Skip to main content
Version: 3.2

DELETE

Deletes data rows from a table based on the specified conditions. The table can be a partitioned or non-partitioned table.

For Duplicate Key tables, Aggregate tables, and Unique Key tables, you can delete data from specified partitions. From v2.3, Primary Key tables support complete DELETE...WHERE semantics, which allows you to delete data rows based on the primary key, any column, or the results of a subquery. From v3.0, StarRocks enriches the DELETE...WHERE semantics with multi-table joins and common table expressions (CTEs). If you need to join Primary Key tables with other tables in the database, you can reference these other tables in the USING clause or CTE.

Usage notes​

  • You must have privileges on the table and database you want to perform DELETE.
  • Frequent DELETE operations are not recommended. If needed, perform such operations during off-peak hours.
  • The DELETE operation only deletes data in the table. The table remains. To drop the table, run DROP TABLE.
  • To prevent misoperations from deleting data in the entire table, you must specify the WHERE clause in the DELETE statement.
  • The deleted rows are not immediately cleaned. They are marked as "deleted" and will be temporarily saved in Segment. Physically, the rows are removed only after data version merge (compaction) is completed.
  • This operation also deletes data of the materialized views that reference this table.

Duplicate Key tables, Aggregate tables, and Unique Key tables​

Syntax​

DELETE FROM  [<db_name>.]<table_name> [PARTITION <partition_name>]
WHERE
column_name1 op { value | value_list } [ AND column_name2 op { value | value_list } ...]

Parameters​

ParameterRequiredDescription
db_nameNoThe database to which the destination table belongs. If this parameter is not specified, the current database is used by default.
table_nameYesThe table from which you want to delete data.
partition_nameNoThe partition from which you want to delete data.
column_nameYesThe column you want to use as the DELETE condition. You can specify one or more columns.
opYesThe operator used in the DELETE condition. The supported operators are =, >, <, >=, <=, !=, IN, and NOT IN.

Limits and usage notes​

  • For Duplicate Key tables, you can use any column as the DELETE condition. For Aggregate tables and Unique Key tables, only key columns can be used as the DELETE condition.

  • The conditions that you specify must be in the AND relation. If you want to specify conditions in OR relation, you must specify the conditions in separate DELETE statements.

  • For Duplicate Key tables, Aggregate tables, and Unique Key tables, the DELETE statement does not support using subquery results as conditions.

Impacts​

After you execute the DELETE statement, the query performance of your cluster may deteriorate for a period of time (before compaction is completed). The degree of deterioration varies based on the number of conditions that you specify. A larger number of conditions indicates a higher degree of deterioration.

Examples​

Create a table and insert data​

The following example creates a partitioned Duplicate Key table.

CREATE TABLE `my_table` (
    `date` date NOT NULL,
    `k1` int(11) NOT NULL COMMENT "",
    `k2` varchar(65533) NULL DEFAULT "" COMMENT "")
DUPLICATE KEY(`date`)
PARTITION BY RANGE(`date`)
(
    PARTITION p1 VALUES [('2022-03-11'), ('2022-03-12')),
    PARTITION p2 VALUES [('2022-03-12'), ('2022-03-13'))
)
DISTRIBUTED BY HASH(`date`)
PROPERTIES
("replication_num" = "3");

INSERT INTO `my_table` VALUES
('2022-03-11', 3, 'abc'),
('2022-03-11', 2, 'acb'),
('2022-03-11', 4, 'abc'),
('2022-03-12', 2, 'bca'),
('2022-03-12', 4, 'cba'),
('2022-03-12', 5, 'cba');

Query data​

select * from my_table order by date;
+------------+------+------+
| date       | k1   | k2   |
+------------+------+------+
| 2022-03-11 |    3 | abc  |
| 2022-03-11 |    2 | acb  |
| 2022-03-11 |    4 | abc  |
| 2022-03-12 |    2 | bca  |
| 2022-03-12 |    4 | cba  |
| 2022-03-12 |    5 | cba  |
+------------+------+------+

Delete data​

Delete data from a specified partition​

Delete rows whose k1 values are 3 from the p1 partition.

