如何insert/copy一个分区的数据到hive中的多个分区?
How to insert/copy one partition's data to multiple partitions in hive?
我的配置单元 table 中有 day='2019-01-01'
的数据,我想将相同的数据复制到整个 Jan-2019 月份。 (即 '2019-01-02'
、'2019-01-03'
...'2019-01-31'
)
我正在尝试关注,但数据仅插入到“2019-01-02”中,而不是“2019-01-03”中。
INSERT OVERWRITE TABLE db_t.students PARTITION(dt='2019-01-02', dt='2019-01-03')
SELECT id, name, marks FROM db_t.students WHERE dt='2019-01-01';
将您的所有数据与所需日期范围内的日历日期交叉联接。使用动态分区:
set hivevar:start_date=2019-01-02;
set hivevar:end_date=2019-01-31;
set hive.exec.dynamic.partition=true;
set hive.exec.dynamic.partition.mode=nonstrict;
with date_range as
(--this query generates date range
select date_add ('${hivevar:start_date}',s.i) as dt
from ( select posexplode(split(space(datediff('${hivevar:end_date}','${hivevar:start_date}')),' ')) as (i,x) ) s
)
INSERT OVERWRITE TABLE db_t.students PARTITION(dt)
SELECT id, name, marks, r.dt --partition column is the last one
FROM db_t.students s
CROSS JOIN date_range r
WHERE s.dt='2019-01-01'
DISTRIBUTE BY r.dt;
另一种可能的解决方案是使用 hadoop fs -cp
或 hadoop distcp
复制分区数据(对每个分区重复或在 shell 中使用循环):
hadoop fs -cp '/usr/warehouse/students/dt=2019-01-01' '/usr/warehouse/students/dt=2019-01-02'
还有一个使用 UNION ALL 的解决方案:
set hive.exec.dynamic.partition=true;
set hive.exec.dynamic.partition.mode=nonstrict;
INSERT OVERWRITE TABLE db_t.students PARTITION(dt)
SELECT id, name, marks, '2019-01-02' as dt FROM db_t.students s WHERE s.dt='2019-01-01'
UNION ALL
SELECT id, name, marks, '2019-01-03' as dt FROM db_t.students s WHERE s.dt='2019-01-01'
UNION ALL
SELECT id, name, marks, '2019-01-04' as dt FROM db_t.students s WHERE s.dt='2019-01-01'
UNION ALL
...
;
我的配置单元 table 中有 day='2019-01-01'
的数据,我想将相同的数据复制到整个 Jan-2019 月份。 (即 '2019-01-02'
、'2019-01-03'
...'2019-01-31'
)
我正在尝试关注,但数据仅插入到“2019-01-02”中,而不是“2019-01-03”中。
INSERT OVERWRITE TABLE db_t.students PARTITION(dt='2019-01-02', dt='2019-01-03')
SELECT id, name, marks FROM db_t.students WHERE dt='2019-01-01';
将您的所有数据与所需日期范围内的日历日期交叉联接。使用动态分区:
set hivevar:start_date=2019-01-02;
set hivevar:end_date=2019-01-31;
set hive.exec.dynamic.partition=true;
set hive.exec.dynamic.partition.mode=nonstrict;
with date_range as
(--this query generates date range
select date_add ('${hivevar:start_date}',s.i) as dt
from ( select posexplode(split(space(datediff('${hivevar:end_date}','${hivevar:start_date}')),' ')) as (i,x) ) s
)
INSERT OVERWRITE TABLE db_t.students PARTITION(dt)
SELECT id, name, marks, r.dt --partition column is the last one
FROM db_t.students s
CROSS JOIN date_range r
WHERE s.dt='2019-01-01'
DISTRIBUTE BY r.dt;
另一种可能的解决方案是使用 hadoop fs -cp
或 hadoop distcp
复制分区数据(对每个分区重复或在 shell 中使用循环):
hadoop fs -cp '/usr/warehouse/students/dt=2019-01-01' '/usr/warehouse/students/dt=2019-01-02'
还有一个使用 UNION ALL 的解决方案:
set hive.exec.dynamic.partition=true;
set hive.exec.dynamic.partition.mode=nonstrict;
INSERT OVERWRITE TABLE db_t.students PARTITION(dt)
SELECT id, name, marks, '2019-01-02' as dt FROM db_t.students s WHERE s.dt='2019-01-01'
UNION ALL
SELECT id, name, marks, '2019-01-03' as dt FROM db_t.students s WHERE s.dt='2019-01-01'
UNION ALL
SELECT id, name, marks, '2019-01-04' as dt FROM db_t.students s WHERE s.dt='2019-01-01'
UNION ALL
...
;