ElasticSearch在项目中做搜索的应用
2022/10/5大约 5 分钟
ElasticSearch 在项目中做搜索的应用
本篇记录商品站内搜索的落地过程：设计 ES 索引域、初始化索引库数据、拆解并开发关键字搜索功能（含分页与高亮），最后与前端联调。
1. 需求设计
站内搜索的原型效果如下，用户输入关键字即可搜索商品，并支持过滤面板、分页等能力：
2. 域设计
设计索引的域（Field）时，从四个维度考虑：
- 要不要分词？
- 要不要建立索引？
- 要不要存储？
- 使用哪种分词器？
据此设计出的商品索引实体：
package com.portal.entity;
import lombok.Data;
import org.springframework.data.annotation.Id;
import org.springframework.data.elasticsearch.annotations.DateFormat;
import org.springframework.data.elasticsearch.annotations.Document;
import org.springframework.data.elasticsearch.annotations.Field;
import org.springframework.data.elasticsearch.annotations.FieldType;
import java.util.Date;
@Data
@Document(indexName = "es-goods", shards = 1, replicas = 0)
public class ESGoods {
@Id
private Long spuId; // spuId
@Field(type = FieldType.Text, analyzer = "ik_max_word", store = true)
private String goodsName; // 商品名称，Text 类型，需要分词参与检索
@Field(type = FieldType.Long, index = true, store = true)
private Long brandId; // 品牌 id
@Field(type = FieldType.Keyword, index = true, store = true)
private String brandName; // 品牌名称，不分词，用于过滤与分桶
@Field(type = FieldType.Long, index = true, store = true)
private Long cid1id; // 1 级分类 id
@Field(type = FieldType.Keyword, index = true, store = true)
private String cat1name; // 1 级分类名称
@Field(type = FieldType.Long, index = true, store = true)
private Long cid2id; // 2 级分类 id
@Field(type = FieldType.Keyword, index = true, store = true)
private String cat2name; // 2 级分类名称
@Field(type = FieldType.Long, index = true, store = true)
private Long cid3id; // 3 级分类 id
@Field(type = FieldType.Keyword, index = true, store = true)
private String cat3name; // 3 级分类名称
@Field(type = FieldType.Date, index = true, store = true, format = DateFormat.date_time)
private Date createTime; // 创建时间
@Field(type = FieldType.Double, index = true, store = true)
private Double price; // 价格，spu 默认取 sku 的 price
@Field(type = FieldType.Keyword, index = false, store = true)
private String imageUrl; // 图片地址，只存储不参与检索
}3. 索引库数据初始化
3.1 spu 信息接口开发（商品微服务提供）
List<WxbGoods> findGoodsSpuInfo();Mapper XML：查询审核通过的商品 spu，并关联查出品牌与三级分类：
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE mapper PUBLIC "-//mybatis.org//DTD Mapper 3.0//EN" "http://mybatis.org/dtd/mybatis-3-mapper.dtd">
<mapper namespace="com.portal.goods.mapper.MallGoodsMapper">
<!-- List<WxbGoods> findGoodsSpuInfo(); -->
<resultMap id="findGoodsSpuInfoMap" type="MallGoods">
<result property="spuId" column="spuId"/>
<result property="goodsName" column="goods_name"/>
<result property="price" column="default_price"/>
<result property="typeTemplateId" column="template_id"/>
<result property="albumPics" column="album_pics"/>
<!-- 品牌 -->
<association property="brand" javaType="MallGoodsBrand">
<id property="id" column="brand_id"/>
<result property="name" column="brand_name"/>
</association>
<!-- 一级分类 -->
<association property="cat1" javaType="MallGoodsCat">
<id property="id" column="cat1id"/>
<result property="name" column="cat1name"/>
</association>
<!-- 二级分类 -->
<association property="cat2" javaType="MallGoodsCat">
<id property="id" column="cat2id"/>
<result property="name" column="cat2name"/>
</association>
<!-- 三级分类 -->
<association property="cat3" javaType="MallGoodsCat">
<id property="id" column="cat3id"/>
<result property="name" column="cat3name"/>
</association>
</resultMap>
<select id="findGoodsSpuInfo" resultMap="findGoodsSpuInfoMap">
SELECT
goods.spu_id as spuId,
goods.goods_name,
goods.price as default_price,
goods.album_pics,
brand.id as brand_id,
brand.`name` as brand_name,
cat1.id as cat1id,
cat1.`name` as cat1name,
cat2.id as cat2id,
cat2.`name` as cat2name,
cat3.id as cat3id,
cat3.`name` as cat3name,
cat3.template_id
from mall_goods goods
LEFT JOIN mall_goods_brand brand on brand.id = goods.brand_id
LEFT JOIN mall_goods_cat cat1 on cat1.id = goods.category1_id
LEFT JOIN mall_goods_cat cat2 on cat2.id = goods.category2_id
LEFT JOIN mall_goods_cat cat3 on cat3.id = goods.category3_id
where goods.audit_status = 1
</select>
</mapper>service、controller 代码【略】。
3.2 搜索微服务搭建
搜索微服务基于 spring-data-elasticsearch 搭建，注册到 Nacos，通过 Feign 调用商品微服务获取商品数据（依赖与配置同 01 篇的通用配置，此处略）。
3.3 初始化索引库数据
冷启动同步：先删除并重建索引库，再查商品数据、组装成 ESGoods 批量保存：
@Override
public ResultVO goods2es() {
// 删除索引库
template.deleteIndex(ESGoods.class);
// 创建索引库
template.createIndex(ESGoods.class);
