ClickHouse 使用EXPLAIN 分析 SQL 执行计划

Posted 2022-04-09 东海陈光剑

tags:

篇首语：本文由小常识网(cha138.com)小编为大家整理，主要介绍了ClickHouse 使用EXPLAIN 分析 SQL 执行计划相关的知识，希望对你有一定的参考价值。

1.1. 使用 EXPLAIN 分析 SQL 执行计划

本节介绍如何使用EXPLAIN命令分析SQL语句的执行计划。

1.1.1. EXPLAIN概述

执行计划是进行SQL查询调优的重要参考。在ClickHouse 中，可以使用EXPLAIN语句查看SQL查询的执行计划。EXPLAIN 的语法如下：

EXPLAIN [AST | SYNTAX | PLAN | PIPELINE | TABLE OVERRIDE]
[setting = value, ...]
[
SELECT ... | tableFunction(...) [COLUMNS (...)]
[ORDER BY ...]
[PARTITION BY ...]
[PRIMARY KEY]
[SAMPLE BY ...]
[TTL ...]
]
[FORMAT ...]

例如，执行 EXPLAIN 查看如下 SQL 的执行计划：

EXPLAIN
SELECT sum(number)
FROM numbers(10)
UNION ALL
SELECT sum(number)
FROM numbers(10)
ORDER BY sum(number) DESC
FORMAT TSV
 
Query id: 6c7d7f0f-d875-4e55-b557-0088c3c6b563
 
Union
  Expression ((Projection + Before ORDER BY))
    Aggregating
      Expression (Before GROUP BY)
        SettingQuotaAndLimits (Set limits and quota after reading from storage)
          ReadFromStorage (SystemNumbers)
  Expression (Projection)
    Sorting (Sorting for ORDER BY)
      Expression (Before ORDER BY)
        Aggregating
          Expression (Before GROUP BY)
            SettingQuotaAndLimits (Set limits and quota after reading from storage)
              ReadFromStorage (SystemNumbers)
 
13 rows in set. Elapsed: 0.002 sec.

还可以可视化查看 SQL 执行管道流水线 Pipeline，如下图（在后面的章节中，我们会介绍如何生成这样的图）：

1.1.2. EXPLAIN类型

EXPLAIN 常用语句类型如下：

1. EXPLAIN AST：用于查看抽象语法树（Abstract Syntax Tree）。支持查看所有类型的语句，不光是SELECT语句。

2. EXPLAIN SYNTAX：查询经过AST层优化后的 SQL 语句；

3. EXPLAIN PLAN：用于查看执行计划（Execution Plan），等价于直接使用 EXPLAIN。可以指定五个参数；

①　header：是否打印计划中各个步骤的 header 说明。1：开启，0：关闭（默认值）；

②　description：是否打印计划中各个步骤的描述。1：开启（默认值），0：关闭；

③　indexes：是否显示索引使用情况。1：开启，0：关闭（默认值）；

④　actions：是否打印计划中各个步骤的详细信息。1：开启，0：关闭（默认值）；

⑤　json：是否以 JSON 格式打印执行计划的详细信息。1：开启，0：关闭（默认值）；

4. EXPLAIN PIPELINE：用于查看 Pipeline 计划，可以指定三个参数；

①　header：是否打印计划中各个步骤的 header 说明。1：开启，0：关闭（默认值）；

②　graph：是否用DOT纯文本图形语言描述管道图。1：开启，0：关闭（默认值）；

③　compact：如果开启了 graph，紧凑打印行开关。1：开启（默认值），0：关闭；

5. EXPLAIN ESTIMATE：显示在处理查询时要从表中读取的行、标记和部分的估计数量。适用于 MergeTree系列表引擎。

1.1.3. EXPLAIN AST

查看查询的抽象语法树（AST）。

一个最简单的 SELECT 查询的AST

EXPLAIN AST
SELECT 1
 
Query id: 6823b212-e89e-40ef-a6a1-2dc36ca7b1f6
 
┌─explain───────────────────────────┐
│ SelectWithUnionQuery (children 1) │
│  ExpressionList (children 1)      │
│   SelectQuery (children 1)        │
│    ExpressionList (children 1)    │
│     Literal UInt64_1              │
└───────────────────────────────────┘
 
5 rows in set. Elapsed: 0.002 sec.

