Tricky SQL query - need to get time frames [closed] - mysql

Question

It's difficult to tell what is being asked here. This question is ambiguous, vague, incomplete, overly broad, or rhetorical and cannot be reasonably answered in its current form. For help clarifying this question so that it can be reopened, visit the help center.
Closed 10 years ago.
I am stumbled upon a problem, when I need a query which will produce a list of speeding time frames.
Here is the data example
[idgps_unit_location] [dt] [idgps_unit] [lat] [long] [speed_kmh]
26 10/18/2012 18:53 2 47 56 30
27 10/18/2012 18:53 2 49 58 31
28 10/18/2012 18:53 2 28 37 15
29 10/18/2012 18:54 2 56 65 33
30 10/18/2012 18:54 2 152 161 73
31 10/18/2012 18:55 2 134 143 64
32 10/18/2012 18:56 2 22 31 12
36 10/18/2012 18:59 2 98 107 47
37 10/18/2012 18:59 2 122 131 58
38 10/18/2012 18:59 2 91 100 44
39 10/18/2012 19:00 2 190 199 98
40 10/18/2012 19:01 2 194 203 101
41 10/18/2012 19:02 2 182 191 91
42 10/18/2012 19:03 2 162 171 78
43 10/18/2012 19:03 2 174 183 83
44 10/18/2012 19:04 2 170 179 81
45 10/18/2012 19:05 2 189 198 97
46 10/18/2012 19:06 2 20 29 10
47 10/18/2012 19:07 2 158 167 76
48 10/18/2012 19:08 2 135 144 64
49 10/18/2012 19:08 2 166 175 79
50 10/18/2012 19:09 2 9 18 5
51 10/18/2012 19:09 2 101 110 48
52 10/18/2012 19:09 2 10 19 7
53 10/18/2012 19:10 2 32 41 20
54 10/18/2012 19:10 1 54 63 85
55 10/19/2012 19:11 2 55 64 50
I need a query that would convert this table into the following report that shows frames of time when speed was >80:
[idgps_unit] [dt_start] [lat_start] [long_start] [speed_start] [dt_end] [lat_end] [long_end] [speed_end] [speed_average]
2 10/18/2012 19:00 190 199 98 10/18/2012 19:02 182 191 91 96.66666667
2 10/18/2012 19:03 174 183 83 10/18/2012 19:05 189 198 97 87
1 10/18/2012 19:10 54 63 85 10/18/2012 19:10 54 63 85 85
Now, what have I tried? I tried putting this into separate tables, queries and do some joins... Nothing works and I am very frustrated... I am not even sure if this could be done via the query. Asking for the expert help!

