elasticsearch的termvectors包括了term的位置、词频等信息。这些信息用于相应的数据统计或开发其他功能，本文介绍termvecters如何使用，如何通过java客户端获取termvectors相关信息。



要使用termvctor首先要配置mapping中field的"term_vector"属性，默认状态es不开启termvector，因为这样会增加索引的体积，毕竟多存了不少元数据。

<br /> PUT test<br /> {<br /> "mappings": {<br /> "qa_test": {<br /> "dynamic": "strict",<br /> "_all": {<br /> "enabled": false<br /> },<br /> "properties": {<br /> "question": {<br /> "properties": {<br /> "cate": {<br /> "type": "keyword"<br /> },<br /> "desc": {<br /> "type": "text",<br /> "store": true,<br /> "term_vector": "with_positions_offsets_payloads",<br /> "analyzer": "ik_smart"<br /> },<br /> "time": {<br /> "type": "date",<br /> "store": true,<br /> "format": "strict_date_optional_time||epoch_millis||yyyy-MM-dd HH:mm:ss"<br /> },<br /> "title": {<br /> "type": "text",<br /> "store": true,<br /> "term_vector": "with_positions_offsets_payloads",<br /> "analyzer": "ik_smart"<br /> }<br /> }<br /> },<br /> "updatetime": {<br /> "type": "date",<br /> "store": true,<br /> "format": "strict_date_optional_time||epoch_millis||yyyy-MM-dd HH:mm:ss"<br /> }<br /> }<br /> }<br /> },<br /> "settings": {<br /> "index": {<br /> "number_of_shards": "1",<br /> "requests": {<br /> "cache": {<br /> "enable": "true"<br /> }<br /> },<br /> "number_of_replicas": "1"<br /> }<br /> }<br /> }<br />
注意示例中的"title"的"term_vector"属性。

接下来为索引创建一条数据

<br /> PUT qa_test_02/qa_test/1<br /> {<br /> "question": {<br /> "cate": [<br /> "装修流程",<br /> "其它"<br /> ],<br /> "desc": "筒灯，大洋和索正这两个牌子，哪个好？希望内行的朋友告知一下，谢谢！",<br /> "time": "2016-07-02 19:59:00",<br /> "title": "筒灯大洋和索正这两个牌子哪个好"<br /> },<br /> "updatetime": 1467503940000<br /> }<br />

下面我们看看这条数据上question.title字段的termvector信息

<br /> GET qa_test_02/qa_test/1/_termvectors<br /> {<br /> "fields": [<br /> "question.title"<br /> ],<br /> "offsets": true,<br /> "payloads": true,<br /> "positions": true,<br /> "term_statistics": true,<br /> "field_statistics": true<br /> }<br />

结果大概这个样子

<br /> {<br /> "_index": "qa_test_02",<br /> "_type": "qa_test",<br /> "_id": "1",<br /> "_version": 1,<br /> "found": true,<br /> "took": 0,<br /> "term_vectors": {<br /> "question.title": {<br /> "field_statistics": {<br /> "sum_doc_freq": 9,<br /> "doc_count": 1,<br /> "sum_ttf": 9<br /> },<br /> "terms": {<br /> "和": {<br /> "doc_freq": 1,<br /> "ttf": 1,<br /> "term_freq": 1,<br /> "tokens": [<br /> {<br /> "position": 2,<br /> "start_offset": 4,<br /> "end_offset": 5<br /> }<br /> ]<br /> },<br /> "哪个": {<br /> "doc_freq": 1,<br /> "ttf": 1,<br /> "term_freq": 1,<br /> "tokens": [<br /> {<br /> "position": 7,<br /> "start_offset": 12,<br /> "end_offset": 14<br /> }<br /> ]<br /> },<br /> "大洋": {<br /> "doc_freq": 1,<br /> "ttf": 1,<br /> "term_freq": 1,<br /> "tokens": [<br /> {<br /> "position": 1,<br /> "start_offset": 2,<br /> "end_offset": 4<br /> }<br /> ]<br /> },<br /> "好": {<br /> "doc_freq": 1,<br /> "ttf": 1,<br /> "term_freq": 1,<br /> "tokens": [<br /> {<br /> "position": 8,<br /> "start_offset": 14,<br /> "end_offset": 15<br /> }<br /> ]<br /> },<br /> "正": {<br /> "doc_freq": 1,<br /> "ttf": 1,<br /> "term_freq": 1,<br /> "tokens": [<br /> {<br /> "position": 4,<br /> "start_offset": 6,<br /> "end_offset": 7<br /> }<br /> ]<br /> },<br /> "牌子": {<br /> "doc_freq": 1,<br /> "ttf": 1,<br /> "term_freq": 1,<br /> "tokens": [<br /> {<br /> "position": 6,<br /> "start_offset": 10,<br /> "end_offset": 12<br /> }<br /> ]<br /> },<br /> "筒灯": {<br /> "doc_freq": 1,<br /> "ttf": 1,<br /> "term_freq": 1,<br /> "tokens": [<br /> {<br /> "position": 0,<br /> "start_offset": 0,<br /> "end_offset": 2<br /> }<br /> ]<br /> },<br /> "索": {<br /> "doc_freq": 1,<br /> "ttf": 1,<br /> "term_freq": 1,<br /> "tokens": [<br /> {<br /> "position": 3,<br /> "start_offset": 5,<br /> "end_offset": 6<br /> }<br /> ]<br /> },<br /> "这两个": {<br /> "doc_freq": 1,<br /> "ttf": 1,<br /> "term_freq": 1,<br /> "tokens": [<br /> {<br /> "position": 5,<br /> "start_offset": 7,<br /> "end_offset": 10<br /> }<br /> ]<br /> }<br /> }<br /> }<br /> }<br /> }<br />
下面我们说说如何通过java代码实现termvector的获取，不说废话直接上代码

<br /> TermVectorsResponse termVectorResponse = client.prepareTermVectors().setIndex(sourceindexname).setType(sourceindextype)<br /> .setId(id).setSelectedFields(fieldname).setTermStatistics(true).execute()<br /> .actionGet();<br /> XContentBuilder builder = XContentFactory.contentBuilder(XContentType.JSON);<br /> termVectorResponse.toXContent(builder, null);<br /> System.out.println(builder.string());<br /> Fields fields = termVectorResponse.getFields();<br /> Iterator<String> iterator = fields.iterator();<br /> while (iterator.hasNext()) {<br /> String field = iterator.next();<br /> Terms terms = fields.terms(field);<br /> TermsEnum termsEnum = terms.iterator();<br /> while (termsEnum.next() != null) {<br /> BytesRef term = termsEnum.term();<br /> if (term != null) {<br /> System.out.println(term.utf8ToString() + termsEnum.totalTermFreq());<br /> }<br /> }<br /> }<br /> <br />
获取TermVectorsResponse的代码很好理解，主要是设置索引名称、索引type、索引id以及需要展示的若干属性。

接下来是如何获取某一term的termvector，有两种方案第一种是通过TermVectorsResponse的toXContent方法直接生成XContentBuilder，这种方法可以直接获取和上面通过DSL查询一样的json结果；第二种是通过Fields的iterator遍历fields，获取TermsEnum，熟悉lucene的同学应会更熟悉第二种方法。