From 129db3cbee704e71cff9f90f62691488c8ef768d Mon Sep 17 00:00:00 2001
From: Tom Lane <tgl@sss.pgh.pa.us>
Date: Sun, 7 Feb 2016 12:29:17 -0500
Subject: [PATCH] ExecHashRemoveNextSkewBucket must physically copy tuples to
 main hashtable.

Commit 45f6240a8fa9d355 added an assumption in ExecHashIncreaseNumBatches
and ExecHashIncreaseNumBuckets that they could find all tuples in the main
hash table by iterating over the "dense storage" introduced by that patch.
However, ExecHashRemoveNextSkewBucket continued its old practice of simply
re-linking deleted skew tuples into the main table's hashchains.  Hence,
such tuples got lost during any subsequent increase in nbatch or nbuckets,
and would never get joined, as reported in bug #13908 from Seth P.

I (tgl) think that the aforesaid commit has got multiple design issues
and should be reworked rather completely; but there is no time for that
right now, so band-aid the problem by making ExecHashRemoveNextSkewBucket
physically copy deleted skew tuples into the "dense storage" arena.

The added test case is able to exhibit the problem by means of fooling the
planner with a WHERE condition that it will underestimate the selectivity
of, causing the initial nbatch estimate to be too small.

Tomas Vondra and Tom Lane.  Thanks to David Johnston for initial
investigation into the bug report.
---
 src/backend/executor/nodeHash.c    | 15 +++++++++++++--
 src/test/regress/expected/join.out | 28 ++++++++++++++++++++++++++++
 src/test/regress/sql/join.sql      | 16 ++++++++++++++++
 3 files changed, 57 insertions(+), 2 deletions(-)

diff --git a/src/backend/executor/nodeHash.c b/src/backend/executor/nodeHash.c
index e7e51d1717..ab766b84dc 100644
--- a/src/backend/executor/nodeHash.c
+++ b/src/backend/executor/nodeHash.c
@@ -1574,8 +1574,19 @@ ExecHashRemoveNextSkewBucket(HashJoinTable hashtable)
 		if (batchno == hashtable->curbatch)
 		{
 			/* Move the tuple to the main hash table */
-			hashTuple->next = hashtable->buckets[bucketno];
-			hashtable->buckets[bucketno] = hashTuple;
+			HashJoinTuple copyTuple;
+
+			/*
+			 * We must copy the tuple into the dense storage, else it will not
+			 * be found by, eg, ExecHashIncreaseNumBatches.
+			 */
+			copyTuple = (HashJoinTuple) dense_alloc(hashtable, tupleSize);
+			memcpy(copyTuple, hashTuple, tupleSize);
+			pfree(hashTuple);
+
+			copyTuple->next = hashtable->buckets[bucketno];
+			hashtable->buckets[bucketno] = copyTuple;
+
 			/* We have reduced skew space, but overall space doesn't change */
 			hashtable->spaceUsedSkew -= tupleSize;
 		}
diff --git a/src/test/regress/expected/join.out b/src/test/regress/expected/join.out
index a0c3461e38..bade392394 100644
--- a/src/test/regress/expected/join.out
+++ b/src/test/regress/expected/join.out
@@ -2331,6 +2331,34 @@ select tt1.*, tt2.* from tt2 right join tt1 on tt1.joincol = tt2.joincol;
 reset enable_hashjoin;
 reset enable_nestloop;
 --
+-- regression test for bug #13908 (hash join with skew tuples & nbatch increase)
+--
+set work_mem to '64kB';
+set enable_mergejoin to off;
+explain (costs off)
+select count(*) from tenk1 a, tenk1 b
+  where a.hundred = b.thousand and (b.fivethous % 10) < 10;
+                         QUERY PLAN                         
+------------------------------------------------------------
+ Aggregate
+   ->  Hash Join
+         Hash Cond: (a.hundred = b.thousand)
+         ->  Index Only Scan using tenk1_hundred on tenk1 a
+         ->  Hash
+               ->  Seq Scan on tenk1 b
+                     Filter: ((fivethous % 10) < 10)
+(7 rows)
+
+select count(*) from tenk1 a, tenk1 b
+  where a.hundred = b.thousand and (b.fivethous % 10) < 10;
+ count  
+--------
+ 100000
+(1 row)
+
+reset work_mem;
+reset enable_mergejoin;
+--
 -- regression test for 8.2 bug with improper re-ordering of left joins
 --
 create temp table tt3(f1 int, f2 text);
diff --git a/src/test/regress/sql/join.sql b/src/test/regress/sql/join.sql
index 560c64fce4..c3902d9e9f 100644
--- a/src/test/regress/sql/join.sql
+++ b/src/test/regress/sql/join.sql
@@ -463,6 +463,22 @@ select tt1.*, tt2.* from tt2 right join tt1 on tt1.joincol = tt2.joincol;
 reset enable_hashjoin;
 reset enable_nestloop;
 
+--
+-- regression test for bug #13908 (hash join with skew tuples & nbatch increase)
+--
+
+set work_mem to '64kB';
+set enable_mergejoin to off;
+
+explain (costs off)
+select count(*) from tenk1 a, tenk1 b
+  where a.hundred = b.thousand and (b.fivethous % 10) < 10;
+select count(*) from tenk1 a, tenk1 b
+  where a.hundred = b.thousand and (b.fivethous % 10) < 10;
+
+reset work_mem;
+reset enable_mergejoin;
+
 --
 -- regression test for 8.2 bug with improper re-ordering of left joins
 --