Post review changes for the optimization of IN predicates into IN sub…

…queries.

mysql-test/r/opt_tvc.result

@@ -6,29 +6,26 @@ create table t2 (a int, b int, c int);
 insert into t2
 values (1,2,3), (5,1,2), (4,3,7),
 (8,9,0), (10,7,1), (5,5,1);
+create table t3 (a int, b varchar(16), index idx(a));
+insert into t3 values
+(1, "abc"), (3, "egh"), (8, "axxx"), (10, "abc"),
+(2, "ccw"), (8, "wqqe"), (7, "au"), (9, "waa"),
+(3, "rass"), (9, "ert"), (9, "lok"), (8, "aww"),
+(1, "todd"), (3, "rew"), (8, "aww"), (3, "sw"),
+(11, "llk"), (7, "rbw"), (1, "sm"), (2, "jyp"),
+(4, "yq"), (5, "pled"), (12, "ligin"), (12, "toww"),
+(6, "mxm"), (15, "wanone"), (9, "sunqq"), (2, "abe");
 # optimization is not used
 select * from t1 where a in (1,2);
 a	b
 1	2
 1	1
 2	5
-explain select * from t1 where a in (1,2);
-id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
-1	SIMPLE	t1	ALL	NULL	NULL	NULL	NULL	6	Using where
-explain format=json select * from t1 where a in (1,2);
-EXPLAIN
-{
-  "query_block": {
-    "select_id": 1,
-    "table": {
-      "table_name": "t1",
-      "access_type": "ALL",
-      "rows": 6,
-      "filtered": 100,
-      "attached_condition": "t1.a in (1,2)"
-    }
-  }
-}
+explain extended select * from t1 where a in (1,2);
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	filtered	Extra
+1	SIMPLE	t1	ALL	NULL	NULL	NULL	NULL	6	100.00	Using where
+Warnings:
+Note	1003	select `test`.`t1`.`a` AS `a`,`test`.`t1`.`b` AS `b` from `test`.`t1` where `test`.`t1`.`a` in (1,2)
 # set minimum number of values in VALUEs list when optimization works to 2
 set @@in_subquery_conversion_threshold= 2;
 # single IN-predicate in WHERE-part
@@ -481,5 +478,68 @@ a	b
 1	1
 2	5
 deallocate prepare stmt;
-drop table t1, t2;
+# use inside out access from tvc rows
+set @@in_subquery_conversion_threshold= default;
+select * from t3 where a in (1,4,10);
+a	b
+1	abc
+1	todd
+1	sm
+4	yq
+10	abc
+explain extended select * from t3 where a in (1,4,10);
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	filtered	Extra
+1	SIMPLE	t3	range	idx	idx	5	NULL	5	100.00	Using index condition
+Warnings:
+Note	1003	select `test`.`t3`.`a` AS `a`,`test`.`t3`.`b` AS `b` from `test`.`t3` where `test`.`t3`.`a` in (1,4,10)
+set @@in_subquery_conversion_threshold= 2;
+select * from t3 where a in (1,4,10);
+a	b
+1	abc
+1	todd
+1	sm
+4	yq
+10	abc
+explain extended select * from t3 where a in (1,4,10);
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	filtered	Extra
+1	PRIMARY	<subquery2>	ALL	distinct_key	NULL	NULL	NULL	2	100.00	
+1	PRIMARY	t3	ref	idx	idx	5	tvc_0.1	3	100.00	
+2	MATERIALIZED	<derived3>	ALL	NULL	NULL	NULL	NULL	2	100.00	
+3	DERIVED	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	No tables used
+Warnings:
+Note	1003	/* select#1 */ select `test`.`t3`.`a` AS `a`,`test`.`t3`.`b` AS `b` from `test`.`t3` semi join ((values (1),(4),(10)) `tvc_0`) where `test`.`t3`.`a` = `tvc_0`.`1`
+# use vectors in IN predeicate
+set @@in_subquery_conversion_threshold= 4;
+select * from t1 where (a,b) in ((1,2),(3,4));
+a	b
+1	2
+explain extended select * from t1 where (a,b) in ((1,2),(3,4));
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	filtered	Extra
+1	PRIMARY	<subquery2>	ALL	distinct_key	NULL	NULL	NULL	2	100.00	
+1	PRIMARY	t1	ALL	NULL	NULL	NULL	NULL	6	100.00	Using where; Using join buffer (flat, BNL join)
+2	MATERIALIZED	<derived3>	ALL	NULL	NULL	NULL	NULL	2	100.00	
+3	DERIVED	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	No tables used
+Warnings:
+Note	1003	/* select#1 */ select `test`.`t1`.`a` AS `a`,`test`.`t1`.`b` AS `b` from `test`.`t1` semi join ((values (1,2),(3,4)) `tvc_0`) where `test`.`t1`.`a` = `tvc_0`.`1` and `test`.`t1`.`b` = `tvc_0`.`2`
+set @@in_subquery_conversion_threshold= 2;
+# trasformation works for the one IN predicate and doesn't work for the other
+set @@in_subquery_conversion_threshold= 5;
+select * from t2
+where (a,b) in ((1,2),(8,9)) and
+(a,c) in ((1,3),(8,0),(5,1));
+a	b	c
+1	2	3
+8	9	0
+explain extended select * from t2
+where (a,b) in ((1,2),(8,9)) and
+(a,c) in ((1,3),(8,0),(5,1));
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	filtered	Extra
+1	PRIMARY	<subquery2>	ALL	distinct_key	NULL	NULL	NULL	2	100.00	
+1	PRIMARY	t2	ALL	NULL	NULL	NULL	NULL	6	100.00	Using where; Using join buffer (flat, BNL join)
+2	MATERIALIZED	<derived3>	ALL	NULL	NULL	NULL	NULL	2	100.00	
+3	DERIVED	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	No tables used
+Warnings:
+Note	1003	/* select#1 */ select `test`.`t2`.`a` AS `a`,`test`.`t2`.`b` AS `b`,`test`.`t2`.`c` AS `c` from `test`.`t2` semi join ((values (1,3),(8,0),(5,1)) `tvc_0`) where `test`.`t2`.`a` = `tvc_0`.`1` and `test`.`t2`.`c` = `tvc_0`.`3` and (`tvc_0`.`1`,`test`.`t2`.`b`) in (<cache>((1,2)),<cache>((8,9)))
+set @@in_subquery_conversion_threshold= 2;
+drop table t1, t2, t3;
 set @@in_subquery_conversion_threshold= default;

