Mocking

When writing tests for code that takes a *loglayer.LogLayer, you usually want one of three things:

1. Silence the logger so test output stays clean and behaviors aren't accidentally tied to log writes.
2. Assert on what was logged: verify your code emits the right entry under the right conditions.
3. Use the real logger but quiet it: keep the production wiring, just suppress output.

LogLayer ships a primitive for each.

1. Silent Mock: loglayer.NewMock()

Use this when logs aren't part of what you're testing. It's a drop-in *loglayer.LogLayer backed by a discard transport. Every call is accepted but produces no output.

import "go.loglayer.dev/v3"

func TestSomething(t *testing.T) {
    log := loglayer.NewMock()

    result, err := DoWork(log, "input")
    if err != nil {
        t.Fatal(err)
    }
    if result != "expected" {
        t.Errorf("got %q", result)
    }
}

NewMock() returns the same concrete *loglayer.LogLayer type a real construction returns, so it satisfies any signature that takes *loglayer.LogLayer (no interface needed). All instance state still works (fields, level filtering, child loggers, prefixes); only the emit step is a no-op.

It also sets DisableFatalExit: true so test code that exercises fatal paths doesn't crash the test runner.

log := loglayer.NewMock()
log = log.WithFields(loglayer.Fields{"requestId": "abc"})
log.SetLevel(loglayer.LogLevelWarn)

log.Info("dropped: below threshold AND silent")
log.Warn("silent: emit step does nothing")

log.GetFields()                            // {"requestId": "abc"}
log.IsLevelEnabled(loglayer.LogLevelInfo)   // false

This is the right default for unit tests of business logic.

Need to assert on the rendered output?

NewMock() emits nothing, and there is no NewMockWithWriter. When the test's purpose is to assert what a transport renders (exact level prefixes, logfmt, JSON shape), construct a real logger with the transports/testing capture transport or a bytes.Buffer writer on a renderer transport (see Testing Transports), and assert against the captured lines or buffer.

2. Capturing Mock: transports/testing

Use this when the test's purpose is to verify what was logged. The transports/testing package provides a transport that captures every entry into an in-memory library, exposed as typed LogLine values.

import (
    "go.loglayer.dev/v3"
    lltest "go.loglayer.dev/transports/testing/v3"
)

func TestRequestLogging(t *testing.T) {
    lib := &lltest.TestLoggingLibrary{}
    log := loglayer.New(loglayer.Config{
        Transport: lltest.New(lltest.Config{Library: lib}),
    })

    handleRequest(log, "abc-123")

    line := lib.PopLine()
    if line == nil {
        t.Fatal("expected a log entry")
    }
    if line.Level != loglayer.LogLevelInfo {
        t.Errorf("level = %s, want info", line.Level)
    }
    if line.Data["requestId"] != "abc-123" {
        t.Errorf("requestId not in fields data: %v", line.Data)
    }
    meta, _ := line.Metadata.(loglayer.Metadata)
    if meta["status"] != 200 {
        t.Errorf("status not in metadata: %v", line.Metadata)
    }
}

LogLine shape

Every captured entry is a lltest.LogLine with typed fields. No parsing flat arg lists or string scraping.

type LogLine struct {
    Level    loglayer.LogLevel
    Messages []any
    Data     loglayer.Data   // assembled fields + error map; nil when neither were set
    Metadata any             // raw value passed to WithMetadata
    Ctx      context.Context // per-call context attached via WithContext; nil if not set
    Prefix   string          // value from WithPrefix / Config.Prefix; empty when unset
}

Library API

Method	Purpose
Lines()	Snapshot copy of all captured lines
GetLastLine()	Most recent line (does not remove); nil if empty
PopLine()	Most recent line (removes it); nil if empty
ClearLines()	Drop all captured lines
Len()	Number of captured lines

All methods are safe for concurrent use.

Asserting on struct metadata

Because WithMetadata(any) passes structs through untouched, you can assert on the original type:

type orderEvent struct {
    OrderID string
    Total   int
}

log.WithMetadata(orderEvent{OrderID: "o-1", Total: 42}).Info("order placed")

line := lib.PopLine()
event, ok := line.Metadata.(orderEvent)
if !ok {
    t.Fatalf("expected orderEvent, got %T", line.Metadata)
}
if event.OrderID != "o-1" {
    t.Errorf("OrderID: %s", event.OrderID)
}

See the transports/testing page for the full transport reference.

3. Quiet the Real Logger: DisableLogging()

When you want the production wiring intact (real transports, real config) but no output during a particular test:

log := buildProductionLogger() // your normal construction
log.DisableLogging()           // master kill switch: no entries emitted

This is rarer than the first two patterns but useful for integration tests that exercise startup/shutdown paths where the log call sites matter (won't panic, won't deadlock) but you don't want the noise.

EnableLogging() restores the previous per-level state. See Adjusting Log Levels.

Choosing a Pattern

Want to…	Use
Test business logic; ignore log output	loglayer.NewMock()
Verify a specific log was emitted with the right fields	transports/testing
Run real transports but silence them for one test	log.DisableLogging()

Why no Logger interface?

Go's idiomatic pattern is the consumer defines the interface. Your application code declares the methods it needs:

type RequestLogger interface {
    WithFields(loglayer.Fields) *loglayer.LogLayer
    Info(...any)
    WithError(error) *loglayer.LogBuilder
}

Both real *loglayer.LogLayer and the mock from NewMock() implicitly satisfy that. Shipping a loglayer.Logger interface would push a one-size-fits-all shape on every consumer (over-broad for most call sites and too narrow for some). Keep the concrete type, swap with NewMock() in tests.