hive sql order by 与sort by

hive> select * from lpx_partition_test;
OK
99001    xxxcompany_name1    2011-06-08    GuangDong
99001    xxxcompany_name1    2011-06-08    ZheJiang
99002    xxxcompany_name1    2011-06-08    ZheJiang
99003    xxxcom2    2011-06-08    ZheJiang
99001    xxxcompany_name1    2011-06-09    GuangDong
99001    xxxcompany_name1    2011-06-09    ZheJiang
99002    xxxcompany_name1    2011-06-09    ZheJiang
99003    xxxcom2    2011-06-09    ZheJiang
99001    xxxcompany_name1    20110608    GuangDong
99001    xxxcompany_name1    20110608    ZheJiang
99002    xxxcompany_name1    20110609    ZheJiang
99003    xxxcom2    20110609    ZheJiang
Time taken: 0.66 seconds
hive> select * from lpx_partition_test order by global_id;
Total MapReduce jobs = 1
Launching Job 1 out of 1
Number of reduce tasks determined at compile time: 1
In order to change the average load for a reducer (in bytes):
  set hive.exec.reducers.bytes.per.reducer=<number>
In order to limit the maximum number of reducers:
  set hive.exec.reducers.max=<number>
In order to set a constant number of reducers:
  set mapred.reduce.tasks=<number>
Execution log at: /tmp/root/root_20110802143737_e2a261fb-af38-45f6-9c36-8fdf7fc57d4b.log
Job running in-process (local Hadoop)
2011-08-02 14:37:46,552 null map = 0%,  reduce = 0%
2011-08-02 14:37:50,761 null map = 100%,  reduce = 0%
2011-08-02 14:37:56,129 null map = 100%,  reduce = 100%
Ended Job = job_local_0001
OK
99001    xxxcompany_name1    20110608    ZheJiang
99001    xxxcompany_name1    2011-06-09    ZheJiang
99001    xxxcompany_name1    20110608    GuangDong
99001    xxxcompany_name1    2011-06-08    ZheJiang
99001    xxxcompany_name1    2011-06-09    GuangDong
99001    xxxcompany_name1    2011-06-08    GuangDong
99002    xxxcompany_name1    2011-06-09    ZheJiang
99002    xxxcompany_name1    2011-06-08    ZheJiang
99002    xxxcompany_name1    20110609    ZheJiang
99003    xxxcom2    20110609    ZheJiang
99003    xxxcom2    2011-06-08    ZheJiang
99003    xxxcom2    2011-06-09    ZheJiang
Time taken: 16.342 seconds

hive> set hive.exec.dynamic.partition.mode=strict;
hive> select * from lpx_partition_test order by global_id limit 8;
Total MapReduce jobs = 1
Launching Job 1 out of 1
Number of reduce tasks determined at compile time: 1
In order to change the average load for a reducer (in bytes):
  set hive.exec.reducers.bytes.per.reducer=<number>
In order to limit the maximum number of reducers:
  set hive.exec.reducers.max=<number>
In order to set a constant number of reducers:
  set mapred.reduce.tasks=<number>
Execution log at: /tmp/root/root_20110802144040_8bba22bc-968f-4588-bcdc-0519c2a03115.log
Job running in-process (local Hadoop)
2011-08-02 14:40:45,156 null map = 0%,  reduce = 0%
2011-08-02 14:40:46,202 null map = 100%,  reduce = 0%
2011-08-02 14:40:49,213 null map = 100%,  reduce = 100%
Ended Job = job_local_0001
OK
99001    xxxcompany_name1    20110608    ZheJiang
99001    xxxcompany_name1    2011-06-09    ZheJiang
99001    xxxcompany_name1    20110608    GuangDong
99001    xxxcompany_name1    2011-06-08    ZheJiang
99001    xxxcompany_name1    2011-06-09    GuangDong
99001    xxxcompany_name1    2011-06-08    GuangDong
99002    xxxcompany_name1    2011-06-09    ZheJiang
99002    xxxcompany_name1    2011-06-08    ZheJiang
Time taken: 12.169 seconds

