MySQL difference between two rows of a SELECT Statement
I am trying to make the difference of two rows in an mysql database.
I have this table containing ID, kilometers, date, car_id, car_driver etc...
Since I don't always enter the information in the table in the correct order, I may end up with information like this:
ID | Kilometers | date | car_id | car_driver | ...
1 | 100 | 2012-05-04 | 1 | 1
2 | 200 | 2012-05-08 | 1 | 1
3 | 1000 | 2012-05-25 | 1 | 1
4 | 600 | 2012-05-16 | 1 | 1
With a select statement I am able to sort my table correctly:
SELECT * FROM mytable ORDER BY car_driver ASC, car_id ASC, date ASC
I will obtain this:
ID | Kilometers | date | car_id | car_driver | ...
1 | 100 | 2012-05-04 | 1 | 1
2 | 200 | 2012-05-08 | 1 | 1
4 | 600 | 2012-05-16 | 1 | 1
3 | 1000 | 2012-05-25 | 1 | 1
Now I would like to make a view where basically I have this extra information: Number of kilometers since last date and I would like to obtain something like this:
ID | Kilometers | date | car_id | car_driver | number_km_since_last_date
1 | 100 | 2012-05-04 | 1 | 1 | 0
2 | 200 | 2012-05-08 | 1 | 1 | 100
4 | 600 | 2012-05-16 | 1 | 1 | 400
3 | 1000 | 2012-05-25 | 1 | 1 | 400
I thought of doing an INNER JOIN to perform what I wanted, but I have the feeling I can't do the join on my ID since they are not sorted correctly.
Is there a way to achieve what I want?
Shall I create a view with a sort of row_number that I can then used in my INNER JOIN?
Solution 1:
SELECT
mt1.ID,
mt1.Kilometers,
mt1.date,
mt1.Kilometers - IFNULL(mt2.Kilometers, 0) AS number_km_since_last_date
FROM
myTable mt1
LEFT JOIN myTable mt2
ON mt2.Date = (
SELECT MAX(Date)
FROM myTable mt3
WHERE mt3.Date < mt1.Date
)
ORDER BY mt1.date
Sql Fiddle
Or, by emulating a lag()
function through MySql hackiness...
SET @kilo=0;
SELECT
mt1.ID,
mt1.Kilometers - @kilo AS number_km_since_last_date,
@kilo := mt1.Kilometers Kilometers,
mt1.date
FROM myTable mt1
ORDER BY mt1.date
Sql Fiddle
Solution 2:
In Postgres, Oracle and SQL-Server 2012, this is plain simple, using the LAG()
function:
SELECT
id, kilometers, date,
kilometers
- COALESCE( LAG(kilometers) OVER (ORDER BY date ASC, car_driver ASC, id ASC)
, kilometers)
AS number_km_since_last_date
FROM
mytable ;
In MySQL, we have to do some nasty constructions. Either an inline subquery (with probably not very good performance):
SELECT
id, kilometers, date,
kilometers - COALESCE(
( SELECT p.kilometers
FROM mytable AS p
WHERE ( p.date = m.date AND p.car_driver = m.car_driver
AND p.id < m.id
OR p.date = m.date AND p.car_driver < m.car_driver
OR p.date < m.date
)
ORDER BY p.date DESC, p.car_driver DESC
LIMIT 1
), kilometers)
AS number_km_since_last_date
FROM
mytable AS m ;
or a self-join (already provided by @Michael Fredrickson) or using MySQL variables (already provided as well).
If you want the counter to start again from 0 for every car_id
, which would be done with PARTITION BY
in many other DBMS:
SELECT
id, kilometers, date,
kilometers
- COALESCE( LAG(kilometers) OVER (PARTITION BY car_id
ORDER BY date ASC, car_driver ASC, id ASC)
, kilometers)
AS number_km_since_last_date
FROM
mytable ;
it could be done in MySQL like this:
SELECT
id, kilometers, date,
kilometers - COALESCE(
( SELECT p.kilometers
FROM mytable AS p
WHERE p.car_id = m.car_id
AND ( p.date = m.date AND p.car_driver = m.car_driver
AND p.id < m.id
OR p.date = m.date AND p.car_driver < m.car_driver
OR p.date < m.date
)
ORDER BY p.date DESC, p.car_driver DESC
LIMIT 1
), kilometers)
AS number_km_since_last_date
FROM
mytable AS m ;