NEST似乎不支持此处描述的pattern replace char filter：
我在https://github.com/elasticsearch/elasticsearch-net/issues/543创建了一个问题。
我的大部分索引都在工作，所以我想继续使用NEST。有没有办法在索引配置期间使用某些手动json注入来解决这个问题？我是NEST的新手，所以不确定这是否可行。
具体而言，我希望在通过自定义分析器运行之前，使用pattern replace char filter从街道地址中删除单元号（即＃205 - 1260 Broadway 变为< em> 1260 Broadway ）。由于自定义分析器，我相信我需要使用此char过滤器来完成此任务。
我目前的配置如下：
elasticClient.CreateIndex("geocoding", c => c
.Analysis(ad => ad
.Analyzers(ab => ab
.Add("address-index", new CustomAnalyzer()
{
Tokenizer = "whitespace",
Filter = new List<string>() { "lowercase", "synonym" }
})
.Add("address-search", new CustomAnalyzer()
{
Tokenizer = "whitespace",
Filter = new List<string>() { "lowercase" },
CharFilter = new List<string>() { "drop-unit" }
})
)
.CharFilters(cfb => cfb
.Add("drop-unit", new CharFilter()) //missing char filter here
)
.TokenFilters(tfb => tfb
.Add("synonym", new SynonymTokenFilter()
{
Expand = true,
SynonymsPath = "analysis/synonym.txt"
})
)
)
更新：
截至2014年5月，NEST现在支持pattern replace char filter：https://github.com/elasticsearch/elasticsearch-net/pull/637
答案 0 :(得分：1)
您可以使用Settings.Add方法以更加手动的方式添加到FluentDictionary，而不是在索引创建过程中使用流畅设置，但可以完全控制传入的设置。例如，显示在Create Index的NEST Documentation中。我使用这种方法的原因非常相似。
您的配置看起来类似于以下内容：
elasticClient.CreateIndex("geocoding", c => c.
.Settings(s => s.
.Add("analysis.analyzer.address-index.type", "custom")
.Add("analysis.analyzer.address-index.tokenizer", "whitespace")
.Add("analysis.analyzer.address-index.filter.0", "lowercase")
.Add("analysis.analyzer.address-index.filter.1", "synonym")
.Add("anaylsis.analyzer.address-search.type", "custom")
.Add("analysis.analyzer.address-search.tokenizer", "whitespace")
.Add("analysis.analyzer.address-search.filter.0", "lowercase")
.Add("analysis.analyzer.address-search.char_filter.0", "drop-unit")
.Add("analysis.char_filter.drop-unit.type", "mapping")
.Add("analysis.char_filter.drop-unit.mappings.0", "<mapping1>")
.Add("analysis.char_filter.drop-unit.mappings.1", "<mapping2>")
...
)
);
您需要将上面的<mapping1>和<mapping2>替换为您要使用的实际char_filter映射。请注意，之前我没有使用过char_filter，所以设置值可能有点偏，但应该让你朝着正确的方向前进。
答案 1 :(得分：0)
为了提供Paige非常有用的答案的后续行动，看起来您可以结合流利和手动Settings.Add方法。以下对我有用：
elasticClient.CreateIndex("geocoding", c => c
.Settings(s => s
.Add("analysis.char_filter.drop_unit.type", "pattern_replace")
.Add("analysis.char_filter.drop_unit.pattern", @"#\d+\s-\s")
.Add("analysis.char_filter.drop_unit.replacement", "")
)
.Analysis(ad => ad
.Analyzers(ab => ab
.Add("address_index", new CustomAnalyzer()
{
Tokenizer = "whitespace",
Filter = new List<string>() { "lowercase", "synonym" }
})
.Add("address_search", new CustomAnalyzer()
{
CharFilter = new List<string> { "drop_unit" },
Tokenizer = "whitespace",
Filter = new List<string>() { "lowercase" }
})
)
.TokenFilters(tfb => tfb
.Add("synonym", new SynonymTokenFilter()
{
Expand = true,
SynonymsPath = "analysis/synonym.txt"
})
)
)
答案 2 :(得分：0)
EsClient.CreateIndex("universal_de", c => c
.NumberOfReplicas(1)
.NumberOfShards(5)
.Settings(s => s //just as an example
.Add("merge.policy.merge_factor", "10")
