Sum array length elasticsearch

Question

I've got indexed documents that look like this:

{
  "fun_field": [...]
}

I'd very much like to sum the array lengths to get the total number of elements in all fun_field arrays. I've been trying lots of combinations and haven't been able to get it to work.

Does anyone speak the cryptic elasticsearch dialect fluently enough to describe what I'm asking?

Answer 1

Well I found a roundabout way of answering this question without using scripts.

{
    "size": 0,
    "aggs" : {
        "outer_agg" : { 
            "nested" : { 
                "path" : "elements"   
            },
            "aggs": {
                "inner_agg": {
                    "top_hits": {
                        "from": 0,
                        "size": 1
                    }
                }
            }
        }
    }
}

This will return a doc_count which is what I'm looking for. ES's DSL will be the death of me.

Answer 2

You can define your own scripts for aggregations (or script_fields), such that they're already summing the array up before the aggregation happens. Syntax may vary for different scripting languages but for Painless in Elasticsearch 5. You'll also need to enable inline scripts.

"aggs": {
  "array_sums" : {
    "sum": {
       "script": {
         "lang": "painless",
         "inline": "doc['fun_field'].length"
       }
    }
  }
}

There are other scripting languages, you can read about them here. There are some examples of scripts in aggregations here.