Question

Consider the following table (portfolio). It is a transaction log for a stock-market investor. Each day, he either buys, sells or holds (previously bought stocks which are not yet being sold) stocks (identified by sp100_id):

_date       sp100_id  action  price
-----------------------------------
2011-03-21  11        buy     10.50
2011-03-21  55        buy     60.00
2011-03-21  99        buy      5.15
2011-03-22  11        sell     9.80
2011-03-22  55        sell    61.50
2011-03-22  99        hold     5.60
2011-03-23   1        buy     95.00
2011-03-23   2        buy     25.60
2011-03-23  99        hold
2011-03-24   1        sell    96.00
2011-03-24   2        hold
2011-03-24  99        hold
2011-03-25  11        buy      8.90
2011-03-25   2        sell    28.00
2011-03-25  99        hold

The log stops at 2011-03-25. For 2011-03-26, I want to know:
– what stocks are still left in the portfolio
– for what price and on what date these stocks were orginally bought

If we do this manually:
– stock 11 is bought on 2011-03-21, sold on 2011-03-22, but bought again on 2011-3-25 for 8.90 and we haven’t sold it since, so it is still in portfolio on 2011-03-26
– stock 55 is bought on 2011-03-21 and sold on 2011-03-22 so not in portfolio anymore
– stock 99 is bought on 2011-03-21 and we have held it and never sold so it is still in portfolio on 2011-03-26 for a price of 5.15
– stock 1 and 2 are both bought and sold before 2011-03-26

So the portfolio on 2011-03-26 consists of:

sp100_id  buy_date    buy_price
-------------------------------
11        2011-03-25  8.90
99        2011-03-21  5.15

My question is: with what query can the above output be returned from the table?

SQLFiddle here

Answer 1

select t1.sp100_id, t1._date as buy_date, t1.price
from (select * from portfolio where action='buy')  t1
    left join (select * from portfolio where action='sell') t2 
      on t1.sp100_id=t2.sp100_id
    and t1._date<t2._date
where t2.sp100_id is null