65

note: this question has been updated to reflect that we are currently using MySQL, having done so, I would like to see a how much easier it would be if we switched to a CTE-supporting database.

I have a self-referencing table with a primary key, id and a foreign key parent_id.

+------------+--------------+------+-----+---------+----------------+
| Field      | Type         | Null | Key | Default | Extra          |
+------------+--------------+------+-----+---------+----------------+
| id         | int(11)      | NO   | PRI | NULL    | auto_increment | 
| parent_id  | int(11)      | YES  |     | NULL    |                | 
| name       | varchar(255) | YES  |     | NULL    |                | 
| notes      | text         | YES  |     | NULL    |                | 
+------------+--------------+------+-----+---------+----------------+

Given a name, how can I query the top-level parent?

Given a name, how can I query all of the id's associated with a record of name = 'foo'?

context: I am not a dba, but am planning to ask a dba to implement this type of hierarchical structure and would like to test some queries. The motivation for doing so is described by Kattge et al 2011.


Here is an example of the relationships among ids in the table:

enter image description here

-- -----------------------------------------------------
-- Create a new database called 'testdb'
-- -----------------------------------------------------
SET @OLD_UNIQUE_CHECKS=@@UNIQUE_CHECKS, UNIQUE_CHECKS=0;
SET @OLD_FOREIGN_KEY_CHECKS=@@FOREIGN_KEY_CHECKS, FOREIGN_KEY_CHECKS=0;
SET @OLD_SQL_MODE=@@SQL_MODE, SQL_MODE='TRADITIONAL';

CREATE SCHEMA IF NOT EXISTS `testdb` DEFAULT CHARACTER SET latin1 COLLATE latin1_swedish_ci ;
USE `testdb` ;

-- -----------------------------------------------------
-- Table `testdb`.`observations`
-- -----------------------------------------------------
CREATE  TABLE IF NOT EXISTS `testdb`.`observations` (
  `id` INT NOT NULL ,
  `parent_id` INT NULL ,
  `name` VARCHAR(45) NULL ,
  PRIMARY KEY (`id`) )
ENGINE = InnoDB;

SET SQL_MODE=@OLD_SQL_MODE;
SET FOREIGN_KEY_CHECKS=@OLD_FOREIGN_KEY_CHECKS;
SET UNIQUE_CHECKS=@OLD_UNIQUE_CHECKS;

-- -----------------------------------------------------
-- Add Example Data Set
-- -----------------------------------------------------


INSERT INTO observations VALUES (1,3), (2,5), (3,NULL), (4,10), 
   (5,NULL), (6,1), (7,5), (8,10), (9,10), (10,3);
0

3 Answers 3

72

You definitely have to script this via MySQL Stored Procedure Language

Here is a Stored Function called GetParentIDByIDto Retrieve a ParentID given an ID to Search For

DELIMITER $$
DROP FUNCTION IF EXISTS `junk`.`GetParentIDByID` $$
CREATE FUNCTION `junk`.`GetParentIDByID` (GivenID INT) RETURNS INT
DETERMINISTIC
BEGIN
    DECLARE rv INT;

    SELECT IFNULL(parent_id,-1) INTO rv FROM
    (SELECT parent_id FROM pctable WHERE id = GivenID) A;
    RETURN rv;
END $$
DELIMITER ;

Here is a Stored Function called GetAncestryto Retrieve a List of ParentIDs starting from the First Generation all the up the hierarchy given an ID to start with:

DELIMITER $$
DROP FUNCTION IF EXISTS `junk`.`GetAncestry` $$
CREATE FUNCTION `junk`.`GetAncestry` (GivenID INT) RETURNS VARCHAR(1024)
DETERMINISTIC
BEGIN
    DECLARE rv VARCHAR(1024);
    DECLARE cm CHAR(1);
    DECLARE ch INT;

    SET rv = '';
    SET cm = '';
    SET ch = GivenID;
    WHILE ch > 0 DO
        SELECT IFNULL(parent_id,-1) INTO ch FROM
        (SELECT parent_id FROM pctable WHERE id = ch) A;
        IF ch > 0 THEN
            SET rv = CONCAT(rv,cm,ch);
            SET cm = ',';
        END IF;
    END WHILE;
    RETURN rv;
END $$
DELIMITER ;

Here is something to generate sample data:

