Mockito.spy(List.class) 和 Mockito.spy(new ArrayList()) 有什么区别？

Question

The first test case is failed due to IndexOutOfBoundsException. The second one is successful. What is difference between Mockito.spy(List.class) and Mockito.spy(new ArrayList())?

@Test
public void thenReturnWithSpy1() {

    List<String> list = Mockito.spy(List.class);

    when(list.get(0)).thenReturn("abc");

    assertEquals("abc", list.get(0));
}

@Test
public void thenReturnWithSpy2() {

    List<String> list = Mockito.spy(new ArrayList());

    when(list.get(0)).thenReturn("abc");

    assertEquals("abc", list.get(0));
}

Answer 1

Mockito spies delegate calls to the actual underlying object. This is true even when stubbing. So :

when(list.get(0)).thenReturn("abc");

already does an invocation on the spy.

new ArrayList() is an empty list. Calling list.get(0) on it, even during stubbing, will result in a IndexOutOfBoundsException .

When you use a spy on a class, there is no actual instance to delegate to, and as List is an interface with no implementation for get() , Mockito will not delegate to any 'actual code'.

So, it's the stubbing that fails.

There is an easy solution, though. By using the doReturn() style of stubbing you avoid actual invocations during stubbing, at the expense of reading a little less fluently.

doReturn("abc").when(list).get(0);