IMPORTANT: No additional bug fixes or documentation updates will be released for this version. For the latest information, see the current release documentation.

« Search multiple data streams and indices Filter search results »

› › ›

Collapse search results

edit

Collapse search results

edit

You can use the collapse parameter to collapse search results based on field values. The collapsing is done by selecting only the top sorted document per collapse key.

For example, the following search collapses results by user.id and sorts them by http.response.bytes.

resp = client.search(
    index="my-index-000001",
    query={
        "match": {
            "message": "GET /search"
        }
    },
    collapse={
        "field": "user.id"
    },
    sort=[
        {
            "http.response.bytes": {
                "order": "desc"
            }
        }
    ],
    from_=0,
)
print(resp)

const response = await client.search({
  index: "my-index-000001",
  query: {
    match: {
      message: "GET /search",
    },
  },
  collapse: {
    field: "user.id",
  },
  sort: [
    {
      "http.response.bytes": {
        order: "desc",
      },
    },
  ],
  from: 0,
});
console.log(response);

GET my-index-000001/_search
{
  "query": {
    "match": {
      "message": "GET /search"
    }
  },
  "collapse": {
    "field": "user.id"         
  },
  "sort": [
    {
      "http.response.bytes": { 
        "order": "desc"
      }
    }
  ],
  "from": 0                    
}

	Collapse the result set using the `user.id` field
	Sort the results by `http.response.bytes`
	Define the offset of the first collapsed result

The total number of hits in the response indicates the number of matching documents without collapsing. The total number of distinct group is unknown.

The field used for collapsing must be a single valued keyword or numeric field with doc_values activated.

Collapsing is applied to the top hits only and does not affect aggregations.

Expand collapse results

edit

It is also possible to expand each collapsed top hits with the inner hits option.

resp = client.search(
    index="my-index-000001",
    query={
        "match": {
            "message": "GET /search"
        }
    },
    collapse={
        "field": "user.id",
        "inner_hits": {
            "name": "most_recent",
            "size": 5,
            "sort": [
                {
                    "@timestamp": "desc"
                }
            ]
        },
        "max_concurrent_group_searches": 4
    },
    sort=[
        {
            "http.response.bytes": {
                "order": "desc"
            }
        }
    ],
)
print(resp)

const response = await client.search({
  index: "my-index-000001",
  query: {
    match: {
      message: "GET /search",
    },
  },
  collapse: {
    field: "user.id",
    inner_hits: {
      name: "most_recent",
      size: 5,
      sort: [
        {
          "@timestamp": "desc",
        },
      ],
    },
    max_concurrent_group_searches: 4,
  },
  sort: [
    {
      "http.response.bytes": {
        order: "desc",
      },
    },
  ],
});
console.log(response);

GET /my-index-000001/_search
{
  "query": {
    "match": {
      "message": "GET /search"
    }
  },
  "collapse": {
    "field": "user.id",                       
    "inner_hits": {
      "name": "most_recent",                  
      "size": 5,                              
      "sort": [ { "@timestamp": "desc" } ]    
    },
    "max_concurrent_group_searches": 4        
  },
  "sort": [
    {
      "http.response.bytes": {
        "order": "desc"
      }
    }
  ]
}

	Collapse the result set using the `user.id` field
	The name used for the inner hit section in the response
	The number of `inner_hits` to retrieve per collapse key
	How to sort the document inside each group
	The number of concurrent requests allowed to retrieve the `inner_hits` per group

See inner hits for the complete list of supported options and the format of the response.

It is also possible to request multiple inner hits for each collapsed hit. This can be useful when you want to get multiple representations of the collapsed hits.

resp = client.search(
    index="my-index-000001",
    query={
        "match": {
            "message": "GET /search"
        }
    },
    collapse={
        "field": "user.id",
        "inner_hits": [
            {
                "name": "largest_responses",
                "size": 3,
                "sort": [
                    {
                        "http.response.bytes": {
                            "order": "desc"
                        }
                    }
                ]
            },
            {
                "name": "most_recent",
                "size": 3,
                "sort": [
                    {
                        "@timestamp": {
                            "order": "desc"
                        }
                    }
                ]
            }
        ]
    },
    sort=[
        "http.response.bytes"
    ],
)
print(resp)

const response = await client.search({
  index: "my-index-000001",
  query: {
    match: {
      message: "GET /search",
    },
  },
  collapse: {
    field: "user.id",
    inner_hits: [
      {
        name: "largest_responses",
        size: 3,
        sort: [
          {
            "http.response.bytes": {
              order: "desc",
            },
          },
        ],
      },
      {
        name: "most_recent",
        size: 3,
        sort: [
          {
            "@timestamp": {
              order: "desc",
            },
          },
        ],
      },
    ],
  },
  sort: ["http.response.bytes"],
});
console.log(response);

