Common Table Expressions

A Common Table Expression (CTE) is a temporary result set that is used only within the statement that contains it. A CTE does not store data as a separate object, but returns data as if it were a table. Unlike derived tables, CTEs can reference themselves and can be reused multiple times in the same query. A CTE can be recursive.

Scenarios:

• You can reuse the same CTE multiple times in the same statement without rewriting the same logic multiple times.
• You can use a CTE to simplify recursive queries, such as those used to process hierarchical data.
• You can break a complex query into smaller parts by using a CTE, thus simplifying the query logic.

OceanBase Database supports both recursive and non-recursive CTEs.

CTE syntax

A common table expression is an optional part of the DML statement syntax. It is defined by using the WITH clause. Multiple WITH clauses can be separated by commas. Each WITH clause contains a subquery that generates a result set and associates the result set with a name. The syntax is as follows:

 WITH [RECURSIVE]
      cte_name [(column_name [, column_name] ...)] AS (subquery)
      [, cte_name [(column_name [, column_name] ...)] AS (subquery)] ...

Arguments

If the CTE name is followed by a name list in parentheses, the names in the list are column names. The number of names must be the same as the number of columns in the SELECT statement of the CTE. If no column names are specified, the column names are from the first SELECT list in AS(subquery).

Scenarios supported by the WITH clause

You can use the WITH clause in the following scenarios:

• At the beginning of a SELECT statement.

  WITH ... SELECT ...
  

• At the beginning of a subquery (including a derived table subquery).

  SELECT ... WHERE id IN (WITH ... SELECT ...) ...
  SELECT * FROM (WITH ... SELECT ...) AS dt ...
  

• Immediately before the SELECT clause in a statement that contains a SELECT statement.

  INSERT ... WITH ... SELECT ...
  REPLACE ... WITH ... SELECT ...
  CREATE TABLE ... WITH ... SELECT ...
  CREATE VIEW ... WITH ... SELECT ...
  

You can use at most one WITH clause at the same level. If multiple WITH clauses are included in the query, separate them with commas.

WITH cte1 AS (...), cte2 AS (...) SELECT ...

The WITH clause can define one or more common table expressions. However, each CTE name must be unique within the WITH clause. The following example is illegal:

WITH cte1 AS (...), cte1 AS (...) SELECT ...

Structure of a recursive CTE

A recursive CTE has the following structure:

• If the CTE in the WITH clause references itself, the WITH clause must start with WITH RECURSIVE. Otherwise, it does not need to start with RECURSIVE.

• The recursive CTE's subquery consists of two parts that are separated by UNION [ALL]:

  SELECT ...      -- Returns the initial row set.
  UNION ALL
  SELECT ...      -- Returns additional row sets.
  

  The first SELECT generates one or more initial rows for the CTE and does not reference the CTE name. The second SELECT generates additional rows through recursion by referencing the CTE name in its FROM clause. The recursion ends when the second SELECT does not generate new rows. Therefore, a recursive CTE consists of a non-recursive SELECT part and a recursive SELECT part. Each SELECT part can be composed of multiple SELECT statements.

• The column types of the CTE are inferred from the column types of the non-recursive SELECT part. By default, all columns can be null. The types of the recursive SELECT part are ignored.

• The rows generated by the recursive part are processed only based on the preceding iteration. If the recursive part contains multiple query blocks, the iterations of the query blocks are scheduled in an unspecified order, and a query block operates on the rows generated by the preceding iteration or by other query blocks after the preceding iteration.

Here is an example:

WITH RECURSIVE cte1 (n) AS
(
  SELECT 1                /*Non-recursive part. It retrieves a single row to generate the initial row set.*/
  UNION ALL
  SELECT n + 2 FROM cte1 WHERE n < 10     /*Recursive part. It generates a new value that is 2 greater than the value of n in the preceding row set. The recursion continues until n is no less than 10.*/
)
SELECT * FROM cte1;

Limitations

The recursive SELECT part of a recursive CTE must meet the following requirements:

• It must not contain the following structures:

  • Aggregate functions such as SUM()

  • Window functions

  • GROUP BY

  • ORDER BY

  • DISTINCT

• It must reference a table only through a subquery in its FROM clause. It can join the CTE with a table or another CTE. If it uses the CTE in a JOIN clause, the LEFT JOIN cannot be on the right side of the JOIN clause.

The cost estimation displayed in the EXPLAIN statement for a recursive CTE represents the cost of each iteration and can differ significantly from the total cost. The optimizer cannot estimate the number of iterations because it cannot determine when the condition in the WHERE clause becomes False.

• The recursive SELECT part cannot contain the UNION [DISTINCT] or LIMIT structure.

Examples

The following example shows how to create a student hierarchy table that contains the IDs, names, and mentors of students to demonstrate the difference between recursive and non-recursive CTEs.

First, create the student hierarchy table and insert some data into it:

obclient> CREATE TABLE student (
    student_id INT PRIMARY KEY,
    name VARCHAR(100),
    mentor_id INT,
    FOREIGN KEY (mentor_id) REFERENCES student(student_id)
);
Query OK, 0 rows affected

obclient> INSERT INTO student (student_id, name, mentor_id) VALUES
(1, 'Alice', NULL),
(2, 'Bob', 1),
(3, 'Charlie', 1),
(4, 'David', 2),
(5, 'Eve', 3);
Query OK, 5 rows affected

In this data model, Alice is a top-level student who does not have a mentor. Bob and Charlie are Alice's students. David is Bob's student, and Eve is Charlie's student.

Non-recursive CTE example

A non-recursive CTE does not reference itself. For example, if you want to select all students directly mentored by Alice (i.e., first-level students), you can use the following non-recursive CTE:

WITH Alice_Students AS (
    SELECT * FROM student WHERE mentor_id = 1
)
SELECT * FROM Alice_Students;

The execution result is as follows:

+------------+---------+-----------+
| student_id | name    | mentor_id |
+------------+---------+-----------+
|          2 | Bob     |         1 |
|          3 | Charlie |         1 |
+------------+---------+-----------+
2 rows in set

Recursive CTE example

If you want to find all students directly and indirectly mentored by Alice (i.e., all students at all levels), you must use a recursive CTE. A recursive CTE references itself to query students at the next level.

WITH RECURSIVE Student_Hierarchy AS (
    -- Anchor member: Select Alice as the starting point.
    SELECT student_id, name, mentor_id FROM student WHERE mentor_id IS NULL
    UNION ALL
    -- Recursive member: Select direct students of students at the previous level.
    SELECT s.student_id, s.name, s.mentor_id
    FROM student s
    INNER JOIN Student_Hierarchy sh ON s.mentor_id = sh.student_id
)
SELECT * FROM Student_Hierarchy;

This recursive CTE selects Alice as the starting point and then recursively selects direct students of each student until no more students are found.

The execution result is as follows:

+------------+---------+-----------+
| student_id | name    | mentor_id |
+------------+---------+-----------+
|          1 | Alice   |      NULL |
|          3 | Charlie |         1 |
|          2 | Bob     |         1 |
|          5 | Eve     |         3 |
|          4 | David   |         2 |
+------------+---------+-----------+
5 rows in set

In this example, the recursive CTE allows you to traverse the hierarchical structure of students, while the non-recursive CTE can only select students at a fixed level.