Beschreibung

Implementiere Benchmarks für kritische Code-Paths (Sync-Engine, Mermaid-Export) mit Regression Detection in CI. Ermöglicht Performance-Tracking über Releases hinweg.

Motivation:

• Go Tool: go test -bench ist built-in, keine Dependencies
• Verhindert Performance-Degradation
• Baseline für Performance-SLOs
• Data-driven optimization decisions

Implementation Plan

Phase 1: Benchmark-Suite erstellen (2h)

1. internal/sync/benchmark_test.go

   • BenchmarkSyncSmallModel (100 elements)
   • BenchmarkSyncLargeModel (5000 elements)
   • BenchmarkDetectChanges
   • BenchmarkApplyForward
   • BenchmarkApplyReverse

2. internal/diagram/benchmark_test.go

   • BenchmarkExportMermaid (10 views)
   • BenchmarkExportMermaid (100 views)
   • BenchmarkWrapDiagramsInMarkdown

3. internal/model/benchmark_test.go

   • BenchmarkLoad (JSONC parsing)
   • BenchmarkFlattenElements
   • BenchmarkResolveView
   • BenchmarkValidate

Deliverable: benchmarks run locally go test -bench=./internal/sync/

Phase 2: CI Integration & Baselines (1.5h)

1. .github/workflows/benchmark.yml

   • Run benchmarks on every PR
   • Store results in benchmarks/ directory
   • Generate summary comment on PR

2. Create baseline results

   go test -bench=. -benchmem ./internal/sync/ > benchmarks/main.txt

3. Add benchstat tool

   • Compare PR results vs main
   • Flag >5% regressions

Deliverable: CI workflow, baseline file in repo

Phase 3: Documentation & SLOs (1h)

1. BENCHMARKS.md

   • Expected runtimes per benchmark
   • Hardware specs (CPU, RAM used for baseline)
   • How to run locally
   • Regression interpretation

2. Performance SLOs

   Sync 1000 elements: <500ms
   Sync 10000 elements: <5s
   Export 100 views: <2s
   Model.Load 50MB: <1s
   

Deliverable: BENCHMARKS.md, SLO document

Akzeptanzkriterien

• ✅ Mindestens 10 Benchmarks implementiert (sync, diagram, model)
• ✅ Benchmarks laufen in CI auf jedem PR
• ✅ Baseline-Ergebnisse gespeichert (benchmarks/main.txt)
• ✅ Regression-Detection: >5% Abweichung flagged
• ✅ Dokumentation: How-to-run, SLOs, expected values
• ✅ All benchmarks stable (<5% variance in 3 runs)

Aufwand: 3-4 Stunden