资讯详情

ClickHouse 数据插入、更新与删除操作 SQL

f5a290eafca4532cc7ba12bc985f2166.png

1.1.

数据操作语言( DML，Data Manipulation Language) 用于在数据库表中添加(插入)、删除和修改(更新)数据。本节主要介绍ClickHouse插入、更新和删除数据。

1.1.1.

数据操作语言DML，包括SQL修改存储的数据，但不修改数据模型，如数据库模式或数据库表结构。DML常见的语法模式如下：

INSERT INTO ... VALUES ... UPDATE ... SET ... WHERE ... DELETE FROM ... WHERE ...

但是，在 ClickHouse 中，UPDATE 与DELETE 是设计在了 ALTER 在指令系统中。

1.1.2.

INSERT INTO clickhouse_tutorial.user_tag (user_id, gender, age, active_level, date) VALUES (1, 'male', '18', '1', '2022-03-21'); INSERT INTO clickhouse_tutorial.user_tag (user_id, gender, age, active_level, date) VALUES (2, 'female', '16', '2', '2022-03-21');

INSERT INTO clickhouse_tutorial.user_tag (user_id, gender, age, active_level, date) VALUES (3, 'female', '20', '3', '2022-03-21'),        (4, 'female', '22', '4', '2022-03-21');

INSERT INTO clickhouse_tutorial.user_tag (UserID, WatchID, EventTime, Sex, Age, OS, RegionID, RequestNum, EventDate) SELECT     UserID,     WatchID,     EventTime,     Sex,     Age,     OS,     RegionID,     RequestNum,     EventDate FROM tutorial.hits_v1   Query id: bfed9d12-b838-4125-9ee2-f61049bf0a56   ↙ Progress: 30.91 million rows, 835.62 MB (6.66 million rows/s., 180.00 MB/s.) (0.0 CPU, 172.08 MB RAM) ████████████████████████████████████████████████ 99%   Ok.   0 rows in set. Elapsed: 7.552 sec. Processed 53.24 million rows, 1.37 GB (7.05 million rows/s., 180.96 MB/s.)

1.1.3.

ALTER TABLE [db.]table UPDATE column1 = expr1 [, ...] WHERE filter_expr

更新表数据。ClickHouse 中ALTER TABLE 前缀语法和大多数其他支持 SQL 不同的数据库系统。它旨在表明与 OLTP 数据库中的类似查询不同，这是一项并非为频繁使用而设计的繁重操作。ALTER 查询是通过一种叫做突变的方式进行的（Mutation）实现了机制。

关于 ALTERTABLE ... UPDATE详细说明如下：

1.WHERE过滤表达式在句子中filter_expr的值是UInt8类型，指定要更新的数据行。

2.用于计算主键或分区键的列不支持更新。

3.一个UPDATE操作可以包含多个用逗号分隔的命令，例如column1 = expr1,column2= expr2。 UPDATE由系统配置的操作数据处理是同步还是异步 mutation_sync 可取值为：

0 - execute asynchronously。

1 - wait current server。

2 - wait all replicas if they exist。默认为0，异步后台流程执行类似 *MergeTree 表中的合并操作。

4.对于 *MergeTree 表，Mutation重写整个数据Part来执行，Mutation没有原子性。数据 Part一旦准备好了MutationPart而且在Mutation执行期间，SELECT从查询结果可以看出，已经变异Part来自未变异的数据Part的数据。