mysql-test/t/opt_tvc.test

@@ -10,12 +10,21 @@ insert into t2
 values (1,2,3), (5,1,2), (4,3,7),
        (8,9,0), (10,7,1), (5,5,1);
 
+create table t3 (a int, b varchar(16), index idx(a));
+insert into t3 values
+  (1, "abc"), (3, "egh"), (8, "axxx"), (10, "abc"),
+  (2, "ccw"), (8, "wqqe"), (7, "au"), (9, "waa"),
+  (3, "rass"), (9, "ert"), (9, "lok"), (8, "aww"),
+  (1, "todd"), (3, "rew"), (8, "aww"), (3, "sw"),
+  (11, "llk"), (7, "rbw"), (1, "sm"), (2, "jyp"),
+  (4, "yq"), (5, "pled"), (12, "ligin"), (12, "toww"),
+  (6, "mxm"), (15, "wanone"), (9, "sunqq"), (2, "abe");
+
 --echo # optimization is not used
 
 let $query= select * from t1 where a in (1,2);
 eval $query;
-eval explain $query;
-eval explain format=json $query;
+eval explain extended $query;
 
 --echo # set minimum number of values in VALUEs list when optimization works to 2
 
@@ -239,5 +248,39 @@ execute stmt;
 execute stmt;
 deallocate prepare stmt;
 
