GaussDB-伪列
ROWNUM是一个伪列，它返回一个数字，表示从查询中获取结果的行编号。第一行的ROWNUM为1，第二行的为2，以此类推，使用ROWNUM来限制查询返回的行数，如以下示例所示：
| ``` gaussdb=# CREATE TABLE Students (name varchar(20), id int) with (STORAGE_TYPE = USTORE); gaussdb=# INSERT INTO Students VALUES ('Jack', 35); gaussdb=# INSERT INTO Students VALUES ('Leon', 15); gaussdb=# INSERT INTO Students VALUES ('James', 24); gaussdb=# INSERT INTO Students VALUES ('Taker', 81); gaussdb=# INSERT INTO Students VALUES ('Mary', 25); gaussdb=# INSERT INTO Students VALUES ('Rose', 64); gaussdb=# INSERT INTO Students VALUES ('Perl', 18); gaussdb=# INSERT INTO Students VALUES ('Under', 57); gaussdb=# INSERT INTO Students VALUES ('Angel', 101); gaussdb=# INSERT INTO Students VALUES ('Frank', 20); gaussdb=# INSERT INTO Students VALUES ('Charlie', 40); --输出表Students前10行。 gaussdb=# SELECT * FROM Students WHERE rownum <= 10; name | id -------+----- Jack | 35 Leon | 15 James | 24 Taker | 81 Mary | 25 Rose | 64 Perl | 18 Under | 57 Angel | 101 Frank | 20 (10 rows)
| ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
如果有子句跟在同一查询语句中，则结果输出的行将按照子句重新排序:
| ```
gaussdb=# SELECT * FROM Students WHERE rownum < 5 order by 1; name | id -------+---- Jack | 35 James | 24 Leon | 15 Taker | 81 (4 rows)
``` |
| -------------------------------------------------------------------------------------------------------------------------------------------------------- |
如果将子句嵌入到子查询中并将条件放在最外层的查询中，则能够在排序后使用ROWNUM条件：
| ```
gaussdb=# SELECT rownum, * FROM (SELECT * FROM Students order by 1) WHERE rownum <= 2; rownum | name | id --------+---------+----- 1 | Angel | 101 2 | Charlie | 40 (2 rows)
``` |
| -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
当ROWNUM大于正整数的值，认为条件始终为 false。例如以下所示，该语句不会返回表中任何结果：
| ```
gaussdb=# SELECT * FROM Students WHERE rownum > 1; name | id ------+---- (0 rows)
``` |
| ------------------------------------------------------------------------------------------- |
使用ROWNUM指定给表的一定范围的每一行分配值:
| ```
gaussdb=# SELECT * FROM Students; name | id ---------+----- Jack | 35 Leon | 15 James | 24 Taker | 81 Mary | 25 Rose | 64 Perl | 18 Under | 57 Angel | 101 Frank | 20 Charlie | 40 (11 rows) gaussdb=# UPDATE Students set id = id + 5 WHERE rownum < 4; UPDATE 3 gaussdb=# SELECT * FROM Students; name | id ---------+----- Jack | 40 Leon | 20 James | 29 Taker | 81 Mary | 25 Rose | 64 Perl | 18 Under | 57 Angel | 101 Frank | 20 Charlie | 40 (11 rows) gaussdb=# DROP TABLE Students; DROP TABLE
``` |
| ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
使用ROWNUM有一定的约束条件：
- ROWNUM不可作为别名，以免SQL语句出现歧义。
- 创建索引时不可使用ROWNUM。
- 创建表时默认值不可为ROWNUM。
- Where子句中不可使用ROWNUM的别名。
- 在插入数据时不可使用ROWNUM。
- 在无表查询中不可以使用ROWNUM。
- ROWNUM不能用于Limit子句。
- ROWNUM不能用于EXECUTE语句的参数。
- UPSERT语句不支持ROWNUM用做update子句更新。
- SELECT ... FOR UPDATE语句不支持ROWNUM用作投影列和WHERE条件。
- 若having子句中含有ROWNUM（且不在聚合函数中）时，group by子句中必须含有ROWNUM（且不在聚合函数中），除非group by子句存在表达式，例如：SELECT a + a FROM t group by a + a having rownum < 5。
