hive中补充月份和日期，使得月份和日期连续

m0_377595902023-07-25 23:38

补充月份，这个需求是计算每个小组前三个月的平均产能，前三个月指的是都是自然月，比如现在是2023年7月，所以现在计算的前三个月的平均产能就是2023年4月到6月的平均产能，但是这些月份可能不是连续的

方法一：把日期补齐

bash 复制代码

WITH t1 AS (
    SELECT group_id,
           group_name,
           all_pert,
           stat_month
    FROM dws_group_user_month_stat
    WHERE dt = '2023-07-10'
    and group_name is not null
    and    group_id = '00a62545-e66f-4961-9f39-8f4b5c894d19'
)
,months_table AS (
    SELECT group_name,
           group_id,
           min_month,
           max_month,
           add_months(min_month, i) as cur_month
    FROM (
        SELECT group_name,
               group_id,
               min(stat_month) AS min_month,
               max(stat_month) AS max_month
        FROM t1
--         WHERE group_id = '508b3e79-6e9a-446a-ac29-1f4428a006af'
        GROUP BY group_name, group_id
    ) t1
    LATERAL VIEW posexplode(split(space(cast(months_between(max_month, min_month) as int)), ' ')) pe AS i, num
)
select group_name,
       group_id,
       cur_month,
       min_month,
       max_month,
       all_pert,
        AVG(all_pert) OVER (PARTITION BY group_name, group_id ORDER BY cur_month ROWS BETWEEN 3 PRECEDING AND 1 PRECEDING) AS average_sales
from (
         SELECT months_table.group_name,
                months_table.group_id,
                cur_month,
                min_month,
                max_month,
                nvl(all_pert, 0) all_pert
         FROM t1
        right JOIN months_table
          ON t1.group_name = months_table.group_name
       and t1.stat_month = cur_month
     )t1;

方法二：

将数据炸开成三份

bash 复制代码

with t1 AS (
select
       month_diff,
       stat_month,
       all_pert,
       group_name
from dws_group_month_stat lateral view explode(array(1,2,3)) tmp as month_diff
where dt = '${datestr}'
)
   ,t2 as (
    select all_pert,
           t1.stat_month,
           group_name,
           add_months(stat_month,month_diff)  cur_month
    from t1
)

insert overwrite table dws_group_three_month_stat  partition (dt='${datestr}')
select
    group_name,
    cur_month,
    three_month_pert/3  three_month_pert
from (
         select  cur_month,
                sum(all_pert)  three_month_pert,
                group_name
         from t2
         group by cur_month,group_name
     )t3;"

补齐日期：

bash 复制代码

select staff_id,
           start_date,
            end_date,
           date_add(start_date, i) real_date,
           dayofweek(date_add(start_date, i)) real_week,
           code
    from t3
     LATERAL VIEW posexplode(split(space(DATEDIFF(end_date, start_date)), ' ')) pe AS i, num
    group by staff_id, start_date, end_date, i, code, date_add(start_date, i),dayofweek(date_add(start_date, i)),code

这里使用的是datediff