5.Mutation按创建顺序排序，并按每个顺序应用MutationPart。

6.在Mutation提交前插入表中的数据将被执行Mutation操作，提交后插入的数据不执行Mutation操作。

7.Mutation操作不会阻塞数据插入。

8.可以查看 system.mutations 跟踪突变进度。

9.即使重新开始 ClickHouse 成功提交的服务器变更将继续执行。一旦提交，就不能回滚突变。

10.如果Mutation由于某种原因被卡住，可以使用 KILL MUTATION 取消查询。

11.已经完成Mutation不会立即删除条目。保留条目的数量由保留条目 finished_mutations_to_keep 存储引擎参数确定。

12.在系统配置表中system.settings中，有关mutation配置如下：

SELECT * FROM system.settings WHERE name LIKE '%mutation%' FORMAT Vertical   Query id: 24f6ca70-7117-41c5-bc3e-dd6615d5ee6d   Row 1: ────── name:        background_merges_mutations_concurrency_ratio value:       2 changed:     0 description: Ratio between a number of how many operations could be processed and a number threads to process them. Only has meaning at server startup. min:         ????? max:         ???? readonly:    0 type:        Float   Row 2: ────── name:        mutations_sync value:       0 changed:     0 description: Wait for synchronous execution of ALTER TABLE UPDATE/DELETE queries (mutations). 0 - execute asynchronously. 1 - wait current server. 2 - wait all replicas if they exist. min:         ????? max:         ???? readonly:    0 type:        UInt64   Row 3: ────── name:        allow_nondeterministic_mutations value:       0 changed:     0 description: Allow non-deterministic functions in ALTER UPDATE/ALTER DELETE statements min:         ????? max:         ???? readonly:    0 type:        Bool   3 rows in set. Elapsed: 0003 sec.

SELECT
    WatchID,
    JavaEnable,
    GoodEvent
FROM tutorial.hits_v1
WHERE WatchID = 7043438415214026105
 
Query id: e0dc9ae5-8f24-48e5-a56d-d107afa1dfe3
 
┌─────────────WatchID─┬─JavaEnable─┬─GoodEvent─┐
│ 7043438415214026105 │          1 │         1 │
└─────────────────────┴────────────┴───────────┘
ALTER TABLE tutorial.hits_v1
    UPDATE JavaEnable = 0, GoodEvent = 0 WHERE WatchID = 7043438415214026105
 
Query id: 32ac8c6b-c78c-4a3c-ab72-29ed38fda687
 
Ok.
SELECT
    WatchID,
    JavaEnable,
    GoodEvent
FROM tutorial.hits_v1
WHERE WatchID = 7043438415214026105
 
Query id: 26591216-0cb4-4c6d-9b89-db9ff588d469
 
┌─────────────WatchID─┬─JavaEnable─┬─GoodEvent─┐
│ 7043438415214026105 │          0 │         0 │
└─────────────────────┴────────────┴───────────┘

可以看到，目标数据已经被更新。

我们可以去服务器端查看 UPDATE 操作提交之后的日志：

2022.03.31 03:06:47.898420 [ 6154029 ] {} <Debug> TCP-Session: fd75399f-85bc-4d6d-a86b-d69fa899f6d9 Creating query context from session context, user_id: 94309d50-4f52-5250-31bd-74fecac179db, parent context user: default

2022.03.31 03:06:47.898553 [ 6154029 ] {32ac8c6b-c78c-4a3c-ab72-29ed38fda687} <Debug> executeQuery: (from 127.0.0.1:52757) ALTER TABLE tutorial.hits_v1 UPDATE JavaEnable = 0, GoodEvent=0 WHERE WatchID=7043438415214026105;

2022.03.31 03:06:47.898600 [ 6154029 ] {32ac8c6b-c78c-4a3c-ab72-29ed38fda687} <Trace> ContextAccess (default): Access granted: ALTER UPDATE(JavaEnable, GoodEvent) ON tutorial.hits_v1

2022.03.31 03:06:47.900244 [ 6154029 ] {32ac8c6b-c78c-4a3c-ab72-29ed38fda687} <Information> tutorial.hits_v1 (0fa45bfe-c9ca-4df7-b7bf-7bd268a6225d):

2022.03.31 03:06:47.900311 [ 6154029 ] {32ac8c6b-c78c-4a3c-ab72-29ed38fda687} <Debug> MemoryTracker: Peak memory usage (for query): 0.00 B.

2022.03.31 03:06:47.900365 [ 6154029 ] {} <Debug> TCPHandler: Processed in 0.00198 sec.

