vam: Fix summarize where

Fix issue with incorrect counts when using where clauses on a
aggregation function in vector runtime. If a value fails the where
clause make the value as null so it is skipped by the aggregation
function.

Closes #5468

runtime/vam/expr/agg/count.go

@@ -10,25 +10,20 @@ type count struct {
 }
 
 func (a *count) Consume(vec vector.Any) {
-	if c, ok := vec.(*vector.Const); ok {
-		val := c.Value()
-		if !val.IsNull() && !val.IsError() {
-			a.count += uint64(vec.Len())
-		}
+	if c, ok := vec.(*vector.Const); ok && c.Value().IsNull() {
 		return
 	}
 	if _, ok := vector.Under(vec).Type().(*super.TypeError); ok {
 		return
 	}
-	nulls := vector.NullsOf(vec)
-	if nulls == nil {
-		a.count += uint64(vec.Len())
-		return
-	}
-	for i := range vec.Len() {
-		if !nulls.Value(i) {
-			a.count++
+	if nulls := vector.NullsOf(vec); nulls != nil {
+		for i := range vec.Len() {
+			if !nulls.Value(i) {
+				a.count++
+			}
 		}
+	} else {
+		a.count += uint64(vec.Len())
 	}
 }
 

runtime/vam/expr/aggregator.go

@@ -41,15 +41,19 @@ func (a *Aggregator) Eval(this vector.Any) vector.Any {
 
 func (a *Aggregator) apply(args ...vector.Any) vector.Any {
 	vec, where := args[0], args[1]
-	var tags []uint32
-	// If type is not bool then we want to filter everything.
-	if where.Type().ID() == super.IDBool {
-		for slot := uint32(0); slot < where.Len(); slot++ {
-			// XXX Feels like we should have a optimzed version of this.
-			if vector.BoolValue(where, slot) {
-				tags = append(tags, slot)
-			}
+	bools, _ := BoolMask(where)
+	if bools.IsEmpty() {
+		// everything is filtered.
+		return vector.NewConst(super.NewValue(vec.Type(), nil), vec.Len(), nil)
+	}
+	bools.Flip(0, uint64(vec.Len()))
+	if !bools.IsEmpty() {
+		nulls := vector.NewBoolEmpty(vec.Len(), nil)
+		bools.WriteDenseTo(nulls.Bits)
+		if origNulls := vector.NullsOf(vec); origNulls != nil {
+			nulls = vector.Or(nulls, origNulls)
 		}
+		vector.SetNulls(vec, nulls)
 	}
-	return vector.NewView(vec, tags)
+	return vec
 }

runtime/ztests/op/summarize/count-where.yaml

@@ -0,0 +1,17 @@
+zed: |
+  summarize
+    num_requests := count()
+    where log_time >= 2012-10-01T00:00:00Z
+    by client_ip
+  | sort client_ip
+
+vector: true
+
+input: |
+  {log_time:2012-01-01T00:00:44Z,client_ip:249.92.17.134}
+  {log_time:2012-10-01T00:24:30Z,client_ip:249.92.17.134}
+  {log_time:2012-05-12T10:23:22Z,client_ip:251.58.48.137}
+
+output: |
+  {client_ip:249.92.17.134,num_requests:1(uint64)}
+  {client_ip:251.58.48.137,num_requests:0(uint64)}