Advanced unit testing patterns in Go

A good developer always tests their code, however, common testing methods may be too simplistic in some cases. Depending on the complexity of a project, you may need to run advanced tests to accurately evaluate the performance of your code.

In this article, we’ll examine a few patterns for testing in Go that will help you write effective tests for any project. We’ll cover concepts like mocking, test fixtures, test helpers, and golden files, and you’ll see how you can apply each technique in a real-world scenario.

To follow along with this article, you should have prior knowledge of unit testing in Go. Let’s get started!

Testing HTTP handlers

First, let’s consider a common scenario, testing HTTP handlers. HTTP handlers should be loosely coupled with their dependencies, making it easy to isolate an element for testing without impacting the rest of the code. If your HTTP handlers are well designed initially, testing should be fairly straightforward.

Checking status code

Let’s consider a basic test that checks the status code of the following HTTP handler:

func index(w http.ResponseWriter, r *http.Request) {
    w.WriteHeader(http.StatusOK)
}

The index() handler above should return a 200 OK response for every request. Let’s verify the handler’s response with the following test:

func TestIndexHandler(t *testing.T) {
    w := httptest.NewRecorder()
    r := httptest.NewRequest(http.MethodGet, "/", nil)

    index(w, r)

    if w.Code != http.StatusOK {
        t.Errorf("Expected status: %d, but got: %d", http.StatusOK, w.Code)
    }
}

In the code snippet above, we use the httptest package to test the index() handler. We returned an httptest.ResponseRecorder, which implements the http.ResponseWriter interface through the NewRecorder() method. http.ResponseWriter records any mutations, allowing us to make assertions in the test.

We can also create an HTTP request using the httptest.NewRequest() method. Doing so specifies the types of requests expected by the handler, like the request method, query parameters, and response body. You can also set request headers after obtaining the http.Request object through the http.Header type.

After calling the index() handler with the http.Request object and response recorder, you can directly inspect the handler’s response using the Code property. To make assertions on other properties on the response, like the headers or the body, you can access the appropriate method or property on the response recorder:

$ go test -v
=== RUN   TestIndexHandler
--- PASS: TestIndexHandler (0.00s)
PASS
ok      github.com/ayoisaiah/random 0.004s

External dependencies

Now, let’s consider another common scenario in which our HTTP handler has a dependency on an external service:

func getJoke(w http.ResponseWriter, r *http.Request) {
    u, err := url.Parse(r.URL.String())
    if err != nil {
        http.Error(w, err.Error(), http.StatusInternalServerError)
        return
    }

    jokeId := u.Query().Get("id")
    if jokeId == "" {
        http.Error(w, "Joke ID cannot be empty", http.StatusBadRequest)
        return
    }

    endpoint := "https://icanhazdadjoke.com/j/" + jokeId

    client := http.Client{
        Timeout: 10 * time.Second,
    }

    req, err := http.NewRequest(http.MethodGet, endpoint, nil)
    if err != nil {
        http.Error(w, err.Error(), http.StatusInternalServerError)
        return
    }

    req.Header.Set("Accept", "text/plain")

    resp, err := client.Do(req)
    if err != nil {
        http.Error(w, err.Error(), http.StatusInternalServerError)
        return
    }

    defer resp.Body.Close()

    b, err := ioutil.ReadAll(resp.Body)
    if err != nil {
        http.Error(w, err.Error(), http.StatusInternalServerError)
        return
    }

    if resp.StatusCode != http.StatusOK {
        http.Error(w, string(b), resp.StatusCode)
        return
    }

    w.Header().Set("Content-Type", "text/plain")
    w.WriteHeader(http.StatusOK)
    w.Write(b)
}

func main() {
    mux := http.NewServeMux()
icanhazdadjoke
Astronomers got tired watching the moon go around the earth for 24 hours. They decided to call it a day.
    mux.HandleFunc("/joke", getJoke)

    http.ListenAndServe(":1212", mux)
}

In the code block above, the getJoke handler expects an id query parameter, which it uses to fetch a joke from the Random dad joke API.

