Bug #43029: FORCE INDEX FOR ORDER BY is ignored when join

buffering is used

FORCE INDEX FOR ORDER BY now prevents the optimizer from 
using join buffering. As a result the optimizer can use
indexed access on the first table and doesn't need to 
sort the complete resultset at the end of the statement.

mysql-test/r/order_by.result

@@ -1557,3 +1557,34 @@ a
 2001
 1991
 DROP TABLE t1;
+#
+# Bug #43029: FORCE INDEX FOR ORDER BY is ignored when join buffering 
+#   is used
+#
+CREATE TABLE t1 (a INT, b INT, KEY (a));
+INSERT INTO t1 VALUES (0, NULL), (1, NULL), (2, NULL), (3, NULL);
+INSERT INTO t1 SELECT a+4, b FROM t1;
+INSERT INTO t1 SELECT a+8, b FROM t1;
+CREATE TABLE t2 (a INT, b INT);
+INSERT INTO t2 VALUES (0,NULL), (1,NULL), (2,NULL), (3,NULL), (4,NULL);
+INSERT INTO t2 SELECT a+4, b FROM t2;
+# shouldn't have "using filesort"
+EXPLAIN 
+SELECT * FROM t1 FORCE INDEX FOR ORDER BY (a), t2 WHERE t1.a < 2 ORDER BY t1.a;
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1	range	a	a	5	NULL	2	Using where
+1	SIMPLE	t2	ALL	NULL	NULL	NULL	NULL	10	
+# should have "using filesort"
+EXPLAIN 
+SELECT * FROM t1 USE INDEX FOR ORDER BY (a), t2 WHERE t1.a < 2 ORDER BY t1.a;
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1	range	a	a	5	NULL	2	Using where; Using temporary; Using filesort
+1	SIMPLE	t2	ALL	NULL	NULL	NULL	NULL	10	Using join buffer
+# should have "using filesort"
+EXPLAIN 
+SELECT * FROM t1 FORCE INDEX FOR JOIN (a), t2 WHERE t1.a < 2 ORDER BY t1.a;
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1	range	a	a	5	NULL	2	Using where; Using temporary; Using filesort
+1	SIMPLE	t2	ALL	NULL	NULL	NULL	NULL	10	Using join buffer
+DROP TABLE t1, t2;
+End of 5.1 tests

mysql-test/t/order_by.test

@@ -1402,3 +1402,35 @@ SELECT DISTINCT a FROM t1 WHERE b = 1 ORDER BY c DESC LIMIT 0, 9;
 SELECT DISTINCT a FROM t1 WHERE b = 1 ORDER BY c DESC LIMIT 0, 9;
 
 DROP TABLE t1;
+
+--echo #
+--echo # Bug #43029: FORCE INDEX FOR ORDER BY is ignored when join buffering 
+--echo #   is used
+--echo #
+
+CREATE TABLE t1 (a INT, b INT, KEY (a));
+
+INSERT INTO t1 VALUES (0, NULL), (1, NULL), (2, NULL), (3, NULL);
+INSERT INTO t1 SELECT a+4, b FROM t1;
+INSERT INTO t1 SELECT a+8, b FROM t1;
+
+CREATE TABLE t2 (a INT, b INT);
+
+INSERT INTO t2 VALUES (0,NULL), (1,NULL), (2,NULL), (3,NULL), (4,NULL);
+INSERT INTO t2 SELECT a+4, b FROM t2;
+
+--echo # shouldn't have "using filesort"
+EXPLAIN 
+SELECT * FROM t1 FORCE INDEX FOR ORDER BY (a), t2 WHERE t1.a < 2 ORDER BY t1.a;
+
+--echo # should have "using filesort"
+EXPLAIN 
+SELECT * FROM t1 USE INDEX FOR ORDER BY (a), t2 WHERE t1.a < 2 ORDER BY t1.a;
+
+--echo # should have "using filesort"
+EXPLAIN 
+SELECT * FROM t1 FORCE INDEX FOR JOIN (a), t2 WHERE t1.a < 2 ORDER BY t1.a;
+
+DROP TABLE t1, t2;
+
+--echo End of 5.1 tests

sql/sql_base.cc

@@ -2305,7 +2305,8 @@ bool reopen_name_locked_table(THD* thd, TABLE_LIST* table_list, bool link_in)
   table->tablenr=thd->current_tablenr++;
   table->used_fields=0;
   table->const_table=0;
-  table->null_row= table->maybe_null= table->force_index= 0;
+  table->null_row= table->maybe_null= 0;
+  table->force_index= table->force_index_order= table->force_index_group= 0;
   table->status=STATUS_NO_RECORD;
   DBUG_RETURN(FALSE);
 }
@@ -2963,7 +2964,8 @@ TABLE *open_table(THD *thd, TABLE_LIST *table_list, MEM_ROOT *mem_root,
   table->tablenr=thd->current_tablenr++;
   table->used_fields=0;
   table->const_table=0;
-  table->null_row= table->maybe_null= table->force_index= 0;
+  table->null_row= table->maybe_null= 0;
+  table->force_index= table->force_index_order= table->force_index_group= 0;
   table->status=STATUS_NO_RECORD;
   table->insert_values= 0;
   table->fulltext_searched= 0;

