Welcome to Software Development on Codidact!

Will you help us build our independent community of developers helping developers? We're small and trying to grow. We welcome questions about all aspects of software development, from design to code to QA and more. Got questions? Got answers? Got code you'd like someone to review? Please join us.

Post History

71%

+3 −0

Q&A How to break infinite loop in CTE

I have not managed to make an answer with standard SQL. I've had to resort to using arrays. And I don't have a Microsoft SQL Server RDBMS at hand. So I've done it instead in PostgreSQL. You'll just...

posted 3y ago by Estela‭ · edited 3y ago by Estela‭

Answer

#4: Post edited by

Estela‭ · 2021-04-19T17:45:18Z (over 3 years ago)

Copy Link

Raw

Markdown

~~> `ON rel.childId = sd.parentId`~~
~~I think you actually want~~
```
ON rel.childId = sd.id
```
As for the infinite loop. There is no such loop. INNER JOIN just does the cartessian product of both sub-queries and filters such product so that only the lines of the product which match the ON condition are present in the result.
So, with my adjustment, you'll get one line for each line in My_Table such that it has a child and there is a line for such child in My_table. In other words, one line for each line which has an existing child.
~~It does not matter whether the child has more children or even if the child of the child happens to be itself.~~

I have not managed to make an answer with standard SQL. I've had to resort to using arrays. And I don't have a Microsoft SQL Server RDBMS at hand. So I've done it instead in PostgreSQL. You'll just need to use whatever equivalent array functions sql-server has.
* ARRAY(SELECT XX) : constructs an array with a single element XX
* ARRAY_CAT( a1, a2) : concatenates two arrays
* ARRAY_POSITION(a, e) : returns the position of e in array a. NULL if e is not in array a.
This is half the answer: how to get the hierarchy for all descendants (children, grandchildren, etc).
For the full answer you just need to do the same for ascendants and then perform a union.
```
WITH RECURSIVE cte AS (
SELECT
childid,
parentid,
ARRAY(SELECT XX) a
FROM my_table
WHERE parentid=XX
UNION ALL
SELECT
m.childid,
m.parentid,
ARRAY_CAT( cte.a, ARRAY(SELECT m.parentid))
FROM
my_table m
INNER JOIN cte
ON cte.childid = m.parentid
WHERE ARRAY_POSITION(cte.a, m.parentid) is null
)
SELECT DISTINCT childid, parentid FROM cte;
```
Replace XX by the id for which you are querying its descendants.
This works by storing in an extra colum an array with all parent ids which have been processed. When doing the recursive CTE we only select rows with a parentid which is not contained in such array. This way we break the loop.
The code as such is hard to understand. I suggest to run the following code and check the result. Which will help in understanding the previous code.
```
WITH RECURSIVE cte AS (
SELECT
1 as iteration,
childid,
parentid,
ARRAY(SELECT XX) a
FROM my_table
WHERE parentid=XX
UNION ALL
SELECT
iteration+1,
m.childid,
m.parentid,
ARRAY_CAT( cte.a, ARRAY(SELECT m.parentid))
FROM
my_table m
INNER JOIN cte
ON cte.childid = m.parentid
WHERE ARRAY_POSITION(cte.a, m.parentid) is null
)
SELECT iteration, childid, parentid, a FROM cte ORDER BY iteration;
```
Which, for XX=1, returns the following table:
```
iteration | childid | parentid | a
--------------------------------------------
1 | 5 | 1 | {1}
2 | 3 | 5 | {1,5}
3 | 1 | 3 | {1,5,3}
```
Which are the descendants of 1, as expected. With their immediate parents and the array of checked parents as it grows.

#3: Post undeleted by

Estela‭ · 2021-04-19T17:44:44Z (over 3 years ago)

Copy Link

Raw

Markdown

#2: Post deleted by

Estela‭ · 2021-04-19T13:23:25Z (over 3 years ago)

Copy Link

Raw

Markdown

#1: Initial revision by

Estela‭ · 2021-04-19T13:02:56Z (over 3 years ago)

Copy Link

Raw

Markdown

> 	`ON rel.childId = sd.parentId`

I think you actually want

```
ON rel.childId = sd.id
```

As for the infinite loop. There is no such loop. INNER JOIN just does the cartessian product of both sub-queries and filters such product so that only the lines of the product which match the ON condition are present in the result.

So, with my adjustment, you'll get one line for each line in My_Table such that it has a child and there is a line for such child in My_table. In other words, one line for each line which has an existing child.

It does not matter whether the child has more children or even if the child of the child happens to be itself.

Communities

Post History