一尘不染

Elasticsearch:聚合,按字段计数

elasticsearch

我将此数据插入了Elasticsearch:

[
  { "name": "Cassandra Irwin",  "location": "Monzon de Campos" ..     },
  { "name": "Gayle Mooney",     "location": "Villarroya del Campo" .. },
  { "name": "Angelita Charles", "location": "Revenga de Campos" ..    }, 
  { "name": "Sheppard Sweet",   "location": "Santiago del Campo" ..   },
  ..
  ..

旁注:重现:
1)下载:http://wmo.co/20160928_es_query/bulk.json
2)执行:卷曲-s -XPOST ‘ 的http://本地主机:9200 /测试/外部/
_bulk漂亮
‘ -数据二进制@
bulk.json

问题: 获取每个“位置”有多少记录的计数。

解决方案1:存储桶聚合..没有得到期望的结果

curl -s -XPOST 'localhost:9200/testing/_search?pretty' -d '
{
  "aggs": {  "location_count": { "terms": { "field":"location",   "size":100 }}}
}' | jq  '.aggregations'

结果:

{"location_count":{"doc_count_error_upper_bound":0,"sum_other_doc_count":0,
 "buckets":[
    {"key":"campo",     "doc_count":47},
    {"key":"del",       "doc_count":47},
    {"key":"campos",    "doc_count":29},
    {"key":"de",        "doc_count":29},
    {"key":"villarroya","doc_count":15},
    {"key":"torre",     "doc_count":12},
    {"key":"monzon",    "doc_count":11},
    {"key":"santiago",  "doc_count":11},
    {"key":"pina",      "doc_count":9},
    {"key":"revenga",   "doc_count":9},
    {"key":"uleila",    "doc_count":9}
]}}

问题 :它将“位置”字段拆分为单词,然后每个单词返回一个文档计数。

解决方案2:期望的结果,但性能令人担忧。

我可以使用此查询来做到这一点,提取所有位置并在jq(每个方便的JSON cli工具)中进行聚合,但是当应用于大量数据时,这可能会成为性能噩梦:

curl -s -XPOST 'localhost:9200/testing/_search?pretty' -d '
 {
   "query": { "wildcard": { "location": "*" } }, "size":1000,
   "_source": ["location"]
 }' | jq  '[.hits.hits[] |
           {location:._source.location,"count":1}] |
           group_by(.location) |
           map({ key: .[0].location, value: map(.count)|add })'

结果:

[
  { "key": "Monzon de Campos",      "value": 11 },
  { "key": "Pina de Campos",        "value": 9  },
  { "key": "Revenga de Campos",     "value": 9  },
  { "key": "Santiago del Campo",    "value": 11 },
  { "key": "Torre del Campo",       "value": 12 },
  { "key": "Uleila del Campo",      "value": 9  },
  { "key": "Villarroya del Campo",  "value": 15 }
]

这是我想要的确切结果。

问题:如何通过elasticsearch查询获得相同的结果? (即通过elasticsearch而不是jq处理聚合)


阅读 1416

收藏
2020-06-22

共1个答案

一尘不染

您需要在not_analyzed您的location字段中添加一个子字段。

首先像这样修改您的映射:

curl -XPOST 'http://localhost:9200/testing/_mapping/external' -d '{
   "properties": {
      "location": {
         "type": "string",
         "fields": {
            "raw": {
               "type": "string",
               "index": "not_analyzed"
            }
         }
      }
   }
}'

然后再次为您的数据重新编制索引:

curl -s -XPOST 'http://localhost:9200/testing/external/_bulk?pretty' --data-binary @bulk.json

最后,您将能够像这样(在location.raw字段上)运行查询并获得您期望的结果:

curl -s -XPOST 'localhost:9200/testing/_search?pretty' -d '
{
  "aggs": {  "location_count": { "terms": { "field":"location.raw",   "size":100 }}}
}' | jq  '.aggregations'
2020-06-22