Improve bit perturbation in TupleHashTableHash.

The changes in b81b5a96f424531b97cdd1dba97d9d1b9c9d372e did not fully
address the issue, because the bit-mixing of the IV into the final
hash-key didn't prevent clustering in the input-data survive in the
output data.

This didn't cause a lot of problems because of the additional growth
conditions added d4c62a6b623d6eef88218158e9fa3cf974c6c7e5. But as we
want to rein those in due to explosive growth in some edges, this
needs to be fixed.

Author: Andres Freund
Discussion: https://postgr.es/m/20171127185700.1470.20362@wrigleys.postgresql.org
Backpatch: 10, where simplehash was introduced

diff --git a/src/backend/executor/execGrouping.c b/src/backend/executor/execGrouping.c
index 07c8852fca8228c4e65e8ab4a3799fc7e90bd33f..cde5f676c1b33e41f6f579f8de1979719f3ea536 100644 (file)
--- a/src/backend/executor/execGrouping.c
+++ b/src/backend/executor/execGrouping.c
@@ -23,6 +23,7 @@
 #include "executor/executor.h"
 #include "miscadmin.h"
 #include "utils/lsyscache.h"
+#include "utils/hashutils.h"
 #include "utils/memutils.h"
 
 static uint32 TupleHashTableHash(struct tuplehash_hash *tb, const MinimalTuple tuple);
@@ -326,7 +327,7 @@ BuildTupleHashTable(int numCols, AttrNumber *keyColIdx,
         * underestimated.
         */
        if (use_variable_hash_iv)
-               hashtable->hash_iv = hash_uint32(ParallelWorkerNumber);
+               hashtable->hash_iv = murmurhash32(ParallelWorkerNumber);
        else
                hashtable->hash_iv = 0;
 
@@ -510,7 +511,13 @@ TupleHashTableHash(struct tuplehash_hash *tb, const MinimalTuple tuple)
                }
        }
 
-       return hashkey;
+       /*
+        * The way hashes are combined above, among each other and with the IV,
+        * doesn't lead to good bit perturbation. As the IV's goal is to lead to
+        * achieve that, perform a round of hashing of the combined hash -
+        * resulting in near perfect perturbation.
+        */
+       return murmurhash32(hashkey);
 }
 
 /*

diff --git a/src/test/regress/expected/groupingsets.out b/src/test/regress/expected/groupingsets.out
index cbfdbfd8563f8704acf96755c0755446509bbfa3..d21a494a9ddfc2d2243154ce6bfc741e887d858b 100644 (file)
--- a/src/test/regress/expected/groupingsets.out
+++ b/src/test/regress/expected/groupingsets.out
@@ -1183,29 +1183,33 @@ explain (costs off)
 -- simple rescan tests
 select a, b, sum(v.x)
   from (values (1),(2)) v(x), gstest_data(v.x)
- group by grouping sets (a,b);
+ group by grouping sets (a,b)
+ order by 1, 2, 3;
  a | b | sum 
 ---+---+-----
- 2 |   |   6
  1 |   |   3
+ 2 |   |   6
+   | 1 |   3
    | 2 |   3
    | 3 |   3
-   | 1 |   3
 (5 rows)
 
 explain (costs off)
   select a, b, sum(v.x)
     from (values (1),(2)) v(x), gstest_data(v.x)
-   group by grouping sets (a,b);
-                QUERY PLAN                
-------------------------------------------
- HashAggregate
-   Hash Key: gstest_data.a
-   Hash Key: gstest_data.b
-   ->  Nested Loop
-         ->  Values Scan on "*VALUES*"
-         ->  Function Scan on gstest_data
-(6 rows)
+   group by grouping sets (a,b)
+   order by 3, 1, 2;
+                             QUERY PLAN                              
+---------------------------------------------------------------------
+ Sort
+   Sort Key: (sum("*VALUES*".column1)), gstest_data.a, gstest_data.b
+   ->  HashAggregate
+         Hash Key: gstest_data.a
+         Hash Key: gstest_data.b
+         ->  Nested Loop
+               ->  Values Scan on "*VALUES*"
+               ->  Function Scan on gstest_data
+(8 rows)
 
 select *
   from (values (1),(2)) v(x),

diff --git a/src/test/regress/sql/groupingsets.sql b/src/test/regress/sql/groupingsets.sql
index b28d8217c127a20c9574ee93171d228254a07277..eb680286030d6d952bc7e88c891503dc54ee1a22 100644 (file)
--- a/src/test/regress/sql/groupingsets.sql
+++ b/src/test/regress/sql/groupingsets.sql
@@ -342,12 +342,13 @@ explain (costs off)
 
 select a, b, sum(v.x)
   from (values (1),(2)) v(x), gstest_data(v.x)
- group by grouping sets (a,b);
+ group by grouping sets (a,b)
+ order by 1, 2, 3;
 explain (costs off)
   select a, b, sum(v.x)
     from (values (1),(2)) v(x), gstest_data(v.x)
-   group by grouping sets (a,b);
-
+   group by grouping sets (a,b)
+   order by 3, 1, 2;
 select *
   from (values (1),(2)) v(x),
        lateral (select a, b, sum(v.x) from gstest_data(v.x) group by grouping sets (a,b)) s;