template.putMapping(ESGoods.class);
// 获取商品信息
List<MallGoods> goodsAudited = goodsApi.findGoodsAudited();
if (goodsAudited == null) {
return new ResultVO(false, "同步失败");
}
List<ESGoods> list = goodsAudited.stream().map(spu -> {
// spu -> esGoods
ESGoods esGoods = new ESGoods();
esGoods.setSpuId(spu.getSpuId());
esGoods.setBrandId(spu.getBrand().getId());
esGoods.setBrandName(spu.getBrand().getName());
esGoods.setCid1id(spu.getCat1().getId());
esGoods.setCat1name(spu.getCat1().getName());
esGoods.setCid2id(spu.getCat2().getId());
esGoods.setCat2name(spu.getCat2().getName());
esGoods.setCid3id(spu.getCat3().getId());
esGoods.setCat3name(spu.getCat3().getName());
esGoods.setCreateTime(new Date());
esGoods.setGoodsName(spu.getGoodsName());
esGoods.setPrice(spu.getPrice().doubleValue());
if (spu.getAlbumPics() != null) {
String[] imgArr = spu.getAlbumPics().split(",");
if (imgArr.length > 0) {
esGoods.setImageUrl(imgArr[0]);
}
}
return esGoods;
}).collect(Collectors.toList());
template.save(list);
return new ResultVO(true, "同步成功");
}4. 商品搜索功能拆解
关键字搜索的需求：
- 需求：关键字匹配商品名称、品牌名称、第三级分类名称；
- 实现：使用 bool 查询【
should】，三个字段任一匹配即命中。
5. 关键字搜索接口开发
@Override
public PageResult<ESGoods> search(Map<String, Object> searchMap) {
// 1. 构建查询条件
BoolQueryBuilder boolQueryBuilder = QueryBuilders.boolQuery();
// match[goodsName]
MatchQueryBuilder goodsNameQueryBuilder = QueryBuilders
.matchQuery("goodsName", searchMap.get("keyword"));
// match[brandName]
MatchQueryBuilder brandNameQueryBuilder = QueryBuilders
.matchQuery("brandName", searchMap.get("keyword"));
// match[cat3name]
MatchQueryBuilder cat3nameQueryBuilder = QueryBuilders
.matchQuery("cat3name", searchMap.get("keyword"));
// should：三个字段任一匹配即命中
boolQueryBuilder.should(goodsNameQueryBuilder);
boolQueryBuilder.should(brandNameQueryBuilder);
boolQueryBuilder.should(cat3nameQueryBuilder);
PageRequest page = null;
// 2. 设置分页
if (!StringUtils.isEmpty(searchMap.get("page")) && !StringUtils.isEmpty(searchMap.get("size"))) {
Integer pageNow = (Integer) searchMap.get("page");
Integer size = (Integer) searchMap.get("size");
page = PageRequest.of(pageNow - 1, size);
}
// 3. 设置高亮域
HighlightBuilder goodsNameHighLight = getHighlightBuilder("goodsName");
NativeSearchQuery build = new NativeSearchQueryBuilder()
.withQuery(boolQueryBuilder)
.withPageable(page)
.withHighlightBuilder(goodsNameHighLight)
.build();
SearchHits<ESGoods> result = template.search(build, ESGoods.class);
// 总记录数
long totalHits = result.getTotalHits();
// 取当前页数据
List<SearchHit<ESGoods>> searchHits = result.getSearchHits();
List<ESGoods> esGoodsList = new ArrayList<>();
searchHits.forEach(hit -> {
// 取文档信息
ESGoods content = hit.getContent();
// 取高亮，覆盖商品名称
Map<String, List<String>> highlightFields = hit.getHighlightFields();
Set<String> keySet = highlightFields.keySet();
keySet.forEach(key -> {
List<String> list = highlightFields.get(key);
if ("goodsName".equals(key)) {
String highlightValue = list.get(0);
content.setGoodsName(highlightValue);
}
});
esGoodsList.add(content);
});
// 组装分页结果
PageResult<ESGoods> pageResult = new PageResult<>();
pageResult.setTotal(totalHits);
pageResult.setData(esGoodsList);
return pageResult;
}
/**
* 设置高亮字段
*/
private HighlightBuilder getHighlightBuilder(String... fields) {
// 高亮条件
HighlightBuilder highlightBuilder = new HighlightBuilder(); // 生成高亮查询器
for (String field : fields) {
highlightBuilder.field(field); // 高亮查询字段
}
highlightBuilder.requireFieldMatch(false); // 如果要多个字段高亮，这项要为 false
highlightBuilder.preTags("<span style=\"color:red\">"); // 高亮前置标签
highlightBuilder.postTags("</span>"); // 高亮后置标签
// 如果要高亮文字内容等很长的字段，必须配置下面两项，否则会导致高亮不全、文章内容缺失
highlightBuilder.fragmentSize(800000); // 最大高亮分片大小
highlightBuilder.numOfFragments(0); // 从第一个分片获取高亮片段
return highlightBuilder;
}6. 前端联调
前端页面在 created 生命周期发起首次搜索，搜索条件 searchMap 中的分页参数变化时重新调用：
<script>
var vm = new Vue({
el: "#app",
data: {
searchMap: {
page: 1,
size: 3
},
goodsList: [],
total: 0,
pages: 0
},
methods: {
search() {
wfx.http.post("/search", this.searchMap).then(res => {
console.log(res);
this.goodsList = res.data;
this.total = res.total;
this.pages = wfx.pageTotal(this.total, this.searchMap.size);
})
}
},
created() {
this.search();
}
})
</script>提示
整条搜索链路可以概括为：冷启动同步 + 审核后增量同步保证 ES 数据与数据库一致；搜索接口用 should 复合查询做多字段匹配，分页用 PageRequest，高亮用 HighlightBuilder 回填；前端只负责传关键字与分页参数、渲染结果。