Wide Events Logging (Canonical Log Lines)

Logging should not be a "debugging diary" where you sprinkle console.log("doing X") across your codebase. This creates log chaos: thousands of fragmented strings that are impossible to correlate or query effectively during an incident.

Instead, when instrumenting applications, implement Wide Events (also known as Canonical Log Lines).

Core Philosophy

1. One Request, One Log Line: Emit exactly one comprehensive structured event per request, per service.
2. Accumulate Context: Initialize an event object at the start of a request (usually in middleware), pass it down (or attach to context), enrich it with business data as the request executes, and log it in a finally block or at the network boundary.
3. High Cardinality: Strongly prefer adding attributes with millions of possible values (e.g., user_id, request_id, cart_id) because these are the most valuable fields for pinpointing specific failures.
4. High Dimensionality: Don't just log HTTP status and duration. Log feature flags, user subscription tiers, database query counts, attempt numbers, and precise decline codes.

1. Building the Wide Event