| name | cobol-insights |
| description | Modernization analysis heuristics, complexity scoring rubrics, migration wave planning guidance, and interpretation of COBOL knowledge graph metrics. Use when graph-query, complexity-scorer, or migration-advisor needs to interpret results or make recommendations. |
COBOL Modernization Insights
Reading Complexity Scores
| Score | Category | Interpretation | Typical Effort |
|---|
| 1-25 | EASY | Leaf programs, low coupling, clean structure | 3-7 days |
| 26-50 | MODERATE | Some coupling, possible REDEFINES, medium size | 1-3 weeks |
| 51-75 | HARD | High coupling, GOTOs, large size, shared copybooks | 1-2 months |
| 76-100 | CRITICAL | ALTER present, circular deps, 2000+ lines | Evaluate rewrite |
Migration Wave Planning
Wave 1 — Foundation (Months 1-2)
Target: EASY programs with zero inbound calls
Goal: Build Java muscle, establish patterns, prove the toolchain
Criteria:
MATCH (p:Program)
WHERE p.migration_category = 'EASY'
AND NOT ()-[:CALLS]->(p)
RETURN p.program_id ORDER BY p.migration_score
Wave 2 — Service Extraction (Months 3-5)
Target: MODERATE programs that are highly called (high fan-in)
Goal: Replace shared subroutines with Java services others can call
Criteria:
MATCH (p:Program)<-[:CALLS]-(callers)
WHERE p.migration_category = 'MODERATE'
RETURN p.program_id, count(callers) AS fan_in
ORDER BY fan_in DESC
Wave 3 — Core Programs (Months 6-10)
Target: HARD programs after their dependencies are migrated
Strategy: Strangler fig pattern — wrap in Java, migrate incrementally
Wave 4 — Legacy Cores (Month 10+)
Target: CRITICAL programs
Decision point: Full rewrite vs. keep-and-wrap
Tools: Anti-corruption layer pattern
Key Metrics to Track
| Metric | Good | Warning | Critical |
|---|
| Average fan-in (incoming calls) | < 3 | 3-8 | > 8 |
| Programs with GOTO | < 5% | 5-15% | > 15% |
| Programs with ALTER | 0 | 1-2 | > 2 |
| Copybook usage breadth | < 5 programs | 5-10 | > 10 |
| Circular dependency count | 0 | 1-3 | > 3 |
Interpreting the Call Graph
- High fan-in programs (many callers) → migrate LAST; they are shared infrastructure
- High fan-out programs (many callees) → migrate AFTER all callees are done
- Isolated programs (no calls in or out) → migrate FIRST; zero risk
- Circular dependencies → must be broken before migration; introduce interface layer
Common Questions and Cypher Answers
"Where should we start migrating?"
→ Run complexity-scorer, then query EASY + zero fan-in programs
"What breaks if we change X?"
→ Run impact-analyzer on X
"Which copybook change would cause the most disruption?"
→ Query copybooks ordered by program usage count (descending)
"Are there any dead programs (never called)?"
MATCH (p:Program)
WHERE NOT ()-[:CALLS]->(p) AND NOT ()-[:EXECUTES]->(p)
RETURN p.program_id, p.line_count
ORDER BY p.line_count DESC
For full complexity scoring heuristics: read COMPLEXITY-HEURISTICS.md
For migration readiness scoring rubrics: read MIGRATION-READINESS.md