DELETE FROM my_table PARTITION p1
WHERE k1 = 3;

-- The query result shows rows whose `k1` values are `3` are deleted.

select * from my_table partition (p1);
+------------+------+------+
| date       | k1   | k2   |
+------------+------+------+
| 2022-03-11 |    2 | acb  |
| 2022-03-11 |    4 | abc  |
+------------+------+------+
Delete data from a specified partition using AND​

Delete rows whose k1 values are greater than or equal to 3 and whose k2 values are "abc" from the p1 partition.

DELETE FROM my_table PARTITION p1
WHERE k1 >= 3 AND k2 = "abc";

select * from my_table partition (p1);
+------------+------+------+
| date       | k1   | k2   |
+------------+------+------+
| 2022-03-11 |    2 | acb  |
+------------+------+------+
Delete data from all partitions​

Delete rows whose k2 values are "abc" or "cba" from all partitions.

DELETE FROM my_table
WHERE  k2 in ("abc", "cba");

select * from my_table order by date;
+------------+------+------+
| date       | k1   | k2   |
+------------+------+------+
| 2022-03-11 |    2 | acb  |
| 2022-03-12 |    2 | bca  |
+------------+------+------+

Primary Key tables​

From v2.3, Primary Key tables support complete DELETE...WHERE semantics, which allows you to delete data rows based on the primary key, any column, or a subquery.

Syntax​

[ WITH <with_query> [, ...] ]
DELETE FROM <table_name>
[ USING <from_item> [, ...] ]
[ WHERE <where_condition> ]

Parameters​

ParameterRequiredDescription
with_queryNoOne or more CTEs that can be referenced by name in a DELETE statement. CTEs are temporary result sets that can improve the readability of complex statements.
table_nameYesThe table from which you want to delete data.
from_itemNoOne or more other tables in the database. These tables can be joined with the table being operated based on the condition specified in the WHERE clause. Based on the result set of the join query, StarRocks deletes the matched rows from the table being operated. For example, if the USING clause is USING t1 WHERE t0.pk = t1.pk;, StarRocks converts the table expression in the USING clause to t0 JOIN t1 ON t0.pk=t1.pk; when executing the DELETE statement.
where_conditionYesThe condition based on which you want to delete rows. Only rows that meet the WHERE condition can be deleted. This parameter is required, because it helps prevent you from accidentally deleting the entire table. If you want to delete the entire table, you can use 'WHERE true'.

Limits and usage notes​

  • Primary Key tables do not support deleting data from a specified partition, for example, DELETE FROM <table_name> PARTITION <partition_id> WHERE <where_condition>.
  • The following comparison operators are supported: =, >, <, >=, <=, !=, IN, NOT IN.
  • The following logical operators are supported: AND and OR.
  • You cannot use the DELETE statement to run concurrent DELETE operations or to delete data at data loading. If you perform such operations, the atomicity, consistency, isolation, and durability (ACID) of transactions may not be ensured.

Examples​

Create a table and insert data​

Create a Primary Key table named score_board:

CREATE TABLE `score_board` (
  `id` int(11) NOT NULL COMMENT "",
  `name` varchar(65533) NULL DEFAULT "" COMMENT "",
  `score` int(11) NOT NULL DEFAULT "0" COMMENT ""
) ENGINE=OLAP
PRIMARY KEY(`id`)
COMMENT "OLAP"
DISTRIBUTED BY HASH(`id`)
PROPERTIES (
"replication_num" = "3",
"storage_format" = "DEFAULT",
"enable_persistent_index" = "false"
);

INSERT INTO score_board VALUES
(0, 'Jack', 21),
(1, 'Bob', 21),
(2, 'Stan', 21),
(3, 'Sam', 22);

Query data​

Execute the following statement to insert data into the score_board table:

select * from score_board;
+------+------+------+
| id   | name | score|
+------+------+------+
|    0 | Jack |   21 |
|    1 | Bob  |   21 |
|    2 | Stan |   21 |
|    3 | Sam  |   22 |
+------+------+------+
4 rows in set (0.00 sec)

Delete data​

Delete data by primary key​

You can specify the primary key in the DELETE statement, so StarRocks does not need to scan the entire table.

