sql - 返回具有 3 个或更多遵循模式的连续值的行

Question

我尝试了 Tababitosan 方法但没有帮助。您能否针对以下情况提出其他解决方案。

使用 Oracle 11g:

请找到下表格式，

Region  Date      Value
East    1/1/2018    1
East    1/2/2018    1
East    1/3/2018    0
East    1/4/2018    1
East    1/6/2018    1
East    1/7/2018    0
West    1/9/2018    0
West    1/10/2018   0
West    2/3/2018    1
West    2/4/2018    1
West    2/5/2018    1
West    2/8/2018    0
West    2/9/2018    0
West    2/10/2018   0
West    2/11/2018   1
West    2/12/2018   0
West    2/13/2018   1
West    2/14/2018   1
West    2/17/2018   0
West    2/18/2018   0
West    2/19/2018   1
West    2/20/2018   0
West    2/21/2018   1
West    2/22/2018   0
West    2/23/2018   1

---

我的输出应包含具有三个或更多连续 1 的行以及 1 之间仅包含一个 0 的行。

注意:按地区分组并按日期排序。 日期列中的日期可能不包含所有日期。假设上面的 1/6/2018 缺失了，这很好。我需要查找按日期排序并按区域分组时遵循模式的“值”列。

East    1/1/2018    1
East    1/2/2018    1
East    1/3/2018    0
East    1/4/2018    1
East    1/6/2018    1

West    2/3/2018    1
West    2/4/2018    1
West    2/5/2018    1

West    2/11/2018   1
West    2/12/2018   0
West    2/13/2018   1
West    2/14/2018   1

West    2/19/2018   1
West    2/20/2018   0
West    2/21/2018   1
West    2/22/2018   0
West    2/23/2018   1

Answer 1

由于要求值被“屏蔽”——在某些条件下 0 变为 1，问题变得更加复杂。否则，这将是直接应用 tababitosan 方法。此外，在最终输出中，您需要原始值，而不是屏蔽值，因此我们必须小心每一步保留的内容和丢弃的内容。

这是解决该问题的一种方法。请注意，DATE 是 Oracle 关键字，因此不应将其用作列名(它可能不会引发语法错误，但会使代码更难理解)。我没有检查 VALUE 是否也是关键字；为了安全起见，我将两个列名称都更改为 DT 和 VAL。 (正如您将看到的，在我的代码中，我分别为组和计数创建列 GRP 和 CT；GROUP 和 COUNT 是 Oracle 关键字，因此同样需要考虑。)

with
  inputs(region, dt, val) as (
    select 'East', to_date('1/1/2018' , 'mm/dd/yyyy'), 1 from dual union all
    select 'East', to_date('1/2/2018' , 'mm/dd/yyyy'), 1 from dual union all
    select 'East', to_date('1/3/2018' , 'mm/dd/yyyy'), 0 from dual union all
    select 'East', to_date('1/4/2018' , 'mm/dd/yyyy'), 1 from dual union all
    select 'East', to_date('1/6/2018' , 'mm/dd/yyyy'), 1 from dual union all
    select 'East', to_date('1/7/2018' , 'mm/dd/yyyy'), 0 from dual union all
    select 'West', to_date('1/9/2018' , 'mm/dd/yyyy'), 0 from dual union all
    select 'West', to_date('1/10/2018', 'mm/dd/yyyy'), 0 from dual union all
    select 'West', to_date('2/3/2018' , 'mm/dd/yyyy'), 1 from dual union all
    select 'West', to_date('2/4/2018' , 'mm/dd/yyyy'), 1 from dual union all
    select 'West', to_date('2/5/2018' , 'mm/dd/yyyy'), 1 from dual union all
    select 'West', to_date('2/8/2018' , 'mm/dd/yyyy'), 0 from dual union all
    select 'West', to_date('2/9/2018' , 'mm/dd/yyyy'), 0 from dual union all
    select 'West', to_date('2/10/2018', 'mm/dd/yyyy'), 0 from dual union all
    select 'West', to_date('2/11/2018', 'mm/dd/yyyy'), 1 from dual union all
    select 'West', to_date('2/12/2018', 'mm/dd/yyyy'), 0 from dual union all
    select 'West', to_date('2/13/2018', 'mm/dd/yyyy'), 1 from dual union all
    select 'West', to_date('2/14/2018', 'mm/dd/yyyy'), 1 from dual union all
    select 'West', to_date('2/17/2018', 'mm/dd/yyyy'), 0 from dual union all
    select 'West', to_date('2/18/2018', 'mm/dd/yyyy'), 0 from dual union all
    select 'West', to_date('2/19/2018', 'mm/dd/yyyy'), 1 from dual union all
    select 'West', to_date('2/20/2018', 'mm/dd/yyyy'), 0 from dual union all
    select 'West', to_date('2/21/2018', 'mm/dd/yyyy'), 1 from dual union all
    select 'West', to_date('2/22/2018', 'mm/dd/yyyy'), 0 from dual union all
    select 'West', to_date('2/23/2018', 'mm/dd/yyyy'), 1 from dual
  )
, prep(region, dt, val, adj_val) as (
    select region, dt, val,
           case when val = 1
                or   lag(val)  over (partition by region order by dt) = 1
                     and
                     lead(val) over (partition by region order by dt) = 1
                then 1 else 0 end
    from   inputs
  )
, tabibitosan(region, dt, val, adj_val, grp) as (
    select region, dt, val, adj_val,
           row_number() over (partition by region order by dt)
           - row_number() over (partition by region, adj_val order by dt)
    from   prep
  )
, group_counts(region, dt, val, ct) as (
    select region, dt, val, count(*) over (partition by region, grp)
    from   tabibitosan
    where  adj_val = 1
  )
select   region, dt, val
from     group_counts
where    ct >= 3
order by region, dt
;

输出:

REGION DT               VAL
------ --------- ----------
East   01-Jan-18          1
East   02-Jan-18          1
East   03-Jan-18          0
East   04-Jan-18          1
East   06-Jan-18          1
West   03-Feb-18          1
West   04-Feb-18          1
West   05-Feb-18          1
West   11-Feb-18          1
West   12-Feb-18          0
West   13-Feb-18          1
West   14-Feb-18          1
West   19-Feb-18          1
West   20-Feb-18          0
West   21-Feb-18          1
West   22-Feb-18          0
West   23-Feb-18          1