Go Table-Driven Tests

Overview

Table-driven tests are a Go testing idiom that reduces code duplication and makes tests more maintainable. Instead of writing separate test functions for each case, you define a table of test cases and iterate over it.

When to Use Table-Driven Tests

Use table-driven tests when:

• You find yourself copying and pasting test code
• You're testing the same function/behavior with multiple inputs
• You want to add more test cases without writing more test functions
• Edge cases and boundary conditions need systematic coverage

Do NOT use for: Completely unrelated test scenarios, or when each test requires substantially different setup/teardown logic.

Basic Template (Slice Pattern)

This is the most common pattern in this codebase:

func TestFunctionName(t *testing.T) {
    cases := []struct {
        name  string
        input string
        want  string
        err   error
    }{
        {
            name:  "simple case",
            input: "a/b/c",
            want:  "a,b,c",
        },
        {
            name:  "empty input",
            input: "",
            want:  "",
        },
        {
            name:  "invalid input",
            input: "!!!",
            want:  "",
            err:   ErrInvalid,
        },
    }

    for _, tt := range cases {
        t.Run(tt.name, func(t *testing.T) {
            got, err := FunctionName(tt.input)
            if !errors.Is(err, tt.err) {
                t.Errorf("FunctionName(%q) error = %v, want %v", tt.input, err, tt.err)
            }
            if got != tt.want {
                t.Errorf("FunctionName(%q) = %q, want %q", tt.input, got, tt.want)
            }
        })
    }
}

Map Pattern (For Non-Deterministic Test Ordering)

Use a map when you want to ensure test independence:

func TestFunctionName(t *testing.T) {
    tests := map[string]struct {
        input string
        want  string
    }{
        "simple case":   {input: "a/b/c", want: "a,b,c"},
        "empty input":   {input: "", want: ""},
        "trailing sep":  {input: "a/b/c/", want: "a,b,c"},
    }

    for name, tc := range tests {
        t.Run(name, func(t *testing.T) {
            got := FunctionName(tc.input)
            if got != tc.want {
                t.Fatalf("%s: expected %q, got %q", name, tc.want, got)
            }
        })
    }
}

Repository-Specific Conventions

Variable Naming

This codebase consistently uses these variable names:

Struct Field Guidelines

// Always use named fields (not anonymous structs in this codebase)
cases := []struct {
    name      string  // Descriptive test name (REQUIRED when using slice pattern)
    input     Type    // Input to function under test
    want      Type    // Expected output
    err       error   // Expected error (use errors.Is for comparison)
    wantErr   bool    // Alternative: true if error is expected
    precondition func(*testing.T)  // Optional setup function
}{ ... }

Error Reporting

This repository uses these patterns:

// For error checking (preferred)
if !errors.Is(err, tt.err) {
    t.Errorf("error = %v, want %v", err, tt.err)
}

// For simple comparisons
if got != tt.want {
    t.Errorf("got %q, want %q", got, tt.want)
}

// Use t.Fatalf only when continuing doesn't make sense
// Use t.Errorf to see all test failures before stopping

Advanced Patterns

With Precondition Functions

When tests need specific setup:

for _, tt := range []struct {
    name         string
    precondition func(*testing.T)
    input        string
    err          error
}{
    {
        name: "with listener",
        precondition: func(t *testing.T) {
            ln, err := net.Listen("tcp", ":8081")
            if err != nil {
                t.Fatal(err)
            }
            t.Cleanup(func() { ln.Close() })
        },
        input: "test",
        err:   nil,
    },
} {
    t.Run(tt.name, func(t *testing.T) {
        if tt.precondition != nil {
            tt.precondition(t)
        }
        // test logic
    })
}

Parallel Tests

For independent tests that can run in parallel:

func TestFunctionName(t *testing.T) {
    cases := []struct {
        name  string
        input string
        want  string
    }{
        // ... test cases
    }

    for _, tt := range cases {
        tt := tt // Capture range variable (Go < 1.22)
        t.Run(tt.name, func(t *testing.T) {
            t.Parallel() // Marks this subtest as parallel

            got := FunctionName(tt.input)
            if got != tt.want {
                t.Errorf("got %q, want %q", got, tt.want)
            }
        })
    }
}

Best Practices

1. Always use t.Run() for subtests - This is 100% consistent in this codebase
2. Use descriptive test names - The name field should clearly describe what is being tested
3. Test one thing per case - Each table entry should test one specific behavior
4. Include edge cases - Empty strings, nil values, maximum values, etc.
5. Use errors.Is for error comparison - Not == or reflect.DeepEqual
6. Prefer t.Errorf over t.Fatalf - See all failures before stopping
7. Keep test data inline - External files only for large golden test sets

Common Pitfalls to Avoid

1. Forgetting t.Run() - Without subtests, all failures appear at the same line
2. Using Fatalf immediately - You won't see other test failures
3. Not capturing range variable - In Go < 1.22, add tt := tt before t.Run
4. Anonymous structs - This codebase prefers named structs for clarity
5. Inconsistent naming - Stick to the conventions (cases, tt, want, got)

Comparison with Traditional Tests

Running Specific Tests

# Run all tests in a function
go test -run TestFunctionName

# Run a specific subtest
go test -run TestFunctionName/descriptive_name

# Run all tests matching a pattern
go test -run TestFunctionName/.*/empty

# Verbose mode (see all test output)
go test -v

# Run with race detector
go test -race

References

• Go Wiki: TableDrivenTests
• Go Testing Package
• Prefer Table Driven Tests - Dave Cheney