feat: change tracking enabled table support replace into

[zhyass]

I hereby agree to the terms of the CLA available at: https://docs.databend.com/dev/policies/cla/

Summary

mysql> create table t(a int);
Query OK, 0 rows affected (0.12 sec)

mysql> insert into t values(1),(2);
Query OK, 2 rows affected (0.22 sec)

mysql> create stream s on table t append_only=false;
Query OK, 0 rows affected (0.13 sec)

mysql> replace into t on(a) values(2),(3);
Query OK, 3 rows affected (0.50 sec)

mysql> select * from s;
+------+---------------+----------------------------------------+------------------+
| a    | change$action | change$row_id                          | change$is_update |
+------+---------------+----------------------------------------+------------------+
|    2 | DELETE        | 84cabcdbe670450c9f0a260a225757ab000001 |                0 |
|    2 | INSERT        | 21550750d91b41298560db28862fc7e1000000 |                0 |
|    3 | INSERT        | 21550750d91b41298560db28862fc7e1000001 |                0 |
+------+---------------+----------------------------------------+------------------+
3 rows in set (0.25 sec)
Read 5 rows, 312.00 B in 0.072 sec., 69.29 rows/sec., 4.22 KiB/sec.

• Fixes #[Link the issue here]

Tests

• Unit Test
• Logic Test
• Benchmark Test
• No Test - Explain why

Type of change

• Bug Fix (non-breaking change which fixes an issue)
• New Feature (non-breaking change which adds functionality)
• Breaking Change (fix or feature that could cause existing functionality not to work as expected)
• Documentation Update
• Refactoring
• Performance Improvement
• Other (please describe):

---

This change is 

[dantengsky]

for append only stream, the "semantic" appears to be inconsistent：

Given table t, insert row r, and then create a stream s based on t:

• after updating row r, s captures nothing
• but, if replacing into row r, one row is captured

mysql >  --- after replace into, one row captured
mysql> create table t (id int, v string);
Query OK, 0 rows affected (1.48 sec)

mysql> insert into t values(1, 'a');
Query OK, 1 row affected (0.25 sec)

mysql> create stream s on table t append_only = true;
Query OK, 0 rows affected (0.35 sec)

mysql> replace into t on(id) values(1, 'b');
Query OK, 2 rows affected (0.51 sec)

mysql> select * from s;
+------+------+---------------+------------------+----------------------------------------+
| id   | v    | change$action | change$is_update | change$row_id                          |
+------+------+---------------+------------------+----------------------------------------+
|    1 | b    | INSERT        |                0 | fcaad8936ed043a6b0c5e02873e91d48000000 |
+------+------+---------------+------------------+----------------------------------------+
1 row in set (0.37 sec)
Read 1 rows, 215.00 B in 0.177 sec., 5.65 rows/sec., 1.19 KiB/sec.

mysql> --- after updating, no rows captured

mysql> create table t1 (id int, v string);
Query OK, 0 rows affected (0.09 sec)

mysql> insert into t1 values(1, 'a');
Query OK, 1 row affected (0.11 sec)

mysql> create stream s1 on table t1 append_only = true;
Query OK, 0 rows affected (0.11 sec)

mysql> update  t1 set v = 'b' where id = 1;
Query OK, 1 row affected (0.56 sec)

mysql> select * from s1;
Empty set (0.27 sec)
Read 1 rows, 58.00 B in 0.061 sec., 16.49 rows/sec., 956.18 B/sec.

mysql> select version();
+----------------------------------------------------------------------------------------+
| version()                                                                              |
+----------------------------------------------------------------------------------------+
| 8.0.26-v1.2.360-nightly-096a909d91(rust-1.77.0-nightly-2024-03-05T08:56:26.362694196Z) |
+----------------------------------------------------------------------------------------+
1 row in set (0.01 sec)
Read 1 rows, 1.00 B in 0.004 sec., 256.96 rows/sec., 256.96 B/sec.

is this intended behavior?

[zhyass]

captured

Yes, it's intended behavior. Here treat the semantics of replace as INSERT and DELETE, not INSERT and UPDATE. Take the following sql as an example.

mysql> create table t(a int, b string);
Query OK, 0 rows affected (0.03 sec)

mysql> insert into t values(1,'a'),(1,'b'),(2,'b');
Query OK, 3 rows affected (0.04 sec)

mysql> replace into t on(a) values(1,'a1'),(3,'c');
Query OK, 4 rows affected (0.11 sec)

mysql> select * from t order by a;
+------+------+
| a    | b    |
+------+------+
|    1 | a1   |
|    2 | b    |
|    3 | c    |
+------+------+
3 rows in set (0.04 sec)
Read 3 rows, 50.00 B in 0.020 sec., 146.61 rows/sec., 2.39 KiB/sec.

We can make a note in the documentation to remind users of this distinction.

[dantengsky]

treat the semantics of replace as INSERT and DELETE, not INSERT and UPDATE
....
We can make a note in the documentation to remind users of this distinction.

looks reasonable to me, let's roll out.