可以看到“Added mutation: mutation_33.txt”这样一行关键日志。去系统表system.mutations中查看mutation_33.txt的详情如下：

SELECT *
FROM system.mutations
WHERE mutation_id = 'mutation_33.txt'
FORMAT Vertical
 
Query id: 3f7c995b-82e3-41fa-bf5b-fc50be8da824
 
Row 1:
──────
database:                   tutorial
table:                      hits_v1
mutation_id:                mutation_33.txt
command:                    UPDATE JavaEnable = 0, GoodEvent = 0 WHERE WatchID = 7043438415214026105
create_time:                2022-03-31 03:06:47
block_numbers.partition_id: ['']
block_numbers.number:       [33]
parts_to_do_names:          []
parts_to_do:                0
is_done:                    1
latest_failed_part:         
latest_fail_time:           1970-01-01 08:00:00
latest_fail_reason:         
 
1 rows in set. Elapsed: 0.003 sec.

1.1.4. 

ALTER TABLE [db.]table [ON CLUSTER cluster] DELETE WHERE filter_expr

删除表数据。

SELECT
    WatchID,
    JavaEnable,
    GoodEvent
FROM tutorial.hits_v1
WHERE WatchID = 7043438415214026105
 
Query id: 1444174c-142b-43ec-8ad6-54da1d871277
 
┌─────────────WatchID─┬─JavaEnable─┬─GoodEvent─┐
│ 7043438415214026105 │          0 │         0 │
└─────────────────────┴────────────┴───────────┘
ALTER TABLE tutorial.hits_v1
    DELETE WHERE WatchID = 7043438415214026105
SELECT
    WatchID,
    JavaEnable,
    GoodEvent
FROM tutorial.hits_v1
WHERE WatchID = 7043438415214026105
 
Query id: fd6a7536-4f2e-4fe3-8b03-bf1aed45302f
 
Ok.
 
0 rows in set. Elapsed: 0.018 sec. Processed 8.87 million rows, 70.99 MB (480.03 million rows/s., 3.84 GB/s.)

根据 query_id: a303b0d4-564d-48e5-9f32-c7e2df554b1f 查询ClickHouse Server 端日志如下：

2022.03.31 03:15:10.905286 [ 6154029 ] {} <Debug> TCP-Session: fd75399f-85bc-4d6d-a86b-d69fa899f6d9 Creating query context from session context, user_id: 94309d50-4f52-5250-31bd-74fecac179db, parent context user: default

2022.03.31 03:15:10.905429 [ 6154029 ] {a303b0d4-564d-48e5-9f32-c7e2df554b1f} <Debug> executeQuery: (from 127.0.0.1:52757) ALTER TABLE tutorial.hits_v1 DELETE WHERE WatchID=7043438415214026105;

2022.03.31 03:15:10.905482 [ 6154029 ] {a303b0d4-564d-48e5-9f32-c7e2df554b1f} <Trace> ContextAccess (default): Access granted: ALTER DELETE ON tutorial.hits_v1

2022.03.31 03:15:10.906794 [ 6154029 ] {a303b0d4-564d-48e5-9f32-c7e2df554b1f} <Debug>

2022.03.31 03:15:10.911000 [ 6154029 ] {a303b0d4-564d-48e5-9f32-c7e2df554b1f} <Information> tutorial.hits_v1 (0fa45bfe-c9ca-4df7-b7bf-7bd268a6225d):

2022.03.31 03:15:10.911081 [ 6154029 ] {a303b0d4-564d-48e5-9f32-c7e2df554b1f} <Debug> MemoryTracker: Peak memory usage (for query): 0.00 B.

2022.03.31 03:15:10.911139 [ 6154029 ] {} <Debug> TCPHandler: Processed in 0.005891 sec.

可以看到关键日志：“Added mutation: mutation_35.txt”。

另外，我们还看到了WHERE过滤自动转为PREWHERE优化的日志：“InterpreterSelectQuery: MergeTreeWhereOptimizer: condition "isZeroOrNull(WatchID = 7043438415214026105)" moved to PREWHERE”。

根据日志内容“Added mutation: mutation_35.txt”，查询 mutation_35.txt详情如下：

SELECT *
FROM system.mutations
WHERE mutation_id = 'mutation_35.txt'
FORMAT Vertical
 
Query id: c7e42d42-feef-4a18-a46f-ea97aa9d7b7e
 
Row 1:
──────
database:                   tutorial
table:                      hits_v1
mutation_id:                mutation_35.txt
command:                    DELETE WHERE WatchID = 7043438415214026105
create_time:                2022-03-31 03:15:10
block_numbers.partition_id: ['']
block_numbers.number:       [35]
parts_to_do_names:          []
parts_to_do:                0
is_done:                    1
latest_failed_part:         
latest_fail_time:           1970-01-01 08:00:00
latest_fail_reason:         
 
1 rows in set. Elapsed: 0.003 sec.