sql/sql_parse.cc

@@ -6258,6 +6258,7 @@ TABLE_LIST *st_select_lex::add_table_to_list(THD *thd,
   ptr->table_name_length=table->table.length;
   ptr->lock_type=   lock_type;
   ptr->updating=    test(table_options & TL_OPTION_UPDATING);
+  /* TODO: remove TL_OPTION_FORCE_INDEX as it looks like it's not used */
   ptr->force_index= test(table_options & TL_OPTION_FORCE_INDEX);
   ptr->ignore_leaves= test(table_options & TL_OPTION_IGNORE_LEAVES);
   ptr->derived=	    table->sel;

sql/sql_select.cc

@@ -1231,13 +1231,22 @@ JOIN::optimize()
       (!group_list && tmp_table_param.sum_func_count))
     order=0;
 
-  // Can't use sort on head table if using row cache
+  // Can't use sort on head table if using join buffering
   if (full_join)
   {
-    if (group_list)
-      simple_group=0;
-    if (order)
-      simple_order=0;
+    TABLE *stable= (sort_by_table == (TABLE *) 1 ? 
+      join_tab[const_tables].table : sort_by_table);
+    /* 
+      FORCE INDEX FOR ORDER BY can be used to prevent join buffering when
+      sorting on the first table.
+    */
+    if (!stable || !stable->force_index_order)
+    {
+      if (group_list)
+        simple_group= 0;
+      if (order)
+        simple_order= 0;
+    }
   }
 
   /*

sql/sql_select.h

@@ -357,6 +357,8 @@ public:
     simple_xxxxx is set if ORDER/GROUP BY doesn't include any references
     to other tables than the first non-constant table in the JOIN.
     It's also set if ORDER/GROUP BY is empty.
+    Used for deciding for or against using a temporary table to compute 
+    GROUP/ORDER BY.
   */
   bool simple_order, simple_group;
   /**

sql/table.cc

@@ -4637,7 +4637,8 @@ Item_subselect *TABLE_LIST::containing_subselect()
     (TABLE_LIST::index_hints). Using the information in this tagged list
     this function sets the members st_table::keys_in_use_for_query, 
     st_table::keys_in_use_for_group_by, st_table::keys_in_use_for_order_by,
-    st_table::force_index and st_table::covering_keys.
+    st_table::force_index, st_table::force_index_order, 
+    st_table::force_index_group and st_table::covering_keys.
 
     Current implementation of the runtime does not allow mixing FORCE INDEX
     and USE INDEX, so this is checked here. Then the FORCE INDEX list 
@@ -4765,14 +4766,28 @@ bool TABLE_LIST::process_index_hints(TABLE *tbl)
     }
 
     /* process FORCE INDEX as USE INDEX with a flag */
+    if (!index_order[INDEX_HINT_FORCE].is_clear_all())
+    {
+      tbl->force_index_order= TRUE;
+      index_order[INDEX_HINT_USE].merge(index_order[INDEX_HINT_FORCE]);
+    }
+
+    if (!index_group[INDEX_HINT_FORCE].is_clear_all())
+    {
+      tbl->force_index_group= TRUE;
+      index_group[INDEX_HINT_USE].merge(index_group[INDEX_HINT_FORCE]);
+    }
+
+    /*
+      TODO: get rid of tbl->force_index (on if any FORCE INDEX is specified) and
+      create tbl->force_index_join instead.
+      Then use the correct force_index_XX instead of the global one.
+    */
     if (!index_join[INDEX_HINT_FORCE].is_clear_all() ||
-        !index_order[INDEX_HINT_FORCE].is_clear_all() ||
-        !index_group[INDEX_HINT_FORCE].is_clear_all())
+        tbl->force_index_group || tbl->force_index_order)
     {
       tbl->force_index= TRUE;
       index_join[INDEX_HINT_USE].merge(index_join[INDEX_HINT_FORCE]);
-      index_order[INDEX_HINT_USE].merge(index_order[INDEX_HINT_FORCE]);
-      index_group[INDEX_HINT_USE].merge(index_group[INDEX_HINT_FORCE]);
     }
 
     /* apply USE INDEX */

sql/table.h

@@ -752,6 +752,18 @@ struct st_table {
     bytes, it would take up 4.
   */
   my_bool force_index;
+
+  /**
+    Flag set when the statement contains FORCE INDEX FOR ORDER BY
+    See TABLE_LIST::process_index_hints().
+  */
+  my_bool force_index_order;
+
+  /**
+    Flag set when the statement contains FORCE INDEX FOR GROUP BY
+    See TABLE_LIST::process_index_hints().
+  */
+  my_bool force_index_group;
   my_bool distinct,const_table,no_rows;
 
   /**