USE junk
DROP TABLE IF EXISTS pctable;
CREATE TABLE pctable
(
    id INT NOT NULL AUTO_INCREMENT,
    parent_id INT,
    PRIMARY KEY (id)
) ENGINE=MyISAM;
INSERT INTO pctable (parent_id) VALUES (0);
INSERT INTO pctable (parent_id) SELECT parent_id+1 FROM pctable;
INSERT INTO pctable (parent_id) SELECT parent_id+2 FROM pctable;
INSERT INTO pctable (parent_id) SELECT parent_id+3 FROM pctable;
INSERT INTO pctable (parent_id) SELECT parent_id+4 FROM pctable;
INSERT INTO pctable (parent_id) SELECT parent_id+5 FROM pctable;
SELECT * FROM pctable;

Here is what it generates:

mysql> USE junk
Database changed
mysql> DROP TABLE IF EXISTS pctable;
Query OK, 0 rows affected (0.00 sec)

mysql> CREATE TABLE pctable
    -> (
    ->     id INT NOT NULL AUTO_INCREMENT,
    ->     parent_id INT,
    ->     PRIMARY KEY (id)
    -> ) ENGINE=MyISAM;
Query OK, 0 rows affected (0.05 sec)

mysql> INSERT INTO pctable (parent_id) VALUES (0);
Query OK, 1 row affected (0.00 sec)

mysql> INSERT INTO pctable (parent_id) SELECT parent_id+1 FROM pctable;
Query OK, 1 row affected (0.00 sec)
Records: 1  Duplicates: 0  Warnings: 0

mysql> INSERT INTO pctable (parent_id) SELECT parent_id+2 FROM pctable;
Query OK, 2 rows affected (0.00 sec)
Records: 2  Duplicates: 0  Warnings: 0

mysql> INSERT INTO pctable (parent_id) SELECT parent_id+3 FROM pctable;
Query OK, 4 rows affected (0.00 sec)
Records: 4  Duplicates: 0  Warnings: 0

mysql> INSERT INTO pctable (parent_id) SELECT parent_id+4 FROM pctable;
Query OK, 8 rows affected (0.01 sec)
Records: 8  Duplicates: 0  Warnings: 0

mysql> INSERT INTO pctable (parent_id) SELECT parent_id+5 FROM pctable;
Query OK, 16 rows affected (0.00 sec)
Records: 16  Duplicates: 0  Warnings: 0

mysql> SELECT * FROM pctable;
+----+-----------+
| id | parent_id |
+----+-----------+
|  1 |         0 |
|  2 |         1 |
|  3 |         2 |
|  4 |         3 |
|  5 |         3 |
|  6 |         4 |
|  7 |         5 |
|  8 |         6 |
|  9 |         4 |
| 10 |         5 |
| 11 |         6 |
| 12 |         7 |
| 13 |         7 |
| 14 |         8 |
| 15 |         9 |
| 16 |        10 |
| 17 |         5 |
| 18 |         6 |
| 19 |         7 |
| 20 |         8 |
| 21 |         8 |
| 22 |         9 |
| 23 |        10 |
| 24 |        11 |
| 25 |         9 |
| 26 |        10 |
| 27 |        11 |
| 28 |        12 |
| 29 |        12 |
| 30 |        13 |
| 31 |        14 |
| 32 |        15 |
+----+-----------+
32 rows in set (0.00 sec)

Here are what the functions generate for each value:

mysql> SELECT id,GetParentIDByID(id),GetAncestry(id) FROM pctable;
+----+---------------------+-----------------+
| id | GetParentIDByID(id) | GetAncestry(id) |
+----+---------------------+-----------------+
|  1 |                   0 |                 |
|  2 |                   1 | 1               |
|  3 |                   2 | 2,1             |
|  4 |                   3 | 3,2,1           |
|  5 |                   3 | 3,2,1           |
|  6 |                   4 | 4,3,2,1         |
|  7 |                   5 | 5,3,2,1         |
|  8 |                   6 | 6,4,3,2,1       |
|  9 |                   4 | 4,3,2,1         |
| 10 |                   5 | 5,3,2,1         |
| 11 |                   6 | 6,4,3,2,1       |
| 12 |                   7 | 7,5,3,2,1       |
| 13 |                   7 | 7,5,3,2,1       |
| 14 |                   8 | 8,6,4,3,2,1     |
| 15 |                   9 | 9,4,3,2,1       |
| 16 |                  10 | 10,5,3,2,1      |
| 17 |                   5 | 5,3,2,1         |
| 18 |                   6 | 6,4,3,2,1       |
| 19 |                   7 | 7,5,3,2,1       |
| 20 |                   8 | 8,6,4,3,2,1     |
| 21 |                   8 | 8,6,4,3,2,1     |
| 22 |                   9 | 9,4,3,2,1       |
| 23 |                  10 | 10,5,3,2,1      |
| 24 |                  11 | 11,6,4,3,2,1    |
| 25 |                   9 | 9,4,3,2,1       |
| 26 |                  10 | 10,5,3,2,1      |
| 27 |                  11 | 11,6,4,3,2,1    |
| 28 |                  12 | 12,7,5,3,2,1    |
| 29 |                  12 | 12,7,5,3,2,1    |
| 30 |                  13 | 13,7,5,3,2,1    |
| 31 |                  14 | 14,8,6,4,3,2,1  |
| 32 |                  15 | 15,9,4,3,2,1    |
+----+---------------------+-----------------+
32 rows in set (0.02 sec)

