perf: do not prevent the reduce-extraction-to-projection optimization in ArangoDB through dangling-edge filter

We currently do not trust edges - if the document they point to does not exist, we ignore the edge.

In the past, this access was done via $node != null. This has the problem that ArangoDB considers it a full document access and no longer applies the reduce-extraction-to-projection optimization rule. This rule is very important to limit query memory limit usage.

Removing the dangling edges filter completely would be breaking, but we can optimize them so the reduce-extraction-to-projection optimization still works.

Author: Yogu

spec/regression/collect/tests/collect-edge-count-access-group/aql/count.aql

@@ -20,7 +20,7 @@ RETURN {
           FOR v_item3
           IN OUTBOUND v_shipment1 @@shipments_deliveries
           FILTER (v_item3.`accessGroup` IN @var5)
-          FILTER v_item3 != null
+          FILTER v_item3._key != null
           RETURN v_item3
         )
         COLLECT WITH COUNT INTO v_count1
@@ -32,7 +32,7 @@ RETURN {
       IN (
         FOR v_node1, v_edge1, v_path1 IN @var6..@var7 OUTBOUND v_shipment1 @@shipments_deliveries
         FILTER (v_node1.`accessGroup` IN @var8)
-        FILTER v_node1 != null
+        FILTER v_node1._key != null
         RETURN v_node1
       )
       COLLECT WITH COUNT INTO v_count2

spec/regression/collect/tests/collect-edge-count/aql/countAfter.aql

@@ -14,7 +14,7 @@ RETURN {
         IN (
           FOR v_node1
           IN OUTBOUND v_shipment1 @@shipments_deliveries
-          FILTER v_node1 != null
+          FILTER v_node1._key != null
           RETURN v_node1
         )
         COLLECT WITH COUNT INTO v_count1
@@ -25,7 +25,7 @@ RETURN {
       FOR v_item2
       IN (
         FOR v_node2, v_edge1, v_path1 IN @var4..@var5 OUTBOUND v_shipment1 @@shipments_deliveries
-        FILTER v_node2 != null
+        FILTER v_node2._key != null
         RETURN v_node2
       )
       COLLECT WITH COUNT INTO v_count2

spec/regression/collect/tests/collect-edge-count/aql/countBefore.aql

@@ -14,7 +14,7 @@ RETURN {
         IN (
           FOR v_node1
           IN OUTBOUND v_shipment1 @@shipments_deliveries
-          FILTER v_node1 != null
+          FILTER v_node1._key != null
           RETURN v_node1
         )
         COLLECT WITH COUNT INTO v_count1
@@ -25,7 +25,7 @@ RETURN {
       FOR v_item2
       IN (
         FOR v_node2, v_edge1, v_path1 IN @var4..@var5 OUTBOUND v_shipment1 @@shipments_deliveries
-        FILTER v_node2 != null
+        FILTER v_node2._key != null
         RETURN v_node2
       )
       COLLECT WITH COUNT INTO v_count2

spec/regression/collect/tests/distinct-aggregation/aql/distinct.aql

@@ -14,7 +14,7 @@ RETURN {
         FOR v_item1
         IN (
           FOR v_node1, v_edge1, v_path1 IN @var3..@var4 OUTBOUND v_delivery1 @@deliveries_handlingUnits
-          FILTER v_node1 != null
+          FILTER v_node1._key != null
           RETURN v_node1.`warehouseSlot`
         )
         FILTER v_item1 != null

spec/regression/collect/tests/recursive-relation-traversal-access-group/aql/direct0to1.aql

@@ -20,7 +20,7 @@ RETURN {
         FILTER (v_node1.`accessGroup` IN @var6)
         FOR v_node2, v_edge2, v_path2 IN @var7..@var8 OUTBOUND v_node1 @@handlingUnits_childHandlingUnits
         FILTER (v_node2.`accessGroup` IN @var9)
-        FILTER v_node2 != null
+        FILTER v_node2._key != null
         RETURN v_node2
       )
       SORT (v_handlingUnit1.`handlingUnitNumber`)

spec/regression/collect/tests/recursive-relation-traversal-access-group/aql/direct0to2.aql

@@ -21,7 +21,7 @@ RETURN {
         FOR v_node2, v_edge2, v_path2 IN @var7..@var8 OUTBOUND v_node1 @@handlingUnits_childHandlingUnits
         PRUNE !((v_node2.`accessGroup` IN @var9))
         FILTER v_path2.vertices[*].`accessGroup` ALL IN @var10
-        FILTER v_node2 != null
+        FILTER v_node2._key != null
         RETURN v_node2
       )
       SORT (v_handlingUnit1.`handlingUnitNumber`)

spec/regression/collect/tests/recursive-relation-traversal-access-group/aql/direct1to1.aql

@@ -20,7 +20,7 @@ RETURN {
         FILTER (v_node1.`accessGroup` IN @var6)
         FOR v_node2, v_edge2, v_path2 IN @var7..@var8 OUTBOUND v_node1 @@handlingUnits_childHandlingUnits
         FILTER (v_node2.`accessGroup` IN @var9)
-        FILTER v_node2 != null
+        FILTER v_node2._key != null
         RETURN v_node2
       )
       SORT (v_handlingUnit1.`handlingUnitNumber`)

spec/regression/collect/tests/recursive-relation-traversal-access-group/aql/direct1to2.aql

@@ -21,7 +21,7 @@ RETURN {
         FOR v_node2, v_edge2, v_path2 IN @var7..@var8 OUTBOUND v_node1 @@handlingUnits_childHandlingUnits
         PRUNE !((v_node2.`accessGroup` IN @var9))
         FILTER v_path2.vertices[*].`accessGroup` ALL IN @var10
-        FILTER v_node2 != null
+        FILTER v_node2._key != null
         RETURN v_node2
       )
       SORT (v_handlingUnit1.`handlingUnitNumber`)

spec/regression/collect/tests/recursive-relation-traversal-access-group/aql/direct2to2.aql

@@ -21,7 +21,7 @@ RETURN {
         FOR v_node2, v_edge2, v_path2 IN @var7..@var8 OUTBOUND v_node1 @@handlingUnits_childHandlingUnits
         PRUNE !((v_node2.`accessGroup` IN @var9))
         FILTER v_path2.vertices[*].`accessGroup` ALL IN @var10
-        FILTER v_node2 != null
+        FILTER v_node2._key != null
         RETURN v_node2
       )
       SORT (v_handlingUnit1.`handlingUnitNumber`)

spec/regression/collect/tests/recursive-relation-traversal-access-group/aql/direct3to3.aql

@@ -21,7 +21,7 @@ RETURN {
         FOR v_node2, v_edge2, v_path2 IN @var7..@var8 OUTBOUND v_node1 @@handlingUnits_childHandlingUnits
         PRUNE !((v_node2.`accessGroup` IN @var9))
         FILTER v_path2.vertices[*].`accessGroup` ALL IN @var10
-        FILTER v_node2 != null
+        FILTER v_node2._key != null
         RETURN v_node2
       )
       SORT (v_handlingUnit1.`handlingUnitNumber`)