使用lucene进行group操作

雨翔河

首页列表关于

使用lucene进行group操作

2016-04-23 23:38

近来用lucene进行查询，需要对结果进行group操作，结果发现核心包里面没有提供这个功能。

如果在内存里面对结果再自己用代码实现的话效率太低。查询了一番文档，发现了lucene－grouping。

jar包位置：

```
<dependency>
    <groupId>org.apache.lucene</groupId>
    <artifactId>lucene-grouping</artifactId>
    <version>4.0.0</version>
</dependency>
```

使用这个包可以实现对搜索结果进行group操作

首先建立索引

```
private Document createDocument(String id, String tag_value,
                                 String score,String city,String time,
                                 String from,String type,String typeId,
                                 String gender,String birthday) {
    Document document = new Document();
    document.add(new TextField("userId", id, Field.Store.YES));
    document.add(new StringField("tagName", tag_value, Field.Store.YES));
    document.add(new StringField("tagScoreValue", score, Field.Store.YES));
    document.add(new StringField("usersCurrentCity", city, Field.Store.YES));
    document.add(new StringField("tagRelatedTime", time, Field.Store.YES));
    document.add(new StringField("dataFrom", from, Field.Store.YES));
    document.add(new StringField("dataType", type, Field.Store.YES));
    document.add(new StringField("dataId", typeId, Field.Store.YES));
    document.add(new StringField("gender", gender, Field.Store.YES));
    document.add(new StringField("birthday", birthday, Field.Store.YES));
    return document;
}
```

这是我的索引结构，测试用的。

下面是我的测试代码

```
public void testCustomSort() throws Exception {
    List<Document> docs = Arrays.asList(
            createDocument("1", "java", "55", "sz", "2015", "www.yuorfei.com", "blog", "0", "1", "1993"),
            createDocument("1", "php", "56", "sz", "2015", "www.yuorfei.com", "blog", "0", "1", "1993"),
            createDocument("2", "docker", "57", "sz", "2015", "www.yuorfei.com", "blog", "0", "1", "1993"),
            createDocument("2", "php", "55", "sz", "2015", "www.yuorfei.com", "blog", "0", "1", "1993"),
            createDocument("2", "java", "56", "sz", "2015", "www.yuorfei.com", "blog", "0", "1", "1993"),
            createDocument("3", "java", "10", "sz", "2015", "www.yuorfei.com", "blog", "0", "1", "1993"),
            createDocument("4", "javascript", "99", "sz", "2015", "www.yuorfei.com", "blog", "0", "1", "1993"),
            createDocument("4", "java", "99999", "szbjbj44444", "20154444", "www.yuorfei.com", "blog", "0", "1", "1993"));
    LuceneTool.writeDocument(indexDirectoryPath, docs);
 
    String groupField = "userId";
    String tagName="java,php";
 
    GroupingSearch groupingSearch = new GroupingSearch(groupField);
    groupingSearch.setFillSortFields(true);
    groupingSearch.setCachingInMB(4.0, true);
    groupingSearch.setAllGroups(true);
    groupingSearch.setGroupDocsLimit(10);
    TagRadarSearchBuilder tagRadarSearchBuilder = new TagRadarSearchBuilder();
    String[] tagNames = StringUtils.split(tagName, ",");
    tagRadarSearchBuilder.tagName(tagNames).sortedBy(TagRadarSortedCategory.SORTED_BY_TAG_SCORE_VALUE_DESC);
    FSDirectory dir = FSDirectory.open(new File(indexDirectoryPath));
    dir.setReadChunkSize(104857600);//100兆
    IndexReader reader = DirectoryReader.open(dir);
    IndexSearcher indexSearch = new IndexSearcher(reader);
    groupingSearch.setGroupSort(tagRadarSearchBuilder.getSort());
    groupingSearch.setSortWithinGroup(tagRadarSearchBuilder.getSort());
    TopGroups<BytesRef> result = groupingSearch.search(indexSearch, tagRadarSearchBuilder.getQuery(), 0, 1000);
    System.out.println("搜索命中数：" + result.totalHitCount);
    System.out.println("搜索结果分组数：" + result.groups.length);
    System.out.println("\n-------------------------\n");
    for (GroupDocs<BytesRef> groupDocs : result.groups) {
        System.out.println("分组用户id：" + groupDocs.groupValue.utf8ToString());
        System.out.println("组内记录数量：" + groupDocs.totalHits);
        for (ScoreDoc scoreDoc : groupDocs.scoreDocs) {
            System.out.println(indexSearch.doc(scoreDoc.doc));
        }
        System.out.println("\n-------------------------\n");
    }
}
```

最主要的核心的地方是：

```
GroupingSearch groupingSearch = new GroupingSearch(groupField);
groupingSearch.setGroupSort(tagRadarSearchBuilder.getSort());
groupingSearch.setSortWithinGroup(tagRadarSearchBuilder.getSort());
TopGroups<BytesRef> result = groupingSearch.search(indexSearch, tagRadarSearchBuilder.getQuery(), 0, 1000);
```

设置的排序方法，调用groupSearch的search进行搜索。

setGroupSort为组间排序

setSortWithinGroup为组内排序

这样就能简单的实现对lucene搜索结果进行group操作，效率非常高。

类型：日常标签：lucene,group,java