hive 分区表和分桶表

1、创建分区表

hive> create table weather_list(year int,data int) partitioned by (createtime string,area string) row format delimited fields terminated by ",";

修改表：

hive> alter table weather_list change data new_data int;
hive> alter table weather_list change year new_year int;

1.1、加载数据

hive> load data local inpath '/home/hadoop/sampler/w2.csv' into table weather_list partition(createtime='2011-01-01',area='bj');
Loading data to table busdata.weather_list partition (createtime=2011-01-01, area=bj)
OK
Time taken: 1.455 seconds
hive> load data local inpath '/home/hadoop/sampler/w3.csv' into table weather_list partition(createtime='2011-01-02',area='sc');
Loading data to table busdata.weather_list partition (createtime=2011-01-02, area=sc)
OK
Time taken: 1.394 seconds
hive> load data local inpath '/home/hadoop/sampler/w4.csv' into table weather_list partition(createtime='2011-01-03',area='tj');
Loading data to table busdata.weather_list partition (createtime=2011-01-03, area=tj)
OK
Time taken: 1.568 seconds
hive> load data local inpath '/home/hadoop/sampler/w4.csv' into table weather_list partition(createtime='2011-01-04',area='sz');
Loading data to table busdata.weather_list partition (createtime=2011-01-04, area=sz)
OK
Time taken: 1.209 seconds
hive> load data local inpath '/home/hadoop/sampler/w5.csv' into table weather_list partition(createtime='2011-01-05',area='gz');
Loading data to table busdata.weather_list partition (createtime=2011-01-05, area=gz)
OK
Time taken: 1.148 seconds
hive> load data local inpath '/home/hadoop/sampler/w5.csv' into table weather_list partition(createtime='2011-01-01',area='gz');
Loading data to table busdata.weather_list partition (createtime=2011-01-01, area=gz)
OK
Time taken: 1.278 seconds

partition的分区字段体现在存储目录上，与文件中的实际存储字段没有关系。

hive> dfs -ls /hive/warehouse/busdata.db/weather_list/createtime=2011-01-01/area=gz;
Found 6 items
-rw-r--r--   1 hadoop supergroup      18018 2019-03-05 22:14 /hive/warehouse/busdata.db/weather_list/createtime=2011-01-01/area=gz/w1.csv
-rw-r--r--   1 hadoop supergroup      18022 2019-03-05 22:14 /hive/warehouse/busdata.db/weather_list/createtime=2011-01-01/area=gz/w2.csv
-rw-r--r--   1 hadoop supergroup      18028 2019-03-05 22:14 /hive/warehouse/busdata.db/weather_list/createtime=2011-01-01/area=gz/w3.csv
-rw-r--r--   1 hadoop supergroup      18022 2019-03-05 22:14 /hive/warehouse/busdata.db/weather_list/createtime=2011-01-01/area=gz/w4.csv
-rw-r--r--   1 hadoop supergroup      18027 2019-03-05 22:12 /hive/warehouse/busdata.db/weather_list/createtime=2011-01-01/area=gz/w5.csv
-rw-r--r--   1 hadoop supergroup      18027 2019-03-05 22:14 /hive/warehouse/busdata.db/weather_list/createtime=2011-01-01/area=gz/w5_copy_1.csv

hive> dfs -ls /hive/warehouse/busdata.db/weather_list/createtime=2011-01-01;
Found 2 items
drwxr-xr-x   - hadoop supergroup          0 2019-03-05 22:09 /hive/warehouse/busdata.db/weather_list/createtime=2011-01-01/area=bj
drwxr-xr-x   - hadoop supergroup          0 2019-03-05 22:14 /hive/warehouse/busdata.db/weather_list/createtime=2011-01-01/area=gz

1.2、显示分区信息

hive> show partitions weather_list;
OK
createtime=2010-01-01/area=bj
createtime=2010-01-01/area=sh
createtime=2010-01-01/area=yn
createtime=2010-01-02/area=sh
createtime=2011-01-01/area=bj
createtime=2011-01-01/area=gz
createtime=2011-01-02/area=sc
createtime=2011-01-03/area=tj
createtime=2011-01-04/area=sz
createtime=2011-01-05/area=gz
Time taken: 0.584 seconds, Fetched: 10 row(s)

1.3、分区列属于表的正式列，但是文件中没有存储分区列信息。分区列的信息是从目录中读取的。

hive> select * from weather_list where area='bj' limit 10;
OK
1999    71    2010-01-01    bj
1994    57    2010-01-01    bj
1995    33    2010-01-01    bj
1993    44    2010-01-01    bj
1994    99    2010-01-01    bj
1994    83    2010-01-01    bj
1995    59    2010-01-01    bj
1991    32    2010-01-01    bj
1992    74    2010-01-01    bj
2000    56    2010-01-01    bj
Time taken: 2.527 seconds, Fetched: 10 row(s)