一个ALTER 查询的AST

EXPLAIN AST
ALTER TABLE tutorial.hits_v1
    DELETE WHERE EventDate = toDate('2014-03-17')
 
Query id: b2452669-68b6-4e5d-8b1e-1a42652f8343
 
┌─explain──────────────────────────────────┐
│ AlterQuery tutorial hits_v1 (children 3) │
│  ExpressionList (children 1)             │
│   AlterCommand DELETE (children 1)       │
│    Function equals (children 1)          │
│     ExpressionList (children 2)          │
│      Identifier EventDate                │
│      Function toDate (children 1)        │
│       ExpressionList (children 1)        │
│        Literal '2014-03-17'              │
│  Identifier tutorial                     │
│  Identifier hits_v1                      │
└──────────────────────────────────────────┘
 
11 rows in set. Elapsed: 0.001 sec.

一个带聚合计算的查询 AST

EXPLAIN AST
SELECT
    uniqExact(UserID) AS cnt,
    sum(RequestNum) / cnt AS rpu,
    EventDate
FROM tutorial.hits_v1 AS DELETE
WHERE JavaEnable = 1
GROUP BY EventDate
ORDER BY rpu DESC
LIMIT 3
 
Query id: 60c3c601-cc57-4a56-bf71-bed4cc7c56e5
 
┌─explain───────────────────────────────────────────────┐
│ SelectWithUnionQuery (children 1)                     │
│  ExpressionList (children 1)                          │
│   SelectQuery (children 6)                            │
│    ExpressionList (children 3)                        │
│     Function uniqExact (alias cnt) (children 1)       │
│      ExpressionList (children 1)                      │
│       Identifier UserID                               │
│     Function divide (alias rpu) (children 1)          │
│      ExpressionList (children 2)                      │
│       Function sum (children 1)                       │
│        ExpressionList (children 1)                    │
│         Identifier RequestNum                         │
│       Identifier cnt                                  │
│     Identifier EventDate                              │
│    TablesInSelectQuery (children 1)                   │
│     TablesInSelectQueryElement (children 1)           │
│      TableExpression (children 1)                     │
│       TableIdentifier tutorial.hits_v1 (alias DELETE) │
│    Function equals (children 1)                       │
│     ExpressionList (children 2)                       │
│      Identifier JavaEnable                            │
│      Literal UInt64_1                                 │
│    ExpressionList (children 1)                        │
│     Identifier EventDate                              │
│    ExpressionList (children 1)                        │
│     OrderByElement (children 1)                       │
│      Identifier rpu                                   │
│    Literal UInt64_3                                   │
└───────────────────────────────────────────────────────┘
 
28 rows in set. Elapsed: 0.002 sec.

JOIN嵌套子查询的 AST

EXPLAIN AST
SELECT count()
FROM
(
    SELECT UserID
    FROM tutorial.hits_v1
    PREWHERE EventDate = '2014-03-18'
) AS t1
INNER JOIN
(
    SELECT UserID
    FROM tutorial.hits_v1
    PREWHERE EventDate = '2014-03-19'
) AS t2 ON t1.UserID = t2.UserID
 
