[原]Lucene系列-facet

1.facet的直观认识

facet：面、切面、方面。个人理解就是维度，在满足query的前提下，观察结果在各维度上的分布（一个维度下各子类的数目）。

如jd上搜“手机”，得到4009个商品。其中品牌、网络、价格就是商品的维度(facet)，点击某个品牌或者网络，获取更细分的结果。

点击品牌小米，获得小米手机的结果，显示27个。

点击移动4G，获得移动4G、小米手机，显示4个。

2.facet特性

• facet counting：返回一个facet下某子类的结果数。如上面的品牌维度下小米子类中满足查询"手机"的结果有27个。
• facet associations：一个文档与某子类的关联度，如一本书30%讲lucene，70%讲solor，这个百分比就是书与分类的关联度(匹配度、信心度)。
• multiple facet requests：支持多facet查询（多维度查询）。如查询品牌为小米、网络为移动4G的手机。

3.实例

一个facet简单使用例子，依赖于lucene-facet-4.10.0。讲述了从搜手机到品牌、到网络向下browser的过程。

public class SimpleFacetsExample {
    private final Directory indexDir = new RAMDirectory();
    private final Directory taxoDir = new RAMDirectory();
    private final FacetsConfig config = new FacetsConfig();

    /** Empty constructor */
    public SimpleFacetsExample() {
        config.setHierarchical("Publish Date", true);
    }

    /** Build the example index. */
    private void index() throws IOException {
        IndexWriter indexWriter = new IndexWriter(indexDir, new IndexWriterConfig(Version.LUCENE_4_10_0,
                new WhitespaceAnalyzer()));
        // Writes facet ords to a separate directory from the main index
        DirectoryTaxonomyWriter taxoWriter = new DirectoryTaxonomyWriter(taxoDir);

        Document doc = new Document();
        doc.add(new TextField("device", "手机", Field.Store.YES));
        doc.add(new TextField("name", "米1", Field.Store.YES));
        doc.add(new FacetField("brand", "小米"));
        doc.add(new FacetField("network", "移动4G"));
        indexWriter.addDocument(config.build(taxoWriter, doc));

        doc = new Document();
        doc.add(new TextField("device", "手机", Field.Store.YES));
        doc.add(new TextField("name", "米4", Field.Store.YES));
        doc.add(new FacetField("brand", "小米"));
        doc.add(new FacetField("network", "联通4G"));
        indexWriter.addDocument(config.build(taxoWriter, doc));

        doc = new Document();
        doc.add(new TextField("device", "手机", Field.Store.YES));
        doc.add(new TextField("name", "荣耀6", Field.Store.YES));
        doc.add(new FacetField("brand", "华为"));
        doc.add(new FacetField("network", "移动4G"));
        indexWriter.addDocument(config.build(taxoWriter, doc));

        doc = new Document();
        doc.add(new TextField("device", "电视", Field.Store.YES));
        doc.add(new TextField("name", "小米电视2", Field.Store.YES));
        doc.add(new FacetField("brand", "小米"));
        indexWriter.addDocument(config.build(taxoWriter, doc));

        taxoWriter.close();
        indexWriter.close();
    }

    private void facetsWithSearch() throws IOException {
        DirectoryReader indexReader = DirectoryReader.open(indexDir);
        IndexSearcher searcher = new IndexSearcher(indexReader);
        TaxonomyReader taxoReader = new DirectoryTaxonomyReader(taxoDir);

        FacetsCollector fc = new FacetsCollector();
        //1.查询手机
        System.out.println("-----手机-----");
        TermQuery query = new TermQuery(new Term("device", "手机"));
        FacetsCollector.search(searcher, query, 10, fc);
        Facets facets = new FastTaxonomyFacetCounts(taxoReader, config, fc);
        List<FacetResult> results = facets.getAllDims(10);
        //手机总共有3个,品牌维度：小米2个，华为1个;网络维度：移动4G 2个，联通4G 1个
        for (FacetResult tmp : results) {
            System.out.println(tmp);
        }
        //2.drill down，品牌选小米
        System.out.println("-----小米手机-----");
        DrillDownQuery drillDownQuery = new DrillDownQuery(config, query);
        drillDownQuery.add("brand", "小米");
        FacetsCollector fc1 = new FacetsCollector();//要new新collector，否则会累加
        FacetsCollector.search(searcher, drillDownQuery, 10, fc1);
        facets = new FastTaxonomyFacetCounts(taxoReader, config, fc1);
        results = facets.getAllDims(10);
        //获得小米手机的分布，总数2个，网络：移动4G 1个，联通4G 1个
        for (FacetResult tmp : results) {
            System.out.println(tmp);
        }

        //3.drill down，小米移动4G手机
        System.out.println("-----移动4G小米手机-----");
        drillDownQuery.add("network", "移动4G");
        FacetsCollector fc2 = new FacetsCollector();
        FacetsCollector.search(searcher, drillDownQuery, 10, fc2);
        facets = new FastTaxonomyFacetCounts(taxoReader, config, fc2);
        results = facets.getAllDims(10);
        for (FacetResult tmp : results) {
            System.out.println(tmp);
        }

        //4.drill sideways，横向浏览
        //如果已经进入了小米手机，但是还想看到其他牌子(华为)的手机数目，就用到了sideways
        System.out.println("-----小米手机drill sideways-----");
        DrillSideways ds = new DrillSideways(searcher, config, taxoReader);
        DrillDownQuery drillDownQuery1 = new DrillDownQuery(config, query);
        drillDownQuery1.add("brand", "小米");
        DrillSidewaysResult result = ds.search(drillDownQuery1, 10);
        results = result.facets.getAllDims(10);
        for (FacetResult tmp : results) {
            System.out.println(tmp);
        }

        indexReader.close();
        taxoReader.close();
    }

    /** Runs the search and drill-down examples and prints the results. */
    public static void main(String[] args) throws Exception {
        SimpleFacetsExample example = new SimpleFacetsExample();
        example.index();
        example.facetsWithSearch();
    }
}

输出：

-----手机-----
//总数3个，2个子类
dim=brand path=[] value=3 childCount=2
  小米 (2)
  华为 (1)

dim=network path=[] value=3 childCount=2
  移动4G (2)
  联通4G (1)

-----小米手机-----
//普通向下浏览，丢失了同一维度，其他子类的统计
dim=brand path=[] value=2 childCount=1
  小米 (2)

dim=network path=[] value=2 childCount=2
  移动4G (1)
  联通4G (1)

-----移动4G小米手机-----
dim=brand path=[] value=1 childCount=1
  小米 (1)

dim=network path=[] value=1 childCount=1
  移动4G (1)

-----小米手机drill sideways-----
//drill sideways, 保留了该drill维度的其他子类统计
dim=brand path=[] value=3 childCount=2
  小米 (2)
  华为 (1)
//小米手机中的网络分布
dim=network path=[] value=2 childCount=2
  移动4G (1)
  联通4G (1)