Elasticsearch: Add generic support for template variables (#32762)

* Elasticsearch: Add generic support for template variables

* format MovingAverage settings as numbers

* Move formatting logic to query builder & forma serial_diff settings as numbers

* modify presence check

* add todo

* minor fixes

* transform string values to numbers

* Move casting logic

* Slightly cleaner implementation

* Add BE tests

* Leverage elastic validation when string doesn't resolve to a numeric value

* move newly introduced test to testify

* add FE query_builder tests

* check error

* Parse values to float instead of int

* Fix tests & ParseFloat bit size

pkg/tsdb/elasticsearch/time_series_query.go

@@ -164,7 +164,7 @@ func (e *timeSeriesQuery) processQuery(q *Query, ms *es.MultiSearchRequestBuilde
 						}
 
 						aggBuilder.Pipeline(m.ID, m.Type, bucketPath, func(a *es.PipelineAggregation) {
-							a.Settings = m.Settings.MustMap()
+							a.Settings = m.generateSettingsForDSL()
 						})
 					}
 				} else {
@@ -181,6 +181,31 @@ func (e *timeSeriesQuery) processQuery(q *Query, ms *es.MultiSearchRequestBuilde
 	return nil
 }
 
+// Casts values to int when required by Elastic's query DSL
+func (metricAggregation MetricAgg) generateSettingsForDSL() map[string]interface{} {
+	setFloatPath := func(path ...string) {
+		if stringValue, err := metricAggregation.Settings.GetPath(path...).String(); err == nil {
+			if value, err := strconv.ParseFloat(stringValue, 64); err == nil {
+				metricAggregation.Settings.SetPath(path, value)
+			}
+		}
+	}
+
+	switch metricAggregation.Type {
+	case "moving_avg":
+		setFloatPath("window")
+		setFloatPath("predict")
+		setFloatPath("settings", "alpha")
+		setFloatPath("settings", "beta")
+		setFloatPath("settings", "gamma")
+		setFloatPath("settings", "period")
+	case "serial_diff":
+		setFloatPath("lag")
+	}
+
+	return metricAggregation.Settings.MustMap()
+}
+
 func addDateHistogramAgg(aggBuilder es.AggBuilder, bucketAgg *BucketAgg, timeFrom, timeTo string) es.AggBuilder {
 	aggBuilder.DateHistogram(bucketAgg.ID, bucketAgg.Field, func(a *es.DateHistogramAgg, b es.AggBuilder) {
 		a.Interval = bucketAgg.Settings.Get("interval").MustString("auto")

pkg/tsdb/elasticsearch/time_series_query_test.go

@@ -8,6 +8,7 @@ import (
 	"github.com/grafana/grafana/pkg/plugins"
 	es "github.com/grafana/grafana/pkg/tsdb/elasticsearch/client"
 	"github.com/grafana/grafana/pkg/tsdb/interval"
+	"github.com/stretchr/testify/assert"
 
 	"github.com/grafana/grafana/pkg/components/simplejson"
 	. "github.com/smartystreets/goconvey/convey"
@@ -856,6 +857,83 @@ func TestExecuteTimeSeriesQuery(t *testing.T) {
 	})
 }
 
+func TestSettingsCasting(t *testing.T) {
+	from := time.Date(2018, 5, 15, 17, 50, 0, 0, time.UTC)
+	to := time.Date(2018, 5, 15, 17, 55, 0, 0, time.UTC)
+
+	t.Run("Correctly transforms moving_average settings", func(t *testing.T) {
+		c := newFakeClient(5)
+		_, err := executeTsdbQuery(c, `{
+			"timeField": "@timestamp",
+			"bucketAggs": [
+				{ "type": "date_histogram", "field": "@timestamp", "id": "2" }
+			],
+			"metrics": [
+				{ "id": "1", "type": "average", "field": "@value" },
+				{
+					"id": "3",
+					"type": "moving_avg",
+					"field": "1",
+					"pipelineAgg": "1",
+					"settings": {
+						"model": "holt_winters",
+						"window": "10",
+						"predict": "5",
+						"settings": {
+							"alpha": "0.5",
+							"beta": "0.7",
+							"gamma": "SHOULD NOT CHANGE",
+							"period": "4"
+						}
+					}
+				}
+			]
+		}`, from, to, 15*time.Second)
+		assert.Nil(t, err)
+		sr := c.multisearchRequests[0].Requests[0]
+
+		movingAvgSettings := sr.Aggs[0].Aggregation.Aggs[1].Aggregation.Aggregation.(*es.PipelineAggregation).Settings
+
+		assert.Equal(t, 10., movingAvgSettings["window"])
+		assert.Equal(t, 5., movingAvgSettings["predict"])
+
+		modelSettings := movingAvgSettings["settings"].(map[string]interface{})
+
+		assert.Equal(t, .5, modelSettings["alpha"])
+		assert.Equal(t, .7, modelSettings["beta"])
+		assert.Equal(t, "SHOULD NOT CHANGE", modelSettings["gamma"])
+		assert.Equal(t, 4., modelSettings["period"])
+	})
+
+	t.Run("Correctly transforms serial_diff settings", func(t *testing.T) {
+		c := newFakeClient(5)
+		_, err := executeTsdbQuery(c, `{
+			"timeField": "@timestamp",
+			"bucketAggs": [
+				{ "type": "date_histogram", "field": "@timestamp", "id": "2" }
+			],
+			"metrics": [
+				{ "id": "1", "type": "average", "field": "@value" },
+				{
+					"id": "3",
+					"type": "serial_diff",
+					"field": "1",
+					"pipelineAgg": "1",
+					"settings": {
+						"lag": "1"
+					}
+				}
+			]
+		}`, from, to, 15*time.Second)
+		assert.Nil(t, err)
+		sr := c.multisearchRequests[0].Requests[0]
+
+		serialDiffSettings := sr.Aggs[0].Aggregation.Aggs[1].Aggregation.Aggregation.(*es.PipelineAggregation).Settings
+
+		assert.Equal(t, 1., serialDiffSettings["lag"])
+	})
+}
+
 type fakeClient struct {
 	version             int
 	timeField           string