multithreading - Java 8 ConcurrentHashMap 合并与 computeIfAbsent

Question

我正在练习 Cay S. Horstmann 所著的“Java SE 8 for the really Impatient”一书的练习。有 2 个练习要求相同算法的不同实现，一个使用 merge ，其他 computeIfAbsent .我已经使用 merge 实现了该程序但不知道如何使用 computeIfAbsent做同样的事情。在我看来 computeIfPresent会更合适，因为 merge仅当 key 存在时才有效，computeIfPresent 也是如此.

问题陈述:

Write an application in which multiple threads read all words from a collection of files. Use a ConcurrentHashMap<String, Set<File>> to track in which files each word occurs. Use the merge method to update the map.

我的代码使用 merge :

public static Map<String, Set<File>> reverseIndexUsingMerge(final Path path)
        throws IOException {
    final ConcurrentHashMap<String, Set<File>> map = new ConcurrentHashMap<>();

    final BiConsumer<? super String, ? super Set<File>> action = (key,
        value) -> map.merge(key, value, (existingValue, newValue) -> {
        LOGGER.info("Received key: {}, existing value: {}, new value: {}.",
            key, existingValue, newValue);

        newValue.addAll(existingValue);

        return newValue;
    });

    commonPool().invokeAll(
        find(path, 1,
            (p, fileAttributes) -> fileAttributes.isRegularFile())
            .map(p -> new ReverseIndex(p, action))
            .collect(toList()));

    return unmodifiableMap(map);
}

private static class ReverseIndex implements Callable<Void> {
    private final Path p;
    private final BiConsumer<? super String, ? super Set<File>> action;

    private static final Pattern AROUND_WHITESPACE = compile("\\s");

    private ReverseIndex(final Path p,
        final BiConsumer<? super String, ? super Set<File>> action) {
        this.p = p;
        this.action = action;
    }

    @Override
    public Void call() throws Exception {
        reverseIndex().forEach(action);

        return null;
    }

    private Map<String, Set<File>> reverseIndex() {
        /* File stream needs to be closed. */
        try (Stream<String> lines = lines(p, UTF_8)) {
        return lines.flatMap(AROUND_WHITESPACE::splitAsStream)
            .collect(
                groupingBy(String::toString,
                    mapping(word -> p.toFile(), toSet())));
        } catch (IOException e) {
        LOGGER.error("Something went wrong. Get the hell outta here.",
            e);

        throw new UncheckedIOException(e);
        }
    }
}

Answer 1

如果没有值(value)，则专注于必须完成的工作。你要做的，是创建一个新的Set缺失条目的值。当然，如果您使用具有原子性保证的操作来创建 Set仅添加到 Set将同时发生，这需要使用并发 Set .您可以使用 ConcurrentHashMap创建事实上的 ConcurrentHashSet (它不存在于那种形式)通过映射到一个固定值，如果你让值信令存在是 Boolean.TRUE，这尤其简单。 :

ConcurrentHashMap<String, Set<File>> map=new ConcurrentHashMap<>();
final BiConsumer<? super String, ? super Set<File>> action =
    (key, value) -> map.computeIfAbsent(key, x->ConcurrentHashMap.newKeySet())
                       .addAll(value);