通过添加Appender进行单元测试并不能真正测试Logger的配置。因此，我认为这是一种独特的情况，在这种情况下，单元测试没有带来那么多价值，而集成测试带来了很多价值(特别是如果您的日志有一些审计目的)。

为了为它创建集成测试，让我们假设您正在运行一个简单的ConsoleAppender，并希望测试它的输出。然后，您应该测试如何将消息从System.out写入到它自己的ByteArrayOutputStream。

从这个意义上说，我会做以下事情(我使用JUnit 5):

public class Slf4jAuditLoggerTest {

    private final ByteArrayOutputStream outContent = new ByteArrayOutputStream();

    @BeforeEach
    public void beforeEach() {
        System.setOut(new PrintStream(outContent));
    }

通过这种方式，你可以简单地测试它的输出:

    @Test
    public void myTest() {
        // Given...
        // When...
        // Then
        assertTrue(outContent.toString().contains("[INFO] My formatted string from Logger"));
    }

如果你这样做了，你将为你的项目带来更多的价值，而不需要使用内存中的实现，创建一个新的Appender，或者其他什么。

模拟Appender可以帮助捕获日志行。 在http://clearqa.blogspot.co.uk/2016/12/test-log-lines.html上找到示例

// Fully working test at: https://github.com/njaiswal/logLineTester/blob/master/src/test/java/com/nj/Utils/UtilsTest.java

@Test
public void testUtilsLog() throws InterruptedException {

    Logger utilsLogger = (Logger) LoggerFactory.getLogger("com.nj.utils");

    final Appender mockAppender = mock(Appender.class);
    when(mockAppender.getName()).thenReturn("MOCK");
    utilsLogger.addAppender(mockAppender);

    final List<String> capturedLogs = Collections.synchronizedList(new ArrayList<>());
    final CountDownLatch latch = new CountDownLatch(3);

    //Capture logs
    doAnswer((invocation) -> {
        LoggingEvent loggingEvent = invocation.getArgumentAt(0, LoggingEvent.class);
        capturedLogs.add(loggingEvent.getFormattedMessage());
        latch.countDown();
        return null;
    }).when(mockAppender).doAppend(any());

    //Call method which will do logging to be tested
    Application.main(null);

    //Wait 5 seconds for latch to be true. That means 3 log lines were logged
    assertThat(latch.await(5L, TimeUnit.SECONDS), is(true));

    //Now assert the captured logs
    assertThat(capturedLogs, hasItem(containsString("One")));
    assertThat(capturedLogs, hasItem(containsString("Two")));
    assertThat(capturedLogs, hasItem(containsString("Three")));
}

在我的案例中，我解决了如下的问题:

Logger root = (Logger) LoggerFactory.getLogger(CSVTasklet.class); //CSVTasklet is my target class
    final Appender mockAppender = mock(Appender.class);
    root.addAppender(mockAppender); 

verify(mockAppender).doAppend(argThat((ArgumentMatcher) argument -> ((LoggingEvent) argument).getMessage().contains("No projects."))); // I checked "No projects." in the log

正如前面提到的，您可以使用mock框架。为此，您必须在类中公开记录器(尽管我可能更倾向于使其包私有而不是创建公共setter)。

另一种解决方案是手工创建一个假记录器。您必须编写伪记录器(更多的fixture代码)，但在这种情况下，我更喜欢使用模拟框架中保存的代码来增强测试的可读性。

我会这样做:

class FakeLogger implements ILogger {
    public List<String> infos = new ArrayList<String>();
    public List<String> errors = new ArrayList<String>();

    public void info(String message) {
        infos.add(message);
    }

    public void error(String message) {
        errors.add(message);
    }
}

class TestMyClass {
    private MyClass myClass;        
    private FakeLogger logger;        

    @Before
    public void setUp() throws Exception {
        myClass = new MyClass();
        logger = new FakeLogger();
        myClass.logger = logger;
    }

    @Test
    public void testMyMethod() {
        myClass.myMethod(true);

        assertEquals(1, logger.infos.size());
    }
}

请注意，在Log4J 2中。X，公共接口org.apache.logging.log4j。Logger不包括setAppender()和removeAppender()方法。

但是，如果您没有做太花哨的事情，您应该能够将它强制转换为实现类org.apache.logging.log4j.core。记录器，它公开了这些方法。

这里有一个Mockito和AssertJ的例子:

// Import the implementation class rather than the API interface
import org.apache.logging.log4j.core.Logger;

// Cast logger to implementation class to get access to setAppender/removeAppender
Logger log = (Logger) LogManager.getLogger(MyClassUnderTest.class);

// Set up the mock appender, stubbing some methods Log4J needs internally
Appender appender = mock(Appender.class);
when(appender.getName()).thenReturn("Mock Appender");
when(appender.isStarted()).thenReturn(true);

log.addAppender(appender);
try {
    new MyClassUnderTest().doSomethingThatShouldLogAnError();
} finally {
    log.removeAppender(appender);
}

// Verify that we got an error with the expected message
ArgumentCaptor<LogEvent> logEventCaptor = ArgumentCaptor.forClass(LogEvent.class);
verify(appender).append(logEventCaptor.capture());
LogEvent logEvent = logEventCaptor.getValue();
assertThat(logEvent.getLevel()).isEqualTo(Level.ERROR);
assertThat(logEvent.getMessage().getFormattedMessage()).contains(expectedErrorMessage);

对于我来说，您可以通过使用JUnit和Mockito来简化您的测试。 我提出以下解决方案:

import org.apache.log4j.Appender;
import org.apache.log4j.Level;
import org.apache.log4j.LogManager;
import org.apache.log4j.spi.LoggingEvent;
import org.junit.After;
import org.junit.Before;
import org.junit.Test;
import org.junit.runner.RunWith;
import org.mockito.ArgumentCaptor;
import org.mockito.Captor;
import org.mockito.InjectMocks;
import org.mockito.Mock;
import org.mockito.runners.MockitoJUnitRunner;

import java.util.List;

import static org.assertj.core.api.Assertions.assertThat;
import static org.assertj.core.api.Assertions.tuple;
import static org.mockito.Mockito.times;

@RunWith(MockitoJUnitRunner.class)
public class MyLogTest {
    private static final String FIRST_MESSAGE = "First message";
    private static final String SECOND_MESSAGE = "Second message";
    @Mock private Appender appender;
    @Captor private ArgumentCaptor<LoggingEvent> captor;
    @InjectMocks private MyLog;

    @Before
    public void setUp() {
        LogManager.getRootLogger().addAppender(appender);
    }

    @After
    public void tearDown() {
        LogManager.getRootLogger().removeAppender(appender);
    }

    @Test
    public void shouldLogExactlyTwoMessages() {
        testedClass.foo();

        then(appender).should(times(2)).doAppend(captor.capture());
        List<LoggingEvent> loggingEvents = captor.getAllValues();
        assertThat(loggingEvents).extracting("level", "renderedMessage").containsExactly(
                tuple(Level.INFO, FIRST_MESSAGE)
                tuple(Level.INFO, SECOND_MESSAGE)
        );
    }
}

这就是为什么我们对不同消息量的测试有很好的灵活性