Answer 1

You were right, it is fairly tricky, but I think I've managed it:
SELECT s.idgps_unit,
MIN(s.dt) AS DT_Start,
MIN(CASE WHEN s.RowNumber = 1 THEN s.Lat END) AS Lat_Start,
MIN(CASE WHEN s.RowNumber = 1 THEN s.Long END) AS Long_Start,
MIN(CASE WHEN s.RowNumber = 1 THEN s.Speed_kmh END) AS Speed_Start,
MAX(s.dt) AS dt_end,
MIN(CASE WHEN s.RowNumber = MaxRowNumber THEN s.Lat END) AS Lat_End,
MIN(CASE WHEN s.RowNumber = MaxRowNumber THEN s.Long END) AS Long_End,
MIN(CASE WHEN s.RowNumber = MaxRowNumber THEN s.Speed_kmh END) AS Speed_End,
AVG(Speed_kmh) AS Speed_Average
FROM ( SELECT T.*,
#i:= CASE WHEN Speed_Kmh > 80 AND #b = 0 THEN #i + 1 ELSE #i END AS IntervalID,
#r:= CASE WHEN Speed_Kmh > 80 AND #b = 0 THEN 1 ELSE #r + 1 END AS RowNumber,
#b:= CASE WHEN Speed_Kmh> 80 THEN 1 ELSE 0 END AS IntervalCheck
FROM T,
(SELECT #i:= 0) i,
(SELECT #r:= 0) r,
(SELECT #b:= 0) b
ORDER BY dt, idgps_unit_location
) s
INNER JOIN
( SELECT IntervalID, MAX(RowNumber) AS MaxRowNumber
FROM ( SELECT T.*,
#i:= CASE WHEN Speed_Kmh > 80 AND #b = 0 THEN #i + 1 ELSE #i END AS IntervalID,
#r:= CASE WHEN Speed_Kmh > 80 AND #b = 0 THEN 1 ELSE #r + 1 END AS RowNumber,
#b:= CASE WHEN Speed_Kmh> 80 THEN 1 ELSE 0 END AS IntervalCheck
FROM T,
(SELECT #i:= 0) i,
(SELECT #r:= 0) r,
(SELECT #b:= 0) b
ORDER BY dt, idgps_unit_location
) d
WHERE IntervalCheck = 1
GROUP BY IntervalID
) MaxInt
ON MaxInt.IntervalID = s.IntervalID
WHERE s.IntervalCheck = 1
GROUP BY s.IntervalID, s.idgps_unit;
The key is in this part:
SELECT T.*,
#i:= CASE WHEN Speed_Kmh > 80 AND #b = 0 THEN #i + 1 ELSE #i END AS IntervalID,
#r:= CASE WHEN Speed_Kmh > 80 AND #b = 0 THEN 1 ELSE #r + 1 END AS RowNumber,
#b:= CASE WHEN Speed_Kmh> 80 THEN 1 ELSE 0 END AS IntervalCheck
FROM T,
(SELECT #i:= 0) i,
(SELECT #r:= 0) r,
(SELECT #b:= 0) b
ORDER BY dt, idgps_unit_location
Each time a row is encountered where the speed is over it sets the variable #b to 1, if this variable was 0 before it assigns the row a new intervalID, if it does this it begins numbering the row at 1 again, so you end up with something like this:
[idgps_unit_location] [dt] [idgps_unit] [lat] [long] [speed_kmh] [IntervalID] RowNumber IntervalCheck
37 10/18/2012 18:59 2 122 131 58 1 1 0
38 10/18/2012 18:59 2 91 100 44 1 2 0
39 10/18/2012 19:00 2 190 199 98 2 1 1
40 10/18/2012 19:01 2 194 203 101 2 2 1
41 10/18/2012 19:02 2 182 191 91 2 3 1
42 10/18/2012 19:03 2 162 171 78 2 4 0
43 10/18/2012 19:03 2 174 183 83 3 1 1
You then need to elimate all rows where the speed is under 80 (WHERE IntervalCheck = 1), and finally you can use aggregate functions along with CASE to find the rows where RowNumber is 1 (the first row of speeding), or the highest rownumber for that interval (the last row of speeding). The join at the end simply repeats the process to find what the maximum rownumber is for each intervalID.
Example on SQL Fiddle

Answer 2

Have you tried something like this (omitting the average speed calculation):
SELECT * FROM (
SELECT
start.idgps_unit,
start.dt dt_start,
...
end.dt dt_end,
...
(...) average_speed
FROM
your_table start,
your_table end
WHERE
start.dt < end.dt
)
WHERE average_speed > 80
This will get you a lot of overlapping timeframes, not sure whether this is desired or not. If not, you could filter with NOT EXISTS:
SELECT *
FROM (query_above) timeframes
WHERE NOT EXISTS (
SELECT *
FROM (query_above) longer_timeframes
WHERE
longer_timeframes.dt_start < timeframes.dt_end OR
longer_timeframes.dt_end > timeframes.dt_end
)
This might still get you some overlap, e.g. if you go 60 from 19:00 to 19:03, 100 from 19:03 to 19:07, and again 60 from 19:07 to 19:10. Then you have two maximum-length time intervals in which average speed was greater than 80, one from 19:00 to 19:07, the other from 19:03 to 19:10.