I want to something which is similar to the scala grouped function. Basically, pick 2 elements at a time and process them. Here is a reference for the same :
Split list into multiple lists with fixed number of elements
Lambdas do provide things like groupingBy and partitioningBy but none of them seem to do the same as the grouped function in Scala. Any pointers would be appreciated.
It sounds like a problem that is better handled like a low-level Stream operation just like the ops provided by the Stream API itself. A (relative) simple solution may look like:
public static <T> Stream<List<T>> chunked(Stream<T> s, int chunkSize) {
if(chunkSize<1) throw new IllegalArgumentException("chunkSize=="+chunkSize);
if(chunkSize==1) return s.map(Collections::singletonList);
Spliterator<T> src=s.spliterator();
long size=src.estimateSize();
if(size!=Long.MAX_VALUE) size=(size+chunkSize-1)/chunkSize;
int ch=src.characteristics();
ch&=Spliterator.SIZED|Spliterator.ORDERED|Spliterator.DISTINCT|Spliterator.IMMUTABLE;
ch|=Spliterator.NONNULL;
return StreamSupport.stream(new Spliterators.AbstractSpliterator<List<T>>(size, ch)
{
private List<T> current;
@Override
public boolean tryAdvance(Consumer<? super List<T>> action) {
if(current==null) current=new ArrayList<>(chunkSize);
while(current.size()<chunkSize && src.tryAdvance(current::add));
if(!current.isEmpty()) {
action.accept(current);
current=null;
return true;
}
return false;
}
}, s.isParallel());
}
Simple test:
chunked(Stream.of(1, 2, 3, 4, 5, 6, 7), 3)
.parallel().forEachOrdered(System.out::println);
The advantage is that you do not need a full collection of all items for subsequent stream processing, e.g.
chunked(
IntStream.range(0, 1000).mapToObj(i -> {
System.out.println("processing item "+i);
return i;
}), 2).anyMatch(list->list.toString().equals("[6, 7]")));
will print:
processing item 0
processing item 1
processing item 2
processing item 3
processing item 4
processing item 5
processing item 6
processing item 7
true
rather than processing a thousand items of IntStream.range(0, 1000). This also enables using infinite source Streams:
chunked(Stream.iterate(0, i->i+1), 2).anyMatch(list->list.toString().equals("[6, 7]")));
If you are interested in a fully materialized collection rather than applying subsequent Stream operations, you may simply use the following operation:
List<Integer> list=Arrays.asList(1, 2, 3, 4, 5, 6, 7);
int listSize=list.size(), chunkSize=2;
List<List<Integer>> list2=
IntStream.range(0, (listSize-1)/chunkSize+1)
.mapToObj(i->list.subList(i*=chunkSize,
listSize-chunkSize>=i? i+chunkSize: listSize))
.collect(Collectors.toList());
You can create your own collector. Something like this:
class GroupingCollector<T> implements Collector<T, List<List<T>>, List<List<T>>> {
private final int elementCountInGroup;
public GroupingCollector(int elementCountInGroup) {
this.elementCountInGroup = elementCountInGroup;
}
@Override
public Supplier<List<List<T>>> supplier() {
return ArrayList::new;
}
@Override
public BiConsumer<List<List<T>>, T> accumulator() {
return (lists, integer) -> {
if (!lists.isEmpty()) {
List<T> integers = lists.get(lists.size() - 1);
if (integers.size() < elementCountInGroup) {
integers.add(integer);
return;
}
}
List<T> list = new ArrayList<>();
list.add(integer);
lists.add(list);
};
}
@Override
public BinaryOperator<List<List<T>>> combiner() {
return (lists, lists2) -> {
List<List<T>> r = new ArrayList<>();
r.addAll(lists);
r.addAll(lists2);
return r;
};
}
@Override
public Function<List<List<T>>, List<List<T>>> finisher() {
return lists -> lists;
}
@Override
public Set<Characteristics> characteristics() {
return Collections.emptySet();
}
}
And then you can use it in a way like this:
List<List<Integer>> collect = Stream.of(1, 2, 3, 4, 5, 6, 7, 8, 9, 10).collect(new GroupingCollector<>(3));
System.out.println(collect);
Will print:
[[1, 2, 3], [4, 5, 6], [7, 8, 9], [10]]
A recursive solution to transform the list to a list-of-lists would also be possible
int chunkSize = 2;
private <T> List<List<T>> process(List<T> list) {
if (list.size() > chunkSize) {
List<T> chunk = list.subList(0, chunkSize);
List<T> rest = list.subList(chunkSize, list.size());
List<List<T>> lists = process(rest);
return concat(chunk, lists);
} else {
ArrayList<List<T>> retVal = new ArrayList<>();
retVal.add(list);
return retVal;
}
}
private <T> List<List<T>> concat(List<T> chunk, List<List<T>> rest) {
rest.add(0, chunk);
return rest;
}
You could write your own collector finisher, similar to
final List<String> strings = Arrays.asList("Hello", "World", "I", "Am", "You");
final int size = 3;
final List<List<String>> stringLists = strings.stream()
.collect(Collectors.collectingAndThen(Collectors.toList(), new Function<List<String>, List<List<String>>>() {
@Override
public List<List<String>> apply(List<String> strings) {
final List<List<String>> result = new ArrayList<>();
int counter = 0;
List<String> stringsToAdd = new ArrayList<>();
for (final String string : strings) {
if (counter == 0) {
result.add(stringsToAdd);
} else {
if (counter == size) {
stringsToAdd = new ArrayList<>();
result.add(stringsToAdd);
counter = 0;
}
}
++counter;
stringsToAdd.add(string);
}
return result;
}
}));
System.out.println("stringLists = " + stringLists); // stringLists = [[Hello, World, I], [Am, You]]
A simple version with java 8 streams api:
static <T> List<List<T>> partition(List<T> list, Integer partitionSize) {
int numberOfLists = BigDecimal.valueOf(list.size())
.divide(BigDecimal.valueOf(partitionSize), 0, CEILING)
.intValue();
return IntStream.range(0, numberOfLists)
.mapToObj(it -> list.subList(it * partitionSize, Math.min((it+1) * partitionSize, list.size())))
.collect(Collectors.toList());
}
