Creating Custom Scripted Aggregations across all the buckets in a pipeline

[alexsanjoseph (Alex Joseph)]

Here is some example data pertaining to my domain:

PUT test_index 
{
  "mappings": {
    "test_type": { 
      "_all":       { "enabled": false  }, 
      "properties": { 
        "dummy":    { "type": "keyword"  }, 
        "timestamp":     { "type": "date"  }, 
        "state":      { "type": "keyword" } ,
        "duration":  {"type": "integer"}
      }
    }
  }
}


POST test_index/test_type
{
  "dummy": "dummy",
  "timestamp": "2015-03-25T17:49:00.000Z",
  "state": "State_1",
  "duration": 50
}

POST test_index/test_type
{
  "dummy": "dummy",
  "timestamp": "2015-03-25T17:49:00.050Z",
  "state": "State_2",
  "duration": 100
}

POST test_index/test_type
{
  "dummy": "dummy",
  "timestamp": "2015-03-25T17:49:00.150Z",
  "state": "State_2",
  "duration": 1000
}

POST test_index/test_type
{
  "dummy": "dummy",
  "timestamp": "2015-03-25T17:49:01.150Z",
  "state": "State_1",
  "duration": 2000
}

I want to know the percentage duration of each state. For a Single state I can get it with the following script.

GET test_index/_search?size=0
{
  "aggs": {
    "dummy": {
      "terms": {
        "field": "dummy",
        "size": 10
      },
      "aggs": {
        "state": {
          "filter": {
            "term": {
              "state": "State_1"
            }
          },
          "aggs": {
            "state_duration": {
              "sum": {
                "field": "duration"
              }
            }
          }
        },
        "total_duration_1": {
          "sum": {
            "field": "duration"
          }
        },
        "percentage_duration": {
          "bucket_script": {
            "buckets_path": {
              "state_duration": "state>state_duration",
              "total_duration": "total_duration_1"
            },
            "script": "Math.round(params.state_duration / params.total_duration * 100)/100.0"
          }
        }
      }
}

}
}

However what I want to do is to get this for all the states automatically using a terms aggregation.

GET test_index/_search?size=0
{
  "aggs": {
    "dummy": {
      "terms": {
        "field": "dummy",
        "size": 10
      },
      "aggs": {
        "state": {
          "terms": {
            "field": "state",
            "size": 10
          },
          "aggs": {
            "state_duration": {
              "sum": {
                "field": "duration"
              }
            }
          }
        },
        "total_duration_1": {
          "sum": {
            "field": "duration"
          }
        },
        "total_duration_2":{
          "sum_bucket": {
            "buckets_path": "state>state_duration"
          }
        },
        "percentage_duration": {
          "bucket_script": {
            "buckets_path": {
              "state_duration": "state>state_duration",
              "total_duration": "total_duration_1"
            },
            "script": "Math.round(params.state_duration / params.total_duration * 100)/100.0"
          }
        }
      }
    }
  }
}

However, this fails with the following error ""buckets_path must reference either a number value or a single value numeric metric aggregation, got: java.lang.Object[]" because the bucket script cannot handle an array which comes out of the terms aggregation.

Is there any way to run a script on all the terms in a bucket rather than choosing individual ones manually? (or doing the math in the client side?)