MORAL OF THE STORY : Recursive data retrieval must be scripted in MySQL

UPDATE 2011-10-24 17:17 EDT

Here is the reverse of GetAncestry. I call it GetFamilyTree.

Here is the algorithm:

  • Place Given ID in Queue
  • Loop
    • Dequeue into front_id
    • Retrieve all ids into queue_children whose parent_id = front_id
    • Append queue_children to retval_list (rv)
    • Enqueue queue_children
    • Repeat until queue and queue_children are concurrently empty

I believe from my Data Structures and Algorithms classes in College, this is called something like preorder/prefix tree traversal.

Here is the code:

DELIMITER $$

DROP FUNCTION IF EXISTS `junk`.`GetFamilyTree` $$
CREATE FUNCTION `junk`.`GetFamilyTree` (GivenID INT) RETURNS varchar(1024) CHARSET latin1
DETERMINISTIC
BEGIN

    DECLARE rv,q,queue,queue_children VARCHAR(1024);
    DECLARE queue_length,front_id,pos INT;

    SET rv = '';
    SET queue = GivenID;
    SET queue_length = 1;

    WHILE queue_length > 0 DO
        SET front_id = FORMAT(queue,0);
        IF queue_length = 1 THEN
            SET queue = '';
        ELSE
            SET pos = LOCATE(',',queue) + 1;
            SET q = SUBSTR(queue,pos);
            SET queue = q;
        END IF;
        SET queue_length = queue_length - 1;

        SELECT IFNULL(qc,'') INTO queue_children
        FROM (SELECT GROUP_CONCAT(id) qc
        FROM pctable WHERE parent_id = front_id) A;

        IF LENGTH(queue_children) = 0 THEN
            IF LENGTH(queue) = 0 THEN
                SET queue_length = 0;
            END IF;
        ELSE
            IF LENGTH(rv) = 0 THEN
                SET rv = queue_children;
            ELSE
                SET rv = CONCAT(rv,',',queue_children);
            END IF;
            IF LENGTH(queue) = 0 THEN
                SET queue = queue_children;
            ELSE
                SET queue = CONCAT(queue,',',queue_children);
            END IF;
            SET queue_length = LENGTH(queue) - LENGTH(REPLACE(queue,',','')) + 1;
        END IF;
    END WHILE;

    RETURN rv;