1.1.5. 

EXCHANGE TABLES|DICTIONARIES [db0.]name_A AND [db1.]name_B
EXCHANGE DICTIONARIES [db0.]dict_A AND [db1.]dict_B

1. EXCHANGE 操作以原子操作的方式交换两个表或字典的名称。

2. EXCHANGE 操作也可以通过使用RENAME 操作来完成，区别是RENAME不是原子操作。 例如，RENAME重命名交换两张表new_table、old_table：

RENAME TABLE new_table TO tmp, old_table TO new_table, tmp TO old_table;

直接使用EXCHANGE 命令实现如下：

EXCHANGE TABLES new_table AND old_table;

3. EXCHANGE 底层是通过系统调用renameat2() 实现，Linux Kernel 3.15+版本才支持。

4. 只有Atomic 数据库引擎支持EXCHANGE操作。

在Atomic 数据库引擎下创建的数据表，支持无锁原子CREATE/DROP/RENAME 操作，并且支持原子EXCHANGE TABLES A and B 直接交换两张表。

EXCHANGE 命令可以实现 AB 两张表的快速切换。AB 表切换的使用场景很广泛，比如历史表归档、批量抽数、数据同步过程写临时表等等，都可以采用 AB 表切换的思路来实现。

drop table if exists tutorial.hits_v2;
drop table if exists tutorial.hits_v3;
CREATE TABLE tutorial.hits_v2
(
    `WatchID` UInt64,
    `UserID` UInt64,
    `JavaEnable` UInt8,
    `Title` String,
    `GoodEvent` Int16,
    `EventTime` DateTime,
    `EventDate` Date,
    `RequestNum` UInt32,
    `RequestTry` UInt8
)
    ENGINE = MergeTree()
        PARTITION BY toYYYYMM(EventDate)
        ORDER BY (WatchID, EventDate, intHash32(UserID))
        SAMPLE BY intHash32(UserID);
 
CREATE TABLE tutorial.hits_v3
(
    `WatchID` UInt64,
    `UserID` UInt64,
    `JavaEnable` UInt8,
    `Title` String,
    `GoodEvent` Int16,
    `EventTime` DateTime,
    `EventDate` Date,
    `RequestNum` UInt32,
    `RequestTry` UInt8
)
    ENGINE = MergeTree()
        PARTITION BY toYYYYMM(EventDate)
        ORDER BY (WatchID, EventDate, intHash32(UserID))
        SAMPLE BY intHash32(UserID);
EXCHANGE TABLES tutorial.hits_v1 AND tutorial.hits_v2;

上面的命令，在笔者的 Mac OS电脑上是报错的：

Received exception from server (version 22.4.1):
Code: 48. DB::Exception: Received from 127.0.0.1:9009. DB::Exception: RENAME EXCHANGE is not supported. (NOT_IMPLEMENTED)

查看内核版本：

$uname -a
Darwin C02FJ0KMMD6V 20.3.0 Darwin Kernel Version 20.3.0: Thu Jan 21 00:07:06 PST 2021; root:xnu-7195.81.3~1/RELEASE_X86_64 x86_64

可见， Mac OS Darwin Kernel Version 20.3.0内核版本，还不支持renameat2函数系统调用。

Ø 

ClickHouse源代码实现在renameat2.cpp中，相关代码行如下：

#if !defined(__NR_renameat2)
    #if defined(__x86_64__)
        #define __NR_renameat2 316
    #elif defined(__aarch64__)
        #define __NR_renameat2 276
    #elif defined(__ppc64__)
        #define __NR_renameat2 357
    #elif defined(__riscv)
        #define __NR_renameat2 276
    #else
        #error "Unsupported architecture"
    #endif
#endif
...
static bool renameat2(const std::string & old_path, const std::string & new_path, int flags)
{
    if (!supportsRenameat2())
        return false;
    if (old_path.empty() || new_path.empty())
        throw Exception(ErrorCodes::LOGICAL_ERROR, "Cannot rename {} to {}: path is empty", old_path, new_path);
 
