arraylist 和 parallelStream 的奇怪情况

Question

I have a parallel stream because the task is really slow, I will paste the code below. The situation is this.

I have an arrayList, I need to do something with each object in that list (this is slow) and add the object to a temporal list, the process in the stream ends ok, I think, because I can see each object processed with logs.

When the stream ends, sometimes, the temporal list has n-1 objects or one as null.

Any idea?

With this sample code the errors are not happening, but the logic is the same but without the business logic.

public class SampleCode {
    public List<SomeObject> example(List<SomeObject> someObjectList) {
        List<SomeObject> someObjectListTemp = new ArrayList<>();
        someObjectList.parallelStream().forEach(someObject -> {
            List<ExtraData> extraDataList = getExtraData(someObject.getId());
            if (extraDataList.isEmpty()) {
                someObjectListTemp.add(someObject);
            } else {
                for (ExtraData extraData : extraDataList) {
                    SomeObject someObjectTemp = null;
                    someObjectTemp = (SomeObject) cloneObject(someObject);
                    if (extraData != null) {
                        someObjectTemp.setDate(extraData.getDate());
                        someObjectTemp.setData2(extraData.getData2());
                    }
                    if (someObjectTemp == null) {
                        System.out.println("Warning null object"); //I NEVER see this
                    }
                    someObjectListTemp.add(someObjectTemp);
                    System.out.println("Added object to list"); //I Always see this the same times as elements in original list
                }
            }
        });

        if (someObjectListTemp.size() < 3) {
            System.out.println("Error: There should be at least 3 elements"); //Some times one object is missing in the list
        }

        for (SomeObject someObject : someObjectListTemp) {
            if (someObject == null) {
                System.out.println("Error: null element in list"); //Some times one object is null in the list
            }
        }

        return someObjectListTemp;
    }

Answer 1

Could you try to use the flatMap method instead of foreach ? flatMap takes a list of lists and put all their elements in a single list.

This way you do not use another ArrayList to store your temporary objects. I feel that this might be the issue, because parallelStream is multi threading and ArrayList is not synchronised

List<SomeObject> someObjectListTemp = someObjectList.parallelStream()
    .map(so -> processSomeObject(so)) // makes a stream of lists (Stream<List<SomeObject>>)
    .flatMap(Collection::stream) // groups all the elements of all the lists in one stream (Stream<Someobject>)
    .collect(Collectors.toList()); // transforms the stream into a list (List<SomeObject>)

And stick your code in a separate method processSomeObject which returns a list of SomeObject :

static List<SomeObject> processSomeObject(SomeObject someObject) {
    List<ExtraData> extraDataList = getExtraData(someObject.getId());
    List<SomeObject> someObjectListTemp = new ArrayList<>();
    if (extraDataList.isEmpty()) {
        someObjectListTemp.add(someObject);
    } else {
        for (ExtraData extraData : extraDataList) {
            SomeObject someObjectTemp = (SomeObject) cloneObject(someObject);
            if (extraData != null) {
                someObjectTemp.setDate(extraData.getDate());
                someObjectTemp.setData2(extraData.getData2());
            }
            someObjectListTemp.add(someObjectTemp);
            System.out.println("Added object to list");
        }
    }

    return someObjectListTemp;
}

Answer 2

A small example would be

public static void main(String[] args) {
    List<Object> test = new ArrayList<>();
    IntStream.range(0, 100000).parallel().forEach(i -> test.add(new Object()));
    for(Object o : test) {
        System.out.println(o.getClass());
    }
}

i'ts because ArrayList is not threadsafe and the internal array gets screwed