Unexpected Estimated Rows by INNER JOIN
- Dominant language
- Go
- Stars
- 40.5k
- Forks
- 6.2k
- PR merge metrics
- PR metrics pending
Description
## Bug Report
### 1. Minimal reproduce step (Required)
```sql
CREATE TABLE t0(c0 CHAR);
CREATE TABLE t1(c0 INT);
INSERT INTO t0 VALUES ('a'), ('b'), ('c'), ('d'), ('e'), ('f'), ('g'), ('h'), ('i'), ('j');
INSERT INTO t1 VALUES (1), (NULL);
ANALYZE TABLE t0;
ANALYZE TABLE t1;
EXPLAIN SELECT * FROM t1 RIGHT JOIN t0 ON t1.c0!=t0.c0 WHERE t1.c0 OR t0.c0; -- estimated row count: 16
EXPLAIN SELECT * FROM t1 INNER JOIN t0 ON t1.c0!=t0.c0 WHERE t1.c0 OR t0.c0; -- estimated row count: 20
```
### 2. What did you expect to see? (Required)
I expect the second SELECT returns no more estimated rows than the first SELECT.
The query plan of the first SELECT:
```
id estRows task access object operator info
Selection_7 16.00 root or(database4.t1.c0, istrue_with_null(cast(database4.t0.c0, double BINARY)))
└─HashJoin_9 20.00 root CARTESIAN right outer join, other cond:ne(cast(database4.t1.c0, double BINARY), cast(database4.t0.c0, double BINARY))
├─TableReader_11(Build) 2.00 root data:TableFullScan_10
│ └─TableFullScan_10 2.00 cop[tikv] table:t1 keep order:false
└─TableReader_13(Probe) 10.00 root data:TableFullScan_12
└─TableFullScan_12 10.00 cop[tikv] table:t0 keep order:false
```
The query plan of the second SELECT:
```
id estRows task access object operator info
HashJoin_10 20.00 root CARTESIAN inner join, other cond:ne(cast(database4.t1.c0, double BINARY), cast(database4.t0.c0, double BINARY)), or(database4.t1.c0, istrue_with_null(cast(database4.t0.c0, double BINARY)))
├─TableReader_12(Build) 2.00 root data:TableFullScan_11
│ └─TableFullScan_11 2.00 cop[tikv] table:t1 keep order:false
└─TableReader_14(Probe) 10.00 root data:TableFullScan_13
└─TableFullScan_13 10.00 cop[tikv] table:t0 keep order:false
```
Both query plans have similar structure below HashJoin and exact the same number of estimated rows after the HashJoin. After that, applying the same `WHERE` clause incur different estimated rows.
### 3. What did you see instead (Required)
The second SELECT returns more estimated rows than the first SELECT.
### 4. What is your TiDB version? (Required)
```
| Release Version: v6.4.0-alpha-133-g6c55faf03
Edition: Community
Git Commit Hash: 6c55faf034e8c205ffd23126829c637fb8a47451
Git Branch: master
UTC Build Time: 2022-10-26 07:47:53
GoVersion: go1.19.1
Race Enabled: false
TiKV Min Version: 6.2.0-alpha
Check Table Before Drop: true
Store: unistore |
```
Contributor guide
Assessment
This issue has not been assessed yet.