    /// int olddirfd (ignored for absolute oldpath), const char *oldpath,
    /// int newdirfd (ignored for absolute newpath), const char *newpath,
    /// unsigned int flags
    if (0 == syscall(__NR_renameat2, AT_FDCWD, old_path.c_str(), AT_FDCWD, new_path.c_str(), flags))
        return true;
 
    /// EINVAL means that filesystem does not support one of the flags.
    /// It also may happen when running clickhouse in docker with Mac OS as a host OS.
    /// supportsRenameat2() with uname is not enough in this case, because virtualized Linux kernel is used.
    /// Other cases when EINVAL can be returned should never happen.
    if (errno == EINVAL)
        return false;
    /// We should never get ENOSYS on Linux, because we check kernel version in supportsRenameat2Impl().
    /// However, we can get in on WSL.
    if (errno == ENOSYS)
        return false;
 
    if (errno == EEXIST)
        throwFromErrno(fmt::format("Cannot rename {} to {} because the second path already exists", old_path, new_path), ErrorCodes::ATOMIC_RENAME_FAIL);
    if (errno == ENOENT)
        throwFromErrno(fmt::format("Paths cannot be exchanged because {} or {} does not exist", old_path, new_path), ErrorCodes::ATOMIC_RENAME_FAIL);
    throwFromErrnoWithPath(fmt::format("Cannot rename {} to {}", old_path, new_path), new_path, ErrorCodes::SYSTEM_ERROR);
}
...
bool supportsRenameat2()
{
    static bool supports = supportsRenameat2Impl();
    return supports;
}
...
static bool supportsRenameat2Impl()
{
    VersionNumber renameat2_minimal_version(3, 15, 0); // since linux kernel 3.15
    VersionNumber linux_version(Poco::Environment::osVersion());
    return linux_version >= renameat2_minimal_version;
}

1.1.6. 

OPTIMIZE TABLE [db.]name [ON CLUSTER cluster]
[PARTITION partition | PARTITION ID 'partition_id'] [FINAL]
[DEDUPLICATE [BY expression]]

1. OPTIMIZE操作尝试为数据库表db.table ，初始化一个调度计划外的数据part合并操作。

2. OPTIMIZE操作仅支持MergeTree系列表引擎、MaterializedView和 Buffer表引擎。

3. 当 OPTIMIZE 与 ReplicatedMergeTree 系列表引擎一起使用时，ClickHouse 创建一个用于合并的任务，并等待在所有副本上执行（replication_alter_partitions_sync=2）或者等待当前副本上执行（replication_alter_partitions_sync=1）。

4. OPTIMIZE无法修复“Too many parts”错误。

Ø 

系统配置项replication_alter_partitions_sync，用来指定等待副本分区变更操作（ALTER、OPTIMIZE 或 TRUNCATE等）执行的sync策略。

replication_alter_partitions_sync可取值：

0 — 不等待，直接异步执行。

1 — 同步等待在当前副本上执行。

2 — 同步等待在所有副本上执行。

执行OPTIMIZE TABLE命令：

OPTIMIZE TABLE clickhouse_tutorial.user_tag

Query id: c03335ce-fee8-42b6-bc36-3e23a8b59b29

Ok.

0 rows in set. Elapsed: 0.210 sec.

