Elasticsearch 建议合并两种类型的结果
Elasticsearch suggestion is merging result of two types
我在一个索引中有 2 种类型，并且都有建议字段（根据需要）
public class LegalAreaSearchModel : LegalAreaModel
{
public SuggestField Suggest
{
get
{
List<string> input = new List<string>();
string[] childArea = !string.IsNullOrEmpty(this.LegalArea) ? this.LegalArea.Split(' ') : new string[] { "" };
string[] ParentArea = !string.IsNullOrEmpty(this.Parent) ? this.Parent.Split(' ') : new string[] { "" };
input.AddRange(childArea);
input.AddRange(ParentArea);
return
new SuggestField
{
Input = input,
Output = this.LegalArea,
Payload = new
{
Id = this.RowId,
Name = !string.IsNullOrEmpty(this.Parent) ? this.Parent + "/" + this.LegalArea : this.LegalArea,
Type = "Specialization"
},
Weight = !string.IsNullOrEmpty(this.LegalArea) ? this.LegalArea.Length : 0
};
}
}
}
和
public class LegalDocumentSearchModel : LegalDocumentModel
{
public LegalDocumentSearchModel()
{
//this = ObjectCopier.Clone<LegalDocumentSearchModel>(legalDocumentSearchModel);
SupplierDetails = new List<LegalDocumentSupplierDetailForSearch>();
CategoryDetails = new List<CategoryDetails>();
}
[Nested()]
public List<LegalDocumentSupplierDetailForSearch> SupplierDetails { get; set; }
[Nested()]
public List<CategoryDetails> CategoryDetails { get; set; }
public SuggestField Suggest
{
get
{
return
new SuggestField
{
Input = new List<string>(!string.IsNullOrEmpty(this.Name) ? this.Name.Split(' ') : new string[] { "" }) { this.Name },
Output = this.Name,
Payload = new
{
Id = this.RowId,
SEOFriendlyURLName = !string.IsNullOrEmpty(this.SEOFriendlyURLName) ? string.Concat(this.SEOFriendlyURLName) : string.Empty,
Name = this.Name,
ProductType = this.ProductType,
Description = this.Description
},
Weight = !string.IsNullOrEmpty(this.Description) ? this.Description.Length : 0
};
}
}
}
public class LegalDocumentSupplierDetailForSearch
{
public string PinCode { get; set; }
public string Lattitude { get; set; }
public string Longitude { get; set; }
}
public class CategoryDetails
{
public int CategoryId { get; set; }
public string CategoryName { get; set; }
}
现在我的搜索如下
List<AdvocateListingSuggestionModel> lsResult = new List<AdvocateListingSuggestionModel>();
var result = _searchProvider.Client.Suggest<LegalAreaSearchModel>(s => s
.Index(SearchConfigurationManager.DefaultSearchIndex)
.Completion("ml-la-suggestions", c => c
.Text(Search)
.Field(p => p.Suggest)
.Fuzzy(fz => fz
.Fuzziness(Fuzziness.Auto))
)
);
我也从其他类型获得结果（混合）我们如何限制仅来自一种类型的结果
Type1.LegalAreaSearchModel
输入 2.LegalDocumentSearchModel
我的索引创建如下
public static bool CheckForIndex(ElasticSearchProvider searchProvider)
{
Nest.IndexExistsRequest idr = new Nest.IndexExistsRequest(Nest.Indices.Index(new Nest.IndexName() { Name = SearchConfigurationManager.DefaultSearchIndex }));
var indxres = searchProvider.Client.IndexExists(idr);
if (!indxres.Exists)
{
searchProvider.Client.CreateIndex(SearchConfigurationManager.DefaultSearchIndex, i => i
.Settings(s => s
.NumberOfShards(2)
.NumberOfReplicas(0)
.Analysis(analysis => analysis
.Tokenizers(tokenizers => tokenizers