Query id: c42553b2-77fc-4657-b83c-9fc767e58b8c
 
┌─explain────────────────────────────────────────────┐
│ SelectWithUnionQuery (children 1)                  │
│  ExpressionList (children 1)                       │
│   SelectQuery (children 2)                         │
│    ExpressionList (children 1)                     │
│     Function count (children 1)                    │
│      ExpressionList                                │
│    TablesInSelectQuery (children 2)                │
│     TablesInSelectQueryElement (children 1)        │
│      TableExpression (children 1)                  │
│       Subquery (alias t1) (children 1)             │
│        SelectWithUnionQuery (children 1)           │
│         ExpressionList (children 1)                │
│          SelectQuery (children 3)                  │
│           ExpressionList (children 1)              │
│            Identifier UserID                       │
│           TablesInSelectQuery (children 1)         │
│            TablesInSelectQueryElement (children 1) │
│             TableExpression (children 1)           │
│              TableIdentifier tutorial.hits_v1      │
│           Function equals (children 1)             │
│            ExpressionList (children 2)             │
│             Identifier EventDate                   │
│             Literal '2014-03-18'                   │
│     TablesInSelectQueryElement (children 2)        │
│      TableExpression (children 1)                  │
│       Subquery (alias t2) (children 1)             │
│        SelectWithUnionQuery (children 1)           │
│         ExpressionList (children 1)                │
│          SelectQuery (children 3)                  │
│           ExpressionList (children 1)              │
│            Identifier UserID                       │
│           TablesInSelectQuery (children 1)         │
│            TablesInSelectQueryElement (children 1) │
│             TableExpression (children 1)           │
│              TableIdentifier tutorial.hits_v1      │
│           Function equals (children 1)             │
│            ExpressionList (children 2)             │
│             Identifier EventDate                   │
│             Literal '2014-03-19'                   │
│      TableJoin (children 1)                        │
│       Function equals (children 1)                 │
│        ExpressionList (children 2)                 │
│         Identifier t1.UserID                       │
│         Identifier t2.UserID                       │
└────────────────────────────────────────────────────┘
 
44 rows in set. Elapsed: 0.001 sec.

IN嵌套子查询 AST

EXPLAIN AST
SELECT count()
FROM
(
    SELECT UserID
    FROM tutorial.hits_v1
    PREWHERE (EventDate = '2014-03-18') AND (UserID IN (
        SELECT UserID
        FROM tutorial.hits_v1
        PREWHERE EventDate = '2014-03-19'
    ))
)
 
Query id: 5885eb3a-ddea-4cf4-a5cf-0bc38c02a02c
 
┌─explain────────────────────────────────────────────────────┐
│ SelectWithUnionQuery (children 1)                          │
│  ExpressionList (children 1)                               │
│   SelectQuery (children 2)                                 │
│    ExpressionList (children 1)                             │
│     Function count (children 1)                            │
│      ExpressionList                                        │
│    TablesInSelectQuery (children 1)                        │
│     TablesInSelectQueryElement (children 1)                │
│      TableExpression (children 1)                          │
│       Subquery (children 1)                                │
│        SelectWithUnionQuery (children 1)                   │
│         ExpressionList (children 1)                        │
│          SelectQuery (children 3)                          │
│           ExpressionList (children 1)                      │
│            Identifier UserID                               │
│           TablesInSelectQuery (children 1)                 │
│            TablesInSelectQueryElement (children 1)         │
│             TableExpression (children 1)                   │
│              TableIdentifier tutorial.hits_v1              │
│           Function and (children 1)                        │
│            ExpressionList (children 2)                     │
│             Function equals (children 1)                   │
│              ExpressionList (children 2)                   │
│               Identifier EventDate                         │
│               Literal '2014-03-18'                         │
│             Function in (children 1)                       │
│              ExpressionList (children 2)                   │
│               Identifier UserID                            │
│               Subquery (children 1)                        │
│                SelectWithUnionQuery (children 1)           │
│                 ExpressionList (children 1)                │
│                  SelectQuery (children 3)                  │
│                   ExpressionList (children 1)              │
│                    Identifier UserID                       │
│                   TablesInSelectQuery (children 1)         │
│                    TablesInSelectQueryElement (children 1) │
│                     TableExpression (children 1)           │
│                      TableIdentifier tutorial.hits_v1      │
│                   Function equals (children 1)             │
│                    ExpressionList (children 2)             │
│                     Identifier EventDate                   │
│                     Literal '2014-03-19'                   │
└────────────────────────────────────────────────────────────┘
 
42 rows in set. Elapsed: 0.001 sec.