查看 Server 端日志：

2022.03.31 04:52:31.277919 [ 6156710 ] {} <Debug> TCP-Session: 610442b9-37d5-49e0-821d-c25984fc7f41 Creating query context from session context, user_id: 94309d50-4f52-5250-31bd-74fecac179db, parent context user: default

2022.03.31 04:52:31.278067 [ 6156710 ] {c03335ce-fee8-42b6-bc36-3e23a8b59b29} <Debug> executeQuery: (from 127.0.0.1:53198) optimize table clickhouse_tutorial.user_tag;

2022.03.31 04:52:31.278128 [ 6156710 ] {c03335ce-fee8-42b6-bc36-3e23a8b59b29} <Trace> ContextAccess (default): Access granted: OPTIMIZE ON clickhouse_tutorial.user_tag

2022.03.31 04:52:31.278239 [ 6156710 ] {c03335ce-fee8-42b6-bc36-3e23a8b59b29} <Debug>

2022.03.31 04:52:31.278294 [ 6156710 ] {c03335ce-fee8-42b6-bc36-3e23a8b59b29} <Debug> DiskLocal: Reserving 11.19 MiB on disk `default`, having unreserved 192.05 GiB.

2022.03.31 04:52:31.278340 [ 6156710 ] {edac2738-6508-46fa-aee7-7c5560f4539c::20140320_46_59_1} <Debug> : from 20140320_46_46_0 to 20140320_59_59_0 into Wide

...

MergeTreeSequentialSource: Reading 2 marks from part pRegionID, total 1443 rows starting from the beginning of the part

...

2022.03.31 04:52:31.470792 [ 6156710 ] {edac2738-6508-46fa-aee7-7c5560f4539c::20140320_46_59_1} <Debug>

2022.03.31 04:52:31.479793 [ 6156710 ] {edac2738-6508-46fa-aee7-7c5560f4539c::20140320_46_59_1} <Trace> MergedBlockOutputStream: filled checksums pRegionID (state Active)

...

clickhouse_tutorial.user_tag (edac2738-6508-46fa-aee7-7c5560f4539c):

2022.03.31 04:52:31.487058 [ 6156710 ] {edac2738-6508-46fa-aee7-7c5560f4539c::20140320_46_59_1} <Trace>

2022.03.31 04:52:31.487648 [ 6156710 ] {c03335ce-fee8-42b6-bc36-3e23a8b59b29} <Debug> MemoryTracker: Peak memory usage Mutate/Merge: 36.46 MiB.

2022.03.31 04:52:31.487756 [ 6156710 ] {c03335ce-fee8-42b6-bc36-3e23a8b59b29} <Debug> MemoryTracker: Peak memory usage (for query): 0.00 B.

2022.03.31 04:52:31.487861 [ 6156710 ] {} <Debug> TCPHandler: Processed in 0.20999 sec.

可以看到执行OPTIMIZE TABLE命名，服务端通过MergerMutator（源码MergeTreeDataMergerMutator.cpp）发起了一个MergeTask（源码MergeTask.cpp）任务，执行了一组数据 Part 的合并（merge）、突变（mutation）和移动（move）操作。

其中，MergerMutator为后台进程选择数据Part，并执行合并、突变和移动等操作。

1.1.7. 

ATTACH操作与 CREATE 的功能相同。

挂载数据库表或者字典：

ATTACH TABLE|DICTIONARY [IF NOT EXISTS] [db.]name [ON CLUSTER cluster] ...

从数据文件和指定表结构挂载表：

ATTACH TABLE name FROM 'path/to/data/' (col1 Type1, ...)

该操作使用提供的结构创建一个新表，并将表数据从提供的目录，挂载到 `user_files` 中。

挂载表或字典，执行ATTACH查询后，服务器将知道表或字典的存在。例如，在将ClickHouse数据库移动到另一台服务器时，可以使用此操作迁移数据。ATTACH 操作不会在磁盘上创建数据，而是假设数据已经在适当的位置，并且只是将有关表或字典的信息添加到服务器。

并将表数据从提供的目录中附加到 `user_files` 中。

1、准备数据文件data.CSV：

DROP TABLE IF EXISTS clickhouse_tutorial.my_test_table;

INSERT INTO TABLE FUNCTION file('/Users/data/clickhouse/user_files/my_test_table/data.CSV', 'CSV', 's String, n UInt8') VALUES ('abc', 3);

2、使用 ATTACH从表数据文件目录 /Users/data/clickhouse/user_files/my_test_table/ 建表：

ATTACH TABLE clickhouse_tutorial.my_test_table

FROM '/Users/data/clickhouse/user_files/my_test_table'  

(s String, n UInt8)

ENGINE = File(CSV);

3、查看表数据：

SELECT *

FROM clickhouse_tutorial.my_test_table

Query id: 33a9e9d9-e95f-4aa0-806c-6362a2f1baeb

┌─s───┬─n─┐

│ abc │ 3 │

└─────┴───┘

1 rows in set. Elapsed: 0.002 sec.