Let’s write a test for this handler:

func TestGetJokeHandler(t *testing.T) {
    table := []struct {
        id         string
        statusCode int
        body       string
    }{
        {"R7UfaahVfFd", 200, "My dog used to chase people on a bike a lot. It got so bad I had to take his bike away."},
        {"173782", 404, `Joke with id "173782" not found`},
        {"", 400, "Joke ID cannot be empty"},
    }

    for _, v := range table {
        t.Run(v.id, func(t *testing.T) {
            w := httptest.NewRecorder()
            r := httptest.NewRequest(http.MethodGet, "/joke?id="+v.id, nil)

            getJoke(w, r)

            if w.Code != v.statusCode {
                t.Fatalf("Expected status code: %d, but got: %d", v.statusCode, w.Code)
            }

            body := strings.TrimSpace(w.Body.String())

            if body != v.body {
                t.Fatalf("Expected body to be: '%s', but got: '%s'", v.body, body)
            }
        })
    }
}

We use table driven tests to test the handler against a range of inputs. The first input is a valid Joke ID that should return a 200 OK response. The second is an invalid ID that should return a 404 response. The final input is an empty ID that should return a 400 bad request response.

When you run the test, it should pass successfully:

$ go test -v
=== RUN   TestGetJokeHandler
=== RUN   TestGetJokeHandler/R7UfaahVfFd
=== RUN   TestGetJokeHandler/173782
=== RUN   TestGetJokeHandler/#00
--- PASS: TestGetJokeHandler (1.49s)
    --- PASS: TestGetJokeHandler/R7UfaahVfFd (1.03s)
    --- PASS: TestGetJokeHandler/173782 (0.47s)
    --- PASS: TestGetJokeHandler/#00 (0.00s)
PASS
ok      github.com/ayoisaiah/random     1.498s

Note that the test in the code block above makes HTTP requests to the real API. Doing so affects the dependencies of the code being tested, which is bad practice for unit testing code.

Instead, we should mock the HTTP client. We have several different methods for mocking in Go, which we’ll explore below.

Mocking in Go

A fairly simple pattern for mocking an HTTP client in Go is to create a custom interface. Our interface will define the methods used in a function and pass different implementations depending on where the function is called from.

The custom interface for our HTTP client above should look like the following code block:

type HTTPClient interface {
    Do(req *http.Request) (*http.Response, error)
}

Our signature for getJoke() will look like the code block below:

func getJoke(client HTTPClient) http.HandlerFunc {
    return func(w http.ResponseWriter, r *http.Request) {
      // rest of the function
    }
}

The original body of the getJoke() handler is moved inside of the return value. The client variable declaration is removed from the body in favor of the HTTPClient interface.

The HTTPClient interface wraps a Do() method, which accepts an HTTP request and returns an HTTP response and an error.

We need to provide a concrete implementation of HTTPClient when we call getJoke() in the main() function:

func main() {
    mux := http.NewServeMux()

    client := http.Client{
        Timeout: 10 * time.Second,
    }

    mux.HandleFunc("/joke", getJoke(&client))

    http.ListenAndServe(":1212", mux)
}

The http.Client type implements the HTTPClient interface, so the program continues to call the Random dad joke API. We need to update the tests with a different HTTPClient implementation that does not make HTTP requests over the network.

