是否有一种方法可以让存根方法在后续调用时返回不同的对象?我希望这样做是为了测试来自ExecutorCompletionService的不确定响应。也就是说,测试不管方法的返回顺序如何,结果都保持不变。

我要测试的代码看起来像这样。

// Create an completion service so we can group these tasks together
ExecutorCompletionService<T> completionService =
        new ExecutorCompletionService<T>(service);

// Add all these tasks to the completion service
for (Callable<T> t : ts)
    completionService.submit(request);

// As an when each call finished, add it to the response set.
for (int i = 0; i < calls.size(); i ++) {
    try {
        T t = completionService.take().get();
        // do some stuff that I want to test
    } catch (...) { }        
}

当前回答

我已经实现了一个MultipleAnswer类,它可以帮助我在每个调用中存根不同的答案。下面是一段代码:

private final class MultipleAnswer<T> implements Answer<T> {

    private final ArrayList<Answer<T>> mAnswers;

    MultipleAnswer(Answer<T>... answer) {
        mAnswers = new ArrayList<>();
        mAnswers.addAll(Arrays.asList(answer));
    }

    @Override
    public T answer(InvocationOnMock invocation) throws Throwable {
        return mAnswers.remove(0).answer(invocation);
    }
}

其他回答

BDD风格:

import static org.mockito.BDDMockito.given;
        ...

        given(yourMock.yourMethod()).willReturn(1, 2, 3);

经典的风格:

import static org.mockito.Mockito.when;
        ...

        when(yourMock.yourMethod()).thenReturn(1, 2, 3);

明确的风格:

        ...

        when(yourMock.yourMethod())
            .thenReturn(1)
            .thenReturn(2)
            .thenReturn(3);

取决于参数

假设我们有2个参数,检查第2个(列表)参数的大小:

        ...

        when(yourMock.yourMethod(any(), anyList()))
             .thenAnswer(args -> ((List) args.getArgument(1)).size() < 2
                                 ? 1
                                 : 3);

arg是对象,所以我们必须将arg转换为我们的类型。在我的例子中,我转换了^^^ to (List)。

BDD

        ...

        given(yourMock.yourMethod(any(), anyList()))
             .willAnswer(args -> ((List) args.getArgument(1)).size() < 2
                                 ? 1
                                 : 3);

你可以使用thenAnswer方法(当与when链接时):

when(someMock.someMethod()).thenAnswer(new Answer() {
    private int count = 0;

    public Object answer(InvocationOnMock invocation) {
        if (count++ == 1)
            return 1;

        return 2;
    }
});

或者使用等效的静态doAnswer方法:

doAnswer(new Answer() {
    private int count = 0;

    public Object answer(InvocationOnMock invocation) {
        if (count++ == 1)
            return 1;

        return 2;
    }
}).when(someMock).someMethod();

正如前面指出的,几乎所有的调用都是可链的。

所以你可以打电话

when(mock.method()).thenReturn(foo).thenReturn(bar).thenThrow(new Exception("test"));

//OR if you're mocking a void method and/or using spy instead of mock

doReturn(foo).doReturn(bar).doThrow(new Exception("Test").when(mock).method();

更多信息在Mockito的文档。

这可能是基本的/明显的,但如果像我一样,你试图模拟一个方法的多次调用,每次调用要测试的方法被调用未知次数,例如:

public String method(String testArg) {
    //...
    while(condition) {
        someValue = someBean.nestedMethod(); // This is called unknown number of times
        //...
    }
    //...
}

你可以这样做:

@Test
public void testMethod() {
    mockNestedMethodForValue("value1");
    assertEquals(method("arg"), "expected1");
    mockNestedMethodForValue("value2");
    assertEquals(method("arg"), "expected2");
    mockNestedMethodForValue("value3");
    assertEquals(method("arg"), "expected3");
}

private void mockNestedMethodForValue(String value) {
    doReturn(value).when(someBeanMock).nestedMethod();
}

doReturn(value1, value2, value3)。When(方法调用)