Oinone引入搜索引擎(增强模型)

场景描述

在碰到大数据量并且需要全文检索的场景，我们在分布式架构中基本会架设ElasticSearch来作为一个常规解决方案。在oinone体系中增强模型就是应对这类场景，其背后也是整合了ElasticSearch;

使用前你应该

• 了解ElasticSearch，包括不限于：Index(索引)、分词、Node(节点)、Document(文档)、Shards(分片) & Replicas(副本)。参考官方网站：https://www.elastic.co/cn/
• 有一个可用的ElasticSearch环境(本地项目能引用到)

前置约束

增强模型增量依赖数据变更实时消息，因此确保项目的event是开启的，mq配置正确。

项目引入搜索步骤

1、boot工程加入相关依赖包

• boot工程需要指定ES客户端包版本，不指定版本会隐性依赖顶层spring-boot依赖管理指定的低版本
• boot工程加入pamris-channel的工程依赖

<dependency>
    <groupId>org.elasticsearch.client</groupId>
    <artifactId>elasticsearch-rest-client</artifactId>
    <version>8.4.1</version>
</dependency>
<dependency>
    <groupId>jakarta.json</groupId>
    <artifactId>jakarta.json-api</artifactId>
    <version>2.1.1</version>
</dependency>

<dependency>
    <groupId>pro.shushi.pamirs.core</groupId>
    <artifactId>pamirs-sql-record-core</artifactId>
</dependency>
<dependency>
    <groupId>pro.shushi.pamirs.core</groupId>
    <artifactId>pamirs-channel-core</artifactId>
</dependency>

2、api工程加入相关依赖包

在XXX-api中增加入pamirs-channel-api的依赖

<dependency>
    <groupId>pro.shushi.pamirs.core</groupId>
    <artifactId>pamirs-channel-api</artifactId>
</dependency>

3、yml文件配置

在pamirs-demo-boot的application-dev.yml文件中增加配置pamirs.boot.modules增加channel，即在启动模块中增加channel模块。同时注意es的配置，是否跟es的服务一致

pamirs:
  record:
    sql:
      #改成自己本地路径(或服务器路径)
      store: /Users/oinone/record
  boot:
    modules:
      - channel
      ## 确保也安装了sql_record
      - sql_record
  channel:
    packages:
      # 增强模型扫描包配置
      - com.xxx.xxx
  elastic:
    url: 127.0.0.1:9200

4、项目的模块增加模块依赖

XXXModule增加对ChannelModule的依赖

@Module(dependencies = {ChannelModule.MODULE_MODULE})

5、增加增强模型(举例)

package pro.shushi.pamirs.demo.api.enhance;

import pro.shushi.pamirs.channel.enmu.IncrementEnum;
import pro.shushi.pamirs.channel.meta.Enhance;
import pro.shushi.pamirs.channel.meta.EnhanceModel;
import pro.shushi.pamirs.demo.api.model.ShardingModel;
import pro.shushi.pamirs.meta.annotation.Model;
import pro.shushi.pamirs.meta.enmu.ModelTypeEnum;

@Model(displayName = "测试EnhanceModel")
@Model.model(ShardingModelEnhance.MODEL_MODEL)
@Model.Advanced(type = ModelTypeEnum.PROXY, inherited = {EnhanceModel.MODEL_MODEL})
@Enhance(shards = "3", replicas = "1", reAlias = true,increment= IncrementEnum.OPEN)
public class ShardingModelEnhance extends ShardingModel {
    public static final String MODEL_MODEL="demo.ShardingModelEnhance";
}

