hive中简单介绍分区表(partition table)，含动态分区(dynamic partition)与静态分区(static partition)

简介：

hive中简单介绍分区表(partition table)，含动态分区(dynamic partition)与静态分区(static partition)

hive中创建分区表没有什么复杂的分区类型(范围分区、列表分区、hash分区、混合分区等)。分区列也不是表中的一个实际的字段，而是一个或者多个伪列。意思是说在表的数据文件中实际上并不保存分区列的信息与数据。

（1）下面的语句创建了一个简单的分区表

create table partition_test(
member_id string,
name string
)
partitioned by (
stat_date string,
province string)
ROW FORMAT DELIMITED FIELDS TERMINATED BY '\t';

（2）这个例子中创建了stat_date和province两个字段作为分区列。通常情况下需要先预先创建好分区，然后才能使用该分区，例如：

hive> alter table partition_test add partition (stat_date='20110728',province='zhejiang');
OK
Time taken: 0.39 seconds

2.1 这样就创建好了一个分区。这时我们会看到hive在HDFS存储中创建了一个相应的文件夹：

localhost:result_data a6$ hadoop fs -ls /user/hive/warehouse/yyz_workdb.db/partition_test
17/11/04 19:13:27 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Found 1 items
drwxr-xr-x   - a6 supergroup          0 2017-11-04 19:11 /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110728
localhost:result_data a6$ hadoop fs -ls /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110728
17/11/04 19:14:48 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Found 1 items
drwxr-xr-x   - a6 supergroup          0 2017-11-04 19:11 /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110728/province=zhejiang
localhost:result_data a6$

2.2 每一个分区都会有一个独立的文件夹，下面是该分区所有的数据文件。在这个例子中stat_date是主层次，province是副层次，所stat_date='20110728'，而province不同的分区都会在/user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110728 下面，而stat_date不同的分区都会在/user/hive/warehouse/yyz_workdb.db/partition_test下面，如：

假设我们在创建一个分区，代码如下：

hive> alter table partition_test add partition (stat_date='20110728',province='henan');
OK
Time taken: 0.256 seconds
hive> alter table partition_test add partition (stat_date='20110730',province='beijing');
OK
Time taken: 0.113 seconds

2.3 则在该hive表的hdfs存储目录下，如下结构：

localhost:result_data a6$ hadoop fs -ls /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110728
17/11/04 19:24:29 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Found 2 items
drwxr-xr-x   - a6 supergroup          0 2017-11-04 19:20 /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110728/province=henan
drwxr-xr-x   - a6 supergroup          0 2017-11-04 19:11 /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110728/province=zhejiang
localhost:result_data a6$ hadoop fs -ls /user/hive/warehouse/yyz_workdb.db/partition_test
17/11/04 19:24:34 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Found 2 items
drwxr-xr-x   - a6 supergroup          0 2017-11-04 19:20 /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110728
drwxr-xr-x   - a6 supergroup          0 2017-11-04 19:21 /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110730
localhost:result_data a6$ hadoop fs -ls /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110730
17/11/04 19:24:39 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Found 1 items
drwxr-xr-x   - a6 supergroup          0 2017-11-04 19:21 /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110730/province=beijing
localhost:result_data a6$

注意，因为分区列的值要转化为文件夹的存储路径，所以如果分区列的值中包含特殊值，如 '%', ':', '/', '#',它将会被使用%加上2字节的ASCII码进行转义，如：

hive>  alter table partition_test add partition (stat_date='2011/07/28',province='zhejiang');
OK
Time taken: 0.135 seconds

存储该hive表的分区的hdfs目录如下：

localhost:result_data a6$ hadoop fs -ls /user/hive/warehouse/yyz_workdb.db/partition_test
17/11/04 19:26:08 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Found 3 items
drwxr-xr-x   - a6 supergroup          0 2017-11-04 19:26 /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=2011%2F07%2F28
drwxr-xr-x   - a6 supergroup          0 2017-11-04 19:20 /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110728
drwxr-xr-x   - a6 supergroup          0 2017-11-04 19:21 /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110730

（3）下面准备使用一个辅助的非分区表partition_test_input准备向partition_test中插入数据：
3.1 创建非分区表格命令如下：

create table partition_test_input
(member_id string,
name string,
stat_date string,
province string)
ROW FORMAT DELIMITED FIELDS TERMINATED BY '\t';

