仅针对具有过滤器的排序查询结果聚合脚本，而不是完整数据集

debugcn 发表于 Dev

亚当

仅供参考-elasticsearch @ v1.5; npm elasticsearch @ 4.0.2

对于我的特定用例，我需要找到其他点附近的五个最近点，并计算这五个结果的最大距离。由于某种原因，我在下面的查询返回的是所有过滤数据的最大距离，而不是最近的五个。

到目前为止，这是我的查询：

    elasticsearchAPI = Meteor.npmRequire('elasticsearch');
    esClient = new elasticsearchAPI.Client({
        host: 'myHost'
    });

    var esQueryObject = {
        "index": "ma_homes",
        "size": 5,
        "body": {
            "query": {
                "filtered": {
                    "query": {
                        "match_all": {}
                    },
                    "filter": {
                        "geo_distance": {
                            "LOCATION": {
                                "lat": 42.5125339,
                                "lon": -71.06748
                            },
                            "distance": "3mi",
                            "optimize_bbox": "memory"
                        }
                    }
                }
            },
            "size": 5,
            "sort": [{
                "_geo_distance": {
                    "LOCATION": {
                        "lat": 42.5125339,
                        "lon": -71.06748
                    },
                    "order": "asc",
                    "unit": "mi",
                    "distance_type": "sloppy_arc"
                }
            }],
            "fields": ["F1_V7_2_F1TOWN"],
            "aggs": {
                "max_dist": {
                    "max": {
                        "script": "doc[\u0027LOCATION\u0027].arcDistanceInMiles(lat,lon)",
                        "params" : {
                            "lat" : 42.5125339,
                            "lon" : -71.06748
                        }
                    }
                }
            }
        }
    }



     try {
        esClient.search(esQueryObject, function(err, res) {
            if ( err ) console.log("err: ", err);
            if ( res ) {
                console.log("res: ", JSON.stringify(res, null, "\t"));
            };
        });
    }
    catch(error) {
        console.log("search err: ", error);
    };

