MDEV-38877: Unnecessary filesort on derived table materialization

OmarGamal10 · OmarGamal10 · commit 2d88d09203fb · 2026-03-07T02:19:54.000+02:00
Fixes unnecessary filesort on derived tables when ordered/grouped by a
field in the key. The data is inherently sorted, wrapping the result set
in a filesort is redundant.
diff --git a/mysql-test/main/derived_cond_pushdown.result b/mysql-test/main/derived_cond_pushdown.result
@@ -21164,7 +21164,7 @@ id	select_type	table	type	possible_keys	key	key_len	ref	rows	filtered	Extra
 1	PRIMARY	<derived2>	ALL	NULL	NULL	NULL	NULL	7	100.00	
 2	DERIVED	<derived4>	ALL	NULL	NULL	NULL	NULL	7	100.00	Using temporary; Using filesort
 2	DERIVED	a2	eq_ref	PRIMARY	PRIMARY	4	a1.f	1	100.00	Using index
-4	DERIVED	t1	index	PRIMARY	PRIMARY	4	NULL	7	100.00	Using index; Using temporary; Using filesort
+4	DERIVED	t1	index	PRIMARY	PRIMARY	4	NULL	7	100.00	Using index
 Warnings:
 Note	1003	/* select#1 */ select `s`.`f` AS `f`,`s`.`c` AS `c` from (/* select#2 */ select straight_join `a2`.`f` AS `f`,count(0) AS `c` from (/* select#4 */ select `test`.`t1`.`f` AS `f`,count(0) AS `c` from `test`.`t1` group by `test`.`t1`.`f`) `a1` join `test`.`t1` `a2` where `a2`.`f` = `a1`.`f` group by `a2`.`f`) `s`
 SELECT * FROM ( SELECT STRAIGHT_JOIN f, COUNT(*) as c FROM v1 GROUP BY f ) AS s;
diff --git a/mysql-test/main/derived_split_innodb.result b/mysql-test/main/derived_split_innodb.result
@@ -131,7 +131,7 @@ id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
 1	PRIMARY	t	const	f2	NULL	NULL	NULL	0	Impossible ON condition
 1	PRIMARY	<derived2>	const	key0,key1	NULL	NULL	NULL	0	Impossible ON condition
 1	PRIMARY	t1	ALL	NULL	NULL	NULL	NULL	3	
-2	DERIVED	t2	ALL	PRIMARY	NULL	NULL	NULL	3	Using temporary; Using filesort
+2	DERIVED	t2	index	PRIMARY	PRIMARY	4	NULL	3	
 set statement optimizer_switch='split_materialized=off' for explain select t.f2
 from t1
 left join
@@ -998,7 +998,7 @@ id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
 1	PRIMARY	t1	ALL	NULL	NULL	NULL	NULL	2	
 1	PRIMARY	t2	ref	a	a	5	test.t1.a	1	Using where; Using index
 1	PRIMARY	<derived2>	ref	key0	key0	5	func	1	Using where
-2	DERIVED	t10	index	grp_id	grp_id	5	NULL	10000	Using index; Using temporary; Using filesort
+2	DERIVED	t10	index	grp_id	grp_id	5	NULL	10000	Using index
 drop table t1,t2, t10;
 drop view v1;
 #
diff --git a/mysql-test/main/mdev-38877.result b/mysql-test/main/mdev-38877.result
@@ -0,0 +1,32 @@
+#
+# MDEV-38877: Unnecessary filesort on derived table materialization
+# when derived table is inherently sorted by the grouping columns
+#
+CREATE TABLE t1 (
+groups_20 int NOT NULL,
+groups_20_2 int NOT NULL,
+b int,
+PRIMARY KEY (groups_20, groups_20_2)
+);
+CREATE TABLE t2 (a int, b int, index(a));
+INSERT INTO t1 SELECT seq/1000, seq+1, seq from seq_1_to_10000;
+INSERT INTO t2 SELECT seq, seq from seq_1_to_1000;
+ANALYZE TABLE t1, t2;
+EXPLAIN
+SELECT 
+a, 
+SUM(b) 
+FROM ( 
+SELECT 
+groups_20 
+FROM t1 
+GROUP BY groups_20 
+HAVING COUNT(*) != 1000
+) DT 
+JOIN t2 ON a = groups_20 
+GROUP BY a;
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	PRIMARY	t2	ALL	a	NULL	NULL	NULL	1000	Using where; Using temporary; Using filesort
+1	PRIMARY	<derived2>	ref	key0	key0	4	test.t2.a	1	
+2	DERIVED	t1	index	PRIMARY	PRIMARY	8	NULL	10000	Using index
+DROP TABLE t1, t2;
diff --git a/mysql-test/main/mdev-38877.test b/mysql-test/main/mdev-38877.test
@@ -0,0 +1,39 @@
+--source include/have_sequence.inc
+
+--echo #
+--echo # MDEV-38877: Unnecessary filesort on derived table materialization
+--echo # when derived table is inherently sorted by the grouping columns
+--echo #
+
+CREATE TABLE t1 (
+  groups_20 int NOT NULL,
+  groups_20_2 int NOT NULL,
+  b int,
+  PRIMARY KEY (groups_20, groups_20_2)
+);
+
+CREATE TABLE t2 (a int, b int, index(a));
+
+--disable_result_log
+INSERT INTO t1 SELECT seq/1000, seq+1, seq from seq_1_to_10000;
+
+INSERT INTO t2 SELECT seq, seq from seq_1_to_1000;
+
+ANALYZE TABLE t1, t2;
+--enable_result_log
+
+EXPLAIN
+SELECT 
+    a, 
+    SUM(b) 
+FROM ( 
+    SELECT 
+        groups_20 
+    FROM t1 
+    GROUP BY groups_20 
+    HAVING COUNT(*) != 1000
+) DT 
+JOIN t2 ON a = groups_20 
+GROUP BY a;
+
+DROP TABLE t1, t2;
diff --git a/mysql-test/main/opt_hints_split_materialized.result b/mysql-test/main/opt_hints_split_materialized.result
@@ -763,15 +763,15 @@ one_k T1, (select grp, count(*) from t1000 group by grp) TBL where TBL.grp=T1.a;
 id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
 1	PRIMARY	T1	ALL	NULL	NULL	NULL	NULL	1000	Using where
 1	PRIMARY	<derived2>	ref	key0	key0	5	test.T1.a	1	
-2	DERIVED	t1000	index	grp	grp	5	NULL	1000	Using index; Using temporary; Using filesort
+2	DERIVED	t1000	index	grp	grp	5	NULL	1000	Using index
 explain
 select /*+ SPLIT_MATERIALIZED(TBL) */ *
 from
 one_k T1, (select grp, count(*) from t1000 group by grp) TBL where TBL.grp=T1.a;
 id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
 1	PRIMARY	<derived2>	ALL	NULL	NULL	NULL	NULL	100	
 1	PRIMARY	T1	ALL	NULL	NULL	NULL	NULL	1000	Using where; Using join buffer (flat, BNL join)
-2	DERIVED	t1000	index	grp	grp	5	NULL	1000	Using index; Using temporary; Using filesort
+2	DERIVED	t1000	index	grp	grp	5	NULL	1000	Using index
 drop table one_k, t1000;
 #
 # end case 4
diff --git a/sql/sql_select.cc b/sql/sql_select.cc
@@ -32727,9 +32727,24 @@ void JOIN::save_query_plan(Join_plan_state *save_to)
 
