Skip to content

Commit

Permalink
adderss comments
Browse files Browse the repository at this point in the history 
Signed-off-by: ekexium <eke@fastmail.com>
  • Loading branch information
@ekexium
ekexium committed Sep 7, 2022
1 parent 55cb01f commit 565d3fb
Show file tree
Hide file tree
Showing 2 changed files with 68 additions and 121 deletions.
177 changes: 66 additions & 111 deletions constraints.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -11,8 +11,6 @@ TiDB 支持的约束与 MySQL 的基本相同。

TiDB 支持的非空约束规则与 MySQL 支持的一致。例如：

{{< copyable "sql" >}}

```sql
CREATE TABLE users (
id INT NOT NULL PRIMARY KEY AUTO_INCREMENT,
Expand All @@ -21,8 +19,6 @@ CREATE TABLE users (
);
```

{{< copyable "sql" >}}

```sql
INSERT INTO users (id,age,last_login) VALUES (NULL,123,NOW());
```
Expand All @@ -31,8 +27,6 @@ INSERT INTO users (id,age,last_login) VALUES (NULL,123,NOW());
Query OK, 1 row affected (0.02 sec)
```

{{< copyable "sql" >}}

```sql
INSERT INTO users (id,age,last_login) VALUES (NULL,NULL,NOW());
```
Expand All @@ -41,8 +35,6 @@ INSERT INTO users (id,age,last_login) VALUES (NULL,NULL,NOW());
ERROR 1048 (23000): Column 'age' cannot be null
```

{{< copyable "sql" >}}

```sql
INSERT INTO users (id,age,last_login) VALUES (NULL,123,NULL);
```
Expand All @@ -63,8 +55,6 @@ TiDB 会解析并忽略 `CHECK` 约束。该行为与 MySQL 5.7 的相兼容。

示例如下：

{{< copyable "sql" >}}

```sql
DROP TABLE IF EXISTS users;
CREATE TABLE users (
Expand All @@ -83,8 +73,6 @@ SELECT * FROM users;

在 TiDB 的乐观事务中，默认会对唯一约束进行[惰性检查](/transaction-overview.md#惰性检查)。通过在事务提交时再进行批量检查，TiDB 能够减少网络开销、提升性能。例如：

{{< copyable "sql" >}}

```sql
DROP TABLE IF EXISTS users;
CREATE TABLE users (
Expand All @@ -97,8 +85,6 @@ INSERT INTO users (username) VALUES ('dave'), ('sarah'), ('bill');

乐观事务模式下且 `tidb_constraint_check_in_place=0`：

{{< copyable "sql" >}}

```sql
BEGIN OPTIMISTIC;
INSERT INTO users (username) VALUES ('jane'), ('chris'), ('bill');
Expand All @@ -110,8 +96,6 @@ Query OK, 3 rows affected (0.00 sec)
Records: 3 Duplicates: 0 Warnings: 0
```

{{< copyable "sql" >}}

```sql
INSERT INTO users (username) VALUES ('steve'),('elizabeth');
```
Expand All @@ -121,8 +105,6 @@ Query OK, 2 rows affected (0.00 sec)
Records: 2 Duplicates: 0 Warnings: 0
```

{{< copyable "sql" >}}

```sql
COMMIT;
```
Expand All @@ -145,8 +127,6 @@ CREATE TABLE users (
INSERT INTO users (username) VALUES ('dave'), ('sarah'), ('bill');
```

{{< copyable "sql" >}}

```sql
SET tidb_constraint_check_in_place = 1;
```
Expand All @@ -155,8 +135,6 @@ SET tidb_constraint_check_in_place = 1;
Query OK, 0 rows affected (0.00 sec)
```

{{< copyable "sql" >}}

```sql
BEGIN OPTIMISTIC;
```
Expand All @@ -165,24 +143,20 @@ BEGIN OPTIMISTIC;
Query OK, 0 rows affected (0.00 sec)
```