First, we’ll create a mock implementation of the HTTPClient interface:

type MockClient struct {
    DoFunc func(req *http.Request) (*http.Response, error)
}

func (m *MockClient) Do(req *http.Request) (*http.Response, error) {
    return m.DoFunc(req)
}

In the code block above, the MockClient struct implements the HTTPClient interface through its provision of the Do method, which calls a DoFunc property. Now, we need to implement the DoFunc function when we create an instance of MockClient in the test:

func TestGetJokeHandler(t *testing.T) {
    table := []struct {
        id         string
        statusCode int
        body       string
    }{
        {"R7UfaahVfFd", 200, "My dog used to chase people on a bike a lot. It got so bad I had to take his bike away."},
        {"173782", 404, `Joke with id "173782" not found`},
        {"", 400, "Joke ID cannot be empty"},
    }

    for _, v := range table {
        t.Run(v.id, func(t *testing.T) {
            w := httptest.NewRecorder()
            r := httptest.NewRequest(http.MethodGet, "/joke?id="+v.id, nil)

            c := &MockClient{}

            c.DoFunc = func(req *http.Request) (*http.Response, error) {
                return &http.Response{
                    Body:       io.NopCloser(strings.NewReader(v.body)),
                    StatusCode: v.statusCode,
                }, nil
            }

            getJoke(c)(w, r)

            if w.Code != v.statusCode {
                t.Fatalf("Expected status code: %d, but got: %d", v.statusCode, w.Code)
            }

            body := strings.TrimSpace(w.Body.String())

            if body != v.body {
                t.Fatalf("Expected body to be: '%s', but got: '%s'", v.body, body)
            }
        })
    }
}

In the code snippet above, DoFunc is adjusted for each test case, so it returns a custom response. Now, we’ve avoided all of the network calls, so the test will pass at a much faster rate:

$ go test -v
=== RUN   TestGetJokeHandler
=== RUN   TestGetJokeHandler/R7UfaahVfFd
=== RUN   TestGetJokeHandler/173782
=== RUN   TestGetJokeHandler/#00
--- PASS: TestGetJokeHandler (0.00s)
    --- PASS: TestGetJokeHandler/R7UfaahVfFd (0.00s)
    --- PASS: TestGetJokeHandler/173782 (0.00s)
    --- PASS: TestGetJokeHandler/#00 (0.00s)
PASS
ok      github.com/ayoisaiah/random     0.005s

You can use this same principle when your handler depends on another external system, like a database. Decoupling the handler from any specific implementation allows you to easily mock the dependency in the test while retaining the real implementation in your application’s code.

Using external data in tests

In Go, you should place external data for tests in a directory called testdata. When you build binaries for your programs, the testdata directory is ignored, so you can use this approach to store inputs that you want to test your program against.

For example, let’s write a function that generates the base64 encoding from a binary file:

func getBase64Encoding(b []byte) string {
    return base64.StdEncoding.EncodeToString(b)
}

To test that this function produces the correct output, let’s place some sample files and their corresponding base64 encoding in a testdata directory at the root of our project:

$ ls testdata
img1.jpg img1_base64.txt img2.jpg img2_base64.txt img3.jpg img3_base64.txt

To test our getBase64Encoding() function, run the code below:

func TestGetBase64Encoding(t *testing.T) {
    cases := []string{"img1", "img2", "img3"}

    for _, v := range cases {
        t.Run(v, func(t *testing.T) {
            b, err := os.ReadFile(filepath.Join("testdata", v+".jpg"))
            if err != nil {
                t.Fatal(err)
            }

            expected, err := os.ReadFile(filepath.Join("testdata", v+"_base64.txt"))
            if err != nil {
                t.Fatal(err)
            }

            got := getBase64Encoding(b)

            if string(expected) != got {
                t.Fatalf("Expected output to be: '%s', but got: '%s'", string(expected), got)
            }
        })
    }
}

The bytes for each sample file are read from the file system and then fed into the getBase64Encoding() function. The output is subsequently compared to the expected output, which is also retrieved from the testdata directory.

Let’s make the test easier to maintain by creating a subdirectory inside of testdata. Inside of our subdirectory, we’ll add all of the input files, allowing us to simply iterate over each binary file and compare the actual to the expected output.