1.1.4. EXPLAIN SYNTAX

查看 AST语法优化后的查询SQL。

自动优化PREWHERE

一个WHERE 自动优化成PREWHERE例子：

EXPLAIN SYNTAX
SELECT
    UserID,
    WatchID
FROM tutorial.hits_v1
WHERE JavaEnable = 1
 
Query id: 81dd2b32-8256-4afe-b65d-8017c15daa8a
 
┌─explain─────────────────┐
│ SELECT                  │
│     UserID,             │
│     WatchID             │
│ FROM tutorial.hits_v1   │
│ PREWHERE JavaEnable = 1 │
└─────────────────────────┘
 
5 rows in set. Elapsed: 0.002 sec.

自动谓词下推优化

SQL 里面的过滤条件：

WHERE (JavaEnable = 1) AND (EventDate = '2014-03-18')

自动谓词下推优化成了分区EventDate的过滤条件在前面

PREWHERE (EventDate = '2014-03-18') AND (JavaEnable = 1)

执行EXPLAIN SYNTAX如下：

EXPLAIN SYNTAX
SELECT count(UserID) AS cnt
FROM
(
    SELECT
        UserID,
        WatchID
    FROM tutorial.hits_v1
    WHERE (JavaEnable = 1) AND (EventDate = '2014-03-18')
) AS t1
INNER JOIN
(
    SELECT
        UserID,
        WatchID
    FROM tutorial.hits_v1
    WHERE (EventDate = '2014-03-19') AND (JavaEnable = 1)
) AS t2 ON t1.UserID = t2.UserID
 
Query id: b3b1cc3e-61b8-4ae7-9130-bc1f82fbdb45
 
┌─explain──────────────────────────────────────────────────────┐
│ SELECT count(UserID) AS cnt                                  │
│ FROM                                                         │
│ (                                                            │
│     SELECT                                                   │
│         UserID,                                              │
│         WatchID                                              │
│     FROM tutorial.hits_v1                                    │
│     PREWHERE (EventDate = '2014-03-18') AND (JavaEnable = 1) │
│ ) AS t1                                                      │
│ ALL INNER JOIN                                               │
│ (                                                            │
│     SELECT UserID                                            │
│     FROM tutorial.hits_v1                                    │
│     PREWHERE EventDate = '2014-03-19'                        │
│     WHERE (EventDate = '2014-03-19') AND (JavaEnable = 1)    │
│ ) AS t2 ON UserID = t2.UserID                                │
└──────────────────────────────────────────────────────────────┘
 
16 rows in set. Elapsed: 0.006 sec.

1.1.5. EXPLAIN PLAN

查看 SQL 执行计划。可以配置参数header、indexes、actions、json、description等。

一个最简单的SQL的执行计划

EXPLAIN header = 1, indexes = 1, actions = 1, json = 0, description = 0
SELECT 1
 
Query id: d41efec8-c69e-41bd-bb44-0e2b6781b2f2
 
┌─explain─────────────────────────────────────┐
│ Expression                                  │
│ Header: 1 UInt8                             │
│ Actions: COLUMN Const(UInt8) -> 1 UInt8 : 0 │
│ Positions: 0                                │
│   SettingQuotaAndLimits                     │
│   Header: dummy UInt8                       │
│     ReadFromStorage                         │
│     Header: dummy UInt8                     │
└─────────────────────────────────────────────┘
 
8 rows in set. Elapsed: 0.001 sec.