{{< copyable "sql" >}}

```sql
INSERT INTO users (username) VALUES ('jane'), ('chris'), ('bill');
```

```
ERROR 1062 (23000): Duplicate entry 'bill' for key 'username'
..
```

第一条 `INSERT` 语句导致了重复键错误。这会造成额外的网络通信开销，并可能降低插入操作的吞吐量。

### 悲观事务

在 TiDB 的悲观事务中，默认在每一条语句加锁时对唯一约束进行检查：
在 TiDB 的悲观事务中，默认在每条语句执行时对该语句需要的锁进行加锁，并在加锁时进行该语句需要的唯一约束检查：

{{< copyable "sql" >}}

```sql
DROP TABLE IF EXISTS users;
Expand All @@ -193,7 +167,7 @@ CREATE TABLE users (
);
INSERT INTO users (username) VALUES ('dave'), ('sarah'), ('bill');

BEGIN;
BEGIN PESSIMISTIC;
INSERT INTO users (username) VALUES ('jane'), ('chris'), ('bill');
```

Expand All @@ -203,99 +177,94 @@ ERROR 1062 (23000): Duplicate entry 'bill' for key 'username'

悲观事务可以通过设置变量 [`tidb_constraint_check_in_place_pessimistic`](/system-variables.md#tidb_constraint_check_in_place_pessimistic) 为 `OFF` 来推迟唯一约束的检查到下一次对该唯一索引项加锁时或事务提交时，同时也跳过这个悲观锁加锁，以获得更好的性能。此时需要注意：

1. 由于唯一性约束被推迟检查，读取时可能读到不满足唯一性约束的结果，执行 `COMMIT` 语句时可能返回 `Duplicate entry` 错误。如果返回 `Duplicate entry` 错误，该事务最后一定会回滚。

下面这个例子跳过了对 bill 的加锁，因此可以读到不满足唯一性约束的结果：
- 由于唯一性约束被推迟检查，读取时可能读到不满足唯一性约束的结果，执行 `COMMIT` 语句时可能返回 `Duplicate entry` 错误。如果返回 `Duplicate entry` 错误，该事务最后一定会回滚。

{{< copyable "sql" >}}

```sql
SET tidb_constraint_check_in_place_pessimistic = OFF;
BEGIN PESSIMISTIC;
INSERT INTO users (username) VALUES ('jane'), ('chris'), ('bill'); -- Query OK, 3 rows affected
SELECT * FROM users FOR UPDATE;
```
下面这个例子跳过了对 bill 的加锁，因此可以读到不满足唯一性约束的结果：

读到了不满足唯一性约束的结果：有两个 bill。
```sql
SET tidb_constraint_check_in_place_pessimistic = OFF;
BEGIN PESSIMISTIC;
INSERT INTO users (username) VALUES ('jane'), ('chris'), ('bill'); -- Query OK, 3 rows affected
SELECT * FROM users FOR UPDATE;
```

```sql
+----+----------+
| id | username |
+----+----------+
| 1 | dave |
| 2 | sarah |
| 3 | bill |
| 7 | jane |
| 8 | chris |
| 9 | bill |
+----+----------+
```
读到了不满足唯一性约束的结果：有两个 bill。

```sql
+----+----------+
| id | username |
+----+----------+
| 1 | dave |
| 2 | sarah |
| 3 | bill |
| 7 | jane |
| 8 | chris |
| 9 | bill |
+----+----------+
```

```sql
COMMIT;
```
```sql
COMMIT;
```

```
ERROR 1062 (23000): Duplicate entry 'bill' for key 'username'
```
```
ERROR 1062 (23000): Duplicate entry 'bill' for key 'username'
```

2. 关闭该变量时，`COMMIT` 语句可能会返回 `Write conflict` 错误或 `Duplicate entry` 错误，两种错误都意味着事务回滚。
- 关闭该变量时，`COMMIT` 语句可能会返回 `Write conflict` 错误或 `Duplicate entry` 错误，两种错误都意味着事务回滚。

