Go Testing

The testing package provides support for automated testing of Go packages. Tests are run with go test and require no external libraries — the standard library covers unit tests, benchmarks, fuzz tests, and example functions.

File and Function Conventions

Test files must end in _test.go. They are excluded from normal builds but included by go test.

Test functions must match the signature func TestXxx(*testing.T) where Xxx does not start with a lowercase letter:

package mypackage

import "testing"

func TestAdd(t *testing.T) {
    got := Add(2, 3)
    if got != 5 {
        t.Errorf("Add(2, 3) = %d; want 5", got)
    }
}

White-box vs Black-box Tests

Same package (package mypackage) — accesses unexported identifiers
_test suffix package (package mypackage_test) — tests only the exported API; this is "black-box" testing

Both styles can coexist in the same directory.

Reporting Failures

Method	Behaviour
`t.Errorf(format, args...)`	Marks failed, continues execution
`t.Error(args...)`	Marks failed, continues execution
`t.Fatalf(format, args...)`	Marks failed, stops test immediately
`t.Fatal(args...)`	Marks failed, stops test immediately
`t.Fail()`	Marks failed without logging, continues
`t.FailNow()`	Marks failed without logging, stops immediately

Use Errorf/Error when checking multiple independent conditions so all failures are reported. Use Fatalf/Fatal when a failure makes further checks meaningless (e.g., a nil pointer).

Prefer t.Errorf over t.Fatalf unless subsequent assertions depend on a prior one succeeding.

Table-Driven Tests

Table-driven tests are the idiomatic Go pattern for testing a function against many inputs:

func TestDivide(t *testing.T) {
    tests := []struct {
        name    string
        a, b    float64
        want    float64
        wantErr bool
    }{
        {name: "positive", a: 10, b: 2, want: 5},
        {name: "negative divisor", a: 10, b: -2, want: -5},
        {name: "divide by zero", a: 10, b: 0, wantErr: true},
    }

    for _, tc := range tests {
        t.Run(tc.name, func(t *testing.T) {
            got, err := Divide(tc.a, tc.b)
            if (err != nil) != tc.wantErr {
                t.Fatalf("Divide(%v, %v) error = %v, wantErr %v", tc.a, tc.b, err, tc.wantErr)
            }
            if !tc.wantErr && got != tc.want {
                t.Errorf("Divide(%v, %v) = %v; want %v", tc.a, tc.b, got, tc.want)
            }
        })
    }
}

Name each case descriptively. Use t.Run so each case appears as a named subtest in output and can be run individually.

Subtests

t.Run(name, func(t *testing.T)) creates a named subtest. Subtests:

Appear in output as TestParent/SubName
Can be run individually: go test -run TestParent/SubName
Share setup/teardown with the parent
Can be run in parallel independently of other top-level tests

func TestAPI(t *testing.T) {
    server := startTestServer(t) // shared setup

    t.Run("GET /users", func(t *testing.T) {
        // ...
    })
    t.Run("POST /users", func(t *testing.T) {
        // ...
    })
    // server is cleaned up after all subtests finish
}

Parallel Tests

Call t.Parallel() at the start of a test function to allow it to run concurrently with other parallel tests:

func TestExpensive(t *testing.T) {
    t.Parallel()
    // ...
}

For parallel subtests in a table-driven test, capture the loop variable:

for _, tc := range tests {
    tc := tc // capture range variable (required before Go 1.22)
    t.Run(tc.name, func(t *testing.T) {
        t.Parallel()
        // use tc safely
    })
}

From Go 1.22 onward, loop variable capture is automatic and the tc := tc line is no longer needed.

Cleanup

t.Cleanup(f func()) registers a function to run after the test (and all its subtests) complete. Cleanup functions run in LIFO order.

func TestWithDB(t *testing.T) {
    db := openTestDB(t)
    t.Cleanup(func() { db.Close() })
    // test body — db.Close is called automatically when test ends
}

Prefer t.Cleanup over defer inside test helpers because it runs after all subtests complete, not just when the helper function returns.

