从窗口函数中推导 TopN 或 Limit

窗口函数是一种常见的 SQL 函数。对于 ROW_NUMBER() 或者 RANK() 等编号相关的窗口函数,一种常见的用法是在进行窗口函数求值之后,对求值的结果进行过滤,例如:

  1. SELECT * FROM (SELECT ROW_NUMBER() OVER (ORDER BY a) AS rownumber FROM t) dt WHERE rownumber <= 3

按照正常的 SQL 执行流程,TiDB 需要先对 t 表的所有数据进行排序,然后为每一行都求得相应的 ROW_NUMBER() 结果,最后再进行 rownumber <= 3 的过滤。

从 v7.0.0 开始,TiDB 支持从窗口函数中推导 TopN 或 Limit 算子。通过该优化规则,TiDB 可以将原始 SQL 等价改写成以下形式:

  1. WITH t_topN AS (SELECT a FROM t1 ORDER BY a LIMIT 3) SELECT * FROM (SELECT ROW_NUMBER() OVER (ORDER BY a) AS rownumber FROM t_topN) dt WHERE rownumber <= 3

可以看出,改写后,TiDB 可以从窗口函数与后续的过滤条件中推导出一个 TopN 算子,相比于原始 SQL 中的 Sort 算子(对应 ORDER BY),TopN 算子的运行效率远高于 Sort 算子,而且 TiKV 和 TiFlash 均支持 TopN 算子的下推,因此能进一步提升改写后的 SQL 的性能。

从窗口函数中推导 TopN 或 Limit 默认关闭。你可以通过将 session 变量 tidb_opt_derive_topn 设置为 ON 开启该功能。

开启后,如需关闭,可以进行以下操作之一:

限制

  • 目前仅 ROW_NUMBER() 窗口函数支持 SQL 语句改写。
  • 只有当 SQL 语句的过滤条件是针对 ROW_NUMBER() 结果而且过滤条件为 < 或者 <= 时,TiDB 才支持改写 SQL 语句。

示例

以下通过一些例子对该优化规则进行说明。

不包含 PARTITION BY 的窗口函数

示例 1:不包含 ORDER BY 的窗口函数

  1. CREATE TABLE t(id int, value int);
  2. SET tidb_opt_derive_topn=on;
  3. EXPLAIN SELECT * FROM (SELECT ROW_NUMBER() OVER () AS rownumber FROM t) dt WHERE rownumber <= 3;
  1. +----------------------------------+---------+-----------+---------------+-----------------------------------------------------------------------+
  2. | id | estRows | task | access object | operator info |
  3. +----------------------------------+---------+-----------+---------------+-----------------------------------------------------------------------+
  4. | Projection_9 | 2.40 | root | | Column#5 |
  5. | └─Selection_10 | 2.40 | root | | le(Column#5, 3) |
  6. | └─Window_11 | 3.00 | root | | row_number()->Column#5 over(rows between current row and current row) |
  7. | └─Limit_15 | 3.00 | root | | offset:0, count:3 |
  8. | └─TableReader_26 | 3.00 | root | | data:Limit_25 |
  9. | └─Limit_25 | 3.00 | cop[tikv] | | offset:0, count:3 |
  10. | └─TableFullScan_24 | 3.00 | cop[tikv] | table:t | keep order:false, stats:pseudo |
  11. +----------------------------------+---------+-----------+---------------+-----------------------------------------------------------------------+