3.2 插入数据，命令如下，结果如下：

hive> load data local inpath '/Users/a6/Applications/apache-hive-2.3.0-bin/result_data/partition_test_input.txt' overwrite into table partition_test_input;
Loading data to table yyz_workdb.partition_test_input
OK
Time taken: 0.867 seconds
hive> select * from partition_test_input;
OK
1	liujiannan	20110526	liaoning
2	wangchaoqun	20110526	hubei
3	xuhongxing	20110728	sichuan
4	zhudaoyong	20110728	henan
5	zhouchengyu	20110728	heilongjiang
Time taken: 0.08 seconds, Fetched: 5 row(s)

3.3 然后我向partition_test的分区中插入数据：

hive> select * from partition_test;
OK
Time taken: 0.132 seconds
hive> insert overwrite table partition_test partition(stat_date='20110728',province='henan') select member_id,name from partition_test_input where stat_date='20110728' and province='henan';
WARNING: Hive-on-MR is deprecated in Hive 2 and may not be available in the future versions. Consider using a different execution engine (i.e. spark, tez) or using Hive 1.X releases.
Query ID = a6_20171104200454_cbad387f-05f3-4bba-81a3-0a185396512d
Total jobs = 3
Launching Job 1 out of 3
Number of reduce tasks is set to 0 since there's no reduce operator
Starting Job = job_1509763925736_0007, Tracking URL = http://localhost:8088/proxy/application_1509763925736_0007/
Kill Command = /Users/a6/Applications/hadoop-2.6.5/bin/hadoop job  -kill job_1509763925736_0007
Hadoop job information for Stage-1: number of mappers: 1; number of reducers: 0
2017-11-04 20:05:02,062 Stage-1 map = 0%,  reduce = 0%
2017-11-04 20:05:08,361 Stage-1 map = 100%,  reduce = 0%
Ended Job = job_1509763925736_0007
Stage-4 is selected by condition resolver.
Stage-3 is filtered out by condition resolver.
Stage-5 is filtered out by condition resolver.
Moving data to directory hdfs://localhost:9002/user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110728/province=henan/.hive-staging_hive_2017-11-04_20-04-54_622_5274372985717082223-1/-ext-10000
Loading data to table yyz_workdb.partition_test partition (stat_date=20110728, province=henan)
MapReduce Jobs Launched:
Stage-Stage-1: Map: 1   HDFS Read: 5037 HDFS Write: 128 SUCCESS
Total MapReduce CPU Time Spent: 0 msec
OK
Time taken: 15.232 seconds
hive> select * from partition_test;
OK
4	zhudaoyong	20110728	henan
Time taken: 0.1 seconds, Fetched: 1 row(s)
hive>

3.4 还可以同时向多个分区插入数据，0.7版本以后不存在的分区会自动创建，0.6之前的版本官方文档上说必须要预先创建好分区。

为测试以上说法，首先显示partition_test这个hive表所具有的分区，命令如下：

hive> show partitions partition_test;
OK
stat_date=2011%2F07%2F28/province=zhejiang
stat_date=20110728/province=henan
stat_date=20110728/province=zhejiang
stat_date=20110730/province=beijing
Time taken: 0.101 seconds, Fetched: 4 row(s)

FROM partition_test_input
       INSERT overwrite TABLE partition_test PARTITION(
		stat_date = '20110526' ,
		province = 'liaoning'
	) SELECT
		member_id ,
		name
	WHERE
		stat_date = '20110526' AND province = 'liaoning' 
        INSERT overwrite TABLE partition_test PARTITION(
		stat_date = '20110728' ,
		province = 'sichuan'
	) SELECT
		member_id ,
		name
	WHERE
		stat_date = '20110728' AND province = 'sichuan' 
	INSERT overwrite TABLE partition_test PARTITION(
		stat_date = '20110728' ,
		province = 'heilongjiang'
	) SELECT
		member_id ,
		name
	WHERE
		stat_date = '20110728' AND province = 'heilongjiang';

3.5 执行以上语句后，则发现果然没有的分区被自动创建啦，

hive> show partitions partition_test;
OK
stat_date=2011%2F07%2F28/province=zhejiang
stat_date=20110526/province=liaoning
stat_date=20110728/province=heilongjiang
stat_date=20110728/province=henan
stat_date=20110728/province=sichuan
stat_date=20110728/province=zhejiang
stat_date=20110730/province=beijing
Time taken: 0.077 seconds, Fetched: 7 row(s)
hive>

