contribute 10+ to density instead of flat 3.0)
• Component category saw the largest drops — missing-component (score: -7) was previously treated as risk=2.0, now contributes 7.0
• Token category dropped significantly for material3-52949-27916 (204 issues with raw-font at -4 instead of 2.0)
• Structure on material3-56615-82356 was already at floor (5%) — no change possible
• Scores will need recalibration via /calibrate-loop — but now score adjustments in rule-config.ts will actually affect the output