在该查询中,优化器从窗口函数中推导出来了 Limit 并将它下推给了 TiKV。

示例 2:包含 ORDER BY 的窗口函数

  1. CREATE TABLE t(id int, value int);
  2. SET tidb_opt_derive_topn=on;
  3. EXPLAIN SELECT * FROM (SELECT ROW_NUMBER() OVER (ORDER BY value) AS rownumber FROM t) dt WHERE rownumber <= 3;
  1. +----------------------------------+----------+-----------+---------------+---------------------------------------------------------------------------------------------+
  2. | id | estRows | task | access object | operator info |
  3. +----------------------------------+----------+-----------+---------------+---------------------------------------------------------------------------------------------+
  4. | Projection_10 | 2.40 | root | | Column#5 |
  5. | └─Selection_11 | 2.40 | root | | le(Column#5, 3) |
  6. | └─Window_12 | 3.00 | root | | row_number()->Column#5 over(order by test.t.value rows between current row and current row) |
  7. | └─TopN_13 | 3.00 | root | | test.t.value, offset:0, count:3 |
  8. | └─TableReader_25 | 3.00 | root | | data:TopN_24 |
  9. | └─TopN_24 | 3.00 | cop[tikv] | | test.t.value, offset:0, count:3 |
  10. | └─TableFullScan_23 | 10000.00 | cop[tikv] | table:t | keep order:false, stats:pseudo |
  11. +----------------------------------+----------+-----------+---------------+---------------------------------------------------------------------------------------------+

在该查询中,优化器从窗口函数中推导出来了 TopN 并将它下推给了 TiKV。

包含 PARTITION BY 的窗口函数

从窗口函数中推导 TopN 或 Limit - 图1

注意

当窗口函数包含 PARTITION BY 时,该优化规则仅在 partition 列是主键的前缀且主键是聚簇索引的时候才能生效。

示例 3:不包含 ORDER BY 的窗口函数

  1. CREATE TABLE t(id1 int, id2 int, value1 int, value2 int, primary key(id1,id2) clustered);
  2. SET tidb_opt_derive_topn=on;
  3. EXPLAIN SELECT * FROM (SELECT ROW_NUMBER() OVER (PARTITION BY id1) AS rownumber FROM t) dt WHERE rownumber <= 3;
  1. +------------------------------------+---------+-----------+---------------+-----------------------------------------------------------------------------------------------+
  2. | id | estRows | task | access object | operator info |
  3. +------------------------------------+---------+-----------+---------------+-----------------------------------------------------------------------------------------------+
  4. | Projection_10 | 2.40 | root | | Column#6 |
  5. | └─Selection_11 | 2.40 | root | | le(Column#6, 3) |
  6. | └─Shuffle_26 | 3.00 | root | | execution info: concurrency:2, data sources:[TableReader_24] |
  7. | └─Window_12 | 3.00 | root | | row_number()->Column#6 over(partition by test.t.id1 rows between current row and current row) |
  8. | └─Sort_25 | 3.00 | root | | test.t.id1 |
  9. | └─TableReader_24 | 3.00 | root | | data:Limit_23 |
  10. | └─Limit_23 | 3.00 | cop[tikv] | | partition by test.t.id1, offset:0, count:3 |
  11. | └─TableFullScan_22 | 3.00 | cop[tikv] | table:t | keep order:false, stats:pseudo |
  12. +------------------------------------+---------+-----------+---------------+-----------------------------------------------------------------------------------------------+

在该查询中,优化器从窗口函数中推导出来了 Limit 并将它下推给了 TiKV。值得一提的是这个 Limit 其实是 partition Limit,也就是说对于每个相同 id1 值组成的一组数据上都会进行一次 Limit。

示例 4:包含 ORDER BY 的窗口函数

  1. CREATE TABLE t(id1 int, id2 int, value1 int, value2 int, primary key(id1,id2) clustered);
  2. SET tidb_opt_derive_topn=on;
  3. EXPLAIN SELECT * FROM (SELECT ROW_NUMBER() OVER (PARTITION BY id1 ORDER BY value1) AS rownumber FROM t) dt WHERE rownumber <= 3;
  1. +------------------------------------+----------+-----------+---------------+----------------------------------------------------------------------------------------------------------------------+
  2. | id | estRows | task | access object | operator info |
  3. +------------------------------------+----------+-----------+---------------+----------------------------------------------------------------------------------------------------------------------+
  4. | Projection_10 | 2.40 | root | | Column#6 |
  5. | └─Selection_11 | 2.40 | root | | le(Column#6, 3) |
  6. | └─Shuffle_23 | 3.00 | root | | execution info: concurrency:3, data sources:[TableReader_21] |
  7. | └─Window_12 | 3.00 | root | | row_number()->Column#6 over(partition by test.t.id1 order by test.t.value1 rows between current row and current row) |
  8. | └─Sort_22 | 3.00 | root | | test.t.id1, test.t.value1 |
  9. | └─TableReader_21 | 3.00 | root | | data:TopN_19 |
  10. | └─TopN_19 | 3.00 | cop[tikv] | | partition by test.t.id1 order by test.t.value1, offset:0, count:3 |
  11. | └─TableFullScan_18 | 10000.00 | cop[tikv] | table:t | keep order:false, stats:pseudo |
  12. +------------------------------------+----------+-----------+---------------+----------------------------------------------------------------------------------------------------------------------+

