# Presto

# 一、服务端安装

下载最新版本presto-server-0.247.tar.gz (opens new window)并解压

需要配置以下几项,详细解说见官方文档，这里只是本机快速配置

Node Properties: 每个节点的环境配置
JVM Config: 命令行的Java虚拟机选项
Config Properties: Presto服务器的配置
Catalog Properties: 连接器(数据源)的配置

# 1.1 Node Properties

文件位置etc/node.properties

node.environment=production
node.id=ffffffff-ffff-ffff-ffff-ffffffffffff
node.data-dir=/var/presto/data

1
2
3

# 1.2 JVM Config

文件位置etc/jvm.config

-server
-Xmx16G
-XX:+UseG1GC
-XX:G1HeapRegionSize=32M
-XX:+UseGCOverheadLimit
-XX:+ExplicitGCInvokesConcurrent
-XX:+HeapDumpOnOutOfMemoryError
-XX:+ExitOnOutOfMemoryError

1
2
3
4
5
6
7
8

# 1.3 Config Properties

文件位置etc/config.properties

coordinator=true
node-scheduler.include-coordinator=true
http-server.http.port=8080
query.max-memory=5GB
query.max-memory-per-node=1GB
query.max-total-memory-per-node=2GB
discovery-server.enabled=true
discovery.uri=http://example.net:8080

1
2
3
4
5
6
7
8

# 1.4 Catalog Properties

文件位置etc/catalog/xx.properties

详细内容参考官方Connectors (opens new window)链接

# 1.4.1 Kudu Connector

文件位置 etc/catalog/kudu.properties

内容如下：

connector.name=kudu
kudu.client.master-addresses=localhost
kudu.schema-emulation.enabled=true
kudu.schema-emulation.prefix=

1
2
3
4

查询例子

创建表

CREATE TABLE kudu.default.users (
  user_id int WITH (primary_key = true),
  first_name varchar,
  last_name varchar
) WITH (
  partition_by_hash_columns = ARRAY['user_id'],
  partition_by_hash_buckets = 2
);

1
2
3
4
5
6
7
8

查看表Schema

DESCRIBE kudu.default.users;

插入数据

INSERT INTO kudu.default.users VALUES (1, 'Donald', 'Duck'), (2, 'Mickey', 'Mouse');

查询数据

SELECT * FROM kudu.default.users;

# 1.5 运行

后台运行

bin/launcher start

前台运行

bin/launcher run

停止

bin/launcher stop

# 1.6 命令行客户端安装

下载最新的客户端[Jar]https://repo1.maven.org/maven2/com/facebook/presto/presto-cli/0.247/presto-cli-0.247-executable.jar)包

下载好以后给予执行权限chmod +x 然后执行即可。--catalog 是连接器的名称

./presto --server localhost:8080 --catalog kudu --schema default

# 二、常用SQL

# 2.1 计算中位数

SELECT id, approx_percentile(value, 0.5)
FROM (
    VALUES
        (1, 2),
        (1, 3),
        (1, 7),
        (1, 8),
        (1, 4),
        (2, 20),
        (2, 30),
        (2, 40)
) AS t(id, value)
GROUP BY id;

1
2
3
4
5
6
7
8
9
10
11
12
13

# 2.2 String日期转Date

Date and Time Functions and Operators (opens new window)

select from_iso8601_date('2021-03-03') + interval '2' day
date_diff('second',date_parse(bcc.live_start_date,'%Y-%m-%d %H:%i:%s'),date_parse(bcc.live_end_date,'%Y-%m-%d %H:%i:%s'))

1
2

# 2.3 导出数据

sql="
select distinct username from odl_user_uc where length(username) >3 and not regexp_like(username,'\w+')
"
presto --server namenode1:8093 --catalog hive --schema default --execute "${sql}" --output-format CSV_HEADER > test

1
2
3
4

# 2.4 行专列

    select 
        t8.*
        ,order_id
    from tablexxx t8
    CROSS JOIN UNNEST(split(first_case_order_id,','))  as order_id(order_id)

1
2
3
4
5

# 2.5 date_diff

date_diff('day',date('2021-01-01'),date('2022-01-10'))

# 2.6 UNNEST(LATERAL VIEW explode)

-- hive 
SELECT student,score FROM tests LATERAL VIEW explode(scores)t AS score

-- presto
SELECT student, score
FROM tests
CROSS JOIN UNNEST(scores) AS t (score);

1
2
3
4
5
6
7

# 三、Tips

# 3.1 防止因为和hive不兼容跑数的时候报错

一般都是meta错误

set session hive.collect_column_statistics_on_write = false;