2、分桶表

2.1、检查分桶属性，设置分桶属性是为了使用hive来自动分桶，因为分桶是根据分桶字段和数量进行hash取余，也可以自己分桶后导入。

hive> set hive.enforce.bucketing;
hive.enforce.bucketing=false
 
hive> set hive.enforce.bucketing=true;
 
hive> set hive.enforce.bucketing;
hive.enforce.bucketing=true

2.2、建立分桶表

hive> create table bucket_userinfo(userid int,username string) clustered by (userid) sorted by (userid asc) into 2 buckets row format delimited fields terminated by ",";
 
hive> desc formatted bucket_userinfo;
OK
# col_name                data_type               comment
userid                  int
username                string                                      
 
# Detailed Table Information
Database:               busdata
OwnerType:              USER
Owner:                  hadoop
CreateTime:             Wed Mar 06 23:11:37 CST 2019
LastAccessTime:         UNKNOWN
Retention:              0
Location:               hdfs://bigdata-senior01.home.com:9000/hive/warehouse/busdata.db/bucket_userinfo
Table Type:             MANAGED_TABLE
Table Parameters:
    COLUMN_STATS_ACCURATE    {\"BASIC_STATS\":\"true\",\"COLUMN_STATS\":{\"userid\":\"true\",\"username\":\"true\"}}
    SORTBUCKETCOLSPREFIX    TRUE
    bucketing_version       2
    numFiles                0
    numRows                 0
    rawDataSize             0
    totalSize               0
    transient_lastDdlTime    1551885097          
 
# Storage Information
SerDe Library:          org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe
InputFormat:            org.apache.hadoop.mapred.TextInputFormat
OutputFormat:           org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat
Compressed:             No
Num Buckets:            2
Bucket Columns:         [userid]
Sort Columns:           [Order(col:userid, order:1)]
Storage Desc Params:
    field.delim             ,
    serialization.format    ,
Time taken: 0.379 seconds, Fetched: 34 row(s)

2.3、使用hive自动分桶，这种情况是针对源数据已经导入hive。

hive> insert overwrite table bucket_userinfo select userid,username from userinfo;
 
然后hive启动作业分桶导入数据，本例中分两个桶，所以最终会根据userid的奇偶生成两个文件。

hive> dfs -ls /hive/warehouse/busdata.db/bucket_userinfo;
Found 2 items
-rw-r--r--   1 hadoop supergroup        106 2019-03-06 23:13 /hive/warehouse/busdata.db/bucket_userinfo/000000_0
-rw-r--r--   1 hadoop supergroup        103 2019-03-06 23:13 /hive/warehouse/busdata.db/bucket_userinfo/000001_0

hive> dfs -cat /hive/warehouse/busdata.db/bucket_userinfo/000000_0;
2,xu.dm
4,user123
6,user2
8,user4
10,user6
14,user8
16,user10
18,user12
20,user14
22,soldier2
24,soldier4
hive> dfs -cat /hive/warehouse/busdata.db/bucket_userinfo/000001_0;
1,admin
3,myuser
5,user1
7,user3
9,user5
13,user7
15,user9
17,user11
19,user13
21,soldier1
23,soldier3

hive> select * from bucket_userinfo;
OK
2    xu.dm
4    user123
6    user2
8    user4
10    user6
14    user8
16    user10
18    user12
20    user14
22    soldier2
24    soldier4
1    admin
3    myuser
5    user1
7    user3
9    user5
13    user7
15    user9
17    user11
19    user13
21    soldier1
23    soldier3
Time taken: 0.238 seconds, Fetched: 22 row(s)

2.4、从外部文件导入数据，结果与上面一样

hive> create table bucket_userinfo2(userid int,username string) clustered by (userid) sorted by (userid) into 2 buckets row format delimited fields terminated by ",";

hive> load data local inpath '/home/hadoop/userinfo2.txt' into table bucket_userinfo2;

hive> select * from bucket_userinfo2;
OK
2    xu.dm
4    user123
6    user2
8    user4
10    user6
14    user8
16    user10
18    user12
20    user14
22    soldier2
24    soldier4
1    admin
3    myuser
5    user1
7    user3
9    user5
13    user7
15    user9
17    user11
19    user13
21    soldier1
23    soldier3

hive>dfs -ls /hive/warehouse/busdata.db/bucket_userinfo2;
Found 2 items
-rw-r--r--   1 hadoop supergroup        106 2019-03-07 22:44 /hive/warehouse/busdata.db/bucket_userinfo2/000000_0
-rw-r--r--   1 hadoop supergroup        103 2019-03-07 22:44 /hive/warehouse/busdata.db/bucket_userinfo2/000001_0

2.4、对桶数据采样

hive> select * from bucket_userinfo tablesample(bucket 1 out of 2 on userid);
OK
2    xu.dm
6    user2
10    user6
20    user14
3    myuser
7    user3
17    user11
19    user13
21    soldier1
Time taken: 0.077 seconds, Fetched: 9 row(s)