特别要注意，在其他数据库中，一般向分区表中插入数据时系统会校验数据是否符合该分区，如果不符合会报错。而在hive中，向某个分区中插入什么样的数据完全是由人来控制的，因为分区键是伪列，不实际存储在文件中，如：

hive> insert overwrite table partition_test partition(stat_date='20110527',province='liaoning') select member_id,name from partition_test_input;
WARNING: Hive-on-MR is deprecated in Hive 2 and may not be available in the future versions. Consider using a different execution engine (i.e. spark, tez) or using Hive 1.X releases.
Query ID = a6_20171104201912_291312cf-da20-4b67-a746-1889b297f0bd
Total jobs = 3
Launching Job 1 out of 3
Number of reduce tasks is set to 0 since there's no reduce operator
Starting Job = job_1509763925736_0009, Tracking URL = http://localhost:8088/proxy/application_1509763925736_0009/
Kill Command = /Users/a6/Applications/hadoop-2.6.5/bin/hadoop job  -kill job_1509763925736_0009
Hadoop job information for Stage-1: number of mappers: 1; number of reducers: 0
2017-11-04 20:19:20,577 Stage-1 map = 0%,  reduce = 0%
2017-11-04 20:19:26,869 Stage-1 map = 100%,  reduce = 0%
Ended Job = job_1509763925736_0009
Stage-4 is selected by condition resolver.
Stage-3 is filtered out by condition resolver.
Stage-5 is filtered out by condition resolver.
Moving data to directory hdfs://localhost:9002/user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110527/province=liaoning/.hive-staging_hive_2017-11-04_20-19-12_640_370650644218657006-1/-ext-10000
Loading data to table yyz_workdb.partition_test partition (stat_date=20110527, province=liaoning)
MapReduce Jobs Launched:
Stage-Stage-1: Map: 1   HDFS Read: 4507 HDFS Write: 185 SUCCESS
Total MapReduce CPU Time Spent: 0 msec
OK
Time taken: 15.737 seconds
hive> select * from partition_test;
OK
1	liujiannan	20110526	liaoning
1	liujiannan	20110527	liaoning
2	wangchaoqun	20110527	liaoning
3	xuhongxing	20110527	liaoning
4	zhudaoyong	20110527	liaoning
5	zhouchengyu	20110527	liaoning
5	zhouchengyu	20110728	heilongjiang
4	zhudaoyong	20110728	henan
3	xuhongxing	20110728	sichuan
Time taken: 0.104 seconds, Fetched: 9 row(s)
hive>

可以看到在partition_test_input中的5条数据有着不同的stat_date和province，但是在插入到partition(stat_date='20110527',province='liaoning')这个分区后，5条数据的stat_date和province都变成相同的了，因为这两列的数据是根据文件夹的名字读取来的，而不是实际从数据文件中读取来的：

localhost:result_data a6$ hadoop fs -ls /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110527/province=liaoning
17/11/04 20:22:16 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Found 1 items
-rwxr-xr-x   1 a6 supergroup         67 2017-11-04 20:19 /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110527/province=liaoning/000000_0
localhost:result_data a6$ hadoop dfs -cat /user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110527/province=liaoning/000000_0
DEPRECATED: Use of this script to execute hdfs command is deprecated.
Instead use the hdfs command for it.

17/11/04 20:22:37 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
1	liujiannan
2	wangchaoqun
3	xuhongxing
4	zhudaoyong
5	zhouchengyu
localhost:result_data a6$

（4）下面介绍一下动态分区，因为按照上面的方法向分区表中插入数据，如果源数据量很大，那么针对一个分区就要写一个insert，非常麻烦。况且在之前的版本中，必须先手动创建好所有的分区后才能插入，这就更麻烦了，你必须先要知道源数据中都有什么样的数据才能创建分区。
使用动态分区可以很好的解决上述问题。动态分区可以根据查询得到的数据自动匹配到相应的分区中去。 
4.1 使用动态分区要先设置hive.exec.dynamic.partition参数值为true，默认值为false，即不允许使用：

hive> set hive.exec.dynamic.partition;
hive.exec.dynamic.partition=false
hive> set hive.exec.dynamic.partition=true;
hive> set hive.exec.dynamic.partition;
hive.exec.dynamic.partition=true

动态分区的使用方法很简单，假设我想向stat_date='20110728'这个分区下面插入数据，至于province插入到哪个子分区下面让数据库自己来判断，那可以这样写：

