hive 常用操作

参考：https://www.cnblogs.com/jonban/p/10779938.html

Hive 启动：hive
退出:hive>quit;
show databases;

use analysis;
show tables;
desc tab_name; --查看表的结构及表的路径
show partitions fact_measured_cft_hive ;展示表分区 fact_measured_cft_hive

show create table fact_five_data_cft; --查看创建表的过程

启动pyspark 

pyspark2 --master local[4]   本地运行
==========
HADOOP_CONF_DIR 指hadoop配置文件所在目录
HADOOP_CONF_DIR=/opt/cloudera/parcels/CDH/lib/hadoop/etc/hadoop pyspark2 --master yarn --deploy-mode client  在hadoop yarn上运行pyspark

1、创建表 create-table.sql

create table if not exists db_hive.tb_user
(
    id int,
    username string comment '用户名',
    age int comment '年龄',
    address string comment '地址'
)
comment '用户表'
row format delimited fields terminated by ',' 
stored as textfile
location '/user/hive/warehouse/db_hive.db/db_user'

2、执行创建表

hive -f 'create-table.sql'

3、加载数据到 tb_user 表中

数据文件 /root/files/tb_user.txt

1001,Logan,16,shenzhen
1002,Herry potter,12,Magic school
1003,孙悟空,500,花果山

Hive交互式命令行执行命令 load data local inpath '/root/files/tb_user.txt' into table db_hive.tb_user;

如下所示：

hive (db_hive)> load data local inpath '/root/files/tb_user.txt' into table db_hive.tb_user;

如果要覆盖旧数据，可以加 overwrite，如下所示

hive (db_hive)> load data local inpath '/root/files/tb_user.txt' overwrite into table db_hive.tb_user;

4、查询数据

hive -e "select id,username from db_hive.tb_user"

5、根据已有表创建只有部分字段的子表

create table if not exists db_hive.tb_user_sub 
as 
select id,username from db_hive.tb_user;

6、 like 创建表

create table if not exists db_hive.tb_user_like like db_hive.tb_user;

插入数据

insert into table db_hive.tb_user_like select * from db_hive.tb_user;

7、重命名表

alter table tb_user_like rename to tb_user_rename ;

8、创建外部表,删除时只删除元数据，不会删除表数据

create external table if not exists db_hive.tb_ext(id string);

9、创建分区表

create table if not exists db_hive.tb_logs(
    ip string,
    text string,
    log_time string
)
partitioned by (month string)
row format delimited fields terminated by "\t";

数据文件 /root/files/tb_logs.txt

192.168.32.100  login   20190429072650
192.168.32.100  order   20190429072730
192.168.32.101  browse  20190429072812

载入数据

load data local inpath '/root/files/tb_logs.txt' into table db_hive.tb_logs partition (month = '201904')

查询数据

select ip,text,log_time from tb_logs where month = '201904';

10、手工创建分区数据及修复分区表

创建分区目录

hdfs dfs -mkdir -p /user/hive/warehouse/db_hive.db/tb_logs/month=201905

上传数据文件到分区目录下

hdfs dfs -put /root/files/tb_logs.txt /user/hive/warehouse/db_hive.db/tb_logs/month=201905

此时执行查询

select count(distinct ip) from db_hive.tb_logs where month = '201905';

查询结果为0。

【原因】：数据并未添加到分区中，查看配置的MySQL元数据信息

mysql> use hive_metastore;
mysql> select * from PARTITIONS;

示例配置的Hive元数据存放为MySQL数据库中的 hive_metastore 数据库

查询分区表 PARTITIONS 中的数据，发现只有一条记录，如下所示：

+---------+-------------+------------------+--------------+-------+--------+
| PART_ID | CREATE_TIME | LAST_ACCESS_TIME | PART_NAME    | SD_ID | TBL_ID |
+---------+-------------+------------------+--------------+-------+--------+
|       1 |  1556494255 |                0 | month=201904 |    29 |     28 |
+---------+-------------+------------------+--------------+-------+--------+

【修复方法一】直接执行修复命令

msck repair table tb_logs

此时分区表中的数据如下：

+---------+-------------+------------------+--------------+-------+--------+
| PART_ID | CREATE_TIME | LAST_ACCESS_TIME | PART_NAME    | SD_ID | TBL_ID |
+---------+-------------+------------------+--------------+-------+--------+
|       1 |  1556494255 |                0 | month=201904 |    29 |     28 |
|       2 |  1556495227 |                0 | month=201905 |    30 |     28 |
+---------+-------------+------------------+--------------+-------+--------+

执行查询命令

select count(distinct ip) from db_hive.tb_logs where month = '201905';

返回结果为2，数据已正常加入分区。

【修复方法二】 使用增加分区命令

操作步骤：创建新分区目录并上传数据文件，命令如下：

hive (db_hive)> dfs -mkdir -p /user/hive/warehouse/db_hive.db/tb_logs/month=201906;
hive (db_hive)> dfs -put /root/files/tb_logs.txt /user/hive/warehouse/db_hive.db/tb_logs/month=201906;

执行增加分区命令

alter table tb_logs add partition(month = '201906');

查询数据，测试结果正常。

此时元数据分区表中数据如下：

+---------+-------------+------------------+--------------+-------+--------+
| PART_ID | CREATE_TIME | LAST_ACCESS_TIME | PART_NAME    | SD_ID | TBL_ID |
+---------+-------------+------------------+--------------+-------+--------+
|       1 |  1556494255 |                0 | month=201904 |    29 |     28 |
|       2 |  1556495227 |                0 | month=201905 |    30 |     28 |
|       3 |  1556495635 |                0 | month=201906 |    31 |     28 |
+---------+-------------+------------------+--------------+-------+--------+

查看表分区命令

show partitions db_hive.tb_logs;

11、导出表数据

export table db_hive.tb_logs to '/user/hive/warehouse/export/db_hive/tb_logs';

12、导入表数据

创建表

create table tb_logs_like like tb_logs;

导入数据

import table tb_logs_like from '/user/hive/warehouse/export/db_hive/tb_logs';

13、导出数据到本地文件

insert overwrite local directory '/root/files/hive_out' 
row format delimited fields terminated by '\t' collection items terminated by '\n' 
select * from db_hive.tb_logs;

Hive 常用命令和语句

#!/usr/bin/python
# -*- coding: utf-8 -*-
from sqlalchemy.engine import create_engine
from sqlalchemy import text
import pandas as pd
import datetime

starttime = datetime.datetime.now()

sql = """
    select *
    from fact_five_data_cft
    where source = 'classics' and  wfid='320924' and yyyy='2019'
    limit 10
    """
engine = create_engine('presto://node1:8085/hive/cnyb')
df = pd.read_sql(text(sql),engine)
print(df)
endtime = datetime.datetime.now()
runtime = endtime - starttime
print "presto run time -->"+str(runtime)