Json aggs

[w33ble]

Closes #1731

This had become a larger project than expected, partly because I started going about it the wrong way and created tests along the way for parts that weren't really related. But, at least more things have tests now...

This PR adds a JSON input to the agg builder. The intention is that you can take a JSON ES query from elsewhere, drop it in this input, and get the results in Kibana without having to build it by hand.

Note: Any JSON data put in this input will clobber input from any other fields in the agg builder.

• adds raw_json param type and control, with tests (100%)
• query_input component is now the validate_query directive, with tests (78%)
• debounce service created, with tests (100%), and validate_query uses it instead of _.debounce
• the input-error css class was abstracted out, with hopes to re-use it elsewhere (now in _input.less)
• validate_json directive was added, with tests (100%)

[spalger]

Looks like the json editor only shows up when the agg already has advanced params. Should show up on all aggs.

[spalger]

Besides previous comment 👍

[w33ble]

@spenceralger advanced wasn't being shown because it wasn't always in the params (it was only in terms). The latest commit always appends the JSON type to all param types.

[tomercagan]

I new to Kibana so maybe my assumption is wrong.
From what I understand, if you input some json in the "JSON Input" text area (after clicking advanced) it should replace the aggregation selected (from above: Note: Any JSON data put in this input will clobber input from any other fields in the agg builder.).

I am trying to somehow use scripted_metric so I selected an arbitrary aggregation and field and then pasted my scripted metric into the JSON Input:

But in reality, the json was added into the sum aggregation (I extracted this query from the error report):

{
  "query": {
    "filtered": {
      "query": {
        "query_string": {
          "query": "*"
        }
      },
      "filter": {
        "bool": {
          "must": [
            {
              "range": {
                "mts": {
                  "gte": 1418767200000,
                  "lte": 1418987718938
                }
              }
            }
          ],
          "must_not": []
        }
      }
    }
  },
  "size": 0,
  "aggs": {
    "2": {
      "date_histogram": {
        "field": "mts",
        "interval": "3600000ms",
        "min_doc_count": 1,
        "extended_bounds": {
          "min": 1418767200000,
          "max": 1418987718936
        }
      },
      "aggs": {
        "1": {
          "sum": {
            "field": "M8000C0",
            "aggs": {
              "kpi1": {
                "scripted_metric": {
                  "init_script": "_agg. sumVal1 = 0; _agg.sumVal2 = 0;",
                  "map_script": "_agg.sumVal1 += doc.val1; _agg.sumVal2 += doc.val2;",
                  "reduce_script": "totalVal1 = 0; totalVal2 = 0; for (agg in _aggs) {  totalVal1 += agg.sumVal1; totalVal2 += agg.sumVal2;}; return totalVal2/ totalVal1"
                }
              }
            }
          }
        }
      }
    }
  },
  "fields": [
    "*",
    "_source"
  ],
  "script_fields": {
    "mts": {
      "script": "if (doc['mts'].value == 0) { null } else { doc['mts'].value }"
    }
  }
}

Is this an error in my assumptions about the json actually replacing the aggregation or is this some edge case not covered?

[jvienne]

I have a similar problem trying to visualise the query of the following exemple in Kibana:

http://www.elasticsearch.org/guide/en/elasticsearch/guide/current/children-agg.html

I add a term aggregation with the country field and then try to add the children aggregation using the JSON Input area:

The JSON is not added in the expected place:

Is it a bug or an incorrect use of the JSON Input area?

Thank a lot!!

[spalger]

Hey @tomercagan @jvienne I think you misunderstood the purpose of the JSON input in the agg builder; it is for defining or overriding parameters/arguments for a specific aggregation. It can not add arbitrary aggregations to the visualization.

This is likely to always be the case, as we need to know what an aggregation does in order to read it's results, use its results to create bars/slices/lines/etc, format the values it produces, and label the visual elements it creates. Sorry for the confusion!!

Maybe we could describe the capabilities of this input a bit better in the UI.

[tomercagan]

Hi @spenceralger - thanks for the response.
Is there some way to do something like that?

I am not sure about children aggregation but it seems to me that metric aggregation is very similar to any "arithmetic" aggregation - it results in one value per bucket (though this value could be an object).

Thanks!

[spalger]

@tomercagan no problem. Currently there isn't any way to do that but we are investigating adding the top hits agg once #1536 is complete. Maybe the "children" agg will come around that time.

[jvienne]

@spenceralger thanks for your answer!

There is no way to visualise a children aggregation?

[spalger]

Correct