.Pattern("ml-id-tokenizer", p => p.Pattern(@"\W+"))
)
.TokenFilters(tokenfilters => tokenfilters
.WordDelimiter("ml-id-words", wd => wd
.SplitOnCaseChange()
.PreserveOriginal()
.SplitOnNumerics()
.GenerateNumberParts(false)
.GenerateWordParts()
)
)
.Analyzers(analyzers => analyzers
.Custom("ml-id-analyzer", c => c
.Tokenizer("ml-id-tokenizer")
.Filters("ml-id-words", "lowercase")
)
.Custom("ml-id-keyword", c => c
.Tokenizer("keyword")
.Filters("lowercase")
)
)
)
));
}
return true;
}
正在创建文档类型如下
public bool CreateDocumentIndex()
{
bool retVal = false;
if (Common.CheckForIndex(_searchProvider))
{
var res = _searchProvider.Client.Map<LegalDocumentSearchModel>(m =>
m.Index(SearchConfigurationManager.DefaultSearchIndex)
.Type(this.type)
.AutoMap()
.Properties(ps => ps
.String(s => s
.Name(p => p.Id)
.Analyzer("ml-id-analyzer")
.Fields(f => f
.String(p => p.Name("keyword").Analyzer("ml-id-keyword"))
.String(p => p.Name("raw").Index(FieldIndexOption.NotAnalyzed))
)
)
.Completion(c => c
.Name(p => p.Suggest)
.Payloads()
)
));
retVal = res.IsValid;
}
return retVal;
}
如下创建 LegalArea 类型
public bool CreateLegalAreaIndex()
{
bool retVal = false;
if (Common.CheckForIndex(_searchProvider))
{
var res = _searchProvider.Client.Map<LegalAreaSearchModel>(m =>
m.Index(SearchConfigurationManager.DefaultSearchIndex)
.Type(this.type)
.AutoMap()
.Properties(ps => ps
.String(s => s
.Name(p => p.Id)
.Analyzer("ml-id-analyzer")
.Fields(f => f
.String(p => p.Name("keyword").Analyzer("ml-id-keyword"))
.String(p => p.Name("raw").Index(FieldIndexOption.NotAnalyzed))
)
)
.Completion(c => c
.Name(p => p.Suggest)
.Payloads()
)
));
retVal = res.IsValid;
}
return retVal;
}
现在当我运行合法区域建议如下
public List<AdvocateListingSuggestionModel> LegalAreaSuggestion(string Search)
{
List<AdvocateListingSuggestionModel> lsResult = new List<AdvocateListingSuggestionModel>();
var result = _searchProvider.Client.Suggest<LegalAreaSearchModel>(s => s
.Index(SearchConfigurationManager.DefaultSearchIndex)
.Completion("ml-la-suggestions", c => c
.Text(Search)
.Field(p => p.Suggest)
.Fuzzy(fz => fz
.Fuzziness(Fuzziness.Auto))
)
);
if (result.IsValid)
{
lsResult = result.Suggestions["ml-la-suggestions"]
.FirstOrDefault()
.Options
.Select(suggest => suggest.Payload<AdvocateListingSuggestionModel>()).ToList();
}
return lsResult;
}
我得到 LegalDocumentSearchModel 的结果作为 well.pls 建议
谢谢
建议在索引级别工作，即 one Finite State Transducer (FST) is created for a field per index 因此，如果您在同一索引中有两个具有相同字段名称的类型（两个字段 必须 相同类型），两种类型都将在一个 FST 中表示数据。
我们可以通过一个简单的例子看出这一点。这里我们有两种带有完成字段的类型，我们将它们索引到同一个索引
public class Band
{
public int Id { get; set;}
public CompletionField<object> Suggest { get; set;}
}
public class Sport
{
public int Id { get; set;}
public CompletionField<object> Suggest { get; set; }
}
void Main()
{
var pool = new SingleNodeConnectionPool(new Uri("http://localhost:9200"));
var defaultIndex = "default-index";
var connectionSettings = new ConnectionSettings(pool)
.DefaultIndex(defaultIndex);
var client = new ElasticClient(connectionSettings);
// just so we can re-run this example...
