考虑分数的Elasticsearch汇总

Question

I have a document with the columns:

• fulltext column called 'content'
• employees (one or more)
• ...

I made this query "Michael Seam vacation ", so elasticsearch returned a thousand of results.

The results of the query are OK. First I received the vacations of Michael Seam, and then, results of vacations for the others employees

In the results, there are contents with the term "vacation" for dozens of the employees, like:

• Michael Seam Porter (1 hits)
• Michael Seam Carl (3 hits)
• Lucas (30 hits)
• Maria Fuch (27 hits)
• Jose White (15 hits)
• ...

When I put an aggregation for the employee column, I received Lucas, Maria and others before Michael Seam Porter, and Michael Seam Carl, and sometimes they even appear because of the aggregation size.

How can I show (in the aggregation) only the employees that are contained in the query ? Is it possible?

PS: I'm working with ES 1.7.5

Answer 1

I found one way to do it.

"aggregatePerEmployee" : {
    "terms" : {
        "field" : "employee.raw",
        "order": {
            "top_hit": "desc"
        },
        "size" : 4
    },
    "aggs": {
        "top_hit" : {
            "max": {
                "script": "_score"
            }
        }
    }
}

With this, the aggregation order consider the top score of each employee.

See the results

"aggregatePerEmployee": {
  "doc_count_error_upper_bound": -1,
  "sum_other_doc_count": 1145,
  "buckets": [
    {
      "key": "Michael Seam Carl",
      "doc_count": 3,
      "top_hit": {
        "value": 2.097010612487793
      }
    },
    {
      "key": "Michael Seam Porter ",
      "doc_count": 1,
      "top_hit": {
        "value": 2.0433993339538574
      }
    },
    {
      "key": "Lucas",
      "doc_count": 30,
      "top_hit": {
        "value": 2.0033993339538574
      }
    },
    {
      "key": "Jose White ",
      "doc_count": 15,
      "top_hit": {
        "value": 1.5995635986328125
      }
    }
  ]
}

PS: to enable this groovy script, is necessary to edit the elasticsearch.yml and put this line on it:

script.engine.groovy.inline.aggs: on

after that, restart your elasticsearch node