一个简单SQL查询执行计划

EXPLAIN header = 1, indexes = 1, actions = 1, json = 0, description = 0
SELECT
    UserID,
    WatchID
FROM tutorial.hits_v1
WHERE JavaEnable = 1
 
Query id: 1c3b84e0-87c0-44ef-9736-fcd203d1f9a5
 
┌─explain───────────────────────────────────┐
│ Expression                                │
│ Header: UserID UInt64                     │
│         WatchID UInt64                    │
│ Actions: INPUT :: 0 -> WatchID UInt64 : 0 │
│          INPUT :: 1 -> UserID UInt64 : 1  │
│ Positions: 1 0                            │
│   SettingQuotaAndLimits                   │
│   Header: WatchID UInt64                  │
│           UserID UInt64                   │
│     ReadFromMergeTree                     │
│     Header: WatchID UInt64                │
│             UserID UInt64                 │
│     ReadType: Default                     │
│     Parts: 1                              │
│     Granules: 1084                        │
│     Indexes:                              │
│       MinMax                              │
│         Condition: true                   │
│         Parts: 1/1                        │
│         Granules: 1084/1084               │
│       Partition                           │
│         Condition: true                   │
│         Parts: 1/1                        │
│         Granules: 1084/1084               │
│       PrimaryKey                          │
│         Condition: true                   │
│         Parts: 1/1                        │
│         Granules: 1084/1084               │
└───────────────────────────────────────────┘
 
28 rows in set. Elapsed: 0.003 sec.

一个复杂嵌套子查询JOIN的执行计划

EXPLAIN header = 1, indexes = 1, actions = 1, json = 0, description = 0
SELECT count(UserID) AS cnt
FROM
(
    SELECT
        UserID,
        WatchID
    FROM tutorial.hits_v1
    WHERE (JavaEnable = 1) AND (EventDate = '2014-03-18')
) AS t1
INNER JOIN
(
    SELECT
        UserID,
        WatchID
    FROM tutorial.hits_v1
    WHERE (EventDate = '2014-03-19') AND (JavaEnable = 1)
) AS t2 ON t1.UserID = t2.UserID
FORMAT TSV
 