GET /my-index-000001/_search
{
  "query": {
    "match": {
      "message": "GET /search"
    }
  },
  "collapse": {
    "field": "user.id",                   
    "inner_hits": [
      {
        "name": "largest_responses",      
        "size": 3,
        "sort": [
          {
            "http.response.bytes": {
              "order": "desc"
            }
          }
        ]
      },
      {
        "name": "most_recent",             
        "size": 3,
        "sort": [
          {
            "@timestamp": {
              "order": "desc"
            }
          }
        ]
      }
    ]
  },
  "sort": [
    "http.response.bytes"
  ]
}

	Collapse the result set using the `user.id` field
	Return the three largest HTTP responses for the user
	Return the three most recent HTTP responses for the user

The expansion of the group is done by sending an additional query for each inner_hit request for each collapsed hit returned in the response. This can significantly slow your search if you have too many groups or inner_hit requests.

The max_concurrent_group_searches request parameter can be used to control the maximum number of concurrent searches allowed in this phase. The default is based on the number of data nodes and the default search thread pool size.

collapse cannot be used in conjunction with scroll.

Collapsing with `search_after`

edit

Field collapsing can be used with the search_after parameter. Using search_after is only supported when sorting and collapsing on the same field. Secondary sorts are also not allowed. For example, we can collapse and sort on user.id, while paging through the results using search_after:

resp = client.search(
    index="my-index-000001",
    query={
        "match": {
            "message": "GET /search"
        }
    },
    collapse={
        "field": "user.id"
    },
    sort=[
        "user.id"
    ],
    search_after=[
        "dd5ce1ad"
    ],
)
print(resp)

const response = await client.search({
  index: "my-index-000001",
  query: {
    match: {
      message: "GET /search",
    },
  },
  collapse: {
    field: "user.id",
  },
  sort: ["user.id"],
  search_after: ["dd5ce1ad"],
});
console.log(response);

GET /my-index-000001/_search
{
  "query": {
    "match": {
      "message": "GET /search"
    }
  },
  "collapse": {
    "field": "user.id"
  },
  "sort": [ "user.id" ],
  "search_after": ["dd5ce1ad"]
}

Rescore collapse results

edit

You can use field collapsing alongside the rescore search parameter. Rescorers run on every shard for the top-ranked document per collapsed field. To maintain a reliable order, it is recommended to cluster documents sharing the same collapse field value on one shard. This is achieved by assigning the collapse field value as the routing key during indexing:

resp = client.index(
    index="my-index-000001",
    routing="xyz",
    document={
        "@timestamp": "2099-11-15T13:12:00",
        "message": "You know for search!",
        "user.id": "xyz"
    },
)
print(resp)

const response = await client.index({
  index: "my-index-000001",
  routing: "xyz",
  document: {
    "@timestamp": "2099-11-15T13:12:00",
    message: "You know for search!",
    "user.id": "xyz",
  },
});
console.log(response);

POST /my-index-000001/_doc?routing=xyz      
{
  "@timestamp": "2099-11-15T13:12:00",
  "message": "You know for search!",
  "user.id": "xyz"
}

Assign routing with the collapse field value (user.id).

By doing this, you guarantee that only one top document per collapse key gets rescored globally.

