Fix for BUG#14920 Ordering aggregated result sets corrupts resultset.

The cause of the bug was the use of end_write_group instead of end_write
in the case when ORDER BY required a temporary table, which didn't take
into account the fact that loose index scan already computes the result
of MIN/MAX aggregate functions (and performs grouping).

The solution is to call end_write instead of end_write_group and to add
the MIN/MAX functions to the list of regular functions so that their
values are inserted into the temporary table.


mysql-test/r/group_min_max.result:
  Test for BUG#14920
mysql-test/t/group_min_max.test:
  Test for BUG#14920
sql/sql_class.cc:
  Added new member to TMP_TABLE_PARAM.
sql/sql_class.h:
  Added new member to TMP_TABLE_PARAM.
sql/sql_select.cc:
  Enable result rows generated by loose index scan being written into
  a temporary table. The change is necessary because loose index
  scan already computes the result of GROUP BY and the MIN/MAX aggregate
  functions. This is realized by three changes:
  - create_tmp_table allocates space for aggregate functions in the
    list of regular functions,
  - use end_write instead of end_write group,
  - copy the pointers to the MIN/MAX aggregate functions to the list
    of regular functions TMP_TABLE_PARAM::items_to_copy.
sql/sql_select.h:
  New parameter to create_tmp_table.

6 files changed, 113 insertions, 12 deletions

diff --git a/mysql-test/r/group_min_max.result b/mysql-test/r/group_min_max.result
index 038d0c75f74..636fb719c81 100644
--- a/mysql-test/r/group_min_max.result
+++ b/mysql-test/r/group_min_max.result
@@ -2002,3 +2002,34 @@ a	count(a)
 1	1
 NULL	1
 drop table t1;
+create table t1 (c1 int not null,c2 int not null, primary key(c1,c2));
+insert into t1 (c1,c2) values
+(10,1),(10,2),(10,3),(20,4),(20,5),(20,6),(30,7),(30,8),(30,9);
+select distinct c1, c2 from t1 order by c2;
+c1	c2
+10	1
+10	2
+10	3
+20	4
+20	5
+20	6
+30	7
+30	8
+30	9
+select c1,min(c2) as c2 from t1 group by c1 order by c2;
+c1	c2
+10	1
+20	4
+30	7
+select c1,c2 from t1 group by c1,c2 order by c2;
+c1	c2
+10	1
+10	2
+10	3
+20	4
+20	5
+20	6
+30	7
+30	8
+30	9
+drop table t1;
diff --git a/mysql-test/t/group_min_max.test b/mysql-test/t/group_min_max.test
index 3d751f4a571..f6be6170976 100644
--- a/mysql-test/t/group_min_max.test
+++ b/mysql-test/t/group_min_max.test
@@ -693,3 +693,16 @@ create table t1(a int, key(a)) engine=innodb;
 insert into t1 values(1);
 select a, count(a) from t1 group by a with rollup;
 drop table t1;
+
+#
+# Bug #14920 Ordering aggregated result sets with composite primary keys
+# corrupts resultset
+#
+
+create table t1 (c1 int not null,c2 int not null, primary key(c1,c2));
+insert into t1 (c1,c2) values
+(10,1),(10,2),(10,3),(20,4),(20,5),(20,6),(30,7),(30,8),(30,9);
+select distinct c1, c2 from t1 order by c2;
+select c1,min(c2) as c2 from t1 group by c1 order by c2;
+select c1,c2 from t1 group by c1,c2 order by c2;
+drop table t1;
diff --git a/sql/sql_class.cc b/sql/sql_class.cc
index c52addf3995..4b3d6486ef7 100644
--- a/sql/sql_class.cc
+++ b/sql/sql_class.cc
@@ -1802,6 +1802,7 @@ void TMP_TABLE_PARAM::init()
   group_parts= group_length= group_null_parts= 0;
   quick_group= 1;
   table_charset= 0;
+  precomputed_group_by= 0;
 }
 
 
diff --git a/sql/sql_class.h b/sql/sql_class.h
index ed6f5732ca8..c5f661cebd5 100644
--- a/sql/sql_class.h
+++ b/sql/sql_class.h
@@ -1819,11 +1819,18 @@ public:
   uint  convert_blob_length; 
   CHARSET_INFO *table_charset; 
   bool schema_table;
+  /*
+    True if GROUP BY and its aggregate functions are already computed
+    by a table access method (e.g. by loose index scan). In this case
+    query execution should not perform aggregation and should treat
+    aggregate functions as normal functions.
+  */
+  bool precomputed_group_by;
 
