Is it possible to collect a stream to two different collections using one line?

问题

I have the following code: (simplified for bravity)

public void search(Predicate<String> predicate, Elements elements)
{
    List<SearchResult> searchResults = elements.stream()
        .filter(element -> predicate.test(element.ownText()))
        .map(element -> new SearchResult(element.ownText(), element.baseUri(),element.tagName()))
        .collect(Collectors.toList());
}

But now, I want to have another list which would contain all filtered elements without the mapping. Is it possible to do that with a stream, or should I change to a foreach loop for that?

回答1:

You can collect the elements after filtering and then use this list to map the elements to search results:

public void search(Predicate<String> predicate, Elements elements) {
    List<Element> filteredElements = 
        elements.stream()
                .filter(element -> predicate.test(element.ownText()))
                .collect(Collectors.toList());

    List<SearchResult> searchResults = 
        filteredElements.stream()
                        .map(element -> new SearchResult(element.ownText(),element.baseUri(),element.tagName()))
                        .collect(Collectors.toList());
}

This won't take more time than the solutions of the other answers but doesn't have side effects, is thread safe as well as easy to read and understand.

回答2:

A simple way to do this would be

List<Element> filteredElements = new ArrayList<>();
List<SearchResult> searchResults = elements.stream()
      .filter(element -> predicate.test(element.ownText()))
      .peek(filteredElements::add)
      .map(element -> 
         new SearchResult(element.ownText(),element.baseUri(),element.tagName()))
      .collect(Collectors.toList());

回答3:

One alternative is to define a Supplier<Stream>:

public void search(Predicate<String> predicate, Elements elements)
{
    Supplier<Stream<Element>> supplier = () -> elements.stream()
           .filter(element -> predicate.test(element.ownText()));
    List<SearchResult> searchResults = supplier.get()
           .map(element -> new SearchResult(element.ownText(),element.baseUri(),element.tagName()))
           .collect(Collectors.toList());
    List<Element> elementList = supplier.get().collect(Collectors.toList());
}

Note that using this approach you actually perform the filtering twice.

An alternative (though not very beautiful in this case) solution is use pairing collector from this answer:

Collector<Element, ?, List<SearchResult>> c1 = 
    Collectors.mapping(element -> new SearchResult(element.ownText(),element.baseUri(),element.tagName()),
        Collectors.toList());
Collector<Element, ?, List<Element>> c2 = Collectors.toList();
Collector<Element, ?, Pair<List<SearchResult>, List<Element>>> pairCollector =
    pairing(c1, c2, Pair::new); // Assumes some Pair class existing

Pair<List<SearchResult>, List<Element>> result = elements.stream()
           .filter(element -> predicate.test(element.ownText()))
           .collect(pairing);

These solutions are generic: they allow to do two different operations with single data source. But in your concrete example it's easier to create first list of filtered non-mapped data, then create a second stream on that list:

List<Element> elementList = elements.stream()
           .filter(element -> predicate.test(element.ownText()))
           .collect(Collectors.toList());
List<SearchResult> searchResults = elementList.stream()
           .map(element -> new SearchResult(element.ownText(),element.baseUri(),element.tagName()))
           .collect(Collectors.toList());

回答4:

You cannot use the stream twice, as it is closed when using collect. The following code will throw an "java.lang.IllegalStateException: stream has already been operated upon or closed". So you need to create the stream twice or use a for-each loop.

import java.util.Collections;
import java.util.List;
import java.util.function.Predicate;
import java.util.stream.Collectors;
import java.util.stream.Stream;

class Test
{
    static class SearchResult
    {
        SearchResult(Object o,Object p,Object q) {}
    }

    static class Element
    {
        public String ownText() {return "text";}
        public String tagName() {return "tag";}
        public String baseUri() {return "base";}
    }

    public static void search(Predicate<String> predicate, List<Element> elements)
    {
        Stream<Element> stream = elements.stream().filter(element -> predicate.test(element.ownText()));
        List<Element> filtered = stream.collect(Collectors.toList());
        // next line will throw the IllegalStateException as above line has already consumed the stream
        List<SearchResult> results =    stream.map(element -> 
         new SearchResult(element.ownText(),element.baseUri(),element.tagName()))
                .collect(Collectors.toList());
    }

    public static void main(String[] args)
    {       
        search(s->true,Collections.singletonList(new Element()));
    }
}

来源：https://stackoverflow.com/questions/34134263/is-it-possible-to-collect-a-stream-to-two-different-collections-using-one-line