Go Testing

From NovaOrdis Knowledge Base
Revision as of 22:07, 3 September 2024 by Ovidiu (talk | contribs) (→‎Table (Tabular) Tests)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

External

Internal

TO DO

Overview

Go comes with a lightweight test framework that includes the go test command and the testing package. The tests live in *_test.go files, located in the package directory.

Packages

Test Types in Go

Unit Tests

somepkg_test.go files should contain exclusively unit tests, not integration or system tests. Integration tests should be stored in *_integration_test.go files and system tests in *_system_test.go files. Optionally, unit test files may include the unit_test build tag, which allows controlling test execution based on their type (unit, functional and system).

//go:build unit_test

package somepkg

If no tag is declared, the tests will run fine, but we lose the possibility to control when we don't want them to run. They will always run.

Also see:

Software Testing Concepts | Unit Tests

testing.T

t.Cleanup(func())

Writing a Unit Test

TODO: revisit for Testify.

Write a module as shown here:

Declaring a Module

For each file containing behavior to test (a.go)

package a

func Reverse(s string) string {
 rs := []rune(s)
 var result []rune
 for i := len(rs) - 1; i >= 0; i-- {
  result = append(result, rs[i])
 }
 return string(result)
}

add a <file-name>_test.go test file. In this case a_test.go. These files are ignored by the compiler and only compiled and executed when go test is run. These files will be excluded from regular package builds. For more details on how go test command handles different files, see:

go test File Selection

The test files should be part of the same package. If that is the case, the test has access to unexpected identifiers from the package being tested. It is also possible to declare the test files into a corresponding package with the suffix _test. In this case, the package being tested must be imported explicitly in the test. This is known as "black box" testing. Also see external test packages.

The test files should import testing.

Add individual tests, as functions starting with TestX.., where X... does not start with a lowercase letter, and taking an argument t *testing.T. The function name serves to identify the test routine. Within these functions, use Error, Fail or related methods to signal failure.

func TestX...(t *testing.T) {
  ...
  t.Error("expected this: %q, got that: %q ", ...)
}
package a

import "testing"

func TestReverseEmptyString(t *testing.T) {
 expected := ""
 result := Reverse("")
 if result != expected {
  t.Errorf("expected %q, got %q", expected, result)
 }
}

func TestReverseOneCharString(t *testing.T) {
 expected := "a"
 result := Reverse("a")
 if result != expected {
  t.Errorf("expected %q, got %q", expected, result)
 }
}

func TestReverseTwoCharString(t *testing.T) {
 expected := "ba"
 result := Reverse("ab")
 if result != expected {
  t.Errorf("expected %q, got %q", expected, result)
 }
}

From the module directory, run the tests:

go test

PASS
ok  	example.com/a	0.116s

Table (Tabular) Tests

The "table" ("tabular") format is useful when we need to test a function with a variety of input parameters. The "table" consists in a list of structures (hence the tabular format). Each structure instance provides the test name, the input parameters to be tested, and an assertion function that takes the function result and error and perform assertions correlated to the input. If we simply test for expected results, the assert function can be replaced with the expected results.

Assuming that SomeFunction() is a function to test:

func SomeFunction(arg string) (string, error) {
	return strings.ToUpper(arg), nil
}

then a tabular test would be:

import (
	"testing"
	"github.com/stretchr/testify/require"
)

func TestSomeFunction(t *testing.T) {
	testCases := []struct {
		name   string
		input  string
		assert func(require *require.Assertions, result string, err error) // can be replaced with 'expectedResult'
                                                                           // if no error is expected
        debug  bool // used to stop the debugger on a specific case, by setting a conditional breakpoint on tc.debug
	}{
		{
			name:  "invocation with empty string",
			input: "",
			assert: func(require *require.Assertions, result string, err error) {
				require.NoError(err)
				require.Equal("", result)
			},
		},
		{
			name:  "invocation with non-empty string",
			input: "something",
			assert: func(require *require.Assertions, result string, err error) {
				require.NoError(err)
				require.Equal("SOMETHING", result)
			},
		},
	}
	for _, tc := range testCases {
		t.Run(tc.name, func(t *testing.T) {
			result, err := SomeFunction(tc.input)
			tc.assert(require.New(t), result, err)
		})
	}
}

This is the template:

import (
	testifyrequire "github.com/stretchr/testify/require"
	"testing"
)

func TestSomeMethod(t *testing.T) {
	testCases := []struct {
		name   string
		input  any
		require func(require *testifyrequire.Assertions, result any, err error)
	}{
		{},
	}
	for _, tc := range testCases {
		t.Run(tc.name, func(t *testing.T) {
            require := testifyrequire.New(t)
			result, err := SomeMethod(tc.input)
			tc.require(require, result, err)
		})
	}
}

Integration Tests

We document implementation of integration tests in go. We use this definition of integration test:

Software Testing Concepts | Integration Test

Integration tests, with or without mocks, should be placed in files named *_integration_test.go files. The name of the file, except the ..._test.go part that is mandated by the Go test compiler, has no influence over whether the tests contained in it are executed or not. We control the conditional execution with build tags:

//go:build integration_test

package somepkg

.
└─ internal
    └── mypkg
        ├── mypkg.go
        ├── mypkg_test.go
        └── mypkg_integration_test.go

Code Coverage for Integration Tests

TO PROCESS: https://go.dev/blog/integration-test-coverage

Also see:

go test Test Coverage

System Tests

We use this definition of system test:

Software Testing Concepts | System Test

System tests should be placed in files conventionally named *_system_test.go files and should declare a system_test build tag at the top of the file:

//go:build system_test

package somepkg

Benchmarks

Process this: https://pkg.go.dev/testing#hdr-Benchmarks.

Fuzzing

Process this: https://pkg.go.dev/testing#hdr-Fuzzing.

Skipping

Process this: https://pkg.go.dev/testing#hdr-Skipping.

Subtests and Sub-benchmarks

Process this: https://pkg.go.dev/testing#hdr-Subtests_and_Sub_benchmarks.

Main

Process this: https://pkg.go.dev/testing#hdr-Main.

Mocks

Mock support in Go:

Mocks with Testify
Uber gomock