   TMP_TABLE_PARAM()
     :copy_field(0), group_parts(0),
      group_length(0), group_null_parts(0), convert_blob_length(0),
-     schema_table(0)
+     schema_table(0), precomputed_group_by(0)
   {}
   ~TMP_TABLE_PARAM()
   {
diff --git a/sql/sql_select.cc b/sql/sql_select.cc
index 9c7df461a3e..adebf9b4a52 100644
--- a/sql/sql_select.cc
+++ b/sql/sql_select.cc
@@ -1007,6 +1007,20 @@ JOIN::optimize()
   }
   having= 0;
 
+  /*
+    The loose index scan access method guarantees that all grouping or
+    duplicate row elimination (for distinct) is already performed
+    during data retrieval, and that all MIN/MAX functions are already
+    computed for each group. Thus all MIN/MAX functions should be
+    treated as regular functions, and there is no need to perform
+    grouping in the main execution loop.
+    Notice that currently loose index scan is applicable only for
+    single table queries, thus it is sufficient to test only the first
+    join_tab element of the plan for its access method.
+  */
+  if (join_tab->is_using_loose_index_scan())
+    tmp_table_param.precomputed_group_by= TRUE;
+
   /* Create a tmp table if distinct or if the sort is too complicated */
   if (need_tmp)
   {
@@ -1410,6 +1424,15 @@ JOIN::exec()
       else
       {
 	/* group data to new table */
+
+        /*
+          If the access method is loose index scan then all MIN/MAX
+          functions are precomputed, and should be treated as regular
+          functions. See extended comment in JOIN::exec.
+        */
+        if (curr_join->join_tab->is_using_loose_index_scan())
+          curr_join->tmp_table_param.precomputed_group_by= TRUE;
+
 	if (!(curr_tmp_table=
 	      exec_tmp_table2= create_tmp_table(thd,
 						&curr_join->tmp_table_param,
@@ -8279,6 +8302,7 @@ create_tmp_table(THD *thd,TMP_TABLE_PARAM *param,List<Item> &fields,
   MEM_ROOT *mem_root_save, own_root;
   TABLE *table;
   uint	i,field_count,null_count,null_pack_length;
+  uint  copy_func_count= param->func_count;
   uint  hidden_null_count, hidden_null_pack_length, hidden_field_count;
   uint  blob_count,group_null_items, string_count;
   uint  temp_pool_slot=MY_BIT_NONE;
@@ -8342,6 +8366,16 @@ create_tmp_table(THD *thd,TMP_TABLE_PARAM *param,List<Item> &fields,
   field_count=param->field_count+param->func_count+param->sum_func_count;
   hidden_field_count=param->hidden_field_count;
 
+  /*
+    When loose index scan is employed as access method, it already
+    computes all groups and the result of all aggregate functions. We
+    make space for the items of the aggregate function in the list of
+    functions TMP_TABLE_PARAM::items_to_copy, so that the values of
+    these items are stored in the temporary table.
+  */
+  if (param->precomputed_group_by)
+    copy_func_count+= param->sum_func_count;
+  
   init_sql_alloc(&own_root, TABLE_ALLOC_BLOCK_SIZE, 0);
 
