Logical bug: value changed after adding DISTINCT [v5.3.4, v6.4.0]
- Dominant language
- Go
- Stars
- 40.5k
- Forks
- 6.2k
- PR merge metrics
- PR metrics pending
Description
## Bug Report
### 1. Minimal reproduce step (Required)
```sql
drop table if exists t;
CREATE TABLE t (c1 FLOAT UNSIGNED);
INSERT INTO t VALUES (47),(28.1237);
SELECT (DATE_ADD(f2, INTERVAL 1 YEAR_MONTH)) AS f1 FROM (SELECT 1 FROM t) AS t1 JOIN (SELECT (DATE_ADD(BIN(c1), INTERVAL 1 YEAR)) AS f2 FROM t) AS t2; -- sql1
SELECT (DATE_ADD(f2, INTERVAL 1 YEAR_MONTH)) AS f1 FROM (SELECT DISTINCT 1 FROM t) AS t1 JOIN (SELECT (DATE_ADD(BIN(c1), INTERVAL 1 YEAR)) AS f2 FROM t) AS t2; -- sql2
```
### 2. What did you expect to see? (Required)
In theory, the result of sql2(DISTINCT) ⊆ the result of sql1
### 3. What did you see instead (Required)
However, the value `NULL` changed to ` 2011-12-11` after adding DISTINCT, seems like a logical bug:
```sql
mysql> select version();
+--------------------+
| version() |
+--------------------+
| 5.7.25-TiDB-v6.4.0 |
+--------------------+
1 row in set (0.01 sec)
mysql> SELECT (DATE_ADD(f2, INTERVAL 1 YEAR_MONTH)) AS f1 FROM (SELECT 1 FROM t) AS t1 JOIN (SELECT (DATE_ADD(BIN(c1), INTERVAL 1 YEAR)) AS f2 FROM t) AS t2; -- sql1
+------+
| f1 |
+------+
| NULL |
| NULL |
| NULL |
| NULL |
+------+
4 rows in set, 11 warnings (0.00 sec)
mysql> SELECT (DATE_ADD(f2, INTERVAL 1 YEAR_MONTH)) AS f1 FROM (SELECT DISTINCT 1 FROM t) AS t1 JOIN (SELECT (DATE_ADD(BIN(c1), INTERVAL 1 YEAR)) AS f2 FROM t) AS t2; -- sql2
+------------+
| f1 |
+------------+
| 2011-12-11 |
| NULL |
+------------+
2 rows in set, 5 warnings (0.00 sec)
```
### 4. What is your TiDB version? (Required)
```sql
Release Version: v6.4.0
Edition: Community
Git Commit Hash: cf36a9ce2fe1039db3cf3444d51930b887df18a1
Git Branch: heads/refs/tags/v6.4.0
UTC Build Time: 2022-11-13 05:25:30
GoVersion: go1.19.2
Race Enabled: false
TiKV Min Version: 6.2.0-alpha
Check Table Before Drop: false
Store: unistore
```
### 5. Hope these can be helpful for your debugging:
1. The bug cannot be reproduced after removing BIN
2. We look for the first version of the bug in all docker images (with format x.x.x) of https://hub.docker.com/r/pingcap/tidb/tags
We found that the bug first occurred in pingcap/tidb:v5.3.4, it cannot be reproduced in pingcap/tidb:v5.3.3:
```sql
mysql> select version();
+--------------------+
| version() |
+--------------------+
| 5.7.25-TiDB-v5.3.4 |
+--------------------+
1 row in set (0.00 sec)
mysql> SELECT (DATE_ADD(f2, INTERVAL 1 YEAR_MONTH)) AS f1 FROM (SELECT 1 FROM t) AS t1 JOIN (SELECT (DATE_ADD(BIN(c1), INTERVAL 1 YEAR)) AS f2 FROM t) AS t2; -- sql1
+------------+
| f1 |
+------------+
| NULL |
| NULL |
| NULL |
| NULL |
+------------+
4 rows in set, 11 warnings (0.00 sec)
mysql> SELECT (DATE_ADD(f2, INTERVAL 1 YEAR_MONTH)) AS f1 FROM (SELECT DISTINCT 1 FROM t) AS t1 JOIN (SELECT (DATE_ADD(BIN(c1), INTERVAL 1 YEAR)) AS f2 FROM t) AS t2; -- sql2
+------------------------+
| f1 |
+------------------------+
| 0x323031312D31322D3131 |
| NULL |
+------------------------+
2 rows in set, 5 warnings (0.00 sec)
mysql> select version();
+--------------------+
| version() |
+--------------------+
| 5.7.25-TiDB-v5.3.3 |
+--------------------+
1 row in set (0.00 sec)
mysql> SELECT (DATE_ADD(f2, INTERVAL 1 YEAR_MONTH)) AS f1 FROM (SELECT 1 FROM t) AS t1 JOIN (SELECT (DATE_ADD(BIN(c1), INTERVAL 1 YEAR)) AS f2 FROM t) AS t2; -- sql1
+------------+
| f1 |
+------------+
| NULL |
| 2011-12-11 |
| NULL |
| 2011-12-11 |
+------------+
4 rows in set, 6 warnings (0.00 sec)
mysql> SELECT (DATE_ADD(f2, INTERVAL 1 YEAR_MONTH)) AS f1 FROM (SELECT DISTINCT 1 FROM t) AS t1 JOIN (SELECT (DATE_ADD(BIN(c1), INTERVAL 1 YEAR)) AS f2 FROM t) AS t2; -- sql2
+------------+
| f1 |
+------------+
| 2011-12-11 |
| NULL |
+------------+
2 rows in set, 4 warnings (0.00 sec)
```
Contributor guide
Research direction
Start by running the supplied SQL reproducer against the affected and unaffected TiDB versions, focusing on the interaction between DISTINCT, BIN, and DATE_ADD. Trace the query execution and type or value handling involved, then add a regression test showing that adding DISTINCT does not change the corresponding results.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- go, sql
- Domain
- databases
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 30/100