Hive 进阶

两种情况下不走map-reduce:

1. where ds >' ' //ds 是partition

2. select * from table //后面没有查询条件，什么都没有

1.建表

CREATE TABLE sal(
  id INT,
  name STRING,
  salary INT
  )
partitioned by (city string)
ROW FORMAT DELIMITED
FIELDS TERMINATED BY '\t'
STORED AS TEXTFILE;

1.1 修改表及属性

#把id,name以外的列删除
alter table sal replace columns (id int, name string);
#增加列
alter table sal add columns (remark string);
#修改column
ALTER TABLE table_name
      CHANGE  col_old_name col_new_name
      column_type;

ALTER TABLE sal CHANGE remark city string;

2.导入数据

load data local inpath '/home/hadoop/in/mytable' overwrite into table sal;
1	zuansun	3000	none
2	zuansu2	4000	none
3	zuansu3	3000	none
4	zuansu4	4000	none
5	zuansu5	3000	none
6	zuansu6	4000	none
7	zuansu7	3000	none
8	zuansu8	4000	none
9	zuansu9	10000	none
10	zuansu10	20000	none
11	zuansu11	15000	none
12	zuansu12	25000	none

3.嵌套查询

from (select * from sal) e select e.id,e.name,e.salary  where e.salary>3000;
#case when
select id,name,
 case
    when salary<10000 then '屌丝'
    when salary>=10000 and salary<20000 then '中下等'
    when salary>=20000 and salary<50000 then '高帅富'
    else '外星人'
  end as salarylevel
from sal;

4.group by

select remark,sum(salary) from sal group by remark;

5.动态分区

5.1 创建临时表

CREATE TABLE sal_tmp(
  id INT,
  name STRING,
  salary INT,
  city string
  )
ROW FORMAT DELIMITED
FIELDS TERMINATED BY '\t'
STORED AS TEXTFILE;

5.2将数据导入到临时表中

load data local inpath '/home/hadoop/in/mytable' overwrite into table sal_tmp;

5.3 操作的配置

set hive.exec.dynamic.partition=true; // 允许动态分区
set hive.exec.dynamic.partition.mode=nonstrict;
set hive.exec.dynamic.partitions.pernode=50000;
set hive.exec.dynamic.partitions.partitions=50000;
set hive.exec.max.created.files=500000;
set mapred.reduce.tasks =20000; //每个任务默认的reduce数目
set hive.merge.mapfiles=true;  //在Map-only的任务结束时合并小文件

5.4附partition相关参数：

hive.exec.dynamic.partition（缺省false）： 设置为true允许使用dynamic partition

hive.exec.dynamic.partition.mode（缺省strick）：设置dynamic partition模式（nostrict允许所有partition列都为dynamic partition，strict不允许）

hive.exec.max.dynamic.partitions.pernode （缺省100）：每一个mapreduce job允许创建的分区的最大数量，如果超过了这个数量就会报错

hive.exec.max.dynamic.partitions （缺省1000）：一个dml语句允许创建的所有分区的最大数量

hive.exec.max.created.files （缺省100000）：所有的mapreduce job允许创建的文件的最大数量









5.5

insert into table sal partition (city) select * from sal_tmp;

6. join操作

#建表
create table a(id int,gender string)
row format delimited fields terminated by '\t' stored as textfile;
#加载数据
load data local inpath '/home/hadoop/in/a' overwrite into table a;
#内连接查询
select sal.id,sal.name,sal.salary,sal.city,a.gender from sal join a on(sal.id=a.id);
#左外连接查询
select sal.id,sal.name,sal.salary,sal.city,a.gender from sal left outer join a on(sal.id=a.id);

7.创建索引

create index a_index on table a(id) AS  'org.apache.hadoop.hive.ql.index.compact.CompactIndexHandler' WITH DEFERRED REBUILD ;     

8.桶

#临时表
create table tb_tmp(id int,age int, name string ,timeflag bigint) row format delimited fields terminated by ',';
#带桶的表,4个桶
create table tb_stu(id int,age int, name string,timeflag bigint) clustered by (id) sorted by (age) into 4 buckets row format delimited fields terminated by ',';
#加载数据到临时表
load data local inpath '/home/hadoop/in/tb_tmp' overwrite into table tb_tmp;

1,20,zxm,20140330
2,21,ljz,20140330
3,19,cds,20140330
4,18,mac,20140330
5,22,android,20140330
6,23,symbian,20140330
7,25,wp,20140330
8,20,cxd,20140330
9,21,fvd,20140330
10,19,cvb,20140330
11,18,erd,20140330
12,22,nid,20140330
13,23,fvd,20140330
14,19,cvb,20140330
15,18,e33,20140330
16,22,nid,20140330
#设置执行桶的属性
set hive.enforce.bucketing = true;
#插入到tb_stu表
insert into table tb_stu select * from tb_tmp;
#抽样
select * from tb_stu tablesample(bucket 1 out of 4 on id);
注：tablesample是抽样语句，语法：TABLESAMPLE(BUCKET x OUT OF y),相当于以下语句：
SELECT * FROM numbersflat WHERE number % y = x-1;

