MySQL riddle with subquery - mysql

Question

EDIT: looks like an index issue, update at the bottom of the question
I have the following query + subquery whose results I cannot explain. I am starting with this minimal input data set (the application here is capturing data change, and the PK is the id + the tx_id).
mysql> select * from tag_version;
+----+-------------------+------------+-------+----------------+
| id | name | article_id | tx_id | operation_type |
+----+-------------------+------------+-------+----------------+
| 1 | some tag | 1 | 1 | 0 |
| 1 | updated tag | 1 | 2 | 1 |
| 1 | updated again tag | 1 | 3 | 1 |
| 2 | other tag | 1 | 2 | 0 |
+----+-------------------+------------+-------+----------------+
4 rows in set (0.00 sec)
The subquery, standalone
SELECT max(f.tx_id) as max_tx_id, f.id
from tag_version f
WHERE f.tx_id <= 2
GROUP BY f.id
Result is
+-----------+----+
| max_tx_id | id |
+-----------+----+
| 2 | 1 |
| 2 | 2 |
+-----------+----+
2 rows in set (0.00 sec)
The query, where I manually inject the subquery results, notice how they are equal to the above
select t.*
from tag_version t
where t.article_id = 1
AND (t.tx_id, t.id) IN (
(2,1),
(2,2)
)
With expected results
+----+-------------+------------+-------+----------------+
| id | name | article_id | tx_id | operation_type |
+----+-------------+------------+-------+----------------+
| 1 | updated tag | 1 | 2 | 1 |
| 2 | other tag | 1 | 2 | 0 |
+----+-------------+------------+-------+----------------+
2 rows in set (0.00 sec)
And lastly, using the subquery in place of the tuples...
select t.*
from tag_version t
where t.article_id = 1
AND (t.tx_id, t.id) IN (
SELECT max(f.tx_id) as tx_id, f.id
from tag_version f
WHERE f.tx_id <= 2
GROUP BY f.id
)
The result is Empty set (0.00 sec)! Can someone explain this? I get the same empty results when I re-write the query using EXISTS instead of IN
I noticed than when I remove the line WHERE f.tx_id <= 2 from the subquery, I actually get results (although the wrong ones):
+----+-------------------+------------+-------+----------------+
| id | name | article_id | tx_id | operation_type |
+----+-------------------+------------+-------+----------------+
| 1 | updated again tag | 1 | 3 | 1 |
| 2 | other tag | 1 | 2 | 0 |
+----+-------------------+------------+-------+----------------+
2 rows in set (0.00 sec)
Replacing the subquery with a JOIN actually returns the expected correct results
SELECT t.*
FROM tag_version t
JOIN (
SELECT max(f.tx_id) as max_tx_id, f.id
from tag_version f
WHERE f.tx_id <= 2
GROUP BY f.id
) as max_ids
ON max_ids.max_tx_id = t.tx_id
AND max_ids.id = t.id
where t.article_id = 1
Result:
+----+-------------+------------+-------+----------------+
| id | name | article_id | tx_id | operation_type |
+----+-------------+------------+-------+----------------+
| 1 | updated tag | 1 | 2 | 1 |
| 2 | other tag | 1 | 2 | 0 |
+----+-------------+------------+-------+----------------+
2 rows in set (0.00 sec)
In addition, running the same query+subquery on the same data set with both PostgreSQL and SQLite gives the expected correct results.
My MySQL version is Server version: 5.5.40-0ubuntu0.14.04.1 (Ubuntu).
I think the clue to figuring out what's happening is that I actually get results when I remove the WHERE from the subquery, but I can't make something useful out of it.
EDIT: updated with input data set
EDIT: add table information
The table create statement is as follows
CREATE TABLE `tag_version` (
`id` int(11) NOT NULL,
`name` varchar(255) DEFAULT NULL,
`article_id` int(11) DEFAULT NULL,
`tx_id` bigint(20) NOT NULL,
`operation_type` smallint(6) NOT NULL,
PRIMARY KEY (`id`,`tx_id`),
KEY `ix_tag_version_operation_type` (`operation_type`),
KEY `ix_tag_version_tx_id` (`tx_id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
Data population as well
insert into tag_version (id, name, article_id, tx_id, operation_type) VALUES
(1, 'some tag', 1, 1, 0),
(1, 'updated tag', 1, 2, 1),
(1, 'updated again tag', 1, 3, 1),
(2, 'other tag', 1, 2, 0)
;
When I remove the ix_tag_version_tx_id index, the query returns the correct results... An explanation of why would be useful.

Answer 1

I believe you made a mistake in showing the result of first code(subquery).
The output of this query:
SELECT max(f.tx_id) as qwer, f.id
from tag_version f
WHERE f.tx_id <= 2
GROUP BY f.id
-is not:
+--------------+----+
| max(f.tx_id) | id |
+--------------+----+
| 2 | 1 |
| 2 | 2 |
+--------------+----+
It is:
+------+----+
| qwer | id |
+------+----+
| 2 | 1 |
| 2 | 2 |
+------+----+
(*Notice: the code line max(f.tx_id) as qwer *)
Now try this code for the expexted output.
There is a change when selecting max(f.tx_id).
select t.*
from tag_version t
where t.`article_id` = 1
AND t.`operation_type` != 2
AND (t.`tx_id`, t.`id`) IN (
SELECT max(f.`tx_id`) as `tx_id`, f.`id`
from tag_version f
WHERE f.`tx_id` <= 2
GROUP BY f.`id`
)
Let me know if this gives you the result or any other error.