Query id: 10bcac80-8e24-4c84-99f4-31a81961e978
 
Expression
Header: cnt UInt64
Actions: INPUT : 0 -> count(UserID) UInt64 : 0
         ALIAS count(UserID) :: 0 -> cnt UInt64 : 1
Positions: 1
  Aggregating
  Header: count(UserID) UInt64
  Keys:
  Aggregates:
      count(UserID)
        Function: count(UInt64) → UInt64
        Arguments: UserID
        Argument positions: 0
    Expression
    Header: UserID UInt64
    Actions: INPUT :: 0 -> UserID UInt64 : 0
    Positions: 0
      Join
      Header: UserID UInt64
        Expression
        Header: UserID UInt64
        Actions: INPUT :: 0 -> UserID UInt64 : 0
        Positions: 0
          Filter
          Header: UserID UInt64
          Filter column: and(equals(EventDate, \\'2014-03-18\\'), equals(JavaEnable, 1)) (removed)
          Actions: INPUT : 0 -> equals(EventDate, \\'2014-03-18\\') UInt8 : 0
                   INPUT : 1 -> JavaEnable UInt8 : 1
                   INPUT :: 2 -> UserID UInt64 : 2
                   COLUMN Const(UInt8) -> 1 UInt8 : 3
                   FUNCTION equals(JavaEnable :: 1, 1 :: 3) -> equals(JavaEnable, 1) UInt8 : 4
                   FUNCTION and(equals(EventDate, \\'2014-03-18\\') :: 0, equals(JavaEnable, 1) :: 4) -> and(equals(EventDate, \\'2014-03-18\\'), equals(JavaEnable, 1)) UInt8 : 3
          Positions: 2 3
            SettingQuotaAndLimits
            Header: equals(EventDate, \\'2014-03-18\\') UInt8
                    JavaEnable UInt8
                    UserID UInt64
              ReadFromMergeTree
              Header: equals(EventDate, \\'2014-03-18\\') UInt8
                      JavaEnable UInt8
                      UserID UInt64
              ReadType: Default
              Parts: 1
              Granules: 792
              Indexes:
                MinMax
                  Keys:
                    EventDate
                  Condition: and((EventDate in [16147, 16147]), (EventDate in [16147, 16147]))
                  Parts: 1/1
                  Granules: 1084/1084
                Partition
                  Keys:
                    toYYYYMM(EventDate)
                  Condition: and((toYYYYMM(EventDate) in [201403, 201403]), (toYYYYMM(EventDate) in [201403, 201403]))
                  Parts: 1/1
                  Granules: 1084/1084
                PrimaryKey
                  Keys:
                    EventDate
                  Condition: and((EventDate in [16147, 16147]), (EventDate in [16147, 16147]))
                  Parts: 1/1
                  Granules: 792/1084
        Expression
        Header: t2.UserID UInt64
        Actions: INPUT : 0 -> UserID UInt64 : 0
                 ALIAS UserID :: 0 -> t2.UserID UInt64 : 1
        Positions: 1
          Filter
          Header: UserID UInt64
          Filter column: and(equals(EventDate, \\'2014-03-19\\'), equals(JavaEnable, 1)) (removed)
          Actions: INPUT : 0 -> equals(EventDate, \\'2014-03-19\\') UInt8 : 0
                   INPUT : 1 -> JavaEnable UInt8 : 1
                   INPUT :: 2 -> UserID UInt64 : 2
                   COLUMN Const(UInt8) -> 1 UInt8 : 3
                   FUNCTION equals(JavaEnable :: 1, 1 :: 3) -> equals(JavaEnable, 1) UInt8 : 4
                   FUNCTION and(equals(EventDate, \\'2014-03-19\\') :: 0, equals(JavaEnable, 1) :: 4) -> and(equals(EventDate, \\'2014-03-19\\'), equals(JavaEnable, 1)) UInt8 : 3
          Positions: 2 3
            SettingQuotaAndLimits
            Header: equals(EventDate, \\'2014-03-19\\') UInt8
                    JavaEnable UInt8
                    UserID UInt64
              ReadFromMergeTree
              Header: equals(EventDate, \\'2014-03-19\\') UInt8
                      JavaEnable UInt8
                      UserID UInt64
              ReadType: Default
              Parts: 1
              Granules: 807
              Indexes:
                MinMax
                  Keys:
                    EventDate
                  Condition: and((EventDate in [16148, 16148]), (EventDate in [16148, 16148]))
                  Parts: 1/1
                  Granules: 1084/1084
                Partition
                  Keys:
                    toYYYYMM(EventDate)
                  Condition: and((toYYYYMM(EventDate) in [201403, 201403]), (toYYYYMM(EventDate) in [201403, 201403]))
                  Parts: 1/1
                  Granules: 1084/1084
                PrimaryKey
                  Keys:
                    EventDate
                  Condition: and((EventDate in [16148, 16148]), (EventDate in [16148, 16148]))
                  Parts: 1/1
                  Granules: 807/1084
 
108 rows in set. Elapsed: 0.012 sec.

1.1.6. EXPLAIN PIPELINE

显示查询Pipeline管道流水线图。可以指定header、graph、compact等参数。

通过 graph=1输出DOT 纯文本图形描述语言。如果安装了 graphviz dot 命令行环境，就可以把*.dot 纯文本文件生成图片，然后可视化查看SQL查询管道流水线。例如下面的 SQL

EXPLAIN PIPELINE header = 1, graph = 1, compact = 1
SELECT count(UserID) AS cnt
FROM
(
    SELECT
        UserID,
        WatchID
    FROM tutorial.hits_v1
    WHERE (JavaEnable = 1) AND (EventDate = '2014-03-18')
) AS t1
INNER JOIN
(
    SELECT
        UserID,
        WatchID
    FROM tutorial.hits_v1
    WHERE (EventDate = '2014-03-19') AND (JavaEnable = 1)
) AS t2 ON t1.UserID = t2.UserID
FORMAT TSV
 