6、重启系统看效果

1、进入【传输增强模型】应用，访问增强模型列表我们会发现一条记录，并点击【全量同步】初始化ES，并全量dump数据

2、再次回到Demo应用，进入增强模型页面，可以正常访问并进增删改查操作

个性化dump逻辑

通常dump逻辑是有个性化需求，那么我们可以重写模型的synchronize方法，函数重写特性在“面向对象-继承与多态”部分中已经有详细介绍。

重写ShardingModelEnhance模型的synchronize方法

重写后，如果针对老数据记录需要把新增的字段都自动填充，可以进入【传输增强模型】应用，访问增强模型列表，找到对应的记录并点击【全量同步】

package pro.shushi.pamirs.demo.api.enhance;

import pro.shushi.pamirs.channel.enmu.IncrementEnum;
import pro.shushi.pamirs.channel.meta.Enhance;
import pro.shushi.pamirs.channel.meta.EnhanceModel;
import pro.shushi.pamirs.demo.api.model.ShardingModel;
import pro.shushi.pamirs.meta.annotation.Field;
import pro.shushi.pamirs.meta.annotation.Function;
import pro.shushi.pamirs.meta.annotation.Model;
import pro.shushi.pamirs.meta.enmu.FunctionTypeEnum;
import pro.shushi.pamirs.meta.enmu.ModelTypeEnum;

import java.util.List;

@Model(displayName = "测试EnhanceModel")
@Model.model(ShardingModelEnhance.MODEL_MODEL)
@Model.Advanced(type = ModelTypeEnum.PROXY, inherited = {EnhanceModel.MODEL_MODEL})
@Enhance(shards = "3", replicas = "1", reAlias = true,increment= IncrementEnum.OPEN)
public class ShardingModelEnhance extends ShardingModel {
    public static final String MODEL_MODEL="demo.ShardingModelEnhance";

    @Field(displayName = "nick")
    private String nick;

    @Function.Advanced(displayName = "同步数据", type = FunctionTypeEnum.UPDATE)
    @Function(summary = "数据同步函数")
    public List<ShardingModelEnhance> synchronize(List<ShardingModelEnhance> data) {
        for(ShardingModelEnhance shardingModelEnhance:data){
            shardingModelEnhance.setNick(shardingModelEnhance.getName());
        }
        return data;
    }
}

给搜索增加个性化逻辑

如果我们需要在查询方法中增加逻辑，在前面的教程中一般是重写queryPage函数，但对于增强模型我们需要重写的是search函数。

个性化search函数

@Function(
        summary = "搜索函数",
        openLevel = {FunctionOpenEnum.LOCAL, FunctionOpenEnum.REMOTE, FunctionOpenEnum.API}
)
@pro.shushi.pamirs.meta.annotation.Function.Advanced(
        type = {FunctionTypeEnum.QUERY},
        category = FunctionCategoryEnum.QUERY_PAGE,
        managed = true
)
public  Pagination<ShardingModelEnhance> search(Pagination<ShardingModelEnhance> page, IWrapper<ShardingModelEnhance> queryWrapper) {
    System.out.println("您的个性化搜索逻辑");
    // return ((IElasticRetrieve) CommonApiFactory.getApi(IElasticRetrieve.class)).search(page, queryWrapper);
    return ((ElasticSearchApi) CommonApiFactory.getApi(ElasticSearchApi.class)).search(page, queryWrapper);
}

个性化search函数示例

    @Override
    @SuppressWarnings({"rawtypes"})
    public <T> Pagination<T> search(Pagination<T> page, IWrapper<T> queryWrapper) {
        String modelModel = queryWrapper.getModel();
        if (null == modelModel || modelModel.isEmpty()) {
            return page;
        }
        ModelConfig modelCfg = PamirsSession.getContext().getModelConfig(modelModel);
        if (null == modelCfg) {
            return page;
        }
        String rsql = queryWrapper.getOriginRsql();
        if (StringUtils.isBlank(rsql)) {
            rsql = "id>0";
        }
        BoolQuery.Builder queryBuilder = ElasticRSQLHelper.parseRSQL(modelCfg, rsql);
        TermQuery isDeletedTerm = QueryBuilders.term()
                .queryName(IS_DELETED)
                .field(IS_DELETED).value(0)
                .build();
        BoolQuery.Builder builder = QueryBuilders.bool().must(new Query(queryBuilder.build()));
        builder.must(new Query(isDeletedTerm));
        String alias = IndexNaming.aliasByModel(modelModel);
        Query query = new Query(builder.build());
        log.info("{}", query);

