Deleting records with selection on `_id` is broken

CrateDB version

4.7.0 / 5.0.1 / 5.10

CrateDB setup information

single / multi-node

Steps to Reproduce

cr> CREATE TABLE del_test (val int);                                                                                                                                                                          
CREATE OK, 1 row affected  (0.415 sec)

cr> INSERT INTO del_test VALUES  (1),(1),(2);                                                                                                                                                                 
INSERT OK, 3 rows affected  (0.032 sec)

cr> SELECT arbitrary(_id)  
    FROM del_test  
    WHERE val = 1  
    GROUP BY val  
    HAVING count(*) > 1;                                                                                                                                                                                      
+----------------------+
| arbitrary(_id)       |
+----------------------+
| 6wCcQoQBNMIiChzeUweD |
+----------------------+
SELECT 1 row in set (0.019 sec)

-- here is the problem:
cr> DELETE FROM del_test WHERE _id IN (SELECT arbitrary(_id)  
    FROM del_test  
    WHERE val = 1  
    GROUP BY val  
    HAVING count(*) > 1);                                                                                                                                                                                     
DELETE OK, 0 rows affected  (0.022 sec)

cr> DELETE FROM del_test WHERE _id IN (SELECT '6wCcQoQBNMIiChzeUweD');                                                                                                                                        
DELETE OK, 0 rows affected  (0.029 sec)

-- that works
cr> DELETE FROM del_test WHERE _id IN ('6wCcQoQBNMIiChzeUweD');                                                                                                                                               
DELETE OK, 1 row affected  (0.031 sec)

Expected Result

DELETE OK, 1 row affected  (0.031 sec)

Actual Result

DELETE OK, 0 rows affected  (0.029 sec)

---

edit:

EXPLAIN
SELECT *
FROM del_test
WHERE _id IN (SELECT _id FROM del_test)
-- SELECT OK, 0 records returned (0.033 seconds)

MultiPhase
  └ Collect[doc.del_test | [val] | (_id = ANY((SELECT _id FROM (doc.del_test))))]
  └ OrderBy[_id ASC]
    └ Collect[doc.del_test | [_id] | true]

------------

EXPLAIN
SELECT *
FROM del_test
WHERE _id::CHAR(20) IN (SELECT _id FROM del_test)
-- SELECT OK, 3 records returned (0.041 seconds)

MultiPhase
  └ Collect[doc.del_test | [val] | (_cast(cast(_id AS character), 'text') = ANY((SELECT _id FROM (doc.del_test))))]
  └ OrderBy[_id ASC]
    └ Collect[doc.del_test | [_id] | true]

------------
EXPLAIN
SELECT *
FROM del_test
WHERE _id IN ('58puQ4QBQmJ97UbuHnXp','jxcRQ4QB4KuX8tRunOAg')

Get[doc.del_test | val | DocKeys{'58puQ4QBQmJ97UbuHnXp'; 'jxcRQ4QB4KuX8tRunOAg'} | (_id = ANY(['58puQ4QBQmJ97UbuHnXp', 'jxcRQ4QB4KuX8tRunOAg']))]

---

Seems like the bug first occurred with CrateDB 4.7 (4.6.8 tested to be working)

[matriv]

FYI: That’s not a problem of the DELETE but a problem with _id IN/ANY and subquery.

[hlcianfagna]

Hi, I am the original reporter on this issue, today I tried to add casting to the _id field as suggested. I am on version 4.8.4 and unfortunately the cast does not work here, I get the message The 'char' type doesn't support type parameters when I add ::CHAR(20).

Just logging that on a separate conversation it has been confirmed that on versions of CrateDB < 5.0.0 the CHAR data type is not available, but the issue can still be worked around using VARCHAR instead.