最近，我给我的网站(https://www.xiandanplay.com/)尝试集成了一下es来实现我的一个搜索功能，因为这个是我第一次了解运用elastic，所以如果有不对的地方，大家可以指出来，话不多说，先看看我的一个大致流程
这里我采用的sdk的版本是Elastic.Clients.Elasticsearch, Version=8.0.0.0，官方的网址Installation | Elasticsearch .NET Client [8.0] | Elastic
我的es最开始打算和我的应用程序一起部署到ubuntu上面，结果最后安装kibana的时候，各种问题，虽好无奈，只好和我的SqlServer一起安装到windows上面，对于一个2G内容的服务器来说，属实有点遭罪了。
1、配置es
在es里面，我开启了密码认证。下面是我的配置
"Search": { "IsEnable": "true", "Uri": "http://127.0.0.1:9200/", "User": "123", "Password": "123" }
然后新增一个程序集
然后再ElasticsearchClient里面去写一个构造函数去配置es
using Core.Common;
using Core.CPlatform;
using Core.SearchEngine.Attr;
using Elastic.Clients.Elasticsearch;
using Elastic.Clients.Elasticsearch.IndexManagement;
using Elastic.Transport;
namespace Core.SearchEngine.Client
{
public class ElasticSearchClient : IElasticSearchClient
{
private ElasticsearchClient elasticsearchClient;
public ElasticSearchClient()
{
string uri = ConfigureProvider.configuration.GetSection("Search:Uri").Value;
string username = ConfigureProvider.configuration.GetSection("Search:User").Value;
string password = ConfigureProvider.configuration.GetSection("Search:Password").Value;
var settings = new ElasticsearchClientSettings(new Uri(uri))
.Authentication(new BasicAuthentication(username, password)).DisableDirectStreaming();
elasticsearchClient = new ElasticsearchClient(settings);
}
public ElasticsearchClient GetClient()
{
return elasticsearchClient;
}
}
}
然后，我们看skd的官网有这个这个提示
客户端应用程序应创建一个 该实例，该实例在整个应用程序中用于整个应用程序 辈子。在内部，客户端管理和维护与节点的 HTTP 连接， 重复使用它们以优化性能。如果您使用依赖项注入 容器中，客户端实例应注册到 单例生存期
所以我直接给它来一个AddSingleton
using Core.SearchEngine.Client;
using Microsoft.Extensions.DependencyInjection;
namespace Core.SearchEngine
{
public static class ConfigureSearchEngine
{
public static void AddSearchEngine(this IServiceCollection services)
{
services.AddSingleton<IElasticSearchClient, ElasticSearchClient>();
}
}
}
2、提交文章并且同步到es
然后就是同步文章到es了，我是先写入数据库，再同步到rabbitmq，通过事件总线（基于事件总线EventBus实现邮件推送功能）写入到es
先定义一个es模型
using Core.SearchEngine.Attr;
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using XianDan.Model.BizEnum;
namespace XianDan.Domain.Article
{
[ElasticsearchIndex(IndexName ="t_article")]//自定义的特性，sdk并不包含这个特性
public class Article_ES
{
public long Id { get; set; }
/// <summary>
/// 作者
/// </summary>
public string Author { get; set; }
/// <summary>
/// 标题
/// </summary>
public string Title { get; set; }
/// <summary>
/// 标签
/// </summary>
public string Tag { get; set; }
/// <summary>
/// 简介
/// </summary>
public string Description { get; set; }
/// <summary>
/// 内容
/// </summary>
public string ArticleContent { get; set; }
/// <summary>
/// 专栏
/// </summary>
public long ArticleCategoryId { get; set; }
/// <summary>
/// 是否原创
/// </summary>
public bool? IsOriginal { get; set; }
/// <summary>
/// 评论数
/// </summary>
public int? CommentCount { get; set; }
/// <summary>
/// 点赞数
/// </summary>
public int? PraiseCount { get; set; }
/// <summary>
/// 浏览次数
/// </summary>
public int? BrowserCount { get; set; }
/// <summary>
/// 收藏数量
/// </summary>
public int? CollectCount { get; set; }
/// <summary>
/// 创建时间
/// </summary>
public DateTime CreateTime { get; set; }
}
}
然后创建索引
string index = esArticleClient.GetIndexName(typeof(Article_ES));
await esArticleClient.GetClient().Indices.CreateAsync<Article_ES>(index, s =>
s.Mappings(
x => x.Properties(
t => t.LongNumber(l => l.Id)
.Text(l=>l.Title,z=>z.Analyzer(ik_max_word))
.Keyword(l=>l.Author)
