Edge Case Coverage

Systematically identify and handle boundary conditions.

Decision Tree: Which Edge Case Category?

What kind of edge case are you looking for?
├─ Input validation? → Check null, wrong type, empty, overflow, malformed
├─ Boundary values? → Check 0, 1, min, max, just-above, just-below
├─ State transitions? → Check uninitialized, concurrent, stale, partial
├─ Resource limits? → Check timeout, OOM, disk full, connection pool exhaustion
├─ Network failures? → Check offline, slow, partial failure, retry exhaustion
├─ Permission issues? → Check unauthorized, expired, revoked, insufficient scope
└─ Multi-system interaction? → Check race conditions, ordering, idempotency

Edge Case Categories

Boundary Analysis

Numeric Boundaries

Value: age
├── Below min: -1
├── At min: 0
├── Just above min: 1
├── Normal: 25
├── Just below max: 119
├── At max: 120
└── Above max: 121

String Boundaries

Value: username
├── Empty: ""
├── Single char: "a"
├── Max length: "a" * 255
├── Over max: "a" * 256
├── Special chars: "user@#$"
└── Unicode: "user"

Error Scenario Template

## Scenario: [Name]

**Trigger**: [What causes it]
**Symptoms**: [What user sees]
**Root cause**: [Why it happens]
**Prevention**: [How to avoid]
**Recovery**: [How to fix]

Validation Checklist

Input Validation

• [ ] Required fields present
• [ ] Type correct
• [ ] Format valid (email, URL, etc.)
• [ ] Range within bounds
• [ ] Length within limits
• [ ] Characters allowed

State Validation

• [ ] Object initialized
• [ ] Resources available
• [ ] Permissions granted
• [ ] Dependencies met
• [ ] No conflicts

Coverage Matrix

Anti-Patterns

• Happy path only — designing and testing only the expected flow; every production failure starts as an edge case someone assumed would not happen
• Ignoring nulls — null is the most common edge case in production; always handle explicitly
• Assuming valid input — external input is never trustworthy; validate at the boundary
• Missing timeout handling — every network or I/O operation must have a timeout; without one, the system hangs indefinitely
• Silent failures — swallowing errors without logging or surfacing hides bugs until they cascade
• Testing only one layer — edge cases at the integration boundary (API + DB, UI + API) are the most common production failures
• Forgetting idempotency — retries on failed operations cause duplicates if endpoints are not idempotent

When to Use

• Before implementing a feature — identify edge cases in the spec
• During code review — check for unhandled boundaries
• When hardening an existing system — systematic enumeration of failure modes
• When writing error handling — ensure all categories are covered