How can I unit test Google App Engine Go HTTP handlers?

Local unit testing is supported from version 1.8.6 of the Google App Engine Go SDK. The appengine/aetest package allows me to create a Context to unit test with.

How can I use this with net/http/httptest to test my HTTP handlers?

See the top of goroot/src/pkg/appengine/aetest/context.go (updated source is not yet posted at https://code.google.com/p/appengine-go). At first glance, the new testing app looks to be a slightly beefier/different version of appenginetesting so you can do the same sorts of tests, see here for one way to do it with how sampleHandler(w http.ResponseWriter, r *http.Request) is called.

Alternatively, you can make your http.Handler's ContextHandler like as below:

type ContextHandler struct {
	Real func(*appengine.Context, http.ResponseWriter, *http.Request)
}

func (f ContextHandler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
	c := appengine.NewContext(r)
	f.Real(c, w, r)
}

func myNewHandler(c appengine.Context, w http.ResponseWriter, r *http.Request) {
// do something
}

Then you can do this in init() to support production:

http.Handle("/myNewHandler", ContextHandler{myNewHandler})

This makes testing the function easy:

func TestMyNewHandler(t *testing.T) {
    c := aetest.NewContext()
    r, _ := http.NewRequest("GET", "/tasks/findOverdueSchedules", nil)
    w := httptest.NewRecorder()
    myNewHandler(c, w, r)
    if 200 != w.Code {
        t.Fail()
    }
}

Here's what's from context.go inside appengine/aetest:
>/*
Package aetest provides an appengine.Context for use in tests.

>An example test file:
package foo_test

import (
	"testing"

	"appengine/memcache"
	"appengine/aetest"
)

func TestFoo(t *testing.T) {
	c, err := aetest.NewContext(nil)
	if err != nil {
		t.Fatal(err)
	}
	defer c.Close()

	it := &memcache.Item{
		Key:   "some-key",
		Value: []byte("some-value"),
	}
	err = memcache.Set(c, it)
	if err != nil {
		t.Fatalf("Set err: %v", err)
	}
	it, err = memcache.Get(c, "some-key")
	if err != nil {
		t.Fatalf("Get err: %v; want no error", err)
	}
	if g, w := string(it.Value), "some-value" ; g != w {
		t.Errorf("retrieved Item.Value = %q, want %q", g, w)
	}
}

>The environment variable APPENGINE_API_SERVER specifies the location of the
api_server.py executable to use. If unset, the system PATH is consulted.
*/

If you are not opposed to using Martini, dependency injection is a nice way to solve the problem. Here is how you can set up your test (with Ginkgo):

var _ = Describe("Items", func() {

    var (
	    m *martini.ClassicMartini
    )

    BeforeEach(func() {
	    m = martini.Classic()

	    // injects app engine context into requests
	    m.Use(func(c martini.Context, req *http.Request) {
		     con, _ := aetest.NewContext(nil)
		     c.MapTo(con, (*appengine.Context)(nil))
	    })

	    m.Get("/items", func(c martini.Context){
             // code you want to test
        })
    })

	It("should get items", func() {
		recorder := httptest.NewRecorder()
		r, _ := http.NewRequest("GET", "/items", nil)
		m.ServeHTTP(recorder, r) // martini server used
		Expect(recorder.Code).To(Equal(200))
	})
})

In the production the actual context would be injected:

m.Use(func(c martini.Context, req *http.Request) {
	c.MapTo(appengine.NewContext(req), (*appengine.Context)(nil))
})