The following request utilizes field collapsing on the user.id field and then rescores the top groups with a query rescorer:

resp = client.search(
    index="my-index-000001",
    query={
        "match": {
            "message": "you know for search"
        }
    },
    collapse={
        "field": "user.id"
    },
    rescore={
        "window_size": 50,
        "query": {
            "rescore_query": {
                "match_phrase": {
                    "message": "you know for search"
                }
            },
            "query_weight": 0.3,
            "rescore_query_weight": 1.4
        }
    },
)
print(resp)

const response = await client.search({
  index: "my-index-000001",
  query: {
    match: {
      message: "you know for search",
    },
  },
  collapse: {
    field: "user.id",
  },
  rescore: {
    window_size: 50,
    query: {
      rescore_query: {
        match_phrase: {
          message: "you know for search",
        },
      },
      query_weight: 0.3,
      rescore_query_weight: 1.4,
    },
  },
});
console.log(response);

GET /my-index-000001/_search
{
  "query": {
    "match": {
      "message": "you know for search"
    }
  },
  "collapse": {
    "field": "user.id"
  },
  "rescore" : {
      "window_size" : 50,
      "query" : {
         "rescore_query" : {
            "match_phrase": {
                "message": "you know for search"
            }
         },
         "query_weight" : 0.3,
         "rescore_query_weight" : 1.4
      }
   }
}

Rescorers are not applied to inner hits.

Second level of collapsing

edit

A second level of collapsing is also supported and is applied to inner_hits.

For example, the following search collapses results by geo.country_name. Within each geo.country_name, inner hits are collapsed by user.id.

Second level of collapsing doesn’t allow inner_hits.

resp = client.search(
    index="my-index-000001",
    query={
        "match": {
            "message": "GET /search"
        }
    },
    collapse={
        "field": "geo.country_name",
        "inner_hits": {
            "name": "by_location",
            "collapse": {
                "field": "user.id"
            },
            "size": 3
        }
    },
)
print(resp)

const response = await client.search({
  index: "my-index-000001",
  query: {
    match: {
      message: "GET /search",
    },
  },
  collapse: {
    field: "geo.country_name",
    inner_hits: {
      name: "by_location",
      collapse: {
        field: "user.id",
      },
      size: 3,
    },
  },
});
console.log(response);

GET /my-index-000001/_search
{
  "query": {
    "match": {
      "message": "GET /search"
    }
  },
  "collapse": {
    "field": "geo.country_name",
    "inner_hits": {
      "name": "by_location",
      "collapse": { "field": "user.id" },
      "size": 3
    }
  }
}

{
  "hits" : {
    "hits" : [
      {
        "_index" : "my-index-000001",
        "_id" : "oX9uXXoB0da05OCR3adK",
        "_score" : 0.5753642,
        "_source" : {
          "@timestamp" : "2099-11-15T14:12:12",
          "geo" : {
            "country_name" : "Amsterdam"
          },
          "http" : {
            "request" : {
              "method" : "get"
            },
            "response" : {
              "bytes" : 1070000,
              "status_code" : 200
            },
            "version" : "1.1"
          },
          "message" : "GET /search HTTP/1.1 200 1070000",
          "source" : {
            "ip" : "127.0.0.1"
          },
          "user" : {
            "id" : "kimchy"
          }
        },
        "fields" : {
          "geo.country_name" : [
            "Amsterdam"
          ]
        },
        "inner_hits" : {
          "by_location" : {
            "hits" : {
              "total" : {
                "value" : 1,
                "relation" : "eq"
              },
              "max_score" : 0.5753642,
              "hits" : [
                {
                  "_index" : "my-index-000001",
                  "_id" : "oX9uXXoB0da05OCR3adK",
                  "_score" : 0.5753642,
                  "_source" : {
                    "@timestamp" : "2099-11-15T14:12:12",
                    "geo" : {
                      "country_name" : "Amsterdam"
                    },
                    "http" : {
                      "request" : {
                        "method" : "get"
                      },
                      "response" : {
                        "bytes" : 1070000,
                        "status_code" : 200
                      },
                      "version" : "1.1"
                    },
                    "message" : "GET /search HTTP/1.1 200 1070000",
                    "source" : {
                      "ip" : "127.0.0.1"
                    },
                    "user" : {
                      "id" : "kimchy"
                    }
                  },
                  "fields" : {
                    "user.id" : [
                      "kimchy"
                    ]
                  }
                }
              ]
            }
          }
        }
      }
    ]
  }
}

Track Scores

edit

When collapse is used with sort on a field, scores are not computed. Setting track_scores to true instructs Elasticsearch to compute and track scores.

« Search multiple data streams and indices Filter search results »

Collapse search results

Collapse search results

Expand collapse results

Collapsing with search_after

Rescore collapse results

Second level of collapsing

Track Scores

Collapsing with `search_after`