例如上面的例子，在提交时进行唯一约束检查，该事务报出了 `duplicate entry` 错误并回滚。
例如上面的例子，在提交时进行唯一约束检查，该事务报出了 `Duplicate entry` 错误并回滚。

在下面这个例子中，在有并发事务写入时，跳过悲观锁可能导致提交时报出 write conflict 错误并回滚。
在下面这个例子中，在有并发事务写入时，跳过悲观锁可能导致提交时报出 `Write conflict` 错误并回滚。

```sql
DROP TABLE IF EXISTS users;
CREATE TABLE users (
id INT NOT NULL PRIMARY KEY AUTO_INCREMENT,
username VARCHAR(60) NOT NULL,
UNIQUE KEY (username)
);
```sql
DROP TABLE IF EXISTS users;
CREATE TABLE users (
id INT NOT NULL PRIMARY KEY AUTO_INCREMENT,
username VARCHAR(60) NOT NULL,
UNIQUE KEY (username)
);

SET tidb_constraint_check_in_place_pessimistic = OFF;
BEGIN PESSIMISTIC;
INSERT INTO users (username) VALUES ('jane'), ('chris'), ('bill'); -- Query OK, 3 rows affected
```
SET tidb_constraint_check_in_place_pessimistic = OFF;
BEGIN PESSIMISTIC;
INSERT INTO users (username) VALUES ('jane'), ('chris'), ('bill'); -- Query OK, 3 rows affected
```

然后另一个会话中写入了 bill：
然后另一个会话中写入了 bill：

```sql
INSERT INTO users (username) VALUES ('bill'); -- Query OK, 1 row affected
```
```sql
INSERT INTO users (username) VALUES ('bill'); -- Query OK, 1 row affected
```

在第一个会话中提交，会报出 write conflict 错误。
在第一个会话中提交，会报出 `Write conflict` 错误。

```sql
COMMIT;
```
```sql
COMMIT;
```

```
ERROR 9007 (HY000): Write conflict, txnStartTS=435688780611190794, conflictStartTS=435688783311536129, conflictCommitTS=435688783311536130, key={tableID=74, indexID=1, indexValues={bill, }} primary={tableID=74, indexID=1, indexValues={bill, }} [try again later]
```
```
ERROR 9007 (HY000): Write conflict, txnStartTS=435688780611190794, conflictStartTS=435688783311536129, conflictCommitTS=435688783311536130, key={tableID=74, indexID=1, indexValues={bill, }} primary={tableID=74, indexID=1, indexValues={bill, }} [try again later]
```

3. 关闭该变量会导致悲观事务中可能报出一个新的错误类型 `8174: LazyUniquenessCheckFailure`，返回该错误时当前事务 abort，而不是像其它一些错误仅仅回滚报错的语句。
- 关闭该变量会导致悲观事务中可能报出错误 `8147: LazyUniquenessCheckFailure`。返回该错误时当前事务回滚，而不是像其它一些错误仅仅回滚报错的语句。例如一个普通 DML 语句报出 `1062: Duplicate entry` 错误时，只会回滚该 DML 语句，而不会回滚整个事务。

例如在跳过一次加锁后，在另一个语句中对该唯一索引加锁并检查，即会在该语句报错：
例如在跳过一次加锁后，在另一个语句中对该唯一索引加锁并检查，即会在该语句报错：

```sql
SET tidb_constraint_check_in_place_pessimistic = OFF;
BEGIN PESSIMISTIC;
INSERT INTO users (username) VALUES ('jane'), ('chris'), ('bill'); -- Query OK, 3 rows affected
DELETE FROM users where username = 'bill';
```
```sql
SET tidb_constraint_check_in_place_pessimistic = OFF;
BEGIN PESSIMISTIC;
INSERT INTO users (username) VALUES ('jane'), ('chris'), ('bill'); -- Query OK, 3 rows affected
DELETE FROM users where username = 'bill';
```

```
ERROR 8174 (23000): transaction aborted because lazy uniqueness check is enabled and an error occurred: [kv:1062]Duplicate entry 'bill' for key 'username'
```
```
ERROR 8147 (23000): transaction aborted because lazy uniqueness check is enabled and an error occurred: [kv:1062]Duplicate entry 'bill' for key 'username'
```