   if (!multi_alloc_root(&own_root,
@@ -8349,7 +8383,7 @@ create_tmp_table(THD *thd,TMP_TABLE_PARAM *param,List<Item> &fields,
                         &reg_field, sizeof(Field*) * (field_count+1),
                         &blob_field, sizeof(uint)*(field_count+1),
                         &from_field, sizeof(Field*)*field_count,
-                        &copy_func, sizeof(*copy_func)*(param->func_count+1),
+                        &copy_func, sizeof(*copy_func)*(copy_func_count+1),
                         &param->keyinfo, sizeof(*param->keyinfo),
                         &key_part_info,
                         sizeof(*key_part_info)*(param->group_parts+1),
@@ -9241,11 +9275,13 @@ bool create_myisam_from_heap(THD *thd, TABLE *table, TMP_TABLE_PARAM *param,
 Next_select_func setup_end_select_func(JOIN *join)
 {
   TABLE *table= join->tmp_table;
+  TMP_TABLE_PARAM *tmp_tbl= &join->tmp_table_param;
   Next_select_func end_select;
+
   /* Set up select_end */
   if (table)
   {
-    if (table->group && join->tmp_table_param.sum_func_count)
+    if (table->group && tmp_tbl->sum_func_count)
     {
       if (table->s->keys)
       {
@@ -9258,7 +9294,7 @@ Next_select_func setup_end_select_func(JOIN *join)
 	end_select=end_unique_update;
       }
     }
-    else if (join->sort_and_group)
+    else if (join->sort_and_group && !tmp_tbl->precomputed_group_by)
     {
       DBUG_PRINT("info",("Using end_write_group"));
       end_select=end_write_group;
@@ -9267,19 +9303,27 @@ Next_select_func setup_end_select_func(JOIN *join)
     {
       DBUG_PRINT("info",("Using end_write"));
       end_select=end_write;
+      if (tmp_tbl->precomputed_group_by)
+      {
+        /*
+          A preceding call to create_tmp_table in the case when loose
+          index scan is used guarantees that
+          TMP_TABLE_PARAM::items_to_copy has enough space for the group
+          by functions. It is OK here to use memcpy since we copy
+          Item_sum pointers into an array of Item pointers.
+        */
+        memcpy(tmp_tbl->items_to_copy + tmp_tbl->func_count,
+               join->sum_funcs,
+               sizeof(Item*)*tmp_tbl->sum_func_count);
+        tmp_tbl->items_to_copy[tmp_tbl->func_count+tmp_tbl->sum_func_count]= 0;
+      }
     }
   }
   else
   {
-    /* Test if data is accessed via QUICK_GROUP_MIN_MAX_SELECT. */
-    bool is_using_quick_group_min_max_select=
-      (join->join_tab->select && join->join_tab->select->quick &&
-       (join->join_tab->select->quick->get_type() ==
-        QUICK_SELECT_I::QS_TYPE_GROUP_MIN_MAX));
-
     if ((join->sort_and_group ||
          (join->procedure && join->procedure->flags & PROC_GROUP)) &&
-        !is_using_quick_group_min_max_select)
+        !tmp_tbl->precomputed_group_by)
       end_select= end_send_group;
     else
       end_select= end_send;
@@ -10553,7 +10597,6 @@ end_write(JOIN *join, JOIN_TAB *join_tab __attribute__((unused)),
   {
     copy_fields(&join->tmp_table_param);
     copy_funcs(join->tmp_table_param.items_to_copy);
-
 #ifdef TO_BE_DELETED
     if (!table->uniques)			// If not unique handling
     {
diff --git a/sql/sql_select.h b/sql/sql_select.h
index 2f53c9a3b35..b3abd59f2b4 100644
--- a/sql/sql_select.h
+++ b/sql/sql_select.h
@@ -140,6 +140,12 @@ typedef struct st_join_table {
   nested_join_map embedding_map;
   
   void cleanup();
+  inline bool is_using_loose_index_scan()
+  {
+    return (select && select->quick &&
+            (select->quick->get_type() ==
+             QUICK_SELECT_I::QS_TYPE_GROUP_MIN_MAX));
+  }
 } JOIN_TAB;
 
 enum_nested_loop_state sub_select_cache(JOIN *join, JOIN_TAB *join_tab, bool