MDEV-7836: ANALYZE FORMAT=JSON should provide info about GROUP BY

ANALYZE should also record remove_duplicates() operation.

mysql-test/r/analyze_stmt_orderby.result

@@ -363,21 +363,68 @@ ANALYZE
   }
 }
 drop table t2;
-drop table t0, t1;
 #
 #  MDEV-8282: crash in filesort() with simple ordered delete
 #
-create table t1(a int) engine=innodb;
-delete from t1 order by a;
+create table t3(a int) engine=innodb;
+delete from t3 order by a;
 # EXPLAIN thinks it will use delete_all_rows():
 explain
-delete from t1 order by a;
+delete from t3 order by a;
 id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
 1	SIMPLE	NULL	NULL	NULL	NULL	NULL	NULL	1	Deleting all rows
 # ANALYZE shows that delete_all_rows() didn't work and we deleted rows
 # one-by-one:
 analyze
-delete from t1 order by a;
+delete from t3 order by a;
 id	select_type	table	type	possible_keys	key	key_len	ref	rows	r_rows	filtered	r_filtered	Extra
-1	SIMPLE	t1	ALL	NULL	NULL	NULL	NULL	1	0.00	100.00	100.00	Using filesort
-drop table t1;
+1	SIMPLE	t3	ALL	NULL	NULL	NULL	NULL	1	0.00	100.00	100.00	Using filesort
+drop table t3;
+#
+#
+#
+create table t3 (a int, b int);
+insert into t3 select a, 123 from t0;
+analyze format=json
+select distinct max(t3.b) Q from t0, t3 where t0.a=t3.a group by t0.a order by null;
+ANALYZE
+{
+  "query_block": {
+    "select_id": 1,
+    "r_loops": 1,
+    "r_total_time_ms": "REPLACED",
+    "duplicate_removal": {
+      "temporary_table": {
+        "table": {
+          "table_name": "t0",
+          "access_type": "ALL",
+          "r_loops": 1,
+          "rows": 10,
+          "r_rows": 10,
+          "r_total_time_ms": "REPLACED",
+          "filtered": 100,
+          "r_filtered": 100
+        },
+        "block-nl-join": {
+          "table": {
+            "table_name": "t3",
+            "access_type": "ALL",
+            "r_loops": 1,
+            "rows": 10,
+            "r_rows": 10,
+            "r_total_time_ms": "REPLACED",
+            "filtered": 100,
+            "r_filtered": 100
+          },
+          "buffer_type": "flat",
+          "buffer_size": "128Kb",
+          "join_type": "BNL",
+          "attached_condition": "(t3.a = t0.a)",
+          "r_filtered": 10
+        }
+      }
+    }
+  }
+}
+drop table t3;
+drop table t0,t1;

mysql-test/t/analyze_stmt_orderby.test

@@ -91,22 +91,33 @@ analyze format=json
 select MAX(b) from t2 where mod(a,2)=0 group by c;
 
 drop table t2;
-drop table t0, t1;
 
 --echo #
 --echo #  MDEV-8282: crash in filesort() with simple ordered delete
 --echo #
-create table t1(a int) engine=innodb;
-delete from t1 order by a;
+create table t3(a int) engine=innodb;
+delete from t3 order by a;
 
 --echo # EXPLAIN thinks it will use delete_all_rows():
 explain
-delete from t1 order by a;
+delete from t3 order by a;
 
 --echo # ANALYZE shows that delete_all_rows() didn't work and we deleted rows
 --echo # one-by-one:
 analyze
-delete from t1 order by a;
+delete from t3 order by a;
 
-drop table t1;
+drop table t3;
 
+--echo #
+--echo #
+--echo #
+create table t3 (a int, b int);
+insert into t3 select a, 123 from t0;
+
+--replace_regex /"r_total_time_ms": [0-9]*[.]?[0-9]*/"r_total_time_ms": "REPLACED"/ /"r_buffer_size": "[^"]+"/"r_buffer_size": "REPLACED"/
+analyze format=json
+select distinct max(t3.b) Q from t0, t3 where t0.a=t3.a group by t0.a order by null;
+
+drop table t3;
+drop table t0,t1;

sql/sql_analyze_stmt.cc

@@ -113,3 +113,22 @@ void Sort_and_group_tracker::report_tmp_table(TABLE *tbl)
   cur_action++;
 }
 
+
+void Sort_and_group_tracker::report_duplicate_removal()
+{
+  DBUG_ASSERT(cur_action < MAX_QEP_ACTIONS);
+  if (total_actions)
+  {
+    /* This is not the first execution. Check if the steps match.  */
+    if (qep_actions[cur_action] != EXPL_ACTION_REMOVE_DUPS)
+      varied_executions= true;
+  }
+
+  if (!varied_executions)
+  {
+    qep_actions[cur_action]= EXPL_ACTION_REMOVE_DUPS;
+  }
+
+  cur_action++;
+}
+

sql/sql_analyze_stmt.h

@@ -398,6 +398,12 @@ public:
   */
   Filesort_tracker *report_sorting(THD *thd);
   
+  /*
+    Report that remove_duplicates() is invoked [on a temp. table].
+    We don't collect any statistics on this operation, yet.
+  */
+  void report_duplicate_removal();
+  
   friend class Iterator;
   /*************** Statistics retrieval interface ***************/
   bool had_varied_executions() { return varied_executions; }

sql/sql_select.cc

@@ -21179,6 +21179,7 @@ remove_duplicates(JOIN *join, TABLE *table, List<Item> &fields, Item *having)
   THD *thd= join->thd;
 
   DBUG_ENTER("remove_duplicates");
+  join->explain->ops_tracker.report_duplicate_removal();
 
   table->reginfo.lock_type=TL_WRITE;