Test Helpers

Mark a function as a test helper with t.Helper() so error output points to the call site, not inside the helper:

func assertEqualInts(t *testing.T, got, want int) {
    t.Helper() // makes error line point to the caller
    if got != want {
        t.Errorf("got %d; want %d", got, want)
    }
}

Always call t.Helper() as the first statement in helper functions.

Temporary Directories

t.TempDir() creates a temporary directory that is automatically removed when the test completes:

func TestWriteFile(t *testing.T) {
    dir := t.TempDir()
    path := filepath.Join(dir, "output.txt")
    // write to path — dir is cleaned up automatically
}

Environment Variables

t.Setenv(key, value) sets an environment variable and restores the original value after the test. Cannot be used in parallel tests.

func TestWithEnv(t *testing.T) {
    t.Setenv("MY_CONFIG", "test-value")
    // original value restored after test
}

Skipping Tests

Skip a test conditionally using t.Skip, t.Skipf, or t.SkipNow:

func TestIntegration(t *testing.T) {
    if testing.Short() {
        t.Skip("skipping integration test in short mode")
    }
    // ...
}

func TestRequiresDocker(t *testing.T) {
    if os.Getenv("DOCKER_HOST") == "" {
        t.Skip("DOCKER_HOST not set")
    }
    // ...
}

Example Functions

Example functions serve as documentation and are verified by go test:

func ExampleAdd() {
    fmt.Println(Add(1, 2))
    // Output: 3
}

The // Output: comment is compared against stdout. Examples without an output comment are compiled but not executed. Use // Unordered output: when output order is non-deterministic.

Naming conventions:

func Example() { ... }           // package example
func ExampleAdd() { ... }        // function Add
func ExampleCalc() { ... }       // type Calc
func ExampleCalc_Add() { ... }   // method Calc.Add
func ExampleAdd_second() { ... } // second example for Add (suffix starts lowercase)

TestMain

TestMain controls global test setup and teardown. Define it in any _test.go file in the package:

func TestMain(m *testing.M) {
    // setup
    code := m.Run()
    // teardown
    os.Exit(code)
}

Use TestMain for package-level resources (database connections, server processes). It is not necessary for per-test resources — use t.Cleanup instead.

Running Tests

# Run all tests in the current module
go test ./...

# Run with verbose output
go test -v ./...

# Run tests matching a pattern (regexp)
go test -run TestAdd ./...

# Run a specific subtest
go test -run TestDivide/divide_by_zero ./...

# Run with race detector
go test -race ./...

# Run with coverage
go test -cover ./...

# Generate HTML coverage report
go test -coverprofile=coverage.out ./... && go tool cover -html=coverage.out

# Skip slow tests
go test -short ./...

# Set test timeout
go test -timeout 30s ./...

# Run benchmarks
go test -bench=. ./...

# Run benchmarks with memory allocation stats
go test -bench=. -benchmem ./...

Quick Reference: `testing.T` Methods

Method	Purpose
`t.Run(name, f)`	Create named subtest
`t.Parallel()`	Mark test as parallel
`t.Helper()`	Mark as helper function
`t.Cleanup(f)`	Register teardown function
`t.TempDir()`	Create auto-cleaned temp directory
`t.Setenv(k, v)`	Set env var, auto-restored after test
`t.Chdir(dir)`	Change working dir, auto-restored
`t.Context()`	Context canceled before cleanup runs
`t.Log(args...)`	Log (shown on failure or with `-v`)
`t.Logf(format, args...)`	Log formatted
`t.Error(args...)`	Fail + log, continue
`t.Errorf(format, args...)`	Fail + log formatted, continue
`t.Fatal(args...)`	Fail + log, stop
`t.Fatalf(format, args...)`	Fail + log formatted, stop
`t.Skip(args...)`	Skip + log, stop
`t.Skipf(format, args...)`	Skip + log formatted, stop
`t.Name()`	Return full test name
`t.Failed()`	Reports whether test has failed
`t.Deadline()`	Returns test deadline from `-timeout` flag

Additional Resources

For benchmarks, fuzz testing, and advanced patterns:

references/benchmarks-and-fuzzing.md — testing.B API, b.Loop() style, parallel benchmarks, testing.F fuzz tests, TestMain patterns, and AllocsPerRun

go-testing