在第一个未找到的列中查找值

Question

我必须编写一个 SQL，其中我必须编写 SQL 来计算 RunID 第一次没有看到。让我用一个例子来解释一下

例如：

RunID | RunDate    | ErrorID
----- | ---------- | ---------
101   | 04/11/2017 | 1
101   | 04/11/2017 | 2
101   | 04/11/2017 | 3
102   | 04/22/2017 | 2
102   | 04/22/2017 | 3
103   | 04/26/2017 | 1
104   | 04/27/2017 | 3
105   | 04/28/2017 | 4

在上面的例子中，RunID 101 有错误 1,2,3。 RunID 102 有 2,3。在第二次运行期间找不到 ErrorID 1。所以，RunID 到现在才第一次看到这里是 102 但是在 RunID 103 中再次找到了 ErrorID 1，最后在 RunID 104 中找不到 ErrorID 1。查询应该给出第一次找不到的 RunID，如 104。

我尝试过使用一些窗口函数，例如超前和滞后，但没有帮助。

以下是预期结果：

第一次没有出现 ErrorID 的日期：2

RunID | RunDate    | ErrorID
----- | ---------- | ---------
103   | 04/26/2017 | 2

因为在 RunID-102 之后从未见过 ErrorID 2（第一个实例未见过）

第一次没有出现 ErrorID 的日期：1

RunID | RunDate    | ErrorID
----- | ---------- | ---------
104   | 04/27/2017 | 1

在 RunID-104 之后从未见过 ErrorID 1

第一次没有出现 ErrorID 的日期：3

RunID | RunDate    | ErrorID
----- | ---------- | ---------
105   | 04/28/2017 | 3

在 RunID-105 之后从未见过 ErrorID 3

Answer 1

--Get the last runDate when an errorID was seen
with t1 as (select runId,runDate,errorID
            ,first_value(runDate) over(partition by errorID order by runDate desc 
                                       rows between unbounded preceding and unbounded following) as last_seen
            from tablename
           )
--Get the next runDate based on the previous result
,t2 as (select runid,errorID,runDate
        ,(select min(runDate) from t1 t11 where t11.runDate>t1.last_seen) as date_first_not_seen
        from t1
       )
--Join it to the original table to get the runID information from that runDate in the previous result (t2)
select distinct t.runid,t2.errorid,t.rundate
from t2
join tablename t on t.rundate=t2.date_first_not_seen

或

with t1 as (select runId,runDate,errorID
            ,first_value(runDate) over(partition by errorID order by runDate desc 
                                       rows between unbounded preceding and unbounded following) as last_seen
            from tablename)
select distinct
 t1.errorid
,first_value(t.runDate) over(partition by t1.errorID order by t1.runDate desc rows between unbounded preceding and unbounded following) as rundate
,first_value(t.runID) over(partition by t1.errorID order by t1.runDate desc rows between unbounded preceding and unbounded following) as runid
from t1
join tablename t on t.runDate>t1.last_seen

Sample Demo

Answer 2

so=> with l as (
with m as (
  select distinct max(runid) over(partition by errorid),errorid
  from so80
)
, a as (
  select distinct runid,errorid
  from so80
)
select distinct min(a.runid) over (partition by m.errorid),m.errorid
from m
join a on m.max < a.runid
)
select s.*
from so80 s
join l on l.min=s.runid and s.errorid=s.errorid

;
 runid |   rundate    | errorid
-------+--------------+---------
   104 |  04/27/2017  |       3
   103 |  04/26/2017  |       1
   105 |  04/28/2017  |       4
(3 rows)