hive中行转列、列转行的实现

行转列实现：

表信息：

test_rowcol（dt_m，vaild_num， unvaild_num）

场景一：使用concat_ws和collect_set函数

select dt_m, 'vaild_num', concat_ws(',', collect_set(vaild_num)) from test_rowcol group by dt_m
  union all
select dt_m, 'unvaild_num', concat_ws(',', collect_set(unvaild_num)) from test_rowcol group by dt_m;

说明：collect_set函数可以返回一个Array类型。

concat_ws函数可以拼接数组，如下：

select concat_ws('.', 'www', array('baidu', 'com')) from dual limit 1;

'www.baidu.com'

场景二：有时候如果需要对指标字段求和，则上述sql改写成如下：

select dt_m,'vaild_num',sum(vaild_num) from test_rowcol group by dt_m
 union all
select dt_m,'unvaild_num',sum(unvaild_num) from test_rowcol group by dt_m;

场景三：使用str_to_map和explode函数以及lateral view

select dt_m, adt.data_type, adt.data

from test_rowcol

lateral view

explode(str_to_map(concat('vaild_num=',vaild_num,'&unvaild_num=',unvaild_num),'&','=')) adt as data_type, data;

说明：str_to_map可以将字符串转换为map，如下：

select str_to_map(concat('vaild_num=',vaild_num,'&unvaild_num=',unvaild_num),'&','=') from test_rowcol;

explode是表生成函数（UDTF），入参是一个数组或者是map，将其迭代，返回多行结果，如下：

select explode(str_to_map(concat('vaild_num=',vaild_num,'&unvaild_num=',unvaild_num),'&','=')) from test_rowcol;

lateral view可以将一列拆分成多行数据。

列转行实现：

利用上面的sql，创建一张表：

create table test_colrow as

select dt_m, adt.data_type, adt.data from test_rowcol lateral view

explode(str_to_map(concat('vaild_num=',vaild_num,'&unvaild_num=',unvaild_num),'&','=')) adt as data_type, data;

场景一：

select dt_m,

sum(case when data_type='vaild_num' then data end) as vaild_num,

sum(case when data_type='unvaild_num' then data end) as unvaild_num

from test_colrow

group by dt_m;

场景二：恢复到test_rowcol表样（费了九牛二虎之力，大神们有高招请多指教）

--创建一组临时表

create table temp_vaild

as select explode(collect_set(case when data_type='vaild_num' then concat(dt_m,'&',data) end))  as vaild_col from test_colrow;

create table temp_unvaild as

select explode(collect_set(case when data_type='unvaild_num' then concat(dt_m,'&',data) end)) as unvaild_col from

test_colrow;

--利用split函数和row_number()over()且再次利用临时表

select aa.dt_m, aa.data, bb.data

from

(select split(vaild_col,'&')[0] as dt_m, split(vaild_col,'&')[1] as data, row_number()over() as num from temp_vaild) aa ,

(select split(unvaild_col,'&')[0] as dt_m, split(unvaild_col,'&')[1] as data, row_number()over() as num from temp_unvaild) bb

where aa.num=bb.num and aa.dt_m=bb.dt_m