.Text(l=>l.Tag,z=>z.Analyzer(ik_max_word))
.Text(l=>l.Description,z=>z.Analyzer(ik_max_word))
.Text(l=>l.ArticleContent,z=>z.Analyzer(ik_max_word))
.LongNumber(l=>l.ArticleCategoryId)
.Boolean(l=>l.IsOriginal)
.IntegerNumber(l=>l.BrowserCount)
.IntegerNumber(l=>l.PraiseCount)
.IntegerNumber(l=>l.PraiseCount)
.IntegerNumber(l=>l.CollectCount)
.IntegerNumber(l=>l.CommentCount)
.Date(l=>l.CreateTime)
)
)
);
然后每次增删改文章的时候写入到mq，例如
private async Task SendToMq(Article article, Operation operation)
{
ArticleEventData articleEventData = new ArticleEventData();
articleEventData.Operation = operation;
articleEventData.Article_ES = MapperUtil.Map<Article, Article_ES>(article);
TaskRecord taskRecord = new TaskRecord();
taskRecord.Id = CreateEntityId();
taskRecord.TaskType = TaskRecordType.MQ;
taskRecord.TaskName = "发送文章";
taskRecord.TaskStartTime = DateTime.Now;
taskRecord.TaskStatu = (int)MqMessageStatu.New;
articleEventData.Unique = taskRecord.Id.ToString();
taskRecord.TaskValue = JsonConvert.SerializeObject(articleEventData);
await unitOfWork.GetRepository<TaskRecord>().InsertAsync(taskRecord);
await unitOfWork.CommitAsync();
try
{
eventBus.Publish(GetMqExchangeName(), ExchangeType.Direct, BizKey.ArticleQueueName, articleEventData);
}
catch (Exception ex)
{
var taskRecordRepository = unitOfWork.GetRepository<TaskRecord>();
TaskRecord update = await taskRecordRepository.SelectByIdAsync(taskRecord.Id);
update.TaskStatu = (int)MqMessageStatu.Fail;
update.LastUpdateTime = DateTime.Now;
update.TaskResult = "发送失败";
update.AdditionalData = ex.Message;
await taskRecordRepository.UpdateAsync(update);
await unitOfWork.CommitAsync();
}
}
mq订阅之后写入es，具体的增删改的方法就不写了吧
3、开始查询es
等待写入文章之后，开始查询文章，这里sdk提供的查询的方法比较复杂，全都是通过lmbda一个个链式去拼接的，但是我又没有找到更好的方法，所以就先这样吧
先创建一个集合存放查询的表达式
List<Action<QueryDescriptor<Article_ES>>> querys = new List<Action<QueryDescriptor<Article_ES>>>();
然后定义一个几个需要查询的字段
我这里使用MultiMatch来实现多个字段匹配同一个查询条件，并且指定使用ik_smart分词
Field[] fields =
{
new Field("title"),
new Field("tag"),
new Field("articleContent"),
new Field("description")
};
querys.Add(s => s.MultiMatch(y => y.Fields(Fields.FromFields(fields)).Analyzer(ik_smart).Query(keyword).Type(TextQueryType.MostFields)));
定义查询结果高亮，给查询出来的匹配到的分词的字段添加标签，同时前端需要对这个样式处理，
:deep(.search-words) em { color: #ee0f29; font-style: initial; } Dictionary<Field, HighlightField> highlightFields = new Dictionary<Field, HighlightField>();
highlightFields.Add(new Field("title"), new HighlightField()
{
PreTags = new List<string> { "<em>" },
PostTags = new List<string> { "</em>" },
});
highlightFields.Add(new Field("description"), new HighlightField()
{
PreTags = new List<string> { "<em>" },
PostTags = new List<string> { "</em>" },
});
Highlight highlight = new Highlight()
{
Fields = highlightFields
};
为了提高查询的效率，我只查部分的字段
SourceFilter sourceFilter = new SourceFilter();
sourceFilter.Includes = Fields.FromFields(new Field[] { "title", "id", "author", "description", "createTime", "browserCount", "commentCount" });
SourceConfig sourceConfig = new SourceConfig(sourceFilter);
Action<SearchRequestDescriptor<Article_ES>> configureRequest = s => s.Index(index)
.From((homeArticleCondition.CurrentPage - 1) * homeArticleCondition.PageSize)
.Size(homeArticleCondition.PageSize)
.Query(x => x.Bool(y => y.Must(querys.ToArray())))
.Source(sourceConfig)
.Sort(y => y.Field(ht => ht.CreateTime, new FieldSort() { Order=SortOrder.Desc}))
获取查询的分词结果
