CompletableFuture以异步执行多个数据库查询

Posted 2021-05-05

我想并行执行多个数据库查询，并将结果存储在映射中。我正在尝试这样做，但是访问地图时，地图并未完全填充。

我做错什么了吗？

 public Map<MapKeyEnums, Set<String>> doDBCalls(String phoneNumber, long timestamp) {

         Map<MapKeyEnums, Set<String>> instrumentsEdgesMap = new EnumMap<>(MapKeyEnums.class);

         CompletableFuture.supplyAsync(() -> dbReadService.getCall(phoneNumber, PhoneNumber.class, "ABC", timestamp)).
                    thenApply(x -> instrumentsEdgesMap.put(MapKeyEnums.ABC, x));

         CompletableFuture.supplyAsync(() -> dbReadService.getCall(phoneNumber, PhoneNumber.class, "XYZ", timestamp)).
                    thenApply(x -> instrumentsEdgesMap.put(MapKeyEnums.XYZ, x));

         CompletableFuture.supplyAsync(() -> dbReadService.getCall(phoneNumber, PhoneNumber.class, "DEF", timestamp)).
                    thenApply(x -> instrumentsEdgesMap.put(MapKeyEnums.DEF, x));

         return instrumentsEdgesMap;

}

感谢您的任何帮助，在此先感谢。

答案

返回结果之前，您必须等待期货完成。

尝试类似的东西

    public Map<MapKeyEnums, Set<String>> doDBCalls(String phoneNumber, long timestamp) {

        Map<MapKeyEnums, Set<String>> instrumentsEdgesMap = new EnumMap<>(MapKeyEnums.class);

        CompletableFuture.allOf(
            CompletableFuture.supplyAsync(() -> dbReadService.getCall(phoneNumber, PhoneNumber.class, "ABC", timestamp))
                .thenAccept(x -> instrumentsEdgesMap.put(MapKeyEnums.ABC, x)),

            CompletableFuture.supplyAsync(() -> dbReadService.getCall(phoneNumber, PhoneNumber.class, "XYZ", timestamp))
                .thenAccept(x -> instrumentsEdgesMap.put(MapKeyEnums.XYZ, x)),

            CompletableFuture.supplyAsync(() -> dbReadService.getCall(phoneNumber, PhoneNumber.class, "DEF", timestamp))
                .thenAccept(x -> instrumentsEdgesMap.put(MapKeyEnums.DEF, x)))
        .get();

        return instrumentsEdgesMap;
    }

另一答案

在上述方法中，supplyAsync将由Async中的ForkJoinPool线程执行，但thenApply方法总是通过调用完全不异步的线程来执行

所有没有显式Executor参数的异步方法都是使用ForkJoinPool.commonPool（）执行的（除非它不支持并行度至少为2，在这种情况下，将创建一个新的Thread来运行每个任务）。

这里是示例

CompletableFuture.supplyAsync(()->{
        System.out.println(Thread.currentThread().getName());
        return "SupplyAsync";
    }).thenAccept(i->{
    System.out.println(Thread.currentThread().getName()+"--"+i);
    });

输出：

ForkJoinPool.commonPool-worker-3
main--SupplyAsync

因此，如果您希望您的进程为Async，则首先使用supplyAsync触发所有三个数据库查询，并在CompletableFuture中捕获输出

CompletableFuture<Set<String>> first =  CompletableFuture.supplyAsync(() -> dbReadService.getCall(phoneNumber, PhoneNumber.class, "ABC", timestamp));

CompletableFuture<Set<String>> second =  CompletableFuture.supplyAsync(() -> dbReadService.getCall(phoneNumber, PhoneNumber.class, "XYZ", timestamp));

CompletableFuture<Set<String>> third =  CompletableFuture.supplyAsync(() -> dbReadService.getCall(phoneNumber, PhoneNumber.class, "DEF", timestamp));

然后现在使用其中三个创建流，然后将它们收集到Map

Stream.of(new AbstractMap.SimpleEntry<MapKeyEnums, CompletableFuture<Set<String>>>(MapKeyEnums.ABC, first),
              new AbstractMap.SimpleEntry<MapKeyEnums, CompletableFuture<Set<String>>>(MapKeyEnums.XYZ, second),
              new AbstractMap.SimpleEntry<MapKeyEnums, CompletableFuture<Set<String>>>(MapKeyEnums.DEF, third))
       .forEach(entry->{
           entry.getValue().thenAccept(val-> instrumentsEdgesMap.put(entry.getKey(), val));
       });