Elasticsearch sort by first element of array

Question

I am using Elasticsearch 5.5 and have an index with such mapping:

{
  "my_index": {
    "mappings": {
      "my_type": {
        "properties": {
          "title": {
            "type": "text",
            "fields": {
              "keyword": {
                "type": "keyword",
                "ignore_above": 256
              }
            }
          },
          "my_array": {
            "properties": {
              "array": {
                "type": "float"
              },
              "length": {
                "type": "long"
              }
            }
          }
        }
      }
    }
  }
}

I would like to search by title and sort by first value from array. Also it would be great to set this first value to the _score field. So, I've prepared such request:

GET my_index/my_type/_search
{
    "query": {
      "term": {
        "title.keyword": "Shorts"
      }
    }, 
    "sort" : {
        "_script" : {
            "type" : "number",
            "script" : {
                "lang": "painless",
                "inline": "doc['my_array.array'][0]"
            },
            "order" : "asc"
        }
    }
}

Unfortunately it doesn't work. I feel something is missing or wrong.

May you please provide some example documents and the response? What do you expect to be in the response and what is there actually? Thank you. — Nikolay Vasiliev
– Nikolay Vasiliev, Commented Oct 30, 2017 at 19:16
@NikolayVasiliev With such request: GET my_index/my_type/_search { "sort" : { "_script" : { "type" : "number", "script" : { "lang": "painless", "inline": "doc['embeddings.array'][0]" }, "order" : "asc" } } } The response is jsonblob.com/05e5893a-be1c-11e7-9ae8-cdd8d94a615d — Uladzislau Kaminski
– Uladzislau Kaminski, Commented Oct 31, 2017 at 9:16
Does this valid way to access first element of array doc['my_array.array'][0] ? — Uladzislau Kaminski
– Uladzislau Kaminski, Commented Oct 31, 2017 at 11:13

Andrei Stefan · Accepted Answer · 2017-10-31 16:41:01Z

4

The correct way of using that Painless script is like this:

{
  "query": {
    "term": {
      "title.keyword": "Shorts"
    }
  },
  "sort": {
    "_script": {
      "type": "number",
      "script": {
        "lang": "painless",
        "inline": "params._source.my_array.array[0]"
      },
      "order": "asc"
    }
  }
}

answered Oct 31, 2017 at 16:41

Andrei Stefan

52.5k6 gold badges102 silver badges92 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Nikolay Vasiliev · Accepted Answer · 2017-10-31 17:22:34Z

As Andrei pointed out in his answer, you should refer directly to _source in your painless script.

This happens because in the Lucene index (which is what ElasticSearch is built upon) does not have a notion of original order of the values in your array. Moreover, arrays do not work as you would expect:

Arrays of objects do not work as you would expect: you cannot query each object independently of the other objects in the array.

Basically, you are sorting by a random number from your list.

What Andrei is suggesting is to use the _source, which is, to read the original JSON document, parse it and extract the needed value from there (it will work). But _source is slow (because instead of accessing fast index you read from the disk, every time, every document).

You have 2 other options:

move this first element as a separate field;
use nested data type and explicitly define order.

Hope that helps!

Collectives™ on Stack Overflow

Elasticsearch sort by first element of array

2 Answers 2

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related