From 2d5c6b5bfef5a078279b2f34e1f4ec701a545672 Mon Sep 17 00:00:00 2001
From: Mattias Jonsson <mattias.jonsson@sun.com>
Date: Thu, 4 Mar 2010 12:09:09 +0100
Subject: [PATCH] Bug#48229: group by performance issue of partitioned table

Problem was block_size on partitioned tables was not set,
resulting in keys_per_block was not correct which affects
the cost calculation for read time of indexes (including
cost for group min/max).Which resulted in a bad optimizer
decision.

Fixed by setting stats.block_size correctly.
---
 mysql-test/r/partition_range.result | 14 ++++++++++++++
 mysql-test/t/partition_range.test   | 18 ++++++++++++++++++
 sql/ha_partition.cc                 |  1 +
 3 files changed, 33 insertions(+)

diff --git a/mysql-test/r/partition_range.result b/mysql-test/r/partition_range.result
index 02d2f6359c5..731f2478cc9 100644
--- a/mysql-test/r/partition_range.result
+++ b/mysql-test/r/partition_range.result
@@ -1,4 +1,18 @@
 drop table if exists t1, t2;
+#
+# Bug#48229: group by performance issue of partitioned table
+#
+CREATE TABLE t1 (
+a INT,
+b INT,
+KEY a (a,b)
+)
+PARTITION BY HASH (a) PARTITIONS 1;
+INSERT INTO t1 VALUES (0, 580092), (3, 894076), (4, 805483), (4, 913540), (6, 611137), (8, 171602), (9, 599495), (9, 746305), (10, 272829), (10, 847519), (12, 258869), (12, 929028), (13, 288970), (15, 20971), (15, 105839), (16, 788272), (17, 76914), (18, 827274), (19, 802258), (20, 123677), (20, 587729), (22, 701449), (25, 31565), (25, 230782), (25, 442887), (25, 733139), (25, 851020);
+EXPLAIN SELECT a, MAX(b) FROM t1 WHERE a IN (10, 100, 3) GROUP BY a;
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1	range	a	a	5	NULL	4	Using where; Using index
+DROP TABLE t1;
 create table t1 (a int)
 partition by range (a)
 ( partition p0 values less than (maxvalue));
diff --git a/mysql-test/t/partition_range.test b/mysql-test/t/partition_range.test
index c02d9049f2e..4d011ddc468 100644
--- a/mysql-test/t/partition_range.test
+++ b/mysql-test/t/partition_range.test
@@ -9,6 +9,24 @@
 drop table if exists t1, t2;
 --enable_warnings
 
+--echo #
+--echo # Bug#48229: group by performance issue of partitioned table
+--echo #
+CREATE TABLE t1 (
+ a INT,
+ b INT,
+ KEY a (a,b)
+)
+PARTITION BY HASH (a) PARTITIONS 1;
+
+# insert some rows (i.e. so that rows/blocks > 1)
+INSERT INTO t1 VALUES (0, 580092), (3, 894076), (4, 805483), (4, 913540), (6, 611137), (8, 171602), (9, 599495), (9, 746305), (10, 272829), (10, 847519), (12, 258869), (12, 929028), (13, 288970), (15, 20971), (15, 105839), (16, 788272), (17, 76914), (18, 827274), (19, 802258), (20, 123677), (20, 587729), (22, 701449), (25, 31565), (25, 230782), (25, 442887), (25, 733139), (25, 851020);
+
+# Before the fix the 'Extra' column showed 'Using index for group-by'
+EXPLAIN SELECT a, MAX(b) FROM t1 WHERE a IN (10, 100, 3) GROUP BY a;
+
+DROP TABLE t1;
+
 #
 # BUG 33429: Succeeds in  adding partition when maxvalue on last partition
 #
diff --git a/sql/ha_partition.cc b/sql/ha_partition.cc
index b854e270029..36bd90e3237 100644
--- a/sql/ha_partition.cc
+++ b/sql/ha_partition.cc
@@ -5053,6 +5053,7 @@ int ha_partition::info(uint flag)
 
     file= m_file[handler_instance];
     file->info(HA_STATUS_CONST);
+    stats.block_size= file->stats.block_size;
     stats.create_time= file->stats.create_time;
     ref_length= m_ref_length;
   }