.Add("search.slowlog.threshold.fetch.warn", "1s")
.Add("analysis.char_filter.drop_chars.type", "pattern_replace")
.Add("analysis.char_filter.drop_chars.pattern", @"[^0-9]")
.Add("analysis.char_filter.drop_chars.replacement", "")
.Add("analysis.char_filter.drop_specChars.type", "pattern_replace")
.Add("analysis.char_filter.drop_specChars.pattern", @"[^0-9a-zA-Z]")
.Add("analysis.char_filter.drop_specChars.replacement", "")
)
.Analysis(descriptor => descriptor
.Analyzers(bases => bases
.Add("folded_word", new CustomAnalyzer()
{
Filter = new List<string> { "lowercase", "asciifolding", "trim" },
Tokenizer = "standard"
}
)
.Add("trimmed_number", new CustomAnalyzer()
{
CharFilter = new List<string> { "drop_chars" },
Tokenizer = "standard",
Filter = new List<string>() { "lowercase" }
})
.Add("trimmed_specChars", new CustomAnalyzer()
{
CharFilter = new List<string> { "drop_specChars" },
Tokenizer = "standard",
Filter = new List<string>() { "lowercase" }
})
)
)
.AddMapping<Business>(m => m
//.MapFromAttributes()
.Properties(props => props
.MultiField(mf => mf
.Name(t => t.DirectoryName)
.Fields(fs => fs
.String(s => s.Name(t => t.DirectoryName).Analyzer("standard"))
.String(s => s.Name(t => t.DirectoryName.Suffix("folded")).Analyzer("folded_word"))
)
)
.MultiField(mf => mf
.Name(t => t.Phone)
.Fields(fs => fs
.String(s => s.Name(t => t.Phone).Analyzer("trimmed_number"))
)
)
这是您创建索引并添加映射的方法。 现在搜索我有这样的事情：
var result = _Instance.Search<Business>(q => q
.TrackScores(true)
.Query(qq =>
{
QueryContainer termQuery = null;
if (!string.IsNullOrWhiteSpace(input.searchTerm))
{
var toLowSearchTerm = input.searchTerm.ToLower();
termQuery |= qq.QueryString(qs => qs
.OnFieldsWithBoost(f => f
.Add("directoryName.folded", 5.0)
)
.Query(toLowSearchTerm));
termQuery |= qq.Fuzzy(fz => fz.OnField("directoryName.folded").Value(toLowSearchTerm).MaxExpansions(2));
termQuery |= qq.Term("phone", Regex.Replace(toLowSearchTerm, @"[^0-9]", ""));
}
return termQuery;
})
.Skip(input.skip)
.Take(input.take)
);
新：我设法以更好的方式使用模式替换：
.Analysis(descriptor => descriptor
.Analyzers(bases => bases
.Add("folded_word", new CustomAnalyzer()
{
Filter = new List<string> { "lowercase", "asciifolding", "trim" },
Tokenizer = "standard"
}
)
.Add("trimmed_number", new CustomAnalyzer()
{
CharFilter = new List<string> { "drop_chars" },
Tokenizer = "standard",
Filter = new List<string>() { "lowercase" }
})
.Add("trimmed_specChars", new CustomAnalyzer()
{
CharFilter = new List<string> { "drop_specChars" },
Tokenizer = "standard",
Filter = new List<string>() { "lowercase" }
})
.Add("autocomplete", new CustomAnalyzer()
{
Tokenizer = new WhitespaceTokenizer().Type,
Filter = new List<string>() { "lowercase", "asciifolding", "trim", "engram" }
}
)
)
.TokenFilters(i => i
.Add("engram", new EdgeNGramTokenFilter
{
MinGram = 3,
MaxGram = 15
}
)
)
.CharFilters(cf => cf
.Add("drop_chars", new PatternReplaceCharFilter
{
Pattern = @"[^0-9]",
Replacement = ""
}
)
.Add("drop_specChars", new PatternReplaceCharFilter
{
Pattern = @"[^0-9a-zA-Z]",
Replacement = ""
}
)
)
)