Now, we can add more test cases without touching the source code:

$ go test -v
=== RUN   TestGetBase64Encoding
=== RUN   TestGetBase64Encoding/img1
=== RUN   TestGetBase64Encoding/img2
=== RUN   TestGetBase64Encoding/img3
--- PASS: TestGetBase64Encoding (0.04s)
    --- PASS: TestGetBase64Encoding/img1 (0.01s)
    --- PASS: TestGetBase64Encoding/img2 (0.01s)
    --- PASS: TestGetBase64Encoding/img3 (0.01s)
PASS
ok      github.com/ayoisaiah/random     0.044s

Using golden files

If you’re using a Go template, it’s a good idea to test the generated output against the expected output to confirm that the template is working as intended. Go templates are usually large, so it’s not recommended to hard code the expected output in the source code as we’ve done so far in this tutorial.

Let’s explore an alternative approach to Go templates that simplifies writing and maintaining a test throughout a project’s lifecycle.

A golden file is a special type of file that contains the expected output of a test. The test function reads from the golden file, comparing its contents against a test’s expected output.

In the following example, we’ll use an html/template to generate an HTML table that contains a row for each book in an inventory:

type Book struct {
    Name          string
    Author        string
    Publisher     string
    Pages         int
    PublishedYear int
    Price         int
}

var tmpl = `<table class="table">
  <thead>
    <tr>
      <th>Name</th>
      <th>Author</th>
      <th>Publisher</th>
      <th>Pages</th>
      <th>Year</th>
      <th>Price</th>
    </tr>
  </thead>
  <tbody>
    {{ range . }}<tr>
      <td>{{ .Name }}</td>
      <td>{{ .Author }}</td>
      <td>{{ .Publisher }}</td>
      <td>{{ .Pages }}</td>
      <td>{{ .PublishedYear }}</td>
      <td>${{ .Price }}</td>
    </tr>{{ end }}
  </tbody>
</table>
`

var tpl = template.Must(template.New("table").Parse(tmpl))

func generateTable(books []Book, w io.Writer) error {
    return tpl.Execute(w, books)
}

func main() {
    books := []Book{
        {
            Name:          "The Odessa File",
            Author:        "Frederick Forsyth",
            Pages:         334,
            PublishedYear: 1979,
            Publisher:     "Bantam",
            Price:         15,
        },
    }

    err := generateTable(books, os.Stdout)
    if err != nil {
        log.Fatal(err)
    }
}

The generateTable() function above creates the HTML table from a slice of Book objects. The code above will produce the following output:

$ go run main.go
<table class="table">
  <thead>
    <tr>
      <th>Name</th>
      <th>Author</th>
      <th>Publisher</th>
      <th>Pages</th>
      <th>Year</th>
      <th>Price</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td>The Odessa File</td>
      <td>Frederick Forsyth</td>
      <td>Bantam</td>
      <td>334</td>
      <td>1979</td>
      <td>$15</td>
    </tr>
  </tbody>
</table>

To test the function above, we’ll capture the actual outcome and compare it to the expected outcome. We’ll store the expected result in the testdata directory as we did in the previous section, however, we’ll have to make a few changes.

Assume we have the following list of books in an inventory:

var inventory = []Book{
    {
        Name:          "The Solitare Mystery",
        Author:        "Jostein Gaarder",
        Publisher:     "Farrar Straus Giroux",
        Pages:         351,
        PublishedYear: 1990,
        Price:         12,
    },
    {
        Name:          "Also Known As",
        Author:        "Robin Benway",
        Publisher:     "Walker Books",
        Pages:         208,
        PublishedYear: 2013,
        Price:         10,
    },
    {
        Name:          "Ego Is the Enemy",
        Author:        "Ryan Holiday",
        Publisher:     "Portfolio",
        Pages:         226,
        PublishedYear: 2016,
        Price:         18,
    },
}

The expected output for this list of books will span across many lines, therefore, it is difficult to place it as a string literal inside of the source code:

