meta.json

{
  "slug": "logging-framework",
  "title": "Design a Logging Framework",
  "difficulty": "medium",
  "summary": "Route log records down a handler chain where every link applies its own threshold and format, so one chain can send everything to the console and only errors to a JSON file — and where the chain deliberately does not stop at the first handler that takes it.",
  "tags": ["infrastructure", "observability", "extensibility", "library-design"],
  "patterns": ["chain-of-responsibility", "strategy", "singleton"],
  "concepts": [
    "Adapting a pattern rather than copying it: why a logging chain must not stop",
    "Two thresholds serving different purposes: a cheap gate and per-sink filtering",
    "Keeping format independent of destination to avoid a class per combination",
    "When Singleton is genuinely correct, and what it costs in testability",
    "Bound context for correlating records across a request"
  ],
  "entrypoint": "src/main.py",
  "tests": "tests",
  "estimatedMinutes": 45,
  "askedAt": ["Amazon", "Salesforce", "Oracle", "Zoho"],
  "diagram": { "mermaid": true },
  "extensions": [
    "Make delivery asynchronous with a queue and a writer thread",
    "Rotate log files by size or date",
    "Rate-limit repeated identical messages with a decorating handler",
    "Support hierarchical loggers that inherit handlers from their parent",
    "Make handlers safe for concurrent writers",
    "Redact secrets before records are written"
  ]
}