if (client.IndexExists(defaultIndex).Exists)
client.DeleteIndex(defaultIndex);
client.CreateIndex(defaultIndex, ci => ci
.Mappings(m => m
.Map<Band>(l => l
.AutoMap()
.Properties(p => p
.Completion(c => c
.Name(n => n.Suggest)
)
)
)
.Map<Sport>(l => l
.AutoMap()
.Properties(p => p
.Completion(c => c
.Name(n => n.Suggest)
)
)
)
)
);
var bands = new List<Band>
{
new Band { Id = 1, Suggest = new CompletionField<object> { Input = new [] {"Bowling for Soup"} } },
new Band { Id = 2, Suggest = new CompletionField<object> { Input = new [] {"Fastball"} } },
new Band { Id = 3, Suggest = new CompletionField<object> { Input = new [] {"Dropkick Murphys"} } },
new Band { Id = 4, Suggest = new CompletionField<object> { Input = new [] {"Yellowcard"} } },
new Band { Id = 5, Suggest = new CompletionField<object> { Input = new [] {"American Football"} } },
};
client.IndexMany(bands);
var sports = new List<Sport>
{
new Sport { Id = 1, Suggest = new CompletionField<object> { Input = new [] {"Bowling"} } },
new Sport { Id = 2, Suggest = new CompletionField<object> { Input = new [] {"Football"} } },
new Sport { Id = 3, Suggest = new CompletionField<object> { Input = new [] {"Baseball"} } },
new Sport { Id = 4, Suggest = new CompletionField<object> { Input = new [] {"Table Tennis"} } },
new Sport { Id = 5, Suggest = new CompletionField<object> { Input = new [] {"American Football"} } },
};
client.IndexMany(sports);
client.Refresh(defaultIndex);
var suggestResponse = client.Suggest<Band>(s => s
.Completion("suggestion", cs => cs
.Text("Bo")
.Field(f => f.Suggest)
)
);
}
我们从我们的建议调用中得到以下结果
{
"_shards" : {
"total" : 5,
"successful" : 5,
"failed" : 0
},
"suggestion" : [ {
"text" : "Bo",
"offset" : 0,
"length" : 2,
"options" : [ {
"text" : "Bowling",
"score" : 1.0
}, {
"text" : "Bowling for Soup",
"score" : 1.0
} ]
} ]
}
我们在结果中得到了 Band 类型和 Sport 类型的建议。也许不是我们所期望或想要的，但这就是建议者的工作方式。请注意，在 .Suggest<T>() 调用中，类型 T 用于提供对可能在查询中使用的属性的强类型访问。在此示例中，类型为 Band 的 .Suggest 字段。
这里推荐的解决方案是为不同的类型设置单独的索引；您仍然可以在需要时跨多个索引进行查询，并避免在同一索引中包含多种类型的陷阱。
除此之外，您可以通过对完成类型映射使用 Context Suggester and taking advantage of category context 来实现它。理想情况下，您希望使用类别文档中存在的字段来为您提供类别的灵活性，但如果需要，您可以使用元数据字段。我建议针对您的领域调查和研究方法。
对于我们之前的示例，如果我们现在在映射中包含类别上下文（与以前相同的代码，但具有此更新的映射）
client.CreateIndex(defaultIndex, ci => ci
.Mappings(m => m
.Map<Band>(l => l
.AutoMap()
.Properties(p => p
.Completion(c => c
.Name(n => n.Suggest)
.Context(sc => sc
.Category("type", csc => csc
// recommend that you use another field on
// the document here instead of a metadata field
.Field("_type")
)
)
)
)
)
.Map<Sport>(l => l
.AutoMap()
.Properties(p => p
.Completion(c => c
.Name(n => n.Suggest)
.Context(sc => sc
.Category("type", csc => csc
// recommend that you use another field on
// the document here instead of a metadata field
.Field("_type")
)
)
)
)
)
)
);
并在搜索时使用类别上下文
var suggestResponse = client.Suggest<Band>(s => s
.Completion("suggestion", cs => cs
.Text("Bo")
.Field(f => f.Suggest)
.Context(d => d.Add("type", "band"))
)
);
然后我们得到了想要的结果
{
"_shards" : {
"total" : 5,
"successful" : 5,
"failed" : 0
},
"suggestion" : [ {