var analyzeIndexRequest = new AnalyzeIndexRequest
{
Text = new string[] { keyword },
Analyzer = analyzer
};
var analyzeResponse = await elasticsearchClient.Indices.AnalyzeAsync(analyzeIndexRequest);
if (analyzeResponse.Tokens == null)
return new string[0];
return analyzeResponse.Tokens.Select(s => s.Token).ToArray();
到此，这个就是大致的查询结果，完整的如下
public async Task<Core.SearchEngine.Response.SearchResponse<Article_ES>> SelectArticle(HomeArticleCondition homeArticleCondition)
{
string keyword = homeArticleCondition.Keyword.Trim();
bool isNumber = Regex.IsMatch(keyword, RegexPattern.IsNumberPattern);
List<Action<QueryDescriptor<Article_ES>>> querys = new List<Action<QueryDescriptor<Article_ES>>>();
if (isNumber)
{
querys.Add(s => s.Bool(x => x.Should(
should => should.Term(f => f.Field(z => z.Title).Value(keyword))
, should => should.Term(f => f.Field(z => z.Tag).Value(keyword))
, should => should.Term(f => f.Field(z => z.ArticleContent).Value(keyword))
)));
}
else
{
Field[] fields =
{
new Field("title"),
new Field("tag"),
new Field("articleContent"),
new Field("description")
};
querys.Add(s => s.MultiMatch(y => y.Fields(Fields.FromFields(fields)).Analyzer(ik_smart).Query(keyword).Type(TextQueryType.MostFields)));
}
if (homeArticleCondition.ArticleCategoryId.HasValue)
{
querys.Add(s => s.Term(t => t.Field(f => f.ArticleCategoryId).Value(FieldValue.Long(homeArticleCondition.ArticleCategoryId.Value))));
}
string index = esArticleClient.GetIndexName(typeof(Article_ES));
Dictionary<Field, HighlightField> highlightFields = new Dictionary<Field, HighlightField>();
highlightFields.Add(new Field("title"), new HighlightField()
{
PreTags = new List<string> { "<em>" },
PostTags = new List<string> { "</em>" },
});
highlightFields.Add(new Field("description"), new HighlightField()
{
PreTags = new List<string> { "<em>" },
PostTags = new List<string> { "</em>" },
});
Highlight highlight = new Highlight()
{
Fields = highlightFields
};
SourceFilter sourceFilter = new SourceFilter();
sourceFilter.Includes = Fields.FromFields(new Field[] { "title", "id", "author", "description", "createTime", "browserCount", "commentCount" });
SourceConfig sourceConfig = new SourceConfig(sourceFilter);
Action<SearchRequestDescriptor<Article_ES>> configureRequest = s => s.Index(index)
.From((homeArticleCondition.CurrentPage - 1) * homeArticleCondition.PageSize)
.Size(homeArticleCondition.PageSize)
.Query(x => x.Bool(y => y.Must(querys.ToArray())))
.Source(sourceConfig)
.Sort(y => y.Field(ht => ht.CreateTime, new FieldSort() { Order=SortOrder.Desc})).Highlight(highlight);
var resp = await esArticleClient.GetClient().SearchAsync<Article_ES>(configureRequest);
foreach (var item in resp.Hits)
{
if (item.Highlight == null)
continue;
foreach (var dict in item.Highlight)
{
switch (dict.Key)
{
case "title":
item.Source.Title = string.Join("...", dict.Value);
break;
case "description":
item.Source.Description = string.Join("...", dict.Value);
break;
}
}
}
string[] analyzeWords = await esArticleClient.AnalyzeAsync(homeArticleCondition.Keyword);
List<Article_ES> articles = resp.Documents.ToList();
return new Core.SearchEngine.Response.SearchResponse<Article_ES>(articles, analyzeWords);
}
4、演示效果
搞完之后，发布部署，看看效果，分词这里要想做的像百度那样，估计目前来看非常有难度的
那么这里我也向大家求教一下，如何使用SearchRequest封装多个查询条件，如下
SearchRequest searchRequest = new SearchRequest();
searchRequest.From = 0;
searchRequest.Size = 10;
searchRequest.Query=多个查询条件
因为我觉得这样代码读起来比lambda可读性高些，能更好的动态封装。