Sum of time difference between rows

问题

I have a table which records every status change of an entity

id      recordTime        Status
ID1 2014-03-01 11:33:00 Disconnected  
ID1 2014-03-01 12:13:00 Connected  
ID2 2014-03-01 12:21:00 Connected  
ID1 2014-03-01 12:24:00 Disconnected  
ID1 2014-03-01 12:29:00 Connected  
ID2 2014-03-01 12:40:00 Disconnected  
ID2 2014-03-01 13:03:00 Connected  
ID2 2014-03-01 13:13:00 Disconnected  
ID2 2014-03-01 13:29:00 Connected  
ID1 2014-03-01 13:30:00 Disconnected

I need to calculate the total inactive time i.e time between 'Connected' and last 'Disconnected' status per ID for a given time window.

For above table and time range of 2014-03-01 11:00:00 to 2014-03-01 14:00:00 the output should be:

ID  InactiveTime
ID1  01:15:00
ID2  02:00:00

回答1:

The special difficulty is not to miss the time spans to the outer time frame.
Assuming that the next row for any given id always has the opposite status.
Using the column name ts instead of recordTime:

WITH span AS (
   SELECT '2014-03-01 13:00'::timestamp AS s_from  -- start of time range
        , '2014-03-01 14:00'::timestamp AS s_to    -- end of time range
   )
, cte AS (
   SELECT id, ts, status, s_to
        , lead(ts, 1, s_from) OVER w AS span_start
        , first_value(ts)     OVER w AS last_ts
   FROM   span s
   JOIN   tbl  t ON t.ts BETWEEN s.s_from AND s.s_to
   WINDOW w AS (PARTITION BY id ORDER BY ts DESC)
   )
SELECT id, sum(time_disconnected)::text AS total_disconnected
FROM  (
   SELECT id, ts - span_start AS time_disconnected
   FROM   cte
   WHERE  status = 'Connected'

   UNION  ALL  
   SELECT id, s_to - ts
   FROM   cte
   WHERE  status = 'Disconnected'
   AND    ts = last_ts
   ) sub
GROUP  BY 1
ORDER  BY 1;

Returns intervals as requested.
IDs without entries in the selected time range don't show up. You would have to query them additionally.

SQL Fiddle.
Note: I cast the resulting total_disconnected to text in the fiddle, because the type interval is displayed in a terrible format.

Add IDs without entry in the selected time frame

^{Per request in comment.}
Add to the query above (before the final ORDER BY 1):

...
UNION  ALL
SELECT id, total_disconnected
   FROM  (
   SELECT DISTINCT ON (id)
          t.id, t.status, (s.s_to - s.s_from)::text AS total_disconnected
   FROM   span     s
   JOIN   tbl      t ON t.ts < s.s_from  -- only from before time range
   LEFT   JOIN cte c USING (id)
   WHERE  c.id IS NULL         -- not represented in selected time frame
   ORDER  BY t.id, t.ts DESC   -- only the latest entry
   ) sub
WHERE  status = 'Disconnected' -- only if disconnected
ORDER  BY 1;

SQL Fiddle.

Now, only IDs without entries in or before the selected time range don't show up.

回答2:

This is how I understand your question SQL Fiddle

select id, sum(diff) as inactive
from (
    select
        recordtime,
        recordTime -
            lag(recordTime, 1, recordTime)
            over(
                partition by id
                order by recordTime
            )
        as diff,
        status,
        id
    from t
) s
where status = 'Connected'
group by id
order by id
;
 id | inactive 
----+----------
  1 | 00:45:00
  2 | 00:39:00

Could you explain your desired output?

回答3:

select id , sum(diff) inactif_time
from
(
SELECT id, "recordTime", "Status" ,LEAD("recordTime") OVER(PARTITION BY id order by "recordTime" ),LEAD("recordTime") OVER(PARTITION BY id order by "recordTime" ) - "recordTime" diff
  FROM my_table
) B 
where "Status" = 'Disconnected'
group by id

But it outputs:

"ID1";"00:45:00"
"ID2";"00:39:00"

来源：https://stackoverflow.com/questions/22114645/sum-of-time-difference-between-rows