Question

I have the following structure:

CREATE TABLE IF NOT EXISTS `myTable` (
  `ID` int(11) NOT NULL AUTO_INCREMENT,
  `time` int(11) NOT NULL,
  `item` varchar(255) NOT NULL,
  `price` int(11) NOT NULL,
  PRIMARY KEY (`ID`),
  KEY `time` (`time`),
  KEY `item` (`item`),
  KEY `price` (`price`)
) ENGINE=MyISAM  DEFAULT CHARSET=latin1 AUTO_INCREMENT=8 ;


INSERT INTO `myTable` (`ID`, `time`, `item`, `price`) VALUES
(1, 100, 'cat', 10),
(2, 101, 'cat', 10),
(3, 102, 'cat', 9),
(4, 103, 'cat', 11),
(5, 104, 'cat', 10),
(6, 105, 'cat', 10),
(7, 106, 'cat', 11);

I’m trying to figure out a query that would return the following:

time   item   price
----   ----   -----
100    cat    10
102    cat    9
103    cat    11
104    cat    10
106    cat    11

It’s ordered by time, and shows only rows where the price is different to the price before it (Edit: *According to time*).

EDIT:

Maybe I made the original problem too simple, what if it was more complex with data like this:

INSERT INTO `myTable` (`ID`, `time`, `item`, `price`) VALUES
(1111, 100, 'cat', 10),
(21, 105, 'cat', 10),
(31, 108, 'cat', 9),
(411, 109, 'cat', 11),
(512, 110, 'cat', 10),
(61, 114, 'cat', 10),
(72