Try-catch是用来帮助异常处理的。这意味着它将在某种程度上帮助我们的系统变得更加健壮:尝试从意外事件中恢复。

我们怀疑在执行和指令(发送消息)时可能会发生一些事情,因此它被包含在try中。如果发生了几乎意想不到的事情,我们可以做些什么:我们编写捕获。我不认为我们调用只是为了记录异常。我认为catch块是为了给我们从错误中恢复的机会。

现在,假设我们从错误中恢复,因为我们可以修复错误。如果能再试一次,那就太好了:

try{ some_instruction(); }
catch (NearlyUnexpectedException e){
   fix_the_problem();
   retry;
}

这将很快陷入永恒循环,但让我们假设fix_the_problem返回true,然后重试。假设在Java中没有这样的东西,你将如何解决这个问题?你认为解决这个问题的最佳设计代码是什么?

这就像一个哲学问题,因为我已经知道Java不直接支持我所要求的内容。


当前回答

这是一个老问题,但解决方案仍然重要。以下是我在Java 8中不使用任何第三方库的通用解决方案:

public interface RetryConsumer<T> {
    T evaluate() throws Throwable;
}
public interface RetryPredicate<T> {
    boolean shouldRetry(T t);
}
public class RetryOperation<T> {
    private RetryConsumer<T> retryConsumer;
    private int noOfRetry;
    private int delayInterval;
    private TimeUnit timeUnit;
    private RetryPredicate<T> retryPredicate;
    private List<Class<? extends Throwable>> exceptionList;

    public static class OperationBuilder<T> {
        private RetryConsumer<T> iRetryConsumer;
        private int iNoOfRetry;
        private int iDelayInterval;
        private TimeUnit iTimeUnit;
        private RetryPredicate<T> iRetryPredicate;
        private Class<? extends Throwable>[] exceptionClasses;

        private OperationBuilder() {
        }

        public OperationBuilder<T> retryConsumer(final RetryConsumer<T> retryConsumer) {
            this.iRetryConsumer = retryConsumer;
            return this;
        }

        public OperationBuilder<T> noOfRetry(final int noOfRetry) {
            this.iNoOfRetry = noOfRetry;
            return this;
        }

        public OperationBuilder<T> delayInterval(final int delayInterval, final TimeUnit timeUnit) {
            this.iDelayInterval = delayInterval;
            this.iTimeUnit = timeUnit;
            return this;
        }

        public OperationBuilder<T> retryPredicate(final RetryPredicate<T> retryPredicate) {
            this.iRetryPredicate = retryPredicate;
            return this;
        }

        @SafeVarargs
        public final OperationBuilder<T> retryOn(final Class<? extends Throwable>... exceptionClasses) {
            this.exceptionClasses = exceptionClasses;
            return this;
        }

        public RetryOperation<T> build() {
            if (Objects.isNull(iRetryConsumer)) {
                throw new RuntimeException("'#retryConsumer:RetryConsumer<T>' not set");
            }

            List<Class<? extends Throwable>> exceptionList = new ArrayList<>();
            if (Objects.nonNull(exceptionClasses) && exceptionClasses.length > 0) {
                exceptionList = Arrays.asList(exceptionClasses);
            }
            iNoOfRetry = iNoOfRetry == 0 ? 1 : 0;
            iTimeUnit = Objects.isNull(iTimeUnit) ? TimeUnit.MILLISECONDS : iTimeUnit;
            return new RetryOperation<>(iRetryConsumer, iNoOfRetry, iDelayInterval, iTimeUnit, iRetryPredicate, exceptionList);
        }
    }

    public static <T> OperationBuilder<T> newBuilder() {
        return new OperationBuilder<>();
    }

    private RetryOperation(RetryConsumer<T> retryConsumer, int noOfRetry, int delayInterval, TimeUnit timeUnit,
                           RetryPredicate<T> retryPredicate, List<Class<? extends Throwable>> exceptionList) {
        this.retryConsumer = retryConsumer;
        this.noOfRetry = noOfRetry;
        this.delayInterval = delayInterval;
        this.timeUnit = timeUnit;
        this.retryPredicate = retryPredicate;
        this.exceptionList = exceptionList;
    }

    public T retry() throws Throwable {
        T result = null;
        int retries = 0;
        while (retries < noOfRetry) {
            try {
                result = retryConsumer.evaluate();
                if (Objects.nonNull(retryPredicate)) {
                    boolean shouldItRetry = retryPredicate.shouldRetry(result);
                    if (shouldItRetry) {
                        retries = increaseRetryCountAndSleep(retries);
                    } else {
                        return result;
                    }
                } else {
                    // no retry condition defined, no exception thrown. This is the desired result.
                    return result;
                }
            } catch (Throwable e) {
                retries = handleException(retries, e);
            }
        }
        return result;
    }

    private int handleException(int retries, Throwable e) throws Throwable {
        if (exceptionList.contains(e.getClass()) || (exceptionList.isEmpty())) {
            // exception is excepted, continue retry.
            retries = increaseRetryCountAndSleep(retries);
            if (retries == noOfRetry) {
                // evaluation is throwing exception, no more retry left. Throw it.
                throw e;
            }
        } else {
            // unexpected exception, no retry required. Throw it.
            throw e;
        }
        return retries;
    }

    private int increaseRetryCountAndSleep(int retries) {
        retries++;
        if (retries < noOfRetry && delayInterval > 0) {
            try {
                timeUnit.sleep(delayInterval);
            } catch (InterruptedException ignore) {
                Thread.currentThread().interrupt();
            }
        }
        return retries;
    }
}