-drop table t1, t2;
+--echo # use inside out access from tvc rows
+
+let $query= select * from t3 where a in (1,4,10);
+set @@in_subquery_conversion_threshold= default;
+eval $query;
+eval explain extended $query;
+set @@in_subquery_conversion_threshold= 2;
+eval $query;
+eval explain extended $query;
+
+--echo # use vectors in IN predeicate
+
+set @@in_subquery_conversion_threshold= 4;
+
+let $query=
+select * from t1 where (a,b) in ((1,2),(3,4));
+
+eval $query;
+eval explain extended $query;
+set @@in_subquery_conversion_threshold= 2;
+
+--echo # trasformation works for the one IN predicate and doesn't work for the other
+
+set @@in_subquery_conversion_threshold= 5;
+
+let $query=
+select * from t2
+where (a,b) in ((1,2),(8,9)) and
+      (a,c) in ((1,3),(8,0),(5,1));
+
+eval $query;
+eval explain extended $query;
+set @@in_subquery_conversion_threshold= 2;
+
+drop table t1, t2, t3;
 set @@in_subquery_conversion_threshold= default;

sql/item_cmpfunc.cc

@@ -4334,8 +4334,11 @@ void Item_func_in::mark_as_condition_AND_part(TABLE_LIST *embedding)
   Query_arena *arena, backup;
   arena= thd->activate_stmt_arena_if_needed(&backup);
 
-  if (can_be_transformed_in_tvc(thd))
+  if (to_be_transformed_into_in_subq(thd))
+  {
+    transform_into_subq= true;
     thd->lex->current_select->in_funcs.push_back(this, thd->mem_root);
+  }
 
   if (arena)
     thd->restore_active_arena(arena, &backup);

sql/item_cmpfunc.h

@@ -2128,6 +2128,7 @@ class Item_func_in :public Item_func_opt_neg,
 protected:
   SEL_TREE *get_func_mm_tree(RANGE_OPT_PARAM *param,
                              Field *field, Item *value);
+  bool transform_into_subq;
 public:
   /// An array of values, created when the bisection lookup method is used
   in_vector *array;
@@ -2244,7 +2245,8 @@ class Item_func_in :public Item_func_opt_neg,
     return clone;
   }
   void mark_as_condition_AND_part(TABLE_LIST *embedding);
-  bool can_be_transformed_in_tvc(THD *thd);
+  bool to_be_transformed_into_in_subq(THD *thd);
+  bool create_value_list_for_tvc(THD *thd, List< List<Item> > *values);
   Item *in_predicate_to_in_subs_transformer(THD *thd, uchar *arg);
 };
 

sql/opt_subselect.cc

@@ -1047,7 +1047,7 @@ bool convert_join_subqueries_to_semijoins(JOIN *join)
       DBUG_RETURN(1);
     if (subq_sel->handle_derived(thd->lex, DT_MERGE))
       DBUG_RETURN(TRUE);
-    if (subq_sel->join->transform_in_predicate_into_tvc(thd))
+    if (subq_sel->join->transform_in_predicates_into_in_subq(thd))
       DBUG_RETURN(TRUE);
     subq_sel->update_used_tables();
   }

sql/sql_lex.cc

@@ -2265,7 +2265,7 @@ void st_select_lex::init_select()
   lock_type= TL_READ_DEFAULT;
   tvc= 0;
   in_funcs.empty();
-  cur_tvc= 0;
+  curr_tvc_name= 0;
 }
 
 /*

sql/sql_lex.h

@@ -874,7 +874,7 @@ class st_select_lex: public st_select_lex_node
     transformation of IN-predicate into IN-subquery for this
     st_select_lex.
   */
-  uint cur_tvc;
+  uint curr_tvc_name;
 
   /*
     Needed to correctly generate 'PRIMARY' or 'SIMPLE' for select_type column

sql/sql_select.cc

@@ -1189,7 +1189,7 @@ JOIN::optimize_inner()
   }
 
   if (select_lex->first_cond_optimization &&
-      transform_in_predicate_into_tvc(thd))
+      transform_in_predicates_into_in_subq(thd))
     DBUG_RETURN(1);
 
   // Update used tables after all handling derived table procedures

sql/sql_select.h

@@ -1648,7 +1648,7 @@ class JOIN :public Sql_alloc
                                bool need_order, bool distinct,
                                const char *message);
   JOIN_TAB *first_breadth_first_tab() { return join_tab; }
-  bool transform_in_predicate_into_tvc(THD *thd_arg);
+  bool transform_in_predicates_into_in_subq(THD *thd);
 private:
   /**
     Create a temporary table to be used for processing DISTINCT/ORDER