9.RCfile

#rcfile 格式表
create table tb_rc(id int,age int, name string ,timeflag bigint) row format delimited fields terminated by ',' stored as rcfile;
#插入数据,上表中已经有tb_tmp表，所以直接插入数据即可
insert into table tb_rc select * from tb_tmp;

10.分隔符的多样化（配合正则表达式使用）

#cat /tmp/liuxiaowen/1.txt

000377201207221125^^APPLE IPHONE 4S^^2
132288201210331629^^THINKING IN JAVA^^1
132288201210331629^^THIN ssss^^1111
132288201210331629^^THdd dd ddJAVA^^10

add jar /opt/app/hive-0.7.0-rc1/lib/hive-contrib-0.7.0.jar ;  

create external table tt(times string,
product_name string,
sale_num string
) ROW FORMAT
SERDE 'org.apache.hadoop.hive.contrib.serde2.RegexSerDe'
WITH SERDEPROPERTIES
( 'input.regex' = '([^^]*)\\^\\^([^^]*)\\^\\^([^^]*)',
'output.format.string' = '%1$s %2$s %3$s')
STORED AS TEXTFILE;

load data local inpath '/home/hadoop/in/tt' overwrite into table tt;

hive> select product_name from tt;  

APPLE IPHONE 4S
THINKING IN JAVA
THIN ssss
THdd dd ddJAVA 

11.更加复杂的数据类型

11.1 array

cat login_array.txt
192.168.1.1,3105007010|3105007011|3105007012
192.168.1.2,3105007020|3105007021|3105007022

CREATE TABLE login_array (
  ip  STRING,
  uid  array<BIGINT>
)
PARTITIONED BY (dt STRING)
ROW FORMAT DELIMITED
FIELDS TERMINATED BY ','
COLLECTION ITEMS TERMINATED BY '|'
STORED AS TEXTFILE;
加载数据到hive表
LOAD DATA LOCAL INPATH '/home/hadoop/in/login_array' OVERWRITE INTO TABLE login_array PARTITION (dt='20130101');
#查看数据
select * from login_array;
192.168.1.1	[3105007010,3105007011,3105007012]	20130101
192.168.1.2	[3105007020,3105007021,3105007022]	20130101
select ip,uid[0] from login_array where dt='20130101'; --使用下标访问数组
192.168.1.1	3105007010
192.168.1.2	3105007020
select ip,size(uid) from login_array where dt='20130101'; #查看数组长度
192.168.1.1	3
192.168.1.2	3
select * from login_array where  array_contains(uid,3105007010);#数组查找
192.168.1.1	[3105007010,3105007011,3105007012]	20130101

11.2 使用Map

cat map_test_raw:
2014-03-03 12:22:34#127.0.0.1#get#amap#src=123&code=456&cookie=789#status=success&time=2s
2014-03-03 11:22:34#127.0.0.1#get#autonavi#src=123&code=456#status=success&time=2s&cookie=789
#创建表
create external  table map_test_raw(ts String,ip String,type String,logtype String,request Map<String,String>,response Map<String,String>)
  ROW FORMAT DELIMITED FIELDS TERMINATED BY '#'
  COLLECTION ITEMS TERMINATED BY '&'
  MAP KEYS TERMINATED BY '='
  stored as textfile;
LOAD DATA LOCAL INPATH '/home/hadoop/in/map_test_raw' OVERWRITE INTO TABLE map_test_raw;
#查看数据
select * from map_test_raw;
2014-03-03 12:22:34	127.0.0.1	get	amap	{"src":"123","code":"456","cookie":"789"}	{"status":"success","time":"2s"}
2014-03-03 11:22:34	127.0.0.1	get	autonavi	{"src":"123","code":"456"}	{"status":"success","time":"2s","cookie":"789"}

11.3 使用struct

# cat login_struct.txt
192.168.1.1,zhangsan:40
192.168.1.1,lisi:41
192.168.1.1,gavin:42
192.168.1.1,wangwu:43
192.168.1.1,xiaoming:44
192.168.1.1,xiaojun:45
# 建表
CREATE TABLE login_struct (
  ip  STRING,
  user  struct<name:string,age:int>
)
ROW FORMAT DELIMITED
FIELDS TERMINATED BY ','
COLLECTION ITEMS TERMINATED BY ':'
STORED AS TEXTFILE;
#导入数据
LOAD DATA LOCAL INPATH '/home/hadoop/in/login_struct' OVERWRITE INTO TABLE login_struct;
#查看数据
select ip,user from login_struct;

192.168.1.1	{"name":"zhangsan","age":40}
192.168.1.1	{"name":"lisi","age":41}
192.168.1.1	{"name":"gavin","age":42}
192.168.1.1	{"name":"wangwu","age":43}
192.168.1.1	{"name":"xiaoming","age":44}
192.168.1.1	{"name":"xiaojun","age":45}

版权声明：本文为博主原创文章，未经博主允许不得转载。