Query id: 8eb30d70-ed05-42df-8c50-aee146227e45
 
digraph

  rankdir="LR";
   node [shape = rect]
        n8 [label="FillingRightJoinSide"];
        n4 [label="JoiningTransform × 6"];
        n1 [label="MergeTreeThread × 12"];
        n7 [label="Resize × 2"];
    subgraph cluster_0 
      label ="Aggregating";
      style=filled;
      color=lightgrey;
      node [style=filled,color=white];
       rank = same;
        n11 [label="AggregatingTransform × 6"];
        n12 [label="Resize"];
        n10 [label="StrictResize"];
      
    
    subgraph cluster_1 
      label ="Filter";
      style=filled;
      color=lightgrey;
      node [style=filled,color=white];
       rank = same;
        n2 [label="FilterTransform × 6"];
      
    
    subgraph cluster_2 
      label ="Expression";
      style=filled;
      color=lightgrey;
      node [style=filled,color=white];
       rank = same;
        n3 [label="ExpressionTransform × 6"];
      
    
    subgraph cluster_3 
      label ="Expression";
      style=filled;
      color=lightgrey;
      node [style=filled,color=white];
       rank = same;
        n13 [label="ExpressionTransform"];
      
    
    subgraph cluster_4 
      label ="Expression";
      style=filled;
      color=lightgrey;
      node [style=filled,color=white];
       rank = same;
        n9 [label="ExpressionTransform × 6"];
      
    
    subgraph cluster_5 
      label ="Expression";
      style=filled;
      color=lightgrey;
      node [style=filled,color=white];
       rank = same;
        n6 [label="ExpressionTransform × 6"];
      
    
    subgraph cluster_6 
      label ="Filter";
      style=filled;
      color=lightgrey;
      node [style=filled,color=white];
       rank = same;
        n5 [label="FilterTransform × 6"];
      
    
  
  n8 -> n7 [label=""];
  n4 -> n9 [label="× 6
UserID UInt64 UInt64(size = 0)"];
  n1 -> n2 [label="× 6
equals(EventDate, \\'2014-03-18\\') UInt8 UInt8(size = 0)
JavaEnable UInt8 UInt8(size = 0)
UserID UInt64 UInt64(size = 0)"];
  n1 -> n5 [label="× 6
equals(EventDate, \\'2014-03-19\\') UInt8 UInt8(size = 0)
JavaEnable UInt8 UInt8(size = 0)
UserID UInt64 UInt64(size = 0)"];
  n7 -> n4 [label="× 6"];
  n7 -> n8 [label="
t2.UserID UInt64 UInt64(size = 0)"];
  n11 -> n12 [label="× 6
count(UserID) UInt64 UInt64(size = 0)"];
  n12 -> n13 [label="
count(UserID) UInt64 UInt64(size = 0)"];
  n10 -> n11 [label="× 6
UserID UInt64 UInt64(size = 0)"];
  n2 -> n3 [label="× 6
UserID UInt64 UInt64(size = 0)"];
  n3 -> n4 [label="× 6
UserID UInt64 UInt64(size = 0)"];
  n9 -> n10 [label="× 6
UserID UInt64 UInt64(size = 0)"];
  n6 -> n7 [label="× 6
t2.UserID UInt64 UInt64(size = 0)"];
  n5 -> n6 [label="× 6
UserID UInt64 UInt64(size = 0)"];

 
105 rows in set. Elapsed: 0.011 sec.

把上面的DOT 图形描述语言文本digraph... ，保存一个文本文件 a.dot中，然后执行命令（需要安装 graphviz环境）：

dot -Tpng a.dot -o a.png

渲染出来a.png 图片就是SQL查询管道图，图片打开如下：

为了方便展示，这里把布局改成从上到下布局：rankdir="TD"。

Ø Tips：DOT纯文本图形描述语言

DOT 是一种纯文本图形描述语言。这是一种描述人类和计算机程序都可以阅读的图形的简单方法。

环境安装

graphviz dot 使用步骤

安装graphviz

Mac OS X系统上安装Graphviz可以使用MacPorts 或Homebrew 。

MacPorts 安装Graphviz

sudo port install graphviz

MacPorts提供了Graphviz 和 Mac GUI Graphviz.app的稳定版本和开发版本。这些可以通过port graphviz、 graphviz-devel、 graphviz-gui 和graphviz-gui-devel 获得。