Delete rows whose id values are 0 from the score_board table.

DELETE FROM score_board WHERE id = 0;

select * from score_board;
+------+------+------+
| id   | name | score|
+------+------+------+
|    1 | Bob  |   21 |
|    2 | Stan |   21 |
|    3 | Sam  |   22 |
+------+------+------+
Delete data by condition​

Example 1: Delete rows whose score values are 22 from the score_board table.

DELETE FROM score_board WHERE score = 22;

select * from score_board;
+------+------+------+
| id   | name | score|
+------+------+------+
|    0 | Jack |   21 |
|    1 | Bob  |   21 |
|    2 | Stan |   21 |
+------+------+------+

Example 2: Delete rows whose score values are less than 22 from the score_board table.

DELETE FROM score_board WHERE score < 22;

select * from score_board;
+------+------+------+
| id   | name | score|
+------+------+------+
|    3 | Sam  |   22 |
+------+------+------+

Example 3: Delete rows whose score values are less than 22 and whose name values are not Bob from the score_board table.

DELETE FROM score_board WHERE score < 22 and name != "Bob";

select * from score_board;
+------+------+------+
| id   | name | score|
+------+------+------+
|    1 | Bob  |   21 |
|    3 | Sam  |   22 |
+------+------+------+
2 rows in set (0.00 sec)
Delete data by subquery result​

You can nest one or more subqueries in the DELETE statement and use the subquery results as conditions.

Before you delete data, execute the following statement to create another table named users:

CREATE TABLE `users` (
  `uid` int(11) NOT NULL COMMENT "",
  `name` varchar(65533) NOT NULL COMMENT "",
  `country` varchar(65533) NULL COMMENT ""
) ENGINE=OLAP
PRIMARY KEY(`uid`)
COMMENT "OLAP"
DISTRIBUTED BY HASH(`uid`)
PROPERTIES (
"replication_num" = "3",
"storage_format" = "DEFAULT",
"enable_persistent_index" = "false"
);

Insert data into the users table:

INSERT INTO users VALUES
(0, "Jack", "China"),
(2, "Stan", "USA"),
(1, "Bob", "China"),
(3, "Sam", "USA");
select * from users;
+------+------+---------+
| uid  | name | country |
+------+------+---------+
|    0 | Jack | China   |
|    1 | Bob  | China   |
|    2 | Stan | USA     |
|    3 | Sam  | USA     |
+------+------+---------+
4 rows in set (0.00 sec)

Nest a subquery to find the rows whose country values are China from the users table, and delete the rows, which have the same name values as the rows returned from the subquery, from the score_board table. You can use one of the following methods to fulfill the purpose:

  • Method 1
DELETE FROM score_board
WHERE name IN (select name from users where country = "China");
    
select * from score_board;
+------+------+------+
| id   | name | score|
+------+------+------+
|    2 | Stan |   21 |
|    3 | Sam  |   22 |
+------+------+------+
  • Method 2
DELETE FROM score_board
WHERE EXISTS (select name from users
              where score_board.name = users.name and country = "China");
    
select * from score_board;
+------+------+-------+
| id   | name | score |
+------+------+-------+
|    2 | Stan |    21 |
|    3 | Sam  |    22 |
+------+------+-------+
2 rows in set (0.00 sec)
Delete data by using multi-table join or CTE​

To delete all movies produced by the producer "foo", you can execute the following statement:

DELETE FROM films USING producers
WHERE producer_id = producers.id
    AND producers.name = 'foo';

You can also use a CTE to rewrite the above statement to improve readability.

WITH foo_producers as (
    SELECT * from producers
    where producers.name = 'foo'
)
DELETE FROM films USING foo_producers
WHERE producer_id = foo_producers.id;

References​

SHOW DELETE: queries historical DELETE operations that were successfully performed on Duplicate Key tables, Aggregate tables, and Unique Key tables.