"text" : "Bo",
"offset" : 0,
"length" : 2,
"options" : [ {
"text" : "Bowling for Soup",
"score" : 1.0
} ]
} ]
}
我在一个索引中有 2 种类型，并且都有建议字段（根据需要）
public class LegalAreaSearchModel : LegalAreaModel
{
public SuggestField Suggest
{
get
{
List<string> input = new List<string>();
string[] childArea = !string.IsNullOrEmpty(this.LegalArea) ? this.LegalArea.Split(' ') : new string[] { "" };
string[] ParentArea = !string.IsNullOrEmpty(this.Parent) ? this.Parent.Split(' ') : new string[] { "" };
input.AddRange(childArea);
input.AddRange(ParentArea);
return
new SuggestField
{
Input = input,
Output = this.LegalArea,
Payload = new
{
Id = this.RowId,
Name = !string.IsNullOrEmpty(this.Parent) ? this.Parent + "/" + this.LegalArea : this.LegalArea,
Type = "Specialization"
},
Weight = !string.IsNullOrEmpty(this.LegalArea) ? this.LegalArea.Length : 0
};
}
}
}
和
public class LegalDocumentSearchModel : LegalDocumentModel
{
public LegalDocumentSearchModel()
{
//this = ObjectCopier.Clone<LegalDocumentSearchModel>(legalDocumentSearchModel);
SupplierDetails = new List<LegalDocumentSupplierDetailForSearch>();
CategoryDetails = new List<CategoryDetails>();
}
[Nested()]
public List<LegalDocumentSupplierDetailForSearch> SupplierDetails { get; set; }
[Nested()]
public List<CategoryDetails> CategoryDetails { get; set; }
public SuggestField Suggest
{
get
{
return
new SuggestField
{
Input = new List<string>(!string.IsNullOrEmpty(this.Name) ? this.Name.Split(' ') : new string[] { "" }) { this.Name },
Output = this.Name,
Payload = new
{
Id = this.RowId,
SEOFriendlyURLName = !string.IsNullOrEmpty(this.SEOFriendlyURLName) ? string.Concat(this.SEOFriendlyURLName) : string.Empty,
Name = this.Name,
ProductType = this.ProductType,
Description = this.Description
},
Weight = !string.IsNullOrEmpty(this.Description) ? this.Description.Length : 0
};
}
}
}
public class LegalDocumentSupplierDetailForSearch
{
public string PinCode { get; set; }
public string Lattitude { get; set; }
public string Longitude { get; set; }
}
public class CategoryDetails
{
public int CategoryId { get; set; }
public string CategoryName { get; set; }
}
现在我的搜索如下
List<AdvocateListingSuggestionModel> lsResult = new List<AdvocateListingSuggestionModel>();
var result = _searchProvider.Client.Suggest<LegalAreaSearchModel>(s => s
.Index(SearchConfigurationManager.DefaultSearchIndex)
.Completion("ml-la-suggestions", c => c
.Text(Search)
.Field(p => p.Suggest)
.Fuzzy(fz => fz
.Fuzziness(Fuzziness.Auto))
)
);
我也从其他类型获得结果（混合）我们如何限制仅来自一种类型的结果
Type1.LegalAreaSearchModel 输入 2.LegalDocumentSearchModel
我的索引创建如下
public static bool CheckForIndex(ElasticSearchProvider searchProvider)
{
Nest.IndexExistsRequest idr = new Nest.IndexExistsRequest(Nest.Indices.Index(new Nest.IndexName() { Name = SearchConfigurationManager.DefaultSearchIndex }));
var indxres = searchProvider.Client.IndexExists(idr);
if (!indxres.Exists)
{
searchProvider.Client.CreateIndex(SearchConfigurationManager.DefaultSearchIndex, i => i
.Settings(s => s
.NumberOfShards(2)
.NumberOfReplicas(0)
.Analysis(analysis => analysis
.Tokenizers(tokenizers => tokenizers
