如何在一个sqlite命令中将多个行项目作为不同的列

Question

I have a table looks like this:

Timestamp  direction  John  Mary  car   airplan

 4960        1        1   null   1     null
 5960        2        1     1   null    1  
 6960        1      null  null   1      1
 7960        1        1   null  null    1
 8960        2        1   null  null   null

the record means : john/mary/together go/back to xxx by car/airplan at timestamp and I wish to have one single sqlite query to get the result like this:

          John_t          Mary_t       car_t      airplan_t
          0                0           6960        7960
          8960            5960            0          0

which means the latest timestmap of (john, mary, car, airplan ) (go/back) to xxx.

I try the sql command , but failed...

select
(case when John=1 then max(Timestamp) else 0 end) as John_t, 
(case when Mary=1 then max(Timestamp) else 0 end) as Mary_t,
(case when car=1 then max(Timestamp) else 0 end) as car_t,
(case when airplan=1 then max(Time) else 0 end) as airplan_t
 from table where 4960 < Time < 8960 and group by direction

can anyone help?

Answer 1

Based on the comment by Andriy M , here is the answer:

select John_t, Mary_t, car_t, airplan_t from
    -- John --
    (select direction, (case when Timestamp=max_Timestamp then Timestamp else 0 end) as John_t from
        (
        (select dir.direction as direction, max(Timestamp) as Timestamp from
            (select distinct direction from trips) as dir
            left outer join
            (select * from trips where John = 1) as t
            on dir.direction = t.direction
            group by direction
        )
        cross join
        (select max(Timestamp) as max_Timestamp from trips where John = 1)
        )
    )
    natural join
    -- Mary --
    (select direction, (case when Timestamp=max_Timestamp then Timestamp else 0 end) as Mary_t from
        (
        (select dir.direction as direction, max(Timestamp) as Timestamp from
            (select distinct direction from trips) as dir
            left outer join
            (select * from trips where Mary = 1) as t
            on dir.direction = t.direction
            group by direction
        )
        cross join
        (select max(Timestamp) as max_Timestamp from trips where Mary = 1)
        )
    )
    natural join
    -- car --
    (select direction, (case when Timestamp=max_Timestamp then Timestamp else 0 end) as car_t from
        (
        (select dir.direction as direction, max(Timestamp) as Timestamp from
            (select distinct direction from trips) as dir
            left outer join
            (select * from trips where car = 1) as t
            on dir.direction = t.direction
            group by direction
        )
        cross join
        (select max(Timestamp) as max_Timestamp from trips where car = 1)
        )
    )
    natural join
    -- airplan --
    (select direction, (case when Timestamp=max_Timestamp then Timestamp else 0 end) as airplan_t from
        (
        (select dir.direction as direction, max(Timestamp) as Timestamp from
            (select distinct direction from trips) as dir
            left outer join
            (select * from trips where airplan = 1) as t
            on dir.direction = t.direction
            group by direction
        )
        cross join
        (select max(Timestamp) as max_Timestamp from trips where airplan = 1)
        )
    )
;

A brief explaination (from inner selects to outer):

1. We left outer join all possible directions with all "trips" of John/Mary/car/airplan. This ensures that there is a row for each direction for each John, Mary, car and airplan. This is needed to fill in 0 later even if there was no "trip" in that direction.
2. We "reduce" this to the maximum timestamp ( max(Timestamp) ) for each direction ( group by direction ) since we are not interested in earlier trips in the same direction.

3. The result is cross joined with the overall maximum timestamp of John/Mary/car/airplan to allow replacing the lower timestamps with 0 using

    (case when Timestamp=max_Timestamp then Timestamp else 0) 

4. We do this for each John, Mary, car and airplan and finally natural join everything together.