TSQL：如果记录为空，如何从查询结果中删除最后一条记录

Question

Background: I am using a stored procedure below to retrieve hourly time series data.

SELECT *
   FROM (
      SELECT CAST(LocalDateTime as SmallDateTime) as [DateTime], DataValue, VariableID
         FROM DataValues
         WHERE SiteID = 2 and VariableID IN(1,3,30)
         ) TableDate
   PIVOT (SUM(DataValue) FOR VariableID IN ([1],[3],[30]))
   PivotTable ORDER BY [DateTime]

Usually the data is written to the database at the top of the hour, and the stored procedure is executed at 5 minutes past the hour to retrieve all existing values.

The problem is that occasionally the data retrieval on top of the hour fails for one of many sites and thus no new data is written to the DataValues table for that site. However, the most recent hour is still present in the DataValues table because other sites wrote values to it. This results in a query result that has a record for the most recent hour but all fields are NULL . The code that handles the query results can deal with hourly data that is NULL for whatever reason except in cases where the very last record is NULL for all fields.

Question: Is there a simple way to remove the last record of a query result (once pivoted per procedure above) if all fields of that record are NULL ?

Answer 1

You can do this with by using row_number() :

with cte as (
      SELECT *
      FROM (SELECT CAST(LocalDateTime as SmallDateTime) as [DateTime], DataValue, VariableID
            FROM DataValues
            WHERE SiteID = 2 and VariableID IN(1,3,30)
           ) TableDate
      PIVOT (SUM(DataValue) FOR VariableID IN ([1],[3],[30])) PivotTable
      ORDER BY [DateTime]
     )
select [DateTime], [1], [3], [30]
from (select cte.*, row_number() over (order by [DateTime] desc) as seqnum
      from cte
     ) t
where seqnum = 1 and [1] is null and [3] is null and [30] is null;

However, I think the easiest way is just to filter before the pivot :

      SELECT *
      FROM (SELECT CAST(LocalDateTime as SmallDateTime) as [DateTime], DataValue, VariableID
            FROM DataValues
            WHERE DataValue is not null
            WHERE SiteID = 2 and VariableID IN (1, 3, 30)
           ) TableDate
      PIVOT (SUM(DataValue) FOR VariableID IN ([1],[3],[30])) PivotTable
      ORDER BY [DateTime]

This has the potential disadvantage of removing other rows with NULL s in them. If that is a problem, the first method works better.