如何从 hazelcast map 中获取随机键值

Question

I tried to get a random element from the Map like this

IMap<Integer, Integer> workWaitTasks = hazelcastInstance.getMap(WORK_WAIT_TASKS);
collectionTask = Collections.singleton(workWaitTasks.values().stream()
                .skip(workWaitTasks.isEmpty() ? 0 : new Random().nextInt(workWaitTasks.size()))
                .findFirst()
                .get());
int taskId = collectionTask.iterator().next();

but the best way I think is using predicates

I read this https://docs.hazelcast.com/imdg/4.2/query/how-distributed-query-works.html#querying-with-sql-like-predicates Unfortunately this didn't help me

I couldn't find a way to do this

in sql it is like this

 SELECT column FROM table
 ORDER BY RAND()
 LIMIT 1

how to make correct predicate in hazelcast?

Can you give an example please? Help me please

Answer 1

I think there's no straightforward and effective way to do this using public API. One option is using Jet:

IMap<Object, Object> sourceMap = instance.getMap("table");
IList<Object> targetList = instance.getList("result");
int samplingFactor = sourceMap.size() / 10;
Pipeline p = Pipeline.create();
p.readFrom(Sources.map(sourceMap))
        .filter(item -> item.getKey().hashCode() % samplingFactor == ThreadLocalRandom.current().nextInt(samplingFactor))
        .writeTo(Sinks.list(targetList));

instance.newJob(p).join();

The code above should add approximately 10 elements to the result list, from which it's easy to get a random entry, but it can also end up with an empty list - you can experiment with increasing the multiplier in samplingFactor to get satisfactory probability of getting some result without having to re-run the job.

The same might also be possible using aggregation ( IMap.aggregate ) with a custom aggregator, some colleague might provide answer:-).