SQL subtract two rows based on date and another column

Question

I need to subtract two rows in MySQL by using the most recent date from the previous date:

Starting table:

Stock       Date          Price
GOOG        2012-05-20    402
GOOG        2012-05-21    432
APPL        2012-05-20    553
APPL        2012-05-21    590
FB          2012-05-20     40
FB          2012-05-21     34

This is the result table:

Stock       Date          Price Change
GOOG        2012-05-21    30
APPL        2012-05-21    37
FB          2012-05-21    -6

Right now I just have two dates per company, but bonus upvotes if your query can handle any number of dates.

Answer 1

What I've done was I get two separate queries to get each stock's maximum date and minimum date. Try this,

SELECT  maxList.stock, 
        maxList.Date,
        (maxlist.Price - minlist.Price) totalPrice
FROM
    (
        SELECT  a.*
        FROM    tableName a INNER JOIN
        (
            SELECT      Stock, MAX(date) maxDate
            FROM        tableName
            GROUP BY    Stock
        ) b ON  a.stock = b.stock AND
                a.date = b.maxDate
    ) maxList INNER JOIN
    (
        SELECT  a.*
        FROM    tableName a INNER JOIN
        (
            SELECT      Stock, MIN(date) minDate
            FROM        tableName
            GROUP BY    Stock
        ) b ON  a.stock = b.stock AND
                a.date = b.minDate
    ) minList ON maxList.stock = minList.stock

SQLFiddle Demo

UPDATE 1

seeing your last sentence: Right now I just have two dates per company, but bonus upvotes if your query can handle any number of dates. What if you have records like this?

FB          2012-05-20     40
FB          2012-05-21     34
FB          2012-05-22     42

what would be its result?

Answer 2

I know this was answered already, but the problem caught my interest so I thought I'd give it a try.

I assumed you might want to see the difference between any two consecutive dates.

So for the following data:

GOOG        2012-09-07  42.34
GOOG        2012-09-06  44.56
GOOG        2012-09-01  44.32
FB          2012-09-07  17.82
FB          2012-08-05  12.98

You would get:

GOOG        2012-09-07  2012-09-06  -2.22
GOOG        2012-09-06  2012-09-01  0.24
FB          2012-09-07  2012-08-05  4.84

The query I wrote for this is:

SELECT t1.Stock, t1.Date AS CurrentDate, oldDate.Date AS OldDate, (t1.Price - oldDate.Price) AS PriceChange
FROM MP_StockTable t1
JOIN (SELECT t2.Stock, t2.Date AS date1, (SELECT MAX(t3.Date) FROM MP_StockTable t3 WHERE t3.Date < t2.Date AND t3.Stock = t2.Stock GROUP BY t3.Stock) AS date2 
        FROM MP_StockTable t2) AS StockDates ON StockDates.Stock = t1.Stock AND StockDates.date1 = t1.Date
JOIN MP_StockTable oldDate ON oldDate.Stock = t1.Stock AND oldDate.Date = StockDates.date2 

The inline table basically calculates the closest previous date for each row. So for the data above, the generated table would like this:

GOOG        2012-09-07  2012-09-06
GOOG        2012-09-06  2012-09-01
GOOG        2012-09-01  NULL
FB          2012-09-07  2012-08-05
FB          2012-08-05  NULL

We then use the two dates from that table to get our two price points.

NOTE: I was testing in SQLServer, so apologies if this needs some tweaks for mySql (it's been years since used mySql, so I've forgotten the differences).