feat(influxql): Add hyper log log operators (influxdata#22322)

In addition to helping with normal queries, this can improve the 'SHOW CARDINALITY'
meta-queries.


Co-authored-by: Sam Arnold <[email protected]>

CHANGELOG.md

@@ -49,6 +49,8 @@ This release adds an embedded SQLite database for storing metadata required by t
 1. [21972](https://github.com/influxdata/influxdb/pull/21972): Added support for notebooks and annotations.
 1. [22135](https://github.com/influxdata/influxdb/pull/22135): Added route to return known resources.
 1. [22311](https://github.com/influxdata/influxdb/pull/22311): Add `storage-no-validate-field-size` config to `influxd` to disable enforcement of max field size.
+1. [22316](https://github.com/influxdata/influxdb/pull/22316): Optimize series iteration for queries that can be answered without inspecting TSM data.
+1. [22322](https://github.com/influxdata/influxdb/pull/22322): Add support for `merge_hll`, `sum_hll`, and `count_hll` in InfluxQL.
 
 ### Bug Fixes
 
@@ -75,7 +77,6 @@ Because of the version bump to `go`, the macOS build for this release requires a
 1. [21910](https://github.com/influxdata/influxdb/pull/21910): Added `--ui-disabled` option to `influxd` to allow for running with the UI disabled.
 1. [21958](https://github.com/influxdata/influxdb/pull/21958): Telemetry improvements: Do not record telemetry data for non-existant paths; replace invalid static asset paths with a slug.
 1. [22023](https://github.com/influxdata/influxdb/pull/22023): Upgrade Flux to v0.124.0.
-1. [22316](https://github.com/influxdata/influxdb/pull/22316): Optimize series iteration for queries that can be answered without inspecting TSM data.
 
 ### Bug Fixes
 

influxql/query/call_iterator.go

@@ -53,6 +53,10 @@ func NewCallIterator(input Iterator, opt IteratorOptions) (Iterator, error) {
 		return newLastIterator(input, opt)
 	case "mean":
 		return newMeanIterator(input, opt)
+	case "sum_hll":
+		return NewSumHllIterator(input, opt)
+	case "merge_hll":
+		return NewMergeHllIterator(input, opt)
 	default:
 		return nil, fmt.Errorf("unsupported function call: %s", name)
 	}
@@ -1529,3 +1533,68 @@ func newIntegralIterator(input Iterator, opt IteratorOptions, interval Interval)
 		return nil, fmt.Errorf("unsupported integral iterator type: %T", input)
 	}
 }
+
+// NewSumHllIterator returns an iterator for operating on a distinct() call.
+func NewSumHllIterator(input Iterator, opt IteratorOptions) (Iterator, error) {
+	switch input := input.(type) {
+	case FloatIterator:
+		createFn := func() (FloatPointAggregator, StringPointEmitter) {
+			fn := NewFloatSumHllReducer()
+			return fn, fn
+		}
+		return newFloatReduceStringIterator(input, opt, createFn), nil
+	case IntegerIterator:
+		createFn := func() (IntegerPointAggregator, StringPointEmitter) {
+			fn := NewIntegerSumHllReducer()
+			return fn, fn
+		}
+		return newIntegerReduceStringIterator(input, opt, createFn), nil
+	case UnsignedIterator:
+		createFn := func() (UnsignedPointAggregator, StringPointEmitter) {
+			fn := NewUnsignedSumHllReducer()
+			return fn, fn
+		}
+		return newUnsignedReduceStringIterator(input, opt, createFn), nil
+	case StringIterator:
+		createFn := func() (StringPointAggregator, StringPointEmitter) {
+			fn := NewStringSumHllReducer()
+			return fn, fn
+		}
+		return newStringReduceStringIterator(input, opt, createFn), nil
+	case BooleanIterator:
+		createFn := func() (BooleanPointAggregator, StringPointEmitter) {
+			fn := NewBooleanSumHllReducer()
+			return fn, fn
+		}
+		return newBooleanReduceStringIterator(input, opt, createFn), nil
+	default:
+		return nil, fmt.Errorf("unsupported sum_hll iterator type: %T", input)
+	}
+}
+
+// NewSumHllIterator returns an iterator for operating on a distinct() call.
+func NewMergeHllIterator(input Iterator, opt IteratorOptions) (Iterator, error) {
+	switch input := input.(type) {
+	case StringIterator:
+		createFn := func() (StringPointAggregator, StringPointEmitter) {
+			fn := NewStringMergeHllReducer()
+			return fn, fn
+		}
+		return newStringReduceStringIterator(input, opt, createFn), nil
+	default:
+		return nil, fmt.Errorf("unsupported merge_hll iterator type: %T", input)
+	}
+}
+
+func NewCountHllIterator(input Iterator, opt IteratorOptions) (Iterator, error) {
+	switch input := input.(type) {
+	case StringIterator:
+		createFn := func() (StringPointAggregator, UnsignedPointEmitter) {
+			fn := NewCountHllReducer()
+			return fn, fn
+		}
+		return newStringStreamUnsignedIterator(input, createFn, opt), nil
+	default:
+		return nil, fmt.Errorf("unsupported count_hll iterator type: %T", input)
+	}
+}

influxql/query/compile.go

@@ -308,6 +308,8 @@ func (c *compiledField) compileExpr(expr influxql.Expr) error {
 			return c.compileElapsed(expr.Args)
 		case "integral":
 			return c.compileIntegral(expr.Args)
+		case "count_hll":
+			return c.compileCountHll(expr.Args)
 		case "holt_winters", "holt_winters_with_fit":
 			withFit := expr.Name == "holt_winters_with_fit"
 			return c.compileHoltWinters(expr.Args, withFit)
@@ -393,7 +395,7 @@ func (c *compiledField) compileFunction(expr *influxql.Call) error {
 	switch expr.Name {
 	case "max", "min", "first", "last":
 		// top/bottom are not included here since they are not typical functions.
-	case "count", "sum", "mean", "median", "mode", "stddev", "spread":
+	case "count", "sum", "mean", "median", "mode", "stddev", "spread", "sum_hll":
 		// These functions are not considered selectors.
 		c.global.OnlySelectors = false
 	default:
@@ -784,6 +786,19 @@ func (c *compiledField) compileIntegral(args []influxql.Expr) error {
 	return c.compileSymbol("integral", args[0])
 }
 
+func (c *compiledField) compileCountHll(args []influxql.Expr) error {
+	if exp, got := 1, len(args); exp != got {
+		return fmt.Errorf("invalid number of arguments for count_hll, expected %d, got %d", exp, got)
+	}
+	c.global.OnlySelectors = false
+	switch arg0 := args[0].(type) {
+	case *influxql.Call:
+		return c.compileExpr(arg0)
+	default:
+		return c.compileSymbol("count_hll", arg0)
+	}
+}
+
 func (c *compiledField) compileHoltWinters(args []influxql.Expr, withFit bool) error {
 	name := "holt_winters"
 	if withFit {