END $$

Here is what each row produces

mysql> SELECT id,parent_id,GetParentIDByID(id),GetAncestry(id),GetFamilyTree(id) FROM pctable;
+----+-----------+---------------------+-----------------+--------------------------------------------------------------------------------------+
| id | parent_id | GetParentIDByID(id) | GetAncestry(id) | GetFamilyTree(id)                                                                    |
+----+-----------+---------------------+-----------------+--------------------------------------------------------------------------------------+
|  1 |         0 |                   0 |                 | 2,3,4,5,6,9,7,10,17,8,11,18,15,22,25,12,13,19,16,23,26,14,20,21,24,27,32,28,29,30,31 |
|  2 |         1 |                   1 | 1               | 3,4,5,6,9,7,10,17,8,11,18,15,22,25,12,13,19,16,23,26,14,20,21,24,27,32,28,29,30,31   |
|  3 |         2 |                   2 | 2,1             | 4,5,6,9,7,10,17,8,11,18,15,22,25,12,13,19,16,23,26,14,20,21,24,27,32,28,29,30,31     |
|  4 |         3 |                   3 | 3,2,1           | 6,9,8,11,18,15,22,25,14,20,21,24,27,32,31                                            |
|  5 |         3 |                   3 | 3,2,1           | 7,10,17,12,13,19,16,23,26,28,29,30                                                   |
|  6 |         4 |                   4 | 4,3,2,1         | 8,11,18,14,20,21,24,27,31                                                            |
|  7 |         5 |                   5 | 5,3,2,1         | 12,13,19,28,29,30                                                                    |
|  8 |         6 |                   6 | 6,4,3,2,1       | 14,20,21,31                                                                          |
|  9 |         4 |                   4 | 4,3,2,1         | 15,22,25,32                                                                          |
| 10 |         5 |                   5 | 5,3,2,1         | 16,23,26                                                                             |
| 11 |         6 |                   6 | 6,4,3,2,1       | 24,27                                                                                |
| 12 |         7 |                   7 | 7,5,3,2,1       | 28,29                                                                                |
| 13 |         7 |                   7 | 7,5,3,2,1       | 30                                                                                   |
| 14 |         8 |                   8 | 8,6,4,3,2,1     | 31                                                                                   |
| 15 |         9 |                   9 | 9,4,3,2,1       | 32                                                                                   |
| 16 |        10 |                  10 | 10,5,3,2,1      |                                                                                      |
| 17 |         5 |                   5 | 5,3,2,1         |                                                                                      |
| 18 |         6 |                   6 | 6,4,3,2,1       |                                                                                      |
| 19 |         7 |                   7 | 7,5,3,2,1       |                                                                                      |
| 20 |         8 |                   8 | 8,6,4,3,2,1     |                                                                                      |
| 21 |         8 |                   8 | 8,6,4,3,2,1     |                                                                                      |
| 22 |         9 |                   9 | 9,4,3,2,1       |                                                                                      |
| 23 |        10 |                  10 | 10,5,3,2,1      |                                                                                      |
| 24 |        11 |                  11 | 11,6,4,3,2,1    |                                                                                      |
| 25 |         9 |                   9 | 9,4,3,2,1       |                                                                                      |
| 26 |        10 |                  10 | 10,5,3,2,1      |                                                                                      |
| 27 |        11 |                  11 | 11,6,4,3,2,1    |                                                                                      |
| 28 |        12 |                  12 | 12,7,5,3,2,1    |                                                                                      |
| 29 |        12 |                  12 | 12,7,5,3,2,1    |                                                                                      |
| 30 |        13 |                  13 | 13,7,5,3,2,1    |                                                                                      |
| 31 |        14 |                  14 | 14,8,6,4,3,2,1  |                                                                                      |
| 32 |        15 |                  15 | 15,9,4,3,2,1    |                                                                                      |
+----+-----------+---------------------+-----------------+--------------------------------------------------------------------------------------+
32 rows in set (0.04 sec)

This algorithm works cleanly provided there are no cyclic paths. If there exists any cyclic paths, you would have to add a 'visited' column to the table.

Once you add visited column, here is the algorithm blocking cyclic relationships:

  • Place Given ID in Queue
  • Mark all visited with 0
  • Loop
    • Dequeue into front_id
    • Retrieve all ids into queue_children whose parent_id = front_id and visited = 0
    • Mark all queue_children just retrieved with visited = 1
    • Append queue_children to retval_list (rv)
    • Enqueue queue_children
    • Repeat until queue and queue_children are concurrently empty

UPDATE 2011-10-24 17:37 EDT

I created a new table called observations and populated your sample data. I changed the stored procedures to use observations instead of pctable. Here is your output:

mysql> CREATE TABLE observations LIKE pctable;
Query OK, 0 rows affected (0.04 sec)

mysql> INSERT INTO observations VALUES (1,3), (2,5), (3,0), (4,10),(5,0),(6,1),(7,5),(8,10),(9,10),(10,3);
Query OK, 10 rows affected (0.00 sec)
Records: 10  Duplicates: 0  Warnings: 0

mysql> SELECT * FROM observations;
+----+-----------+
| id | parent_id |
+----+-----------+
|  1 |         3 |
|  2 |         5 |
|  3 |         0 |
|  4 |        10 |
|  5 |         0 |
|  6 |         1 |
|  7 |         5 |
|  8 |        10 |
|  9 |        10 |
| 10 |         3 |
+----+-----------+
10 rows in set (0.00 sec)

mysql> SELECT id,parent_id,GetParentIDByID(id),GetAncestry(id),GetFamilyTree(id) FROM observations;
+----+-----------+---------------------+-----------------+-------------------+
| id | parent_id | GetParentIDByID(id) | GetAncestry(id) | GetFamilyTree(id) |
+----+-----------+---------------------+-----------------+-------------------+
|  1 |         3 |                   3 |                 | 6                 |
|  2 |         5 |                   5 | 5               |                   |
|  3 |         0 |                   0 |                 | 1,10,6,4,8,9      |
|  4 |        10 |                  10 | 10,3            |                   |
|  5 |         0 |                   0 |                 | 2,7               |
|  6 |         1 |                   1 | 1               |                   |
|  7 |         5 |                   5 | 5               |                   |
|  8 |        10 |                  10 | 10,3            |                   |
|  9 |        10 |                  10 | 10,3            |                   |
| 10 |         3 |                   3 | 3               | 4,8,9             |
+----+-----------+---------------------+-----------------+-------------------+
10 rows in set (0.01 sec)

