Question

下面的查询有效，但我想知道是否有更好的方法。

有一个子查询使用两个子查询。这两个子查询是相同的，但返回两个不同的字段。有没有办法只使用一个返回两个字段的子查询？

我检查了类似的问题（this，this和this），但我不认为它们适用于这种情况。

以下是查询：

SELECT *,
       time(strftime('%s', EndTime) - strftime('%s', StartTime), 'unixepoch') AS Duration
  FROM (
           SELECT (
                      SELECT Time
                        FROM Log AS LogStart
                       WHERE LogStart.User = Log.User AND 
                             LogStart.Time <= Log.Time AND 
                             LogStart.[Action] != 'done'
                       ORDER BY LogStart.Time DESC
                       LIMIT 1
                  )
                  AS StartTime,
                  Time AS EndTime,
                  User,
                  (
                      SELECT [Action]
                        FROM Log AS LogStart
                       WHERE LogStart.User = Log.User AND 
                             LogStart.Time <= Log.Time AND 
                             LogStart.[Action] != 'done'
                       ORDER BY LogStart.Time DESC
                       LIMIT 1
                  )
                  AS [Action]
             FROM Log
            WHERE [Action] = 'done'
       )
 ORDER BY duration DESC;

以下是一些测试数据：

CREATE TABLE Log (
    Time     DATETIME,
    User     CHAR,
    [Action] CHAR
);

insert into Log values('2017-01-01 10:00:00', 'Joe', 'Play');
insert into Log values('2017-01-01 10:01:00', 'Joe', 'done');
insert into Log values('2017-01-01 10:02:00', 'Joe', 'Sing');
insert into Log values('2017-01-01 10:03:00', 'Joe', 'done');
insert into Log values('2017-01-01 10:04:00', 'Ann', 'Play');
insert into Log values('2017-01-01 10:04:30', 'Bob', 'Action without corresponding "done" which must be ignored');
insert into Log values('2017-01-01 10:05:00', 'Joe', 'Play');
insert into Log values('2017-01-01 10:06:00', 'Ann', 'done');
insert into Log values('2017-01-01 10:07:00', 'Joe', 'done');
insert into Log values('2017-01-01 10:08:00', 'Ann', 'Play');
insert into Log values('2017-01-01 10:09:00', 'Ann', 'done');

Answer 1

使用自联接...我没有SQLLite，所以语法可能在这里，但你应该明白...

Select e.*, time(strftime('%s', e.[Time]) - 
           strftime('%s', s.[Time]), 'unixepoch') AS Duration
From log e join log s  -- s is for the startevent; e for end event
    on s.[User] = e.[User]
       and s.[Action] != 'done'
       and e.[Action] = 'done'
       and s.[Time] = 
          (Select Max([time] from log
           where [User] = e.[User]
                   and [time] <= e.[Time]
                   and [Action] != 'done')

Answer 2

我做到了：你在寻找什么？（ MSSQL，但我认为应该在SQLLite中工作，因为没有“非标准”SQL命令;我应该在[] 中写入USER。）

SELECT STARTTIME, MIN(ENDTIME) AS ENDTIME, [USER], ACTION 
FROM (
SELECT B.TIME AS STARTTIME, A.TIME AS ENDTIME, A.[USER], B.ACTION
FROM LOG A
INNER JOIN (SELECT * FROM LOG) B ON A.[USER]= B.[USER] AND B.ACTION<>'done' AND B.TIME< A.TIME
WHERE A.Action='done'
) X 
GROUP BY X.STARTTIME, [USER], ACTION ;

输出：

STARTTIME               ENDTIME                 USER       ACTION
----------------------- ----------------------- ---------- ---------------------
2017-01-01 10:00:00.000 2017-01-01 10:01:00.000 Joe        Play
2017-01-01 10:02:00.000 2017-01-01 10:03:00.000 Joe        Sing
2017-01-01 10:04:00.000 2017-01-01 10:06:00.000 Ann        Play
2017-01-01 10:05:00.000 2017-01-01 10:07:00.000 Joe        Play
2017-01-01 10:08:00.000 2017-01-01 10:09:00.000 Ann        Play

仅使用您的数据，比较MSSQL中的执行计划显示您的查询“成本”为87％，而后者成本为13％（他们的总和当然是100％）

如何从子查询返回两个字段

2 个答案: