Lucene NRT -在每次提交后打开一个新的IndexReader？

Question

我需要我在Lucene中的结果在添加文档后立即可用。我读到的所有内容都说我应该缓存IndexSearcher实例，但我不知道如何在每次写入时都重新打开IndexReader以及IndexSearcher的情况下实现我想要的内容……

我做错了什么？

class LuceneStorage {

    private final Directory luceneDirectory;
    private final IndexWriter indexWriter;

    private volatile DirectoryReader indexReader;

    LuceneStorage() {
        try {
            this.luceneDirectory = NIOFSDirectory.open(Paths.get(System.getProperty("user.home")).resolve("lucene"));

            IndexWriterConfig config = new IndexWriterConfig(new KeywordAnalyzer());
            config.setOpenMode(IndexWriterConfig.OpenMode.CREATE_OR_APPEND);

            this.indexWriter = new IndexWriter(luceneDirectory, config);
            this.indexReader = DirectoryReader.open(indexWriter);
        } catch (IOException e) {
            throw new UncheckedIOException(e);
        }
    }

    public Stream<String> read(String id) {
        IndexSearcher indexSearcher = new IndexSearcher(indexReader);
        QueryParser queryParser = new QueryParser("id", new KeywordAnalyzer());
        StreamableSimpleCollector collector = new StreamableSimpleCollector();

        try {
            indexSearcher.search(queryParser.parse(id), collector);
        } catch (Exception e) {
            throw new IllegalStateException(e);
        }

        IntStream docIds = collector.stream();

        return docIds.mapToObj(i -> {
                try {
                    return indexSearcher.doc(i).get("content");
                } catch (IOException e) {
                    throw new UncheckedIOException(e);
                }
            });
    }

    public String write(String content, String... ids) {
        String uuid = UUID.randomUUID().toString();

        Document document = new Document();
        document.add(new StringField("id", uuid, Field.Store.YES));

        Stream.of(ids)
            .forEach(i -> document.add(new StringField("id", i, Field.Store.YES)));

        document.add(new StoredField("content", content));

        try {
            indexWriter.addDocument(document);
            indexReader = DirectoryReader.openIfChanged(indexReader);
        } catch (IOException e) {
            throw new UncheckedIOException(e);
        }

        return uuid;
    }
}

Answer 1

Lucene不是为了提供实时搜索结果而设计的。因此，如果这真的是一个要求，你应该考虑其他东西。

然而Lucene可以提供近乎实时的搜索结果，所以你的用户最多只能等待。在更改可搜索之前的1秒。这样做的方法是使用SearcherManager (http://lucene.apache.org/core/5_5_0/core/org/apache/lucene/search/SearcherManager.html)，并拥有一个定期刷新当前索引读取器的后台线程。