使用 CompletableFuture 分批处理任务

一、无返回值任务函数

// 数据分批
List<List<StatisticsDTO>> batches = Lists.partition(statisticsList, BATCH_SIZE);
List<CompletableFuture<Void>> futures = new ArrayList<>(batches.size());

// 数据处理
for (int i = 0; i < batches.size(); i++) {
   
    logger.info("批次 " + i + " 开始处理...");
    String logId = LogIdThreadLocal.getLogId();  // 传递主线程的 logId
    List<StatisticsDTO> batchData = batches.get(i);
    CompletableFuture<Void> future = CompletableFuture.runAsync(() -> {
   
        try {
   
            LogIdThreadLocal.setLogId(logId);
            processBatch(batchData);
        } finally {
   
            LogIdThreadLocal.clean();
        }
    });
    futures.add(future);
}

// 等待所有的异步任务完成
CompletableFuture<Void> allOf = CompletableFuture.allOf(futures.toArray(new CompletableFuture[0]));
allOf.join();

二、带返回值任务函数

// 数据分批
List<List<StatisticsDTO>> batches = Lists.partition(statisticsList, BATCH_SIZE);
List<CompletableFuture<List<StatisticsDTO>>> futures = new ArrayList<>(batches.size());

// 数据处理
for (int i = 0; i < batches.size(); i++) {
   
    logger.info("批次 " + i + " 开始处理...");
    String logId = LogIdThreadLocal.getLogId();  // 传递主线程的 logId
    List<StatisticsDTO> batchData = batches.get(i);
    CompletableFuture<List<DoctorAvatarAnalysisDTO>> future = CompletableFuture.supplyAsync(() -> {
   
        try {
   
            LogIdThreadLocal.setLogId(logId);
            return processBatch(batchData);
        } finally {
   
            LogIdThreadLocal.clean();
        }
    });
    futures.add(future);
}

// 等待所有 CF 完成并合并结果
CompletableFuture<Void> allOf = CompletableFuture.allOf(futures.toArray(new CompletableFuture[0]));
List<StatisticsDTO> result = allOf.thenApply(
        v -> futures.stream().map(CompletableFuture::join).flatMap(List::stream).collect(Collectors.toList())
).join();