Learn

Go's testing tool is built into the language and toolchain. No external framework needed for the basics.

Test files and the testing.T type

A file foo_test.go next to foo.go:

Run with go test ./.... Test functions:

• Start with Test
• Take *testing.T (and only that)
• Live in _test.go files

t.Errorf logs and continues. t.Fatalf logs and stops the test. t.Helper() marks a function as a test helper so failures point at the caller.

Table-driven tests — Go's idiom

Don't write one TestX per case. Drive a loop:

t.Run(name, func) makes a subtest — failures show the case name. Run a single subtest with go test -run TestAdd/with_zero.

Setup and teardown

Use t.Cleanup(func) to register cleanup that runs at end-of-test:

Cleaner than defer because it runs even on t.Fatal.

For package-level setup, use TestMain:

Benchmarks

A benchmark function:

Run with go test -bench=. -benchmem. Output:

BenchmarkSort-8    10000   125000 ns/op   2048 B/op   2 allocs/op

b.N is chosen by the test framework to run for at least 1 second. Always benchmark the operation, not the setup — use b.ResetTimer().

Examples (and they double as docs)

Run by go test. Verify the actual output matches the // Output: comment. Also rendered in go doc.

Test fuzzing (Go 1.18+)

Run with go test -fuzz=FuzzReverse. Generates random inputs and saves any that crash the test.

Common mistakes

• Not using table-driven tests — repeating Test functions for similar cases is verbose.
• Forgetting t.Helper() in shared assertion functions — failures point at the helper, not the caller's line.
• Including setup in the timed loop — without b.ResetTimer(), your benchmark numbers include setup cost.
• Side effects on test order — tests should be independent. Use t.TempDir(), unique IDs.
• Skipping race detection — go test -race is essential for concurrent code.