planner: rewrite avg for mpp

[fzhedu]

What problem does this PR solve?

Issue Number: close #xxx

Problem Summary: we cannot push down avg to tiflash entalily, so we rewrite it to case count(arg) when 0 then null else sum(arg)/count(arg) end, then the sum(arg) and count(arg) can be pushed down.

What is changed and how it works?

Proposal: xxx

What's Changed:

How it Works:

Related changes

• PR to update pingcap/docs/pingcap/docs-cn:
• Need to cherry-pick to the release branch

Check List

Tests

• Unit test
• Integration test
• Manual test (add detailed scripts or steps below)

mysql>  set @@tidb_allow_mpp=1; set @@tidb_opt_broadcast_join=1;
Query OK, 0 rows affected (0.00 sec)

Query OK, 0 rows affected (0.00 sec)

mysql> select value, avg(id)+1 as b from t1 group by value;
+-------+---------+
| value | b       |
+-------+---------+
|    11 |    NULL |
|    12 |    NULL |
|    10 | 10.0000 |
|     6 |  5.0000 |
|     2 |  2.0000 |
|     5 |  4.0000 |
|     8 |  8.0000 |
|     4 |  4.0000 |
+-------+---------+
8 rows in set (0.01 sec)

mysql> desc select value, avg(id)+1 as b from t1 group by value;
+------------------------------------+---------+--------------+---------------+--------------------------------------------------------------------------------------------------------------------------------------+
| id                                 | estRows | task         | access object | operator info                                                                                                                        |
+------------------------------------+---------+--------------+---------------+--------------------------------------------------------------------------------------------------------------------------------------+
| Projection_4                       | 7.20    | root         |               | test.t1.value, plus(case(eq(Column#4, 0), <nil>, div(Column#5, cast(Column#4, decimal(20,0) BINARY))), 1)->Column#6                  |
| └─TableReader_32                   | 7.20    | root         |               | data:ExchangeSender_31                                                                                                               |
|   └─ExchangeSender_31              | 7.20    | cop[tiflash] |               | ExchangeType: PassThrough                                                                                                            |
|     └─HashAgg_28                   | 7.20    | cop[tiflash] |               | group by:test.t1.value, funcs:sum(Column#14)->Column#4, funcs:sum(Column#15)->Column#5, funcs:firstrow(test.t1.value)->test.t1.value |
|       └─ExchangeReceiver_30        | 7.20    | cop[tiflash] |               |                                                                                                                                      |
|         └─ExchangeSender_29        | 7.20    | cop[tiflash] |               | ExchangeType: HashPartition, Hash Cols: test.t1.value                                                                                |
|           └─HashAgg_10             | 7.20    | cop[tiflash] |               | group by:test.t1.value, funcs:count(test.t1.id)->Column#14, funcs:sum(test.t1.id)->Column#15                                         |
|             └─TableFullScan_27     | 9.00    | cop[tiflash] | table:t1      | keep order:false, stats:pseudo                                                                                                       |
+------------------------------------+---------+--------------+---------------+--------------------------------------------------------------------------------------------------------------------------------------+
8 rows in set (0.00 sec)

mysql> select  avg(id)+1 from t1 where value=-1;
+-----------+
| avg(id)+1 |
+-----------+
|      NULL |
+-----------+
1 row in set (0.01 sec)

mysql> desc select  avg(id)+1 from t1 where value=-1;
+--------------------------------+---------+--------------+---------------+------------------------------------------------------------------------------------------------------+
| id                             | estRows | task         | access object | operator info                                                                                        |
+--------------------------------+---------+--------------+---------------+------------------------------------------------------------------------------------------------------+
| Projection_5                   | 1.00    | root         |               | plus(case(eq(Column#4, 0), <nil>, div(Column#5, cast(Column#4, decimal(20,0) BINARY))), 1)->Column#6 |
| └─StreamAgg_12                 | 1.00    | root         |               | funcs:count(Column#16)->Column#4, funcs:sum(Column#17)->Column#5                                     |
|   └─Projection_36              | 0.01    | root         |               | test.t1.id, cast(test.t1.id, decimal(32,0) BINARY)->Column#17                                        |
|     └─TableReader_28           | 0.01    | root         |               | data:Selection_27                                                                                    |
|       └─Selection_27           | 0.01    | cop[tiflash] |               | eq(test.t1.value, -1)                                                                                |
|         └─TableFullScan_26     | 9.00    | cop[tiflash] | table:t1      | keep order:false, stats:pseudo                                                                       |
+--------------------------------+---------+--------------+---------------+------------------------------------------------------------------------------------------------------+
6 rows in set (0.00 sec)

mysql> select value, avg(id) from t1 where value=-1 group by value;
Empty set (0.01 sec)

mysql> desc select value, avg(id) from t1 where value=-1 group by value;
+--------------------------------------+---------+--------------+---------------+--------------------------------------------------------------------------------------------------------------------------------------+
| id                                   | estRows | task         | access object | operator info                                                                                                                        |
+--------------------------------------+---------+--------------+---------------+--------------------------------------------------------------------------------------------------------------------------------------+
| Projection_5                         | 1.00    | root         |               | test.t1.value, case(eq(Column#4, 0), <nil>, div(Column#5, cast(Column#4, decimal(20,0) BINARY)))->Column#6                           |
| └─TableReader_40                     | 1.00    | root         |               | data:ExchangeSender_39                                                                                                               |
|   └─ExchangeSender_39                | 1.00    | cop[tiflash] |               | ExchangeType: PassThrough                                                                                                            |
|     └─HashAgg_36                     | 1.00    | cop[tiflash] |               | group by:test.t1.value, funcs:sum(Column#14)->Column#4, funcs:sum(Column#15)->Column#5, funcs:firstrow(test.t1.value)->test.t1.value |
|       └─ExchangeReceiver_38          | 1.00    | cop[tiflash] |               |                                                                                                                                      |
|         └─ExchangeSender_37          | 1.00    | cop[tiflash] |               | ExchangeType: HashPartition, Hash Cols: test.t1.value                                                                                |
|           └─HashAgg_11               | 1.00    | cop[tiflash] |               | group by:test.t1.value, funcs:count(test.t1.id)->Column#14, funcs:sum(test.t1.id)->Column#15                                         |
|             └─Selection_35           | 0.01    | cop[tiflash] |               | eq(test.t1.value, -1)                                                                                                                |
|               └─TableFullScan_34     | 9.00    | cop[tiflash] | table:t1      | keep order:false, stats:pseudo                                                                                                       |
+--------------------------------------+---------+--------------+---------------+--------------------------------------------------------------------------------------------------------------------------------------+
9 rows in set (0.00 sec)

Side effects

• Performance regression
  • Consumes more CPU
  • Consumes more MEM
• Breaking backward compatibility

Release note

• No release note