使用Nest创建自定义分析器(用于电子邮件地址)

时间:2016-01-21 13:33:55

标签: c# elasticsearch nest

我有一个包含邮件地址的UserID字段的类,我希望能够在该字段上搜索完全匹配,或者搜索包含一些排除项的邮件地址部分。

例如,如果UserID包含" john.doe@foo.com",我希望能够搜索" john.doe@foo.com" ;,"约翰"和" doe",但" foo"和" com"应该从索引中排除。

我尝试创建一个自定义分析器,它使用带有已排除单词列表的停止过滤器,然后对该属性应用多索引,如下例所示:

using Nest;
using System;
using System.Collections.Generic;

[ElasticType]
public class ElasticUser {
    [ElasticProperty(Index = FieldIndexOption.NotAnalyzed)]
    public string UserID { get; set; }
}

class Program {

    static void Main(string[] args) {
        const string IndexName = "test_index";

        var settings = new ConnectionSettings(uri: new Uri("http://localhost:9200/"), defaultIndex: IndexName);
        var client = new ElasticClient(settings);

        // delete the index for the test
        var deleteIndexResp = client.DeleteIndex(IndexName);

        // create the custom filter and analyzer
        var user_id_stop_filter = new StopTokenFilter {
            Stopwords = new[] { "foo", "bar", "com" }
        };
        var user_id_analyzer = new CustomAnalyzer {
            Filter = new List<string> {
                "user_id_stop_filter"
            },
            Tokenizer = "letter"
        };

        // create the index with the custom filter and analyzer
        var createIndexResp = client.CreateIndex(IndexName, index => index
            .Analysis(analysis => analysis
                .TokenFilters(t => t
                    .Add("user_id_stop_filter", user_id_stop_filter))
                .Analyzers(a => a
                    .Add("user_id_analyzer", user_id_analyzer))));

        // add a mapping for the "ElasticUser" type
        var putMapResp = client.Map<ElasticUser>(
            m => m.MapFromAttributes()
            .Properties(properties => properties
                .MultiField(multi => multi
                    .Name(p => p.UserID)
                    .Fields(fields => fields
                        .String(s => s
                            .Name(p => p.UserID)
                            .Index(FieldIndexOption.NotAnalyzed)
                        )
                        .String(s => s
                            .Name(p => p.UserID.Suffix("searchable"))
                            .Analyzer("user_id_analyzer")
                        )
                    )
                )
            ));

        // add a couple of entries
        client.Index(new ElasticUser {
            UserID = "some.one@foo.com"
        });
        client.Index(new ElasticUser {
            UserID = "another.one@bar.com"
        });
    }

}

但是,这似乎不起作用,因为我只能搜索完全匹配,但电子邮件地址不会在非单词上进行标记化分割。

如何让这个多索引按照描述工作?

当我尝试运行此查询时,我没有得到任何结果:

GET /test_index/elasticuser/_search
{
    "query": {
        "query_string": {
           "query": "one"
        }
    }
}

1 个答案:

答案 0 :(得分:2)

实现目标的最简单方法是在searchable字段上使用simple analyzer

...
.String(s => s
    .Name(p => p.UserID.Suffix("searchable"))
    .Analyzer("simple")                        <---- change this
)
...

电子邮件将被标记为任何非字母字符,您将能够搜索johndoe

<强>更新

如果您想保留您的排除列表,您肯定可以这样做。您可以保留现有的分析器,但需要使用lowercase tokenizer(即与simple分析器中使用的相同)而不是letter分析器。

    var user_id_analyzer = new CustomAnalyzer {
        Filter = new List<string> {
            "user_id_stop_filter"
        },
        Tokenizer = "lowercase"            <--- change this
    };

更新2

在纯JSON中,这就是我所拥有的

curl -XPUT localhost:9200/users -d '{
  "settings": {
    "analysis": {
      "analyzer": {
        "email_analyzer": {
          "type": "custom",
          "tokenizer": "lowercase",
          "filter": [
            "my_stop"
          ]
        }
      },
      "filter": {
        "my_stop": {
          "type": "stop",
          "stopwords": [
            "foo",
            "bar",
            "com"
          ]
        }
      }
    }
  },
  "mappings": {
    "user": {
      "properties": {
        "email": {
          "type": "string",
          "fields": {
            "raw": {
              "type": "string",
              "index": "not_analyzed"
            },
            "parts": {
              "type": "string",
              "analyzer": "email_analyzer"
            }
          }
        }
      }
    }
  }
}'

然后当我分析some.one@foo.com时,这就是我得到的

$ curl -XGET 'localhost:9200/users/_analyze?field=email.parts&pretty' -d 'some.one@foo.com'
{
  "tokens" : [ {
    "token" : "some",
    "start_offset" : 0,
    "end_offset" : 4,
    "type" : "word",
    "position" : 1
  }, {
    "token" : "one",
    "start_offset" : 5,
    "end_offset" : 8,
    "type" : "word",
    "position" : 2
  } ]
}