<table class="table">
  <thead>
    <tr>
      <th>Name</th>
      <th>Author</th>
      <th>Publisher</th>
      <th>Pages</th>
      <th>Year</th>
      <th>Price</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td>The Solitaire Mystery</td>
      <td>Jostein Gaarder</td>
      <td>Farrar Straus Giroux</td>
      <td>351</td>
      <td>1990</td>
      <td>$12</td>
    </tr>
    <tr>
      <td>Also Known As</td>
      <td&gt;Robin Benway</td>
      <td>Walker Books</td>
      <td>308</td>
      <td>2013</td>
      <td>$10</td>
    </tr>
    <tr>
      <td>Ego Is The Enemy</td>
      <td>Ryan Holiday</td>
      <td>Portfolio</td>
      <td>226</td>
      <td>2016</td>
      <td>$18</td>
    </tr>
  </tbody>
</table>

In addition to being practical for larger outputs, a golden file can be automatically updated and generated.

While it’s possible to write a helper function to create and update golden files, we can take advantage of goldie, a utility that was created specifically for golden files.

Install the latest version of goldie with the command below:

$ go get -u github.com/sebdah/goldie/v2

Let’s go ahead and use goldie in a test for the generateTable() function:

func TestGenerateTable(t *testing.T) {
    var buf bytes.Buffer

    err := generateTable(inventory, &buf)
    if err != nil {
        t.Fatal(err)
    }

    actual := buf.Bytes()

    g := goldie.New(t)
    g.Assert(t, "books", actual)
}

The test above captures the output of the generateTable() function in a buffer of bytes. Then, it passes the contents of the buffer to the Assert() method on the goldie instance. The contents on the buffer will be compared to the contents of the books.golden file in the testdata directory.

Initially, running the test will fail because we have not created the books.golden file yet:

$ go test -v
=== RUN   TestGenerateTable
    main_test.go:48: Golden fixture not found. Try running with -update flag.
--- FAIL: TestGenerateTable (0.00s)
FAIL
exit status 1
FAIL    github.com/ayoisaiah/random     0.006s

The error message suggests that we add the -update flag, which will create the books.golden file with the contents of the buffer:

$ go test -v -update
=== RUN   TestGenerateTable
--- PASS: TestGenerateTable (0.00s)
PASS
ok      github.com/ayoisaiah/random     0.006s

On subsequent runs, we should remove the -update flag so that our golden file is not continually updated.

Any changes to the template should cause the test to fail. For example, if you update the price field to Euros instead of USD, you’ll immediately receive an error. These errors occur because the output of the generateTable() function no longer matches the contents of the golden file.

Goldie provides diffing capabilities to help you spot the change when these errors occur:

$ go test -v
=== RUN   TestGenerateTable
    main_test.go:48: Result did not match the golden fixture. Diff is below:

        --- Expected
        +++ Actual
        @@ -18,3 +18,3 @@
               <td>1990&lt;/td>
        -      <td>$12</td>
        +      <td>€12</td>
             </tr><tr>
        @@ -25,3 +25,3 @@
               <td>2013</td>
        -      <td>$10</td>
        +      <td>€10</td>
             </tr><tr>
        @@ -32,3 +32,3 @@
               <td>2016</td>
        -      <td>$18</td>
        +      <td>€18</td>
             </tr>

--- FAIL: TestGenerateTable (0.00s)
FAIL
exit status 1
FAIL    github.com/ayoisaiah/random     0.007s

In the output above, the change is clearly highlighted. These changes are deliberate, so we can make our test pass again by updating the golden file using the -update flag:

$ go test -v -update
=== RUN   TestGenerateTable
--- PASS: TestGenerateTable (0.00s)
PASS
ok      github.com/ayoisaiah/random     0.006s

Conclusion

In this tutorial, we looked at some advanced testing techniques in Go. First, we examined our HTTP packages in depth and learned how to mock our HTTP client with a custom interface. Then, we reviewed how to use external data in tests and create golden files using goldie.

I hope you found this post useful. If have any additional techniques you’d like to share, leave a comment below. Thanks for reading, and happy coding!