.Pattern("ml-id-tokenizer", p => p.Pattern(@"\W+"))
)
.TokenFilters(tokenfilters => tokenfilters
.WordDelimiter("ml-id-words", wd => wd
.SplitOnCaseChange()
.PreserveOriginal()
.SplitOnNumerics()
.GenerateNumberParts(false)
.GenerateWordParts()
)
)
.Analyzers(analyzers => analyzers
.Custom("ml-id-analyzer", c => c
.Tokenizer("ml-id-tokenizer")
.Filters("ml-id-words", "lowercase")
)
.Custom("ml-id-keyword", c => c
.Tokenizer("keyword")
.Filters("lowercase")
)
)
)
));
}
return true;
}
正在创建文档类型如下
public bool CreateDocumentIndex()
{
bool retVal = false;
if (Common.CheckForIndex(_searchProvider))
{
var res = _searchProvider.Client.Map<LegalDocumentSearchModel>(m =>
m.Index(SearchConfigurationManager.DefaultSearchIndex)
.Type(this.type)
.AutoMap()
.Properties(ps => ps
.String(s => s
.Name(p => p.Id)
.Analyzer("ml-id-analyzer")
.Fields(f => f
.String(p => p.Name("keyword").Analyzer("ml-id-keyword"))
.String(p => p.Name("raw").Index(FieldIndexOption.NotAnalyzed))
)
)
.Completion(c => c
.Name(p => p.Suggest)
.Payloads()
)
));
retVal = res.IsValid;
}
return retVal;
}
如下创建 LegalArea 类型
public bool CreateLegalAreaIndex()
{
bool retVal = false;
if (Common.CheckForIndex(_searchProvider))
{
var res = _searchProvider.Client.Map<LegalAreaSearchModel>(m =>
m.Index(SearchConfigurationManager.DefaultSearchIndex)
.Type(this.type)
.AutoMap()
.Properties(ps => ps
.String(s => s
.Name(p => p.Id)
.Analyzer("ml-id-analyzer")
.Fields(f => f
.String(p => p.Name("keyword").Analyzer("ml-id-keyword"))
.String(p => p.Name("raw").Index(FieldIndexOption.NotAnalyzed))
)
)
.Completion(c => c
.Name(p => p.Suggest)
.Payloads()
)
));
retVal = res.IsValid;
}
return retVal;
}
现在当我运行合法区域建议如下
public List<AdvocateListingSuggestionModel> LegalAreaSuggestion(string Search)
{
List<AdvocateListingSuggestionModel> lsResult = new List<AdvocateListingSuggestionModel>();
var result = _searchProvider.Client.Suggest<LegalAreaSearchModel>(s => s
.Index(SearchConfigurationManager.DefaultSearchIndex)
.Completion("ml-la-suggestions", c => c
.Text(Search)
.Field(p => p.Suggest)
.Fuzzy(fz => fz
.Fuzziness(Fuzziness.Auto))
)
);
if (result.IsValid)
{
lsResult = result.Suggestions["ml-la-suggestions"]
.FirstOrDefault()
.Options
.Select(suggest => suggest.Payload<AdvocateListingSuggestionModel>()).ToList();
}
return lsResult;
}
我得到 LegalDocumentSearchModel 的结果作为 well.pls 建议
谢谢
建议在索引级别工作，即 one Finite State Transducer (FST) is created for a field per index 因此，如果您在同一索引中有两个具有相同字段名称的类型（两个字段 必须 相同类型），两种类型都将在一个 FST 中表示数据。
我们可以通过一个简单的例子看出这一点。这里我们有两种带有完成字段的类型，我们将它们索引到同一个索引
public class Band
{
public int Id { get; set;}
public CompletionField<object> Suggest { get; set;}
}
public class Sport
{
public int Id { get; set;}
public CompletionField<object> Suggest { get; set; }
}
void Main()
{
var pool = new SingleNodeConnectionPool(new Uri("http://localhost:9200"));
var defaultIndex = "default-index";
var connectionSettings = new ConnectionSettings(pool)
.DefaultIndex(defaultIndex);
var client = new ElasticClient(connectionSettings);
// just so we can re-run this example...