- having子句中如果存在ROWNUM条件则不允许having子句下推至扫描节点：
| ```
gaussdb=# CREATE TABLE test (a int, b int); NOTICE: The 'DISTRIBUTE BY' clause is not specified. Using 'a' as the distribution column by default. HINT: Please use 'DISTRIBUTE BY' clause to specify suitable data distribution column. CREATE TABLE gaussdb=# INSERT INTO test SELECT generate_series, generate_series FROM generate_series(1, 10); INSERT 0 10 --rownum条件不能下推至seqscan。 gaussdb=# EXPLAIN SELECT a,rownum FROM test group by a,rownum having rownum < 5; QUERY PLAN ------------------------------------------------------------------------- HashAggregate (cost=3.45..3.49 rows=3 width=4) Group By Key: a, ROWNUM Filter: ((ROWNUM) < 5::numeric) -> Rownum (cost=0.19..3.40 rows=10 width=4) -> Streaming (type: GATHER) (cost=0.19..3.40 rows=10 width=4) Node/s: All datanodes -> Seq Scan on test (cost=0.00..3.03 rows=10 width=4) (7 rows)
``` |
| --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
- 子查询中如果存在ROWNUM条件则不允许谓词下推至扫描节点：
| ```
gaussdb=# EXPLAIN SELECT * FROM (SELECT * FROM test WHERE rownum < 5) WHERE b < 5; -- b<5 不能下推至seqscan QUERY PLAN ----------------------------------------------------------------------------------- Streaming (type: GATHER) (cost=0.06..0.63 rows=2 width=8) Node/s: All datanodes -> Subquery Scan on __unnamed_subquery__ (cost=0.00..0.50 rows=2 width=8) Filter: (__unnamed_subquery__.b < 5) -> Rownum (cost=0.00..0.45 rows=4 width=8) StopKey: (ROWNUM < 5::numeric) -> Streaming(type: BROADCAST) (cost=0.00..1.36 rows=12 width=8) Spawn on: All datanodes -> Rownum (cost=0.00..1.35 rows=4 width=8) StopKey: (ROWNUM < 5::numeric) -> Seq Scan on test (cost=0.00..3.03 rows=10 width=8) (11 rows) gaussdb=# DROP TABLE test; DROP TABLE
``` |
| ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
- ROWNUM查询结果将会由于CN接收DN数据的顺序不同导致查询结果不稳定。
| ```
gaussdb=# CREATE TABLE test(a int, b int); gaussdb=# INSERT INTO test VALUES(generate_series(1,10),generate_series(1,10)); --CN先收到DN1数据再收到DN2数据。 gaussdb=# SELECT rownum,* FROM test; rownum | a | b --------+----+---- 1 | 1 | 1 2 | 2 | 2 3 | 5 | 5 4 | 6 | 6 5 | 8 | 8 6 | 9 | 9 7 | 3 | 3 8 | 4 | 4 9 | 7 | 7 10 | 10 | 10 (10 rows) --CN先收到DN2数据再收到DN1数据。 gaussdb=# SELECT rownum,* FROM test; rownum | a | b --------+----+---- 1 | 3 | 3 2 | 4 | 4 3 | 7 | 7 4 | 10 | 10 5 | 1 | 1 6 | 2 | 2 7 | 5 | 5 8 | 6 | 6 9 | 8 | 8 10 | 9 | 9 (10 rows)
``` |
| ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
- 不推荐ROWNUM条件用于JOIN ON子句，GaussDB中ROWNUM条件用于JOIN ON子句时在LEFT JOIN、RIGHT JOIN、FULL JOIN场景下和MERGE INTO场景下与其他数据库行为不一致，直接进行业务迁移存在风险。
当父查询中有rownum限制条件同时子查询的投影列中有rownum时，该限制条件将下推至子查询。其中约束条件如下：
- 只有父查询rownum限制条件为“<”、“<=”、“=”，且子查询直接用rownum作为伪列时，才可以下推。
- 父查询中有多个对子查询中rownum的过滤条件并且满足下推要求时，按顺序仅下推第一个过滤条件。
- 当子查询中包括volatile函数、存储过程时不能下推。
更多详情请参考GaussDB 文档中心：<https://doc.hcs.huawei.com/db/zh-cn/gaussdbqlh/24.1.30/productdesc/qlh_03_0001.html>