1、分离表

DETACH TABLE clickhouse_tutorial.my_test_table

2、查询分离表

SELECT *

FROM clickhouse_tutorial.my_test_table

Query id: 6a16501c-23ca-47ed-970c-1a14cf30a897

0 rows in set. Elapsed: 0.001 sec.

Received exception from server (version 22.4.1):

Code: 60. DB::Exception: Received from 127.0.0.1:9009. DB::Exception: Table clickhouse_tutorial.my_test_table doesn't exist. (UNKNOWN_TABLE)

3、挂载表

ATTACH TABLE clickhouse_tutorial.my_test_table

Query id: 09cbec26-cbb6-41fc-ad58-eb6264ff111f

Ok.

0 rows in set. Elapsed: 0.001 sec.

4、查询表

SELECT *

FROM clickhouse_tutorial.my_test_table

Query id: a9a7cad8-3c06-4397-8762-ad6160091d91

┌─s───┬─n─┐

│ abc │ 3 │

└─────┴───┘

1 rows in set. Elapsed: 0.002 sec.

1.1.8. 

DETACH TABLE|VIEW|DICTIONARY [IF EXISTS] [db.]name [ON CLUSTER cluster] [PERMANENTLY]

DETACH : 分离表、视图或字典。详细说明如下：

分离操作不会删除表、物化视图、字典数据或元数据。

如果实体未“永久”分离，则在下一次服务器启动时，服务器将读取元数据并再次召回该表、视图或字典。如果实体被“永久”分离，则不会自动召回。

无论表或字典是否被永久分离，都可以使用ATTACH操作重新挂载它。

不能RENAME TABLE、DROP TABLE已经分离的表。

不能CREATE TABLE 与永久分离表名称相同。

创建test表：

CREATE TABLE clickhouse_tutorial.test ENGINE = Log AS SELECT * FROM numbers(10);

查询数据：

SELECT *
FROM clickhouse_tutorial.test
 
Query id: 55175f17-9d56-4909-93e3-99c7b945b02c
 
┌─number─┐
│      0 │
│      1 │
│      2 │
│      3 │
│      4 │
│      5 │
│      6 │
│      7 │
│      8 │
│      9 │
└────────┘
 
10 rows in set. Elapsed: 0.002 sec.

DETACH test表：

DETACH TABLE clickhouse_tutorial.test
 
Query id: f16d9541-3640-4556-9466-850b99d131aa
 
Ok.
 
0 rows in set. Elapsed: 0.001 sec.

再次查询 test 表，提示表不存在：：

SELECT *
FROM clickhouse_tutorial.test
 
Query id: 8338bbc9-6d19-4726-a550-c773851ddf58
 
0 rows in set. Elapsed: 0.001 sec.
 
Received exception from server (version 22.4.1):
Code: 60. DB::Exception: Received from 127.0.0.1:9009. DB::Exception: Table clickhouse_tutorial.test doesn't exist. (UNKNOWN_TABLE)

但是，此时也不能使用表名 test 创建另外一张表：

CREATE TABLE clickhouse_tutorial.test
ENGINE = Log AS
SELECT *
FROM numbers(10)
 
Query id: c338b28a-45c9-41a3-a6c0-f6ddb374df52
 
0 rows in set. Elapsed: 0.001 sec.
 
Received exception from server (version 22.4.1):
Code: 57. DB::Exception: Received from 127.0.0.1:9009. DB::Exception: Table `clickhouse_tutorial`.`test` already exists (detached). (TABLE_ALREADY_EXISTS)

标签： 1电容cbb80e贴片二极管丝印f17聚丙烯膜介质电容器cbb80b

锐单商城拥有海量元器件数据手册、 IC替代型号，打造 电子元器件IC百科大全！

锐单商城 - 一站式电子元器件采购平台