if (client.IndexExists(defaultIndex).Exists)
client.DeleteIndex(defaultIndex);
client.CreateIndex(defaultIndex, ci => ci
.Mappings(m => m
.Map<Band>(l => l
.AutoMap()
.Properties(p => p
.Completion(c => c
.Name(n => n.Suggest)
)
)
)
.Map<Sport>(l => l
.AutoMap()
.Properties(p => p
.Completion(c => c
.Name(n => n.Suggest)
)
)
)
)
);
var bands = new List<Band>
{
new Band { Id = 1, Suggest = new CompletionField<object> { Input = new [] {"Bowling for Soup"} } },
new Band { Id = 2, Suggest = new CompletionField<object> { Input = new [] {"Fastball"} } },
new Band { Id = 3, Suggest = new CompletionField<object> { Input = new [] {"Dropkick Murphys"} } },
new Band { Id = 4, Suggest = new CompletionField<object> { Input = new [] {"Yellowcard"} } },
new Band { Id = 5, Suggest = new CompletionField<object> { Input = new [] {"American Football"} } },
};
client.IndexMany(bands);
var sports = new List<Sport>
{
new Sport { Id = 1, Suggest = new CompletionField<object> { Input = new [] {"Bowling"} } },
new Sport { Id = 2, Suggest = new CompletionField<object> { Input = new [] {"Football"} } },
new Sport { Id = 3, Suggest = new CompletionField<object> { Input = new [] {"Baseball"} } },
new Sport { Id = 4, Suggest = new CompletionField<object> { Input = new [] {"Table Tennis"} } },
new Sport { Id = 5, Suggest = new CompletionField<object> { Input = new [] {"American Football"} } },
};
client.IndexMany(sports);
client.Refresh(defaultIndex);
var suggestResponse = client.Suggest<Band>(s => s
.Completion("suggestion", cs => cs
.Text("Bo")
.Field(f => f.Suggest)
)
);
}
我们从我们的建议调用中得到以下结果
{
"_shards" : {
"total" : 5,
"successful" : 5,
"failed" : 0
},
"suggestion" : [ {
"text" : "Bo",
"offset" : 0,
"length" : 2,
"options" : [ {
"text" : "Bowling",
"score" : 1.0
}, {
"text" : "Bowling for Soup",
"score" : 1.0
} ]
} ]
}
我们在结果中得到了 Band 类型和 Sport 类型的建议。也许不是我们所期望或想要的，但这就是建议者的工作方式。请注意，在 .Suggest<T>() 调用中，类型 T 用于提供对可能在查询中使用的属性的强类型访问。在此示例中，类型为 Band 的 .Suggest 字段。
这里推荐的解决方案是为不同的类型设置单独的索引；您仍然可以在需要时跨多个索引进行查询，并避免在同一索引中包含多种类型的陷阱。
除此之外，您可以通过对完成类型映射使用 Context Suggester and taking advantage of category context 来实现它。理想情况下，您希望使用类别文档中存在的字段来为您提供类别的灵活性，但如果需要，您可以使用元数据字段。我建议针对您的领域调查和研究方法。
对于我们之前的示例，如果我们现在在映射中包含类别上下文（与以前相同的代码，但具有此更新的映射）
client.CreateIndex(defaultIndex, ci => ci
.Mappings(m => m
.Map<Band>(l => l
.AutoMap()
.Properties(p => p
.Completion(c => c
.Name(n => n.Suggest)
.Context(sc => sc
.Category("type", csc => csc
// recommend that you use another field on
// the document here instead of a metadata field
.Field("_type")
)
)
)
)
)
.Map<Sport>(l => l
.AutoMap()
.Properties(p => p
.Completion(c => c
.Name(n => n.Suggest)
.Context(sc => sc
.Category("type", csc => csc
// recommend that you use another field on
// the document here instead of a metadata field
.Field("_type")
)
)
)
)
)
)
);
并在搜索时使用类别上下文
var suggestResponse = client.Suggest<Band>(s => s
.Completion("suggestion", cs => cs
.Text("Bo")
.Field(f => f.Suggest)
.Context(d => d.Add("type", "band"))
)
);
然后我们得到了想要的结果
{
"_shards" : {
"total" : 5,
"successful" : 5,
"failed" : 0
},
"suggestion" : [ {
"text" : "Bo",
"offset" : 0,
"length" : 2,
"options" : [ {
"text" : "Bowling for Soup",
"score" : 1.0
} ]
} ]
}