add tck

src/graph/visitor/PropertyTrackerVisitor.cpp

@@ -326,7 +326,7 @@ void PropertyTrackerVisitor::visit(FunctionCallExpression *expr) {
  // length function support `STRING` input too, so we can't ignore it directly
  // TODO add type info to variable to help optimize it.
  static const std::unordered_set<std::string> ignoreFuncs = {
- "src", "dst", "type", "typeid", "id", "rank", /*, "length"*/};
+ "src", "dst", "type", "typeid", "id", "rank" /*, "length"*/};
 
  auto funName = expr->name();
  std::transform(funName.begin(), funName.end(), funName.begin(), ::tolower);

tests/tck/features/optimizer/PrunePropertiesRule.feature

@@ -328,6 +328,31 @@ Feature: Prune Properties rule
  | 6 | Argument | 0 | |
  | 0 | Start | | |
 
+ Scenario: Project after aggregate
+ Given a graph with space named "nba"
+ When profiling query:
+ """
+ MATCH (v1:player)-[e:like]->(v2:player)
+ WHERE id(v1) == "Tony Parker"
+ WITH v1, v2, count(e.likeness) AS cnt
+ RETURN v1.player.age, v2.player.age, cnt
+ """
+ Then the result should be, in order:
+ | v1.player.age | v2.player.age | cnt |
+ | 36 | 42 | 1 |
+ | 36 | 41 | 1 |
+ | 36 | 33 | 1 |
+ And the execution plan should be:
+ | id | name | dependencies | profiling data | operator info |
+ | 8 | Project | 7 | | |
+ | 7 | Aggregate | 6 | | |
+ | 6 | Project | 10 | | |
+ | 10 | AppendVertices | 9 | | { "props": "[{ \"props\":[\"age\", \"_tag\"]}]" } |
+ | 9 | Traverse | 2 | | { "vertexProps": "[{ \"props\":[\"age\"]}]", "edgeProps": "[{\"props\": [\"_src\", \"_dst\", \"_rank\", \"_type\", \"likeness\"]}]" } |
+ | 2 | Dedup | 1 | | |
+ | 1 | PassThrough | 3 | | |
+ | 3 | Start | | | |
+
  # The schema id is not fixed in standalone cluster, so we skip it
  @distonly
  Scenario: Optional Match