在该查询中,优化器从窗口函数中推导出来了 TopN 并将它下推给了 TiKV。需要注意的是,这个 TopN 其实是 partition 的 TopN, 也就是说对于每个相同 id1 值组成的一组数据上都会进行一次 TopN。

示例 5:PARTITION BY 列不是主键的前缀

  1. CREATE TABLE t(id1 int, id2 int, value1 int, value2 int, primary key(id1,id2) clustered);
  2. SET tidb_opt_derive_topn=on;
  3. EXPLAIN SELECT * FROM (SELECT ROW_NUMBER() OVER (PARTITION BY value1) AS rownumber FROM t) dt WHERE rownumber <= 3;
  1. +----------------------------------+----------+-----------+---------------+--------------------------------------------------------------------------------------------------+
  2. | id | estRows | task | access object | operator info |
  3. +----------------------------------+----------+-----------+---------------+--------------------------------------------------------------------------------------------------+
  4. | Projection_9 | 8000.00 | root | | Column#6 |
  5. | └─Selection_10 | 8000.00 | root | | le(Column#6, 3) |
  6. | └─Shuffle_15 | 10000.00 | root | | execution info: concurrency:5, data sources:[TableReader_13] |
  7. | └─Window_11 | 10000.00 | root | | row_number()->Column#6 over(partition by test.t.value1 rows between current row and current row) |
  8. | └─Sort_14 | 10000.00 | root | | test.t.value1 |
  9. | └─TableReader_13 | 10000.00 | root | | data:TableFullScan_12 |
  10. | └─TableFullScan_12 | 10000.00 | cop[tikv] | table:t | keep order:false, stats:pseudo |
  11. +----------------------------------+----------+-----------+---------------+--------------------------------------------------------------------------------------------------+

在该查询中,因为 partition 的列不是主键的前缀,所以 SQL 没有被改写。

示例 6:PARTITION BY 列是主键的前缀,但主键不是聚簇索引

  1. CREATE TABLE t(id1 int, id2 int, value1 int, value2 int, primary key(id1,id2) nonclustered);
  2. SET tidb_opt_derive_topn=on;
  3. EXPLAIN SELECT * FROM (SELECT ROW_NUMBER() OVER (PARTITION BY id1) AS rownumber FROM t use index()) dt WHERE rownumber <= 3;
  1. +----------------------------------+----------+-----------+---------------+-----------------------------------------------------------------------------------------------+
  2. | id | estRows | task | access object | operator info |
  3. +----------------------------------+----------+-----------+---------------+-----------------------------------------------------------------------------------------------+
  4. | Projection_9 | 8000.00 | root | | Column#7 |
  5. | └─Selection_10 | 8000.00 | root | | le(Column#7, 3) |
  6. | └─Shuffle_15 | 10000.00 | root | | execution info: concurrency:5, data sources:[TableReader_13] |
  7. | └─Window_11 | 10000.00 | root | | row_number()->Column#7 over(partition by test.t.id1 rows between current row and current row) |
  8. | └─Sort_14 | 10000.00 | root | | test.t.id1 |
  9. | └─TableReader_13 | 10000.00 | root | | data:TableFullScan_12 |
  10. | └─TableFullScan_12 | 10000.00 | cop[tikv] | table:t | keep order:false, stats:pseudo |
  11. +----------------------------------+----------+-----------+---------------+-----------------------------------------------------------------------------------------------+

在该查询中,即使 PARTITION 的列是主键的前缀,但是因为主键不是聚簇索引,所以 SQL 没被改写。