让我们有一个这样的测试用例:

@Test
public void withPredicateAndException() {
    AtomicInteger integer = new AtomicInteger();
    try {
        Integer result = RetryOperation.<Integer>newBuilder()
                .retryConsumer(() -> {
                    int i = integer.incrementAndGet();
                    if (i % 2 == 1) {
                        throw new NumberFormatException("Very odd exception");
                    } else {
                        return i;
                    }
                })
                .noOfRetry(10)
                .delayInterval(10, TimeUnit.MILLISECONDS)
                .retryPredicate(value -> value <= 6)
                .retryOn(NumberFormatException.class, EOFException.class)
                .build()
                .retry();
        Assert.assertEquals(8, result.intValue());
    } catch (Throwable throwable) {
        Assert.fail();
    }
}

其他回答

解决这个问题的一个简单方法是在while循环中包装try/catch并维护一个计数。通过这种方式,您可以根据其他变量检查计数,同时维护故障日志,从而防止出现无限循环。这不是最精致的解决方案,但它是可行的。

下面是Java 8+的可重用和更通用的方法,不需要外部库:

public interface IUnreliable<T extends Exception>
{
    void tryRun ( ) throws T;
}

public static <T extends Exception> void retry (int retryCount, IUnreliable<T> runnable) throws T {
    for (int retries = 0;; retries++) {
        try {
            runnable.tryRun();
            return;
        } catch (Exception e) {
            if (retries < retryCount) {
                continue;
            } else {
                throw e;
            }
        }
    }
}

用法:

@Test
public void demo() throws IOException {
    retry(3, () -> {
        new File("/tmp/test.txt").createNewFile();
    });
}

通常,最好的设计取决于特定的环境。通常我会这样写:

for (int retries = 0;; retries++) {
    try {
        return doSomething();
    } catch (SomeException e) {
        if (retries < 6) {
            continue;
        } else {
            throw e;
        }
    }
}

Spring AOP和基于注释的解决方案:

用法(@RetryOperation是作业的自定义注释):

@RetryOperation(retryCount = 1, waitSeconds = 10)
boolean someMethod() throws Exception {
}

要做到这一点,我们需要做两件事:1。注释接口;一个春天的方面。下面是实现这些的一种方法:

标注接口:

import java.lang.annotation.*;

@Target(ElementType.METHOD)
@Retention(RetentionPolicy.RUNTIME)
public @interface RetryOperation {
    int retryCount();
    int waitSeconds();
}

春季相位:

import org.aspectj.lang.ProceedingJoinPoint;
import org.aspectj.lang.annotation.Around;
import org.aspectj.lang.annotation.Aspect;
import org.aspectj.lang.reflect.MethodSignature;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.stereotype.Component;
import java.lang.reflect.Method;

@Aspect @Component 
public class RetryAspect {

    private static final Logger LOGGER = LoggerFactory.getLogger(RetryAspect.class);

    @Around(value = "@annotation(RetryOperation)")
    public Object retryOperation(ProceedingJoinPoint joinPoint) throws Throwable {

        Object response = null;
        Method method = ((MethodSignature) joinPoint.getSignature()).getMethod();
        RetryOperation annotation = method.getAnnotation(RetryOperation.class);
        int retryCount = annotation.retryCount();
        int waitSeconds = annotation.waitSeconds();
        boolean successful = false;

        do {
            try {
                response = joinPoint.proceed();
                successful = true;
            } catch (Exception ex) {
                LOGGER.info("Operation failed, retries remaining: {}", retryCount);
                retryCount--;
                if (retryCount < 0) {
                    throw ex;
                }
                if (waitSeconds > 0) {
                    LOGGER.info("Waiting for {} second(s) before next retry", waitSeconds);
                    Thread.sleep(waitSeconds * 1000l);
                }
            }
        } while (!successful);

        return response;
    }
}

下面的代码段执行一些代码段。如果在执行代码段时出现任何错误,请休眠M毫秒并重试。参考链接。

public void retryAndExecuteErrorProneCode(int noOfTimesToRetry, CodeSnippet codeSnippet, int sleepTimeInMillis)
  throws InterruptedException {

 int currentExecutionCount = 0;
 boolean codeExecuted = false;

 while (currentExecutionCount < noOfTimesToRetry) {
  try {
   codeSnippet.errorProneCode();
   System.out.println("Code executed successfully!!!!");
   codeExecuted = true;
   break;
  } catch (Exception e) {
   // Retry after 100 milliseconds
   TimeUnit.MILLISECONDS.sleep(sleepTimeInMillis);
   System.out.println(e.getMessage());
  } finally {
   currentExecutionCount++;
  }
 }

 if (!codeExecuted)
  throw new RuntimeException("Can't execute the code within given retries : " + noOfTimesToRetry);
}