Extended_stats with nested buckets

[Camilo_Salomon (Camilo Salomón)]

Hi everyone,

I'm new to elastic, so this question could be quite trivial for you, but here I go:

I'm using kibana_sample_data_ecommerce, which documents have a mapping like this

{
	...
	"order_date" : <datetime>
	"taxful_total_price" : <double>
	...
}

And, I want to get a basic daily behavior of the data:

As far as I know, I have to split my data two times, first daily and then into the interval I wish (1h in this case). After that I have to sum the ocurrences on that time range (the little one) and to do a cumulative sum over the entire day for each day. To get something like this:

So here is my query code:

POST kibana_sample_data_ecommerce/_search
{
  "size": 0,
  "query": {
    "bool": {
      "must": [
        {
          "range": {
            "order_date": {
              "gt": "now-1M",
              "lte": "now"
            }
          }
        }
      ]
    }
  },
  "aggs": {
    "day_histo": {
      "date_histogram": {
        "field": "order_date",
        "calendar_interval": "day"
      },
      "aggs": {
        "qmin_histo": {
          "date_histogram": {
            "field": "order_date",
            "calendar_interval": "hour"
          },
          "aggs": {
            "qminute_sum": {
              "sum": {
                "field": "taxful_total_price"
              }
            },
            "cumulative_qminute_sum": {
              "cumulative_sum": {
                "buckets_path": "qminute_sum"
              }
            }
          }
        }
      }
    }
  }
}

After doing that, I just have to get "extended stats" iterating the hours list for all days, i.e.,

for hour in hours:
    extended_stats(day1[hour], day2[hour], ... , day30[hour]) 

But this last step is driving me crazy because this process give me nested buckets, for each time interval... So the thing is, how can I do an extended stats aggregation over the month data, but only on each time window across every day?