UPDATE 2011-10-24 18:22 EDT

I changed the code for GetAncestry. There was WHILE ch > 1 it should be WHILE ch > 0

mysql> SELECT id,parent_id,GetParentIDByID(id),GetAncestry(id),GetFamilyTree(id) FROM observations;
+----+-----------+---------------------+-----------------+-------------------+
| id | parent_id | GetParentIDByID(id) | GetAncestry(id) | GetFamilyTree(id) |
+----+-----------+---------------------+-----------------+-------------------+
|  1 |         3 |                   3 | 3               | 6                 |
|  2 |         5 |                   5 | 5               |                   |
|  3 |         0 |                   0 |                 | 1,10,6,4,8,9      |
|  4 |        10 |                  10 | 10,3            |                   |
|  5 |         0 |                   0 |                 | 2,7               |
|  6 |         1 |                   1 | 1,3             |                   |
|  7 |         5 |                   5 | 5               |                   |
|  8 |        10 |                  10 | 10,3            |                   |
|  9 |        10 |                  10 | 10,3            |                   |
| 10 |         3 |                   3 | 3               | 4,8,9             |
+----+-----------+---------------------+-----------------+-------------------+
10 rows in set (0.01 sec)

Try it now !!!

0
33

Getting all parents of a specified node:

WITH RECURSIVE tree AS ( 
   SELECT id, 
          name, 
          parent_id,
          1 as level 
   FROM the_table
   WHERE name = 'foo'

   UNION ALL 

   SELECT p.id,
          p.name,
          p.parent_id, 
          t.level + 1
   FROM the_table p
     JOIN tree t ON t.parent_id = p.id
)
SELECT *
FROM tree

To get the root node, you can e.g. ORDER BY level and take the first row

Getting all children of a specified node:

WITH RECURSIVE tree AS ( 
   SELECT id, 
          name, 
          parent_id,
          1 as level 
   FROM the_table
   WHERE name = 'foo'

   UNION ALL 

   SELECT p.id,
          p.name,
          p.parent_id, 
          t.level + 1
   FROM your_table p
     JOIN tree t ON t.id = p.parent_id
)
SELECT *
FROM tree

(note the swapped condition for the join in the recursive part of the statement)

To my knowledge the following DBMS support recursive CTEs:

  • FirebirdSQL 2.1 (actually the first OpenSource DBMS to implement them)
  • PostgreSQL 8.4
  • DB2 (not sure which exact version)
  • Oracle (since 11.2)
  • SQL Server 2005 and later
  • Teradata
  • H2
  • Sybase (don't know which exact version)

Edit

Based on your sample data, the following would retrieve all subtrees from the table including the complete path for each node as an additional column:

with recursive obs_tree as (
   select id, parent_id, '/'||cast(id as varchar) as tree
   from observations
   where parent_id is null

   union all 

   select t.id, t.parent_id, p.tree||'/'||cast(t.id as varchar)
   from observations t
      join obs_tree p on t.parent_id = p.id
)
select id, parent_id, tree
from obs_tree
order by tree

The output would be this:

 id | parent_id |  tree
----+-----------+---------
  3 |           | /3
  1 |         3 | /3/1
  6 |         1 | /3/1/6
 10 |         3 | /3/10
  4 |        10 | /3/10/4
  8 |        10 | /3/10/8
  9 |        10 | /3/10/9
  5 |           | /5
  2 |         5 | /5/2
  7 |         5 | /5/7
1
  • This is excellent! Regarding your last example where you retrieve the complete tree in the tree column, I assume there is no easy SQL way to split the tree into individual columns, basically dynamically adding columns depending on the tree depth? And a better approach would be just to split the tree in the app with python, java, etc?
    – ruslaniv
    Mar 23, 2022 at 5:47
8

The GetFamilyTree function in Rolando's answer does not work when the given id is more than 4 integer, because the FORMAT MySQL function adds commas for thousand separators. I have modified the stored function GetFamilyTree to work with big integer ids as below:

WHILE queue_length > 0 DO
    IF queue_length = 1 THEN
    SET front_id = queue;
        SET queue = '';
    ELSE
    SET front_id = SUBSTR(queue,1,LOCATE(',',queue)-1);
        SET pos = LOCATE(',',queue) + 1;
        SET q = SUBSTR(queue,pos);
        SET queue = q;
    END IF;

front_id defined inside if else loop.

0

Not the answer you're looking for? Browse other questions tagged or ask your own question.