How to mock a nested client in test - unit-testing

Question

I am building a simple function that calls an API that returns a Post using GraphQL (https://github.com/machinebox/graphql). I wrapped the logic in a service that looks like this:
type Client struct {
gcl graphqlClient
}
type graphqlClient interface {
Run(ctx context.Context, req *graphql.Request, resp interface{}) error
}
func (c *Client) GetPost(id string) (*Post, error) {
req := graphql.NewRequest(`
query($id: String!) {
getPost(id: $id) {
id
title
}
}
`)
req.Var("id", id)
var resp getPostResponse
if err := c.gcl.Run(ctx, req, &resp); err != nil {
return nil, err
}
return resp.Post, nil
}
Now I'd like to add test tables for the GetPost function with a fail case when id is set to empty string which causes an error in the downstream call c.gcl.Run.
What I am struggling with is the way the gcl client can be mocked and forced to return the error (when no real API call happens).
My test so far:
package apiClient
import (
"context"
"errors"
"github.com/aws/aws-sdk-go/aws"
"github.com/google/go-cmp/cmp"
"github.com/machinebox/graphql"
"testing"
)
type graphqlClientMock struct {
graphqlClient
HasError bool
Response interface{}
}
func (g graphqlClientMock) Run(_ context.Context, _ *graphql.Request, response interface{}) error {
if g.HasError {
return errors.New("")
}
response = g.Response
return nil
}
func newTestClient(hasError bool, response interface{}) *Client {
return &Client{
gcl: graphqlClientMock{
HasError: hasError,
Response: response,
},
}
}
func TestClient_GetPost(t *testing.T) {
tt := []struct{
name string
id string
post *Post
hasError bool
response getPostResponse
}{
{
name: "empty id",
id: "",
post: nil,
hasError: true,
},
{
name: "existing post",
id: "123",
post: &Post{id: aws.String("123")},
response: getPostResponse{
Post: &Post{id: aws.String("123")},
},
},
}
for _, tc := range tt {
t.Run(tc.name, func(t *testing.T) {
client := newTestClient(tc.hasError, tc.response)
post, err := client.GetPost(tc.id)
if err != nil {
if tc.hasError == false {
t.Error("unexpected error")
}
} else {
if tc.hasError == true {
t.Error("expected error")
}
if cmp.Equal(post, &tc.post) == false {
t.Errorf("Response data do not match: %s", cmp.Diff(post, tc.post))
}
}
})
}
}
I am not sure if passing the response to the mock like this is the right way to do it. Also, I'm struggling to set the right value to the response, since an interface{} type is passed and I don't know how to convert it to the getPostResponse and set the value to Post there.

Answer 1

Your test cases should not go beyond the implementation. I'm specifically referring to the empty-vs-nonempty input or any kind of input really.
Let's take a look at the code you want to test:
func (c *Client) GetPost(id string) (*Post, error) {
req := graphql.NewRequest(`
query($id: String!) {
getPost(id: $id) {
id
title
}
}
`)
req.Var("id", id)
var resp getPostResponse
if err := c.gcl.Run(ctx, req, &resp); err != nil {
return nil, err
}
return resp.Post, nil
}
Nothing in the implementation above is doing anything based on the id parameter value and therefore nothing in your tests for this piece of code should really care about what input is passed in, if it is irrelevant to the implementation it should also be irrelevant to the tests.
Your GetPost has basically two code branches that are taken based on a single factor, i.e. the "nilness" of the returned err variable. This means that as far as your implementation is concerned there are only two possible outcomes, based on what err value Run returns, and therefore there should only be two test cases, a 3rd or 4th test case would be just a variation, if not an outright copy, of the first two.
Your test client is also doing some unnecessary stuff, the main one being its name, i.e. what you have there is not a mock so calling it that is not helpful. Mocks usually do a lot more than just return predefined values, they ensure that methods are called, in the expected order and with the expected arguments, etc. And actually you don't need a mock here at all so it's a good thing it isn't one.
With that in mind, here's what I would suggest you do with your test client.
type testGraphqlClient struct {
resp interface{} // non-pointer value of the desired response, or nil
err error // the error to be returned by Run, or nil
}
func (g testGraphqlClient) Run(_ context.Context, _ *graphql.Request, resp interface{}) error {
if g.err != nil {
return g.err
}
if g.resp != nil {
// use reflection to set the passed in response value
// (i haven't tested this so there may be a bug or two)
reflect.ValueOf(resp).Elem().Set(reflect.ValueOf(g.resp))
}
return nil
}
... and here are the necessary test cases, all two of them:
func TestClient_GetPost(t *testing.T) {
tests := []struct {
name string
post *Post
err error
client testGraphqlClient
}{{
name: "return error from client",
err: errors.New("bad input"),
client: testGraphqlClient{err: errors.New("bad input")},
}, {
name: "return post from client",
post: &Post{id: aws.String("123")},
client: testGraphqlClient{resp: getPostResponse{Post: &Post{id: aws.String("123")}}},
}}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
client := Client{gql: tt.client}
post, err := client.GetPost("whatever")
if !cmp.Equal(err, tt.err) {
t.Errorf("got error=%v want error=%v", err, tt.err)
}
if !cmp.Equal(post, tt.post) {
t.Errorf("got post=%v want post=%v", post, tt.post)
}
})
}
}
... there's a bit of repetition going on here, the need to spell out the post and err twice but that's a small price to pay when compared to a more sophisticated/complicated test setup that would populate the test client from the test case's expected output fields.
Addendum:
If you were to update GetPost in such a way that it checks for the empty id and returns an error before it sends a request to graphql then your initial setup would make much more sense:
func (c *Client) GetPost(id string) (*Post, error) {
if id == "" {
return nil, errors.New("empty id")
}
req := graphql.NewRequest(`
query($id: String!) {
getPost(id: $id) {
id
title
}
}
`)
req.Var("id", id)
var resp getPostResponse
if err := c.gcl.Run(ctx, req, &resp); err != nil {
return nil, err
}
return resp.Post, nil
}
... and updating the test cases accordingly:
func TestClient_GetPost(t *testing.T) {
tests := []struct {
name string
id string
post *Post
err error
client testGraphqlClient
}{{
name: "return empty id error",
id: "",
err: errors.New("empty id"),
client: testGraphqlClient{},
}, {
name: "return error from client",
id: "nonemptyid",
err: errors.New("bad input"),
client: testGraphqlClient{err: errors.New("bad input")},
}, {
name: "return post from client",
id: "nonemptyid",
post: &Post{id: aws.String("123")},
client: testGraphqlClient{resp: getPostResponse{Post: &Post{id: aws.String("123")}}},
}}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
client := Client{gql: tt.client}
post, err := client.GetPost(tt.id)
if !cmp.Equal(err, tt.err) {
t.Errorf("got error=%v want error=%v", err, tt.err)
}
if !cmp.Equal(post, tt.post) {
t.Errorf("got post=%v want post=%v", post, tt.post)
}
})
}
}