hive> insert overwrite table partition_test partition(stat_date='20110728',province)
    > select member_id,name,province from partition_test_input where stat_date='20110728';
WARNING: Hive-on-MR is deprecated in Hive 2 and may not be available in the future versions. Consider using a different execution engine (i.e. spark, tez) or using Hive 1.X releases.
Query ID = a6_20171104202556_5adab2fb-bc07-4dd9-b61c-2fb87176da63
Total jobs = 3
Launching Job 1 out of 3
Number of reduce tasks is set to 0 since there's no reduce operator
Starting Job = job_1509763925736_0010, Tracking URL = http://localhost:8088/proxy/application_1509763925736_0010/
Kill Command = /Users/a6/Applications/hadoop-2.6.5/bin/hadoop job  -kill job_1509763925736_0010
Hadoop job information for Stage-1: number of mappers: 1; number of reducers: 0
2017-11-04 20:26:05,043 Stage-1 map = 0%,  reduce = 0%
2017-11-04 20:26:11,465 Stage-1 map = 100%,  reduce = 0%
Ended Job = job_1509763925736_0010
Stage-4 is selected by condition resolver.
Stage-3 is filtered out by condition resolver.
Stage-5 is filtered out by condition resolver.
Moving data to directory hdfs://localhost:9002/user/hive/warehouse/yyz_workdb.db/partition_test/stat_date=20110728/.hive-staging_hive_2017-11-04_20-25-56_419_3312543230075088948-1/-ext-10000
Loading data to table yyz_workdb.partition_test partition (stat_date=20110728, province=null)

Loaded : 3/3 partitions.
	 Time taken to load dynamic partitions: 0.359 seconds
	 Time taken for adding to write entity : 0.001 seconds
MapReduce Jobs Launched:
Stage-Stage-1: Map: 1   HDFS Read: 5218 HDFS Write: 322 SUCCESS
Total MapReduce CPU Time Spent: 0 msec
OK
Time taken: 16.814 seconds

4.2 stat_date叫做静态分区列，province叫做动态分区列。select子句中需要把动态分区列按照分区的顺序写出来，静态分区列不用写出来。这样stat_date='20110728'的所有数据，会根据province的不同分别插入到/user/hive/warehouse/partition_test/stat_date=20110728/下面的不同的子文件夹下，如果源数据对应的province子分区不存在，则会自动创建，非常方便，而且避免了人工控制插入数据与分区的映射关系存在的潜在风险。

注意，动态分区不允许主分区采用动态列而副分区采用静态列，这样将导致所有的主分区都要创建副分区静态列所定义的分区：

hive>  insert overwrite table partition_test partition(stat_date,province='liaoning')
    > select member_id,name,province from partition_test_input where province='liaoning';
FAILED: SemanticException [Error 10094]: Line 1:49 Dynamic partition cannot be the parent of a static partition ''liaoning''
hive>

动态分区可以允许所有的分区列都是动态分区列，但是要首先设置一个参数hive.exec.dynamic.partition.mode ：

hive> set hive.exec.dynamic.partition.mode;
hive.exec.dynamic.partition.mode=strict

（5）再介绍3个参数：

5.1 默认参数释义：
hive.exec.max.dynamic.partitions.pernode （缺省值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 （缺省值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 （缺省值100000）：所有的mapreduce job允许创建的文件的最大数量
5.2 举例
当源表数据量很大时，单独一个mapreduce job中生成的数据在分区列上可能很分散，举个简单的例子，比如下面的表要用3个map：
1
1
1
2
2
2
3
3
3

如果数据这样分布，那每个mapreduce只需要创建1个分区就可以了： 
         |1
map1 --> |1 
         |1 

         |2
map2 --> |2 
         |2 

         |3
map3 --> |3 
         |3

但是如果数据按下面这样分布，那第一个mapreduce就要创建3个分区： 

         |1
map1 --> |2 
         |3 

         |1
map2 --> |2 
         |3 

         |1
map3 --> |2 
         |3
5.3 下面给出了一个报错的例子：

hive> set hive.exec.max.dynamic.partitions.pernode=4;
hive> insert overwrite table partition_test partition(stat_date,province)
> select member_id,name,stat_date,province from partition_test_input distribute by stat_date,province;
Total MapReduce jobs = 1
...
[Fatal Error] Operator FS_4 (id=4): Number of dynamic partitions exceeded hive.exec.max.dynamic.partitions.pernode.. Killing the job.
Ended Job = job_201107251641_0083 with errors
FAILED: Execution Error, return code 2 from org.apache.hadoop.hive.ql.exec.MapRedTask

为了让分区列的值相同的数据尽量在同一个mapreduce中，这样每一个mapreduce可以尽量少的产生新的文件夹，可以借助distribute by的功能，将分区列值相同的数据放到一起：

hive> insert overwrite table partition_test partition(stat_date,province)
> select member_id,name,stat_date,province from partition_test_input distribute by stat_date,province;
Total MapReduce jobs = 1
...
18 Rows loaded to partition_test
OK

参考网址：http://blog.sina.com.cn/s/blog_6ff05a2c0100tah0.html