From dd5ecb9deba077d37fb99de2c31dbf53db1e858a Mon Sep 17 00:00:00 2001
From: Dimitris Tsirogiannis <dtsirogiannis@cloudera.com>
Date: Mon, 20 Apr 2015 17:06:26 -0700
Subject: [PATCH] IMPALA-1960: Illegal reference to non-materialized tuple when
 query has an empty select-project-join block

This commit fixes an issue where an aggregation expr may reference a
non-materialized slot if the query contains an empty select-project-join
block. This fix ensures that all the exprs in an aggregation reference
materialized slots/tuples.

Change-Id: Ic2cc9818061b3f06ab1d1cebf4e604352c2df6d1
Reviewed-on: http://gerrit.cloudera.org:8080/348
Reviewed-by: Dimitris Tsirogiannis <dtsirogiannis@cloudera.com>
Tested-by: Internal Jenkins
---
 .../impala/planner/SingleNodePlanner.java     | 19 +++++----
 .../queries/PlannerTest/empty.test            | 41 +++++++++++++++++++
 .../queries/QueryTest/empty.test              | 12 ++++++
 3 files changed, 64 insertions(+), 8 deletions(-)

diff --git a/fe/src/main/java/com/cloudera/impala/planner/SingleNodePlanner.java b/fe/src/main/java/com/cloudera/impala/planner/SingleNodePlanner.java
index 89af0d87d..b56ad3441 100644
--- a/fe/src/main/java/com/cloudera/impala/planner/SingleNodePlanner.java
+++ b/fe/src/main/java/com/cloudera/impala/planner/SingleNodePlanner.java
@@ -487,12 +487,6 @@ public class SingleNodePlanner {
       return createConstantSelectPlan(selectStmt, analyzer);
     }
 
-    // collect output tuples of subtrees
-    ArrayList<TupleId> rowTuples = Lists.newArrayList();
-    for (TableRef tblRef: selectStmt.getTableRefs()) {
-      rowTuples.addAll(tblRef.getMaterializedTupleIds());
-    }
-
     // Slot materialization:
     // We need to mark all slots as materialized that are needed during the execution
     // of selectStmt, and we need to do that prior to creating plans for the TableRefs
@@ -507,11 +501,20 @@ public class SingleNodePlanner {
     //   process repeats itself.
     selectStmt.materializeRequiredSlots(analyzer);
 
-    // return a plan that feeds the aggregation of selectStmt with an empty set,
-    // if the selectStmt's select-project-join portion returns an empty result set
+    ArrayList<TupleId> rowTuples = Lists.newArrayList();
+    // collect output tuples of subtrees
+    for (TableRef tblRef: selectStmt.getTableRefs()) {
+      rowTuples.addAll(tblRef.getMaterializedTupleIds());
+    }
+
+    // If the selectStmt's select-project-join portion returns an empty result set
+    // create a plan that feeds the aggregation of selectStmt with an empty set.
+    // Make sure the slots of the aggregation exprs and the tuples that they reference
+    // are materialized (see IMPALA-1960).
     if (analyzer.hasEmptySpjResultSet()) {
       PlanNode emptySetNode = new EmptySetNode(ctx_.getNextNodeId(), rowTuples);
       emptySetNode.init(analyzer);
+      emptySetNode.setOutputSmap(selectStmt.getBaseTblSmap());
       return createAggregationPlan(selectStmt, analyzer, emptySetNode);
     }
 
diff --git a/testdata/workloads/functional-planner/queries/PlannerTest/empty.test b/testdata/workloads/functional-planner/queries/PlannerTest/empty.test
index 37b3000a1..56259e027 100644
--- a/testdata/workloads/functional-planner/queries/PlannerTest/empty.test
+++ b/testdata/workloads/functional-planner/queries/PlannerTest/empty.test
@@ -278,3 +278,44 @@ WRITE TO HDFS [default.test_1860, OVERWRITE=false]
 |
 00:EMPTYSET
 ====
+# IMPALA-1960: Exprs in the aggregation that reference slots from an inline view when
+# the select stmt has an empty select-project-join portion.
+select sum(T.id), count(T.int_col)
+from
+  (select id, int_col, bigint_col from functional.alltypestiny) T
+where false
+---- PLAN
+01:AGGREGATE [FINALIZE]
+|  output: sum(id), count(int_col)
+|
+00:EMPTYSET
+====
+# IMPALA-1960: Exprs in the aggregation that reference slots from an inline view when
+# the select stmt has an empty select-project-join portion.
+select sum(T1.id + T2.int_col)
+from
+  (select id, bigint_col from functional.alltypestiny) T1 inner join
+  (select id, int_col from functional.alltypestiny) T2 on (T1.id = T2.id)
+where T1.bigint_col < 10 and 1 > 1
+---- PLAN
+01:AGGREGATE [FINALIZE]
+|  output: sum(id + int_col)
+|
+00:EMPTYSET
+====
+# IMPALA-1960: Exprs in the aggregation that reference slots from an inline view when
+# the select stmt has an empty select-project-join portion.
+select count(distinct T1.int_col)
+from
+  (select id, int_col from functional.alltypestiny) T1 inner join
+  functional.alltypessmall T2 on T1.id = T2.id
+where T2.bigint_col < 10 and false
+---- PLAN
+02:AGGREGATE [FINALIZE]
+|  output: count(T1.int_col)
+|
+01:AGGREGATE
+|  group by: int_col
+|
+00:EMPTYSET
+====
diff --git a/testdata/workloads/functional-query/queries/QueryTest/empty.test b/testdata/workloads/functional-query/queries/QueryTest/empty.test
index 014dd67d7..6445bafc2 100644
--- a/testdata/workloads/functional-query/queries/QueryTest/empty.test
+++ b/testdata/workloads/functional-query/queries/QueryTest/empty.test
@@ -117,3 +117,15 @@ WHERE (-339.22) IN (-922.68, -254.84)
 ---- TYPES
 INT
 ====
+---- QUERY
+# IMPALA-1960: Exprs in the aggregation that reference slots from an inline view when
+# the select stmt has an empty select-project-join portion.
+select sum(T.id), count(T.int_col)
+from
+  (select id, int_col, bigint_col from functional.alltypestiny) T
+where false
+---- RESULTS
+NULL,0
+---- TYPES
+BIGINT, BIGINT
+====