嗯。你可以使用窗口功能：

select t.*
from (select t.*,
             sum(case when datatype = 'apparent-power' then 1 else 0 end) over (partition by unitid, timestamp) as cnt_ap,
             sum(case when datatype = 'uncalibrated-current' then 1 else 0 end) over (partition by unitid, timestamp) as cnt_uc
      from t
     ) t
where datatype = 'apparent-power' and cnt_uc > 0 or
      datatype = 'uncalibrated-current' and cnt_ap = 0;

如果我理解正确，你也可以使用 not exists :

select t.*
from t
where t.datatype = 'apparent_power' or
      not exists (select 1
                  from t t2
                  where t2.unitid = t.unitid and
                        t2.timestamp = t.timestamp and
                        t2.datatype = 'apparent_power'
                 );

我认为最好用union将问题分成两部分，all将附加两个表，这样你就可以只关注于获得所有的视在功率，然后只关注未校准的电流
然后就是过滤的任务：

(
    SELECT DISTINCT T1.* 
    FROM T AS T1
    INNER JOIN 
        (SELECT * FROM T WHERE DataType = 'uncalibrated-current') AS T2 
        ON T1.UnitID = T2.UnitID AND T1.TimeStamp = T2.TimeStamp
    WHERE T1.DataType = 'apparent-power'
)
UNION ALL
(
    SELECT DISTINCT T1.*
    FROM T AS T1
    LEFT JOIN 
        (SELECT * FROM T WHERE DataType = 'apparent-power') AS T2 
        ON T1.UnitID = T2.UnitID AND T1.TimeStamp = T2.TimeStamp
    WHERE T1.DataType = 'uncalibrated-current'
        AND T2.DataType IS NULL
)

第一个查询具有 INNER JOIN ，有效地过滤掉任何不匹配的内容。第二个查询称为反连接，过滤掉所有匹配的内容( T2.DataType IS NULL )
最后是 DISTINCT 在这两个表上，都要考虑这样一个事实：如果您的数据类型多于这2个，那么这些连接将复制行