   for (uint i= 0; i < table_count; i++)
   {
+    uint num_keys= join_tab[i].table->s->keys;
     save_to->join_tab_keyuse[i]= join_tab[i].keyuse;
     join_tab[i].keyuse= NULL;
     save_to->join_tab_checked_keys[i]= join_tab[i].checked_keys;
+    if (num_keys > 0)
+    {
+      if (!save_to->const_key_parts[i])
+      {
+        save_to->const_key_parts[i]=
+            (key_part_map *) thd->calloc(sizeof(key_part_map) * num_keys);
+      }
+
+      if (save_to->const_key_parts[i])
+      {
+        memcpy(save_to->const_key_parts[i], join_tab[i].table->const_key_parts,
+               sizeof(key_part_map) * num_keys);
+      }
+    }
     join_tab[i].checked_keys.clear_all();
   }
   memcpy((uchar*) save_to->best_positions, (uchar*) best_positions,
@@ -32777,6 +32792,9 @@ void JOIN::restore_query_plan(Join_plan_state *restore_from)
   {
     join_tab[i].keyuse= restore_from->join_tab_keyuse[i];
     join_tab[i].checked_keys= restore_from->join_tab_checked_keys[i];
+    memcpy(join_tab[i].table->const_key_parts,
+           restore_from->const_key_parts[i],
+           sizeof(key_part_map) * join_tab[i].table->s->keys);
   }
 
   memcpy((uchar*) best_positions, (uchar*) restore_from->best_positions,
diff --git a/sql/sql_select.h b/sql/sql_select.h
@@ -1235,6 +1235,8 @@ class JOIN :public Sql_alloc
     KEYUSE **join_tab_keyuse;
     /* Copies of JOIN_TAB::checked_keys for each JOIN_TAB. */
     key_map *join_tab_checked_keys;
+    /* Copies of TABLE::key_part_map for each JOIN_TAB->TABLE */
+    key_part_map **const_key_parts;
     SJ_MATERIALIZATION_INFO **sj_mat_info;
     my_bool error;
   public:
@@ -1244,7 +1246,7 @@ class JOIN :public Sql_alloc
       keyuse.buffer= NULL;
       keyuse.malloc_flags= 0;
       best_positions= 0;                        /* To detect errors */
-      error= my_multi_malloc(PSI_INSTRUMENT_ME, MYF(MY_WME),
+      error= my_multi_malloc(PSI_INSTRUMENT_ME, MYF(MY_WME | MY_ZEROFILL),
                              &best_positions,
                              sizeof(*best_positions) * (tables + 1),
                              &join_tab_keyuse,
@@ -1253,6 +1255,8 @@ class JOIN :public Sql_alloc
                              sizeof(*join_tab_checked_keys) * tables,
                              &sj_mat_info,
                              sizeof(sj_mat_info) * tables,
+                             &const_key_parts,
+                             sizeof(*const_key_parts) * tables, 
                              NullS) == 0;
     }
     Join_plan_state(JOIN *join);
