Skip to content
/ tester Public

Lightweight test utilities to use with Go's testing package

License

Notifications You must be signed in to change notification settings

voicera/tester

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Tester: Test More, Type Less

Build Status Go Report Card GoDoc Maintainability Test Coverage

Lightweight test utilities to use with Go's testing package.

Features

  • Assertions that make tests easier to read, write, and debug
  • Streamlined data providers for data-driven testing (DDT)
  • Test hooks' hygiene check

Motivations

Testing is an integral part of Go; the language provides strong and opinionated support for testing. However, developers who moved from languages like C#, Java, and Python miss the convenience of test utilities like assertions and data providers for data-driven tests. That's why we started Tester: lightweight test utilities to use with Go's testing package.

Most tests follow the same pattern: set up, invoke the unit under test, assert, then clean up (if need be); said pattern encourages test code reuse and consistency. By using test utilities, you can spend more time thinking about test strategies and less time typing boilerplate code.

Quick Start

Get the latest version (go get -u github.com/voicera/tester) then test away:

package hitchhiker

import (
    "testing"

    "github.com/voicera/tester/assert"
)

func TestDeepThought(t *testing.T) {
    computer := NewDeepThoughtComputer()
    answer, err := computer.AnswerTheUltimateQuestion()
    if assert.For(t).ThatActualError(err).IsNil().Passed() {
        assert.For(t).ThatActual(answer).Equals(42)
    }
}

Learn More

The following can also be found at https://godoc.org/github.com/voicera/tester

Assertions

Package assert provides a more readable way to assert in test cases; for example:

assert.For(t).ThatCalling(fn).PanicsReporting("expected error")

This way, the assert statement reads well; it flows like a proper sentence.

Also, one can easily tell which value is the test case got (actual) and which it wanted (expected); this is key to printing the values correctly to make debugging a bit easier. In Go, the actual value is usually printed first; for example:

assert.For(t).ThatActual(foo).Equals(expected)

The above enforces said order in both reading the code and the assertion failure message (if any).

For convenience (that also improves readability), there are methods for special cases like:

assert.For(t).ThatActual(foo).IsTrue()
assert.For(t).ThatActualString(bar).IsEmpty()

Which are equivalent to:

assert.For(t).ThatActual(foo).Equals(true)
assert.For(t).ThatActual(len(bar)).Equals(0)

To identify a test case in a table-driven test, optional parameters can be specified and will be included in failure messages:

cases := []struct {
    id       string
    actual   interface{}
    expected interface{}
}{
    {"different values", 42, 13},
    {"different types", 42.0, 42},
    {"different containers", [...]int{42}, []int{42}},
}

for _, c := range cases {
    assert.For(t, c.id).ThatActual(c.actual).Equals(c.expected)
}

After an assertion is performed, a ValueAssertionResult is returned to allow for post-assert actions to be performed; for example:

assert.For(t).ThatActual(value).Equals(expected).ThenDiffOnFail()

That will pretty-print a detailed recursive diff of both objects on failure.

It also can be used as a condition to perform extra test steps:

value := GetValue()
if assert.For(t).ThatActual(value).IsNotNil().Passed() {
    assert.For(t).ThatActual(value.GetFoo()).Equals(expected)
}

Or to perform a deeper analysis of the test values:

if !assert.For(t).ThatActual(value).Equals(expected).Passed() {
    analyze(value, expected) // e.g., analyze may look at common bugs
}

Conveniently, the last example above can be rewritten as:

assert.For(t).ThatActual(value).Equals(expected).ThenRunOnFail(analyze)

Another use is to print a custom failure message:

assert.For(t).ThatActual(foo).Equals(bar).ThenRunOnFail(func(actual, expected interface{}) {
    fmt.Printf("JSON: %q != %q", actual.ToJSON(), expected.ToJSON())
})

The above pattern allows for reuse of post-failure analysis and cleanup.

The interfaces in this package are still a work-in-progress, and are subject to change.

Test Hooks

What exists merely for test code to see shall not be exported to the world. You can tag test-hook fields like the following:

type sleeper struct {
    sleep func(time.Duration) `test-hook:"verify-unexported"`
}

Using tags, instead of comments, enables you to search the codebase for test hooks and validate, via reflection, that they're not exported. A test case should be added to verify that test hooks are hidden:

func TestHooksAreHidden(t *testing.T) {
    assert.For(t).ThatType(reflect.TypeOf(sleeper{})).HidesTestHooks()
}

Data-Driven Testing (DDT)

When the number of test cases in a table-driven test gets out of hand and they cannot fit neatly in structs anymore, the use of a data provider is in order. Package ddt provides a way to load test cases from a JSON file whose path is derived from the caller's test function name and file. The file path is <package under test>/_ddt/<name of test function>.json; for example, hitchhiker/_ddt/TestDeepThought.json with the following schema:

{
  "testCases": [
    {
        <properties of the test case to unmarshal>
    },
    ...
  ]
}

For example, the JSON content may look like the following:

{
  "testCases": [
    {
      "id": "The Ultimate Question",
      "input": {
        "question": "What do you get when you multiply six by nine?",
        "timeoutInHours": 65700000000,
        "config": {
          "base": 13
        }
      },
      "expected": {
        "answer": "42",
        "error": null
      }
    }
  ]
}

The details of the test case struct are left for the tester to specify.