Golang无法在同一程序包中进行测试

Having issue creating unit test of one of my source file ( commonutil.go )

package util

import "github.com/nu7hatch/gouuid"

// GenerateUUID Returns generated UUID sequence
func GenerateUniqueID(hostname string) (string, error) {
    var result, err = uuid.NewV4()
    return hostname + ":" + result.String(), err
}

For the above source, I created the test file "commonutil_test.go" ( in the same package )

package util

import "testing" 
func TestCommonUtil(t *testing.T) {
t.Run("Test generate UUID", func(t *testing.T) {

    var uuid, _ = GenerateUniqueID ("test")
    //fmt.Printf("UUID isa %v 
", uuid)
    if uuid == "" {
        t.Errorf("UUID expected, but result is empty ")
    }
})

However when trying executing "go test util/commonutil_test.go" it shows :

util\commonutil_test.go:8: undefined: GenerateUniqueID
FAIL command-line-arguments [build failed]

Changing to util.GenerateUniqueID in the test solve the problem, however when running coverage using Goconvey will cause build failure :

can't load package: import cycle not allowed package rudygunawan.com/MyProject/HFLC-Go/util imports rudygunawan.com/MyProject/HFLC-Go/util

Any idea to solve this issue? I am confused.

Go version is go1.7.1 windows/386

Just realize it is a silly mistake. The package for the test should be "util_test". Putting the test in the separate package ( but still in the same folder) help solve import cycle issue, yet still allow to solve the undefined error.

The way I normally write Go unit tests is to have one (or more) ..._test.go files, in the same package as the code being tested, with one Test... function for each broad set of tests to be done.

package util

import "testing

func TestGenerateUniqueID(t *testing.T) {
   var uuid1, uuid2 string
   uuid1, err = GenerateUniqueID("test")
   if err != nil {
     t.Errorf("Expected no error, got %s", err) // Maybe Fatalf?
   }
   if uuid1 == "" {
     t.Errorf("Expected non-empty string, got empty string (uuid1)")
   }
   uuid2, err = GenerateUniqueID("test")
   if err != nil {
     t.Errorf("Expected no error, got %s", err) // Maybe Fatalf?
   }
   if uuid2 == "" {
     t.Errorf("Expected non-empty string, got empty string (uuid2)")
   }
   if uuid1 == uuid2 {
     t.Errorf("Expected uuid1 and uuid2 to be different, both are %s", uuid1)
   }
}

One of the reasons I tend towards whitebox testing (where I can do "blackbox testing" by carefully not accessing package internals) is that there's usually a whole slew of non-exported code that really should be tested as well. In this specific, small, example, there's no massive argument for one over the other, since all the functionality that can be tested is already exported.