        List<Order> orders = Optional.ofNullable(page.getSort()).map(Sort::getOrders).orElse(new ArrayList<>());
        int currentPage = Optional.ofNullable(page.getCurrentPage()).orElse(1);
        Long size = Optional.ofNullable(page.getSize()).orElse(10L);
        int pageSize = size.intValue();
        List<SortOptions> sortOptions = new ArrayList<>();
        if (CollectionUtils.isEmpty(orders)) {
            orders.add(new Order(SortDirectionEnum.DESC, ID));
            orders.add(new Order(SortDirectionEnum.DESC, CREATE_DATE));
        }
        for (Order order : orders) {
            sortOptions.add(new SortOptions.Builder()
                    .field(SortOptionsBuilders.field()
                            .field(order.getField())
                            .order(SortDirectionEnum.DESC.equals(order.getDirection()) ? SortOrder.Desc : SortOrder.Asc)
                            .build())
                    .build());
        }

        SearchRequest request = new SearchRequest.Builder()
                .index(alias)
                .from((currentPage - 1) * pageSize)
                .size(pageSize)
                .sort(sortOptions)
                .query(query)
                .highlight(_builder ->
                        _builder.numberOfFragments(4)
                                .fragmentSize(50)
                                .type(HighlighterType.Unified)
                                .fields("name", HighlightField.of(_fieldBuilder -> _fieldBuilder.preTags(ElasticsearchConstant.HIGH_LIGHT_PREFIX).postTags(ElasticsearchConstant.HIGH_LIGHT_POSTFIX)))
                                .fields("documentNo", HighlightField.of(_fieldBuilder -> _fieldBuilder.preTags(ElasticsearchConstant.HIGH_LIGHT_PREFIX).postTags(ElasticsearchConstant.HIGH_LIGHT_POSTFIX)))
                                .fields("keywords", HighlightField.of(_fieldBuilder -> _fieldBuilder.preTags(ElasticsearchConstant.HIGH_LIGHT_PREFIX).postTags(ElasticsearchConstant.HIGH_LIGHT_POSTFIX))))
                .build();

        SearchResponse<HashMap> response = null;
        try {
            log.info("ES搜索请求参数：{}", request.toString());
            response = elasticsearchClient.search(request, HashMap.class);
        } catch (ElasticsearchException e) {
            log.error("索引异常", e);
            PamirsSession.getMessageHub()
                    .msg(Message.init()
                            .setLevel(InformationLevelEnum.WARN)
                            .msg("索引异常"));
            return page;
        } catch (IOException e) {
            log.error("ElasticSearch运行状态异常", e);
            PamirsSession.getMessageHub()
                    .msg(Message.init()
                            .setLevel(InformationLevelEnum.WARN)
                            .msg("ElasticSearch运行状态异常"));
            return page;
        }

        if (null == response || response.timedOut()) {
            return page;
        }

        HitsMetadata<HashMap> hits = response.hits();
        if (null == hits) {
            return page;
        }

        TotalHits totalHits = hits.total();
        long total = Optional.ofNullable(totalHits).map(TotalHits::value).orElse(0L);

        List<HashMap> dataMapList = Optional.of(hits)
                .map(HitsMetadata<HashMap>::hits)
                .map(hitsMap ->{
                    hitsMap.stream().forEach(highlightForEach -> {
                        highlightForEach.highlight().forEach((key, value) -> {
                            if(highlightForEach.source().containsKey(key)){
                                highlightForEach.source().put(key,value.get(0));
                            }
                        });

                    });
                    return hitsMap;
                })
                .map(List::stream)
                .orElse(Stream.empty())
                .map(Hit::source)
                .collect(Collectors.toList());

        List<T> context = persistenceDataConverter.out(modelModel, dataMapList);

        page.setSize(size);
        page.setTotalElements(total);
        page.setContent(context);
        log.info("ES搜索请求参数返回total,{}", total);
        return page;
    }