boost-wrapper-log

REQUIRED BACKGROUND: boost-start — log.FromContext returns a no-op until boost.Start() runs.

Always pull the logger from context

import "github.com/xgodev/boost/wrapper/log"

logger := log.FromContext(ctx).
    WithField("component", "orders").
    WithField("order_id", orderID)

logger.Info("processing order")
logger.WithError(err).Warn("validation failed")

The request- or call-scoped context carries enrichment that previous middlewares added (request ID, tracing, etc.). Pulling from context preserves that chain; constructing your own logger throws it away.

API surface

API	When
log.FromContext(ctx)	Always — request- or call-scoped
log.WithField(k, v) / WithError(err)	Enrich the logger with structured fields
log.WithTypeOf(*p)	Inside library/driver code to stamp the type name
log.Fatal/Fatalf	Process-fatal init failures only — never inside a handler

The configured backend (zap, zerolog, logrus) is picked at boost.Start based on boost.factory.<backend>.console.level config. Switching backends doesn't require code changes — the wrapper is the abstraction.

Red flags

Red flag	Fix
log.New(...), zap.NewProduction(), zerolog.New(os.Stdout), slog.New(...)	Replace with log.FromContext(ctx)
Constructing a logger as a struct field, holding across requests	Use log.FromContext per call so per-request enrichment flows through
Calling log.Fatal inside a request handler or worker callback	Return an error instead — fatal kills the whole process
Logging without structured fields (logger.Infof("user %s logged in", id))	Use WithField("user_id", id).Info("user logged in") so log aggregators index it