## 主键约束

与 MySQL 行为一样，主键约束包含了唯一约束，即创建了主键约束相当于拥有了唯一约束。此外，TiDB 其他的主键约束规则也与 MySQL 相似。例如：

{{< copyable "sql" >}}

```sql
CREATE TABLE t1 (a INT NOT NULL PRIMARY KEY);
```
Expand All @@ -304,8 +273,6 @@ CREATE TABLE t1 (a INT NOT NULL PRIMARY KEY);
Query OK, 0 rows affected (0.12 sec)
```
{{< copyable "sql" >}}

```sql
CREATE TABLE t2 (a INT NULL PRIMARY KEY);
```
Expand All @@ -314,8 +281,6 @@ CREATE TABLE t2 (a INT NULL PRIMARY KEY);
ERROR 1171 (42000): All parts of a PRIMARY KEY must be NOT NULL; if you need NULL in a key, use UNIQUE instead
```

{{< copyable "sql" >}}

```sql
CREATE TABLE t3 (a INT NOT NULL PRIMARY KEY, b INT NOT NULL PRIMARY KEY);
```
Expand All @@ -324,8 +289,6 @@ CREATE TABLE t3 (a INT NOT NULL PRIMARY KEY, b INT NOT NULL PRIMARY KEY);
ERROR 1068 (42000): Multiple primary key defined
```

{{< copyable "sql" >}}

```sql
CREATE TABLE t4 (a INT NOT NULL, b INT NOT NULL, PRIMARY KEY (a,b));
```
Expand All @@ -342,8 +305,6 @@ Query OK, 0 rows affected (0.10 sec)

除上述规则外，TiDB 目前仅支持对 `NONCLUSTERED` 的主键进行添加和删除操作。例如：

{{< copyable "sql" >}}

```sql
CREATE TABLE t5 (a INT NOT NULL, b INT NOT NULL, PRIMARY KEY (a,b) CLUSTERED);
ALTER TABLE t5 DROP PRIMARY KEY;
Expand All @@ -353,8 +314,6 @@ ALTER TABLE t5 DROP PRIMARY KEY;
ERROR 8200 (HY000): Unsupported drop primary key when the table is using clustered index
```

{{< copyable "sql" >}}

```sql
CREATE TABLE t5 (a INT NOT NULL, b INT NOT NULL, PRIMARY KEY (a,b) NONCLUSTERED);
ALTER TABLE t5 DROP PRIMARY KEY;
Expand Down Expand Up @@ -387,8 +346,6 @@ CREATE TABLE orders (
);
```

{{< copyable "sql" >}}

```sql
SELECT table_name, column_name, constraint_name, referenced_table_name, referenced_column_name
FROM information_schema.key_column_usage WHERE table_name IN ('users', 'orders');
Expand All @@ -407,8 +364,6 @@ FROM information_schema.key_column_usage WHERE table_name IN ('users', 'orders')

TiDB 也支持使用 `ALTER TABLE` 命令来删除外键 (`DROP FOREIGN KEY`) 和添加外键 (`ADD FOREIGN KEY`)：

{{< copyable "sql" >}}

```sql
ALTER TABLE orders DROP FOREIGN KEY fk_user_id;
ALTER TABLE orders ADD FOREIGN KEY fk_user_id (user_id) REFERENCES users(id);
Expand Down
Loading

0 comments on commit 565d3fb

Please sign in to comment.