hive-sql求最大连续值

• 原始测试数据准备

• 创建测试表

create table if not exists tmp.rows_preced_test_20210526
(
name string  comment '用户名'
,seq_num int comment '使用产品编码'
)partitioned by (pt string comment 'YYMMDD数据入库时间')
ROW FORMAT DELIMITED FIELDS TERMINATED BY '\t' STORED AS orcfile

• 结果计算

select name, max(l_cnt) as m_cnt from 
 (
  select name ,continue ,count(1) as l_cnt
    from 
      (
         select name
         ,seq_num
         -- 伪列排序与原始值差值的绝对值
         ,abs(row_number() over(partition by name order by  name,seq_num  ) - seq_num ) as continue 
         from 
                  (  -- 对原始数据初步处理， 求次数无需group by  , 重复数据算一次需groupby 去重
                   select name,seq_num 
                     from  normal_stg_tmp.rows_preced_test_20210526 
                       group by    name,seq_num 
                ) a 
      ) a group by name,continue
 ) b group by name  

• 预计过程结果

对 伪列与原始值差值 计算每个name 下的count(1) 最大值可得连续最大值
可得