我的问题是返回的max_dist为2.99，但是从点击数中我可以清楚地看到它应该仅为0.02268！

最后，有没有更好的方法来计算最大距离？我不必使用脚本。

查看下面的结果：

I20160729-14:46:08.447(-7)? {
I20160729-14:46:08.447(-7)?     "took": 119,
I20160729-14:46:08.447(-7)?     "timed_out": false,
I20160729-14:46:08.447(-7)?     "_shards": {
I20160729-14:46:08.447(-7)?         "total": 5,
I20160729-14:46:08.448(-7)?         "successful": 5,
I20160729-14:46:08.448(-7)?         "failed": 0
I20160729-14:46:08.448(-7)?     },
I20160729-14:46:08.448(-7)?     "hits": {
I20160729-14:46:08.448(-7)?         "total": 19428,
I20160729-14:46:08.448(-7)?         "max_score": null,
I20160729-14:46:08.452(-7)?         "hits": [
I20160729-14:46:08.452(-7)?             {
I20160729-14:46:08.452(-7)?                 "_index": "ma_homes",
I20160729-14:46:08.452(-7)?                 "_type": "home",
I20160729-14:46:08.453(-7)?                 "_id": "AVY1KqHN5rKRAKXZHxQf",
I20160729-14:46:08.453(-7)?                 "_score": null,
I20160729-14:46:08.453(-7)?                 "fields": {
I20160729-14:46:08.453(-7)?                     "F1_V7_2_F1TOWN": [
I20160729-14:46:08.453(-7)?                         "7WHITECIRWAKEFIELDMA"
I20160729-14:46:08.454(-7)?                     ]
I20160729-14:46:08.454(-7)?                 },
I20160729-14:46:08.454(-7)?                 "sort": [
I20160729-14:46:08.454(-7)?                     0.013847018573431258
I20160729-14:46:08.454(-7)?                 ]
I20160729-14:46:08.455(-7)?             },
I20160729-14:46:08.455(-7)?             {
I20160729-14:46:08.455(-7)?                 "_index": "ma_homes",
I20160729-14:46:08.455(-7)?                 "_type": "home",
I20160729-14:46:08.456(-7)?                 "_id": "AVY1Ewoc5rKRAKXZGhMp",
I20160729-14:46:08.456(-7)?                 "_score": null,
I20160729-14:46:08.456(-7)?                 "fields": {
I20160729-14:46:08.456(-7)?                     "F1_V7_2_F1TOWN": [
I20160729-14:46:08.456(-7)?                         "8WHITECIRWAKEFIELDMA"
I20160729-14:46:08.457(-7)?                     ]
I20160729-14:46:08.457(-7)?                 },
I20160729-14:46:08.457(-7)?                 "sort": [
I20160729-14:46:08.458(-7)?                     0.01675513175670524
I20160729-14:46:08.458(-7)?                 ]
I20160729-14:46:08.458(-7)?             },
I20160729-14:46:08.458(-7)?             {
I20160729-14:46:08.458(-7)?                 "_index": "ma_homes",
I20160729-14:46:08.459(-7)?                 "_type": "home",
I20160729-14:46:08.459(-7)?                 "_id": "AVY1T0cn5rKRAKXZJwC8",
I20160729-14:46:08.459(-7)?                 "_score": null,
I20160729-14:46:08.459(-7)?                 "fields": {
I20160729-14:46:08.459(-7)?                     "F1_V7_2_F1TOWN": [
I20160729-14:46:08.460(-7)?                         "10WHITECIRWAKEFIELDMA"
I20160729-14:46:08.460(-7)?                     ]
I20160729-14:46:08.460(-7)?                 },
I20160729-14:46:08.460(-7)?                 "sort": [
I20160729-14:46:08.461(-7)?                     0.018417500448048605
I20160729-14:46:08.461(-7)?                 ]
I20160729-14:46:08.463(-7)?             },
I20160729-14:46:08.464(-7)?             {
I20160729-14:46:08.464(-7)?                 "_index": "ma_homes",
I20160729-14:46:08.464(-7)?                 "_type": "home",
I20160729-14:46:08.464(-7)?                 "_id": "AVY1Xb2P5rKRAKXZKhUh",
I20160729-14:46:08.464(-7)?                 "_score": null,
I20160729-14:46:08.465(-7)?                 "fields": {
I20160729-14:46:08.465(-7)?                     "F1_V7_2_F1TOWN": [
I20160729-14:46:08.465(-7)?                         "11WHITECIRWAKEFIELDMA"
I20160729-14:46:08.465(-7)?                     ]
I20160729-14:46:08.466(-7)?                 },
I20160729-14:46:08.466(-7)?                 "sort": [
I20160729-14:46:08.466(-7)?                     0.018816876925529115
I20160729-14:46:08.467(-7)?                 ]
I20160729-14:46:08.467(-7)?             },
I20160729-14:46:08.467(-7)?             {
I20160729-14:46:08.468(-7)?                 "_index": "ma_homes",
I20160729-14:46:08.468(-7)?                 "_type": "home",
I20160729-14:46:08.468(-7)?                 "_id": "AVY1TNJh5rKRAKXZJnx0",
I20160729-14:46:08.468(-7)?                 "_score": null,
I20160729-14:46:08.469(-7)?                 "fields": {
I20160729-14:46:08.469(-7)?                     "F1_V7_2_F1TOWN": [
I20160729-14:46:08.470(-7)?                         "6WHITECIRWAKEFIELDMA"
I20160729-14:46:08.470(-7)?                     ]
I20160729-14:46:08.470(-7)?                 },
I20160729-14:46:08.471(-7)?                 "sort": [
I20160729-14:46:08.471(-7)?                     0.022680252269458714
I20160729-14:46:08.471(-7)?                 ]
I20160729-14:46:08.471(-7)?             }
I20160729-14:46:08.471(-7)?         ]
I20160729-14:46:08.472(-7)?     },
I20160729-14:46:08.472(-7)?     "aggregations": {
I20160729-14:46:08.472(-7)?         "max_dist": {
I20160729-14:46:08.472(-7)?             "value": 2.999906924854209,
I20160729-14:46:08.473(-7)?             "value_as_string": "2.999906924854209"
I20160729-14:46:08.473(-7)?         }
I20160729-14:46:08.473(-7)?     }
I20160729-14:46:08.474(-7)? }

pickypg

这里有两件事是错误的，第二件事与第一件事紧密相关：

您假设排序顺序对聚合有任何影响。没有。您可能需要看一下Elasticsearch：确定聚合范围的权威指南。
- 要点是，查询的总结果（包括未返回的命中数）是聚合范围的一部分。在您的确切情况下，它指出存在"total": 19428与您的搜索匹配的文档。您刚得到最近的5。
您正在按升序排序，这意味着它从最小到最大排序。这意味着您只会得到最接近的5个最接近的距离，这就是您想要的，但这并不意味着聚合所看到的就是真实的最大距离。

对于这些问题，您需要弄清楚如何限制前5名，或者根本不进行汇总，我认为这是最简单的方法。只需获得前5名，然后获取最后一个值，就可以完成两个所需的答案。

由于3英里，排序被限制在3英里之内，这很好，但是也许您可以根据需要通过使用更快的搜索来做一些更好的事情distance_type：

{
  "size": 5,
  "_source": "F1_V7_2_F1TOWN",
  "query": {
    "filtered": {
      "filter": [
        {
          "geo_distance": {
            "LOCATION": {
              "lat": 42.5125339,
              "lon": -71.06748
            },
            "distance": "3mi",
            "distance_type": "plane"
          }
        }
      ]
    }
  },
  "sort": [
    {
      "_geo_distance": {
        "LOCATION": {
          "lat": 42.5125339,
          "lon": -71.06748
        },
        "order": "asc",
        "unit": "mi",
        "distance_type": "sloppy_arc"
      }
    }
  ]
}

注意，我不进行汇总，而是使用_source代替fields（fields用于存储的字段，而不是限制源文档的输出），我切换到使用plane过滤器，distance_type因为对于极点以外的短距离它更快。我怀疑太多的房屋会在两极使用距离。为了计分，我将其保留为sloppy_arc，因为它在过滤后可以使用稍微完善一些的方程式。