hive> set hive.exec.dynamic.partition.mode=nonstrict;
hive>  select * from lpx_partition_test order by global_id limit 8;
Total MapReduce jobs = 1
Launching Job 1 out of 1
Number of reduce tasks determined at compile time: 1
In order to change the average load for a reducer (in bytes):
  set hive.exec.reducers.bytes.per.reducer=<number>
In order to limit the maximum number of reducers:
  set hive.exec.reducers.max=<number>
In order to set a constant number of reducers:
  set mapred.reduce.tasks=<number>
Execution log at: /tmp/root/root_20110802144242_5f4ce867-4b96-4dd3-bf65-16588b47bda0.log
Job running in-process (local Hadoop)
2011-08-02 14:42:59,045 null map = 0%,  reduce = 0%
2011-08-02 14:43:02,050 null map = 100%,  reduce = 0%
2011-08-02 14:43:04,052 null map = 100%,  reduce = 100%
Ended Job = job_local_0001
OK
99001    xxxcompany_name1    20110608    ZheJiang
99001    xxxcompany_name1    2011-06-09    ZheJiang
99001    xxxcompany_name1    20110608    GuangDong
99001    xxxcompany_name1    2011-06-08    ZheJiang
99001    xxxcompany_name1    2011-06-09    GuangDong
99001    xxxcompany_name1    2011-06-08    GuangDong
99002    xxxcompany_name1    2011-06-09    ZheJiang
99002    xxxcompany_name1    2011-06-08    ZheJiang
Time taken: 12.595 seconds

在order by 子句中有一些局限，在set hive.exec.dynamic.partition.mode=strict模式下，加limit是有作用的，但是在set hive.exec.dynamic.partition.mode=nonstrict模式下limit不是必须的，因为为了输出所有结果的排序，reducer必须排序所有的结果，如果输出结果非常大，这个reducer将会话费将长的时间。

hive> select * from lpx_partition_test sort by global_id;
Total MapReduce jobs = 1
Launching Job 1 out of 1
Number of reduce tasks not specified. Estimated from input data size: 1
In order to change the average load for a reducer (in bytes):
  set hive.exec.reducers.bytes.per.reducer=<number>
In order to limit the maximum number of reducers:
  set hive.exec.reducers.max=<number>
In order to set a constant number of reducers:
  set mapred.reduce.tasks=<number>
Execution log at: /tmp/root/root_20110802144949_c3fc1b32-bb9e-40fa-af89-de4caefc30e6.log
Job running in-process (local Hadoop)
2011-08-02 14:49:41,150 null map = 0%,  reduce = 0%
2011-08-02 14:49:42,156 null map = 100%,  reduce = 0%
2011-08-02 14:49:44,180 null map = 100%,  reduce = 100%
Ended Job = job_local_0001
OK
99001    xxxcompany_name1    20110608    ZheJiang
99001    xxxcompany_name1    2011-06-09    ZheJiang
99001    xxxcompany_name1    20110608    GuangDong
99001    xxxcompany_name1    2011-06-08    ZheJiang
99001    xxxcompany_name1    2011-06-09    GuangDong
99001    xxxcompany_name1    2011-06-08    GuangDong
99002    xxxcompany_name1    2011-06-09    ZheJiang
99002    xxxcompany_name1    2011-06-08    ZheJiang
99002    xxxcompany_name1    20110609    ZheJiang
99003    xxxcom2    20110609    ZheJiang
99003    xxxcom2    2011-06-08    ZheJiang
99003    xxxcom2    2011-06-09    ZheJiang
Time taken: 10.589 seconds

sort by 子句支持进入reducer前进行排序，order by 和sort by 子句的区别在于，order by 子句保证输出结果的整体顺序，而sort by 子句保证进入reducer前就是有序的，如果有多个reducer任务，则结果可能出现部分最终有序的结果。由于这里只有一个reducerer任务，所以看起来不是很明显。