Homebrew 安装Graphviz

brew install graphviz

如果遇到 Error 报错（通常可能是环境、网络、IO等问题），一般重试几次就会成功。

在其他操作系统（Linux、Windows 等）上安装Graphviz ，请参考： http://www.graphviz.org/download/ 。

编写dot脚本

创建文本文件并命名为a.dot, 编写a.dot脚本。

关于 dot 语法，请参考：http://www.graphviz.org/pdf/dotguide.pdf 。

编译脚本输出图片

编译命令:

dot -Tpng a.dot -o a.png

其中，a 是文件名。这样就可以用脚本渲染出你要绘制的图片了。

简单图形

graph graphname 
rankdir = TD;
a -- b;
    b -- c;
    b -- d;
    d -- a;

图形布局

graph graphname 
    rankdir = LR; //LR: Rank Direction Left to Right; TD: Top to Down
    a -- b;
    b -- c;
    b -- d;
    d -- a;

其中，

rankdir = LR: 左右布局 (Rank Direction Left to Right)

rankdir = TD: 上下布局 (Rank Direction Top to Down)

简单有向图 (Directional Graph)

digraph graphname
    rankdir = LR;
    a -> b;
    b -> c;
    a -> c;

添加标记说明

digraph graphname 
    rankdir = TD;
    T [label = "Teacher"] // node T
    P [label = "Pupil"] // node P
    T -> P [label = "Instructions", fontcolor = darkgreen] // edge T->P

更多内容请参考：https://www.tonyballantyne.com/graphs.html 。

1.1.7. EXPLAIN ESTIMATE

查看在处理查询时要从表中读取的数据行数、Mark和Part的预估数量。这个命令在分析 SQL 查询性能的时候非常有用。SQL 实例如下：

EXPLAIN ESTIMATE
SELECT count(UserID) AS cnt
FROM
(
    SELECT
        UserID,
        WatchID
    FROM tutorial.hits_v1
    WHERE (JavaEnable = 1) AND (EventDate = '2014-03-18')
) AS t1
INNER JOIN
(
    SELECT
        UserID,
        WatchID
    FROM tutorial.hits_v1
    WHERE (EventDate = '2014-03-19') AND (JavaEnable = 1)
) AS t2 ON t1.UserID = t2.UserID
 
Query id: a848f250-9e30-44b3-90eb-8e03c429441f
 
┌─database─┬─table───┬─parts─┬─────rows─┬─marks─┐
│ tutorial │ hits_v1 │     2 │ 13086546 │  1599 │
└──────────┴─────────┴───────┴──────────┴───────┘
 
1 rows in set. Elapsed: 0.011 sec.

1.2. 本章总结

通过对本章的学习，我们对ClickHouse SQL有了系统化的掌握。我们知道了如何使用 SQL 来创建数据库、创建表、创建视图，新增数据库用户账户、角色等，并进行库表权限管理。同时，我们还知道如何向 ClickHouse 表中插入数据，并对表中的数据进行查询。按照对SQL 的解析执行顺序，我们还详细介绍了各种SQL查询子句的用法。为了加深对SQL 查询原理的理解，我们还介绍了 SQL 发展简史、关系代数和关系演算相关的内容。最后，介绍了通过EXPLAIN语句探索查询执行过程。

下一章我们将进入项目实战开发，介绍如何使用 Spring Boot 连接 ClickHouse，实现一个简单的BI数据报表分析系统。

以上是关于ClickHouse 使用EXPLAIN 分析 SQL 执行计划的主要内容，如果未能解决你的问题，请参考以下文章

ClickHouse的介绍（基本sql操作，以及数据库引擎表引擎分片副本explain优化物化视图等）

大数据ClickHouse进阶：ClickHouse的explain查询执行计划

ClickHouse-尚硅谷（8. 高级-Explain 查看执行计划）学习笔记

(转)详解利用EXPLAIN分析sql语句的性能