Log-Driven Debugging

Core Principle

Never assume the cause. Verify with logs.

Debugging Workflow

1. Gather Information

   • Ask: "What is the exact symptom?"
   • Ask: "When does it occur?"
   • Ask: "What is the expected behavior?"

2. Form Hypotheses

   • List 2-3 probable causes ranked by likelihood
   • Consider: API issues, state mutations, async timing, type errors

3. Add Targeted Logs

   • Add console.log at probable cause points
   • Use log patterns from patterns.md
   • Always include context: variable names, values, and execution path

4. Request Console Output

   • Ask user: "Run your code and paste the complete console output"
   • For browser: "Open DevTools (F12) → Console tab → paste output"

5. Analyze & Iterate

   • Compare actual vs expected values
   • If log missing → code path is different than expected
   • If undefined/null → trace upstream
   • Repeat until root cause found

Log Placement Strategy

• Entry logs: Log function/method entry with input parameters
• Exit logs: Log function/method exit with return values
• Before/after critical operations: Log state before and after mutations
• Conditional branches: Log which branch executed and why
• Async operations: Log before and after promises/await points
• Boundary points: Log data entering/leaving modules or components

Log Format Cheatsheet

Example Interaction

User: "My API call returns undefined but should return user data"

Agent adds logs:

async function getUser(id) {
  console.log("[getUser] Called with id:", id);
  const response = await fetch(`/api/users/${id}`);
  console.log("[getUser] Response status:", response.status);
  const data = await response.json();
  console.log("[getUser] Parsed data:", JSON.stringify(data, null, 2));
  return data.user.profile;
}

Console output reveals:

[getUser] Called with id: 123
[getUser] Response status: 200
[getUser] Parsed data: { "user": { "name": "John" } }

Analysis: data.user.profile is undefined because the API returns { user: { name } } without a profile property → fix: return data.user or update API.

When Not to Use

• Build/compilation errors: Clear error messages; no logs needed
• Type-only issues: Use TypeScript compiler diagnostics instead
• Performance profiling: Use browser DevTools Performance tab
• Production debugging: Consider structured logging (Winston, Pino)

Common Patterns

See patterns.md for debugging scenarios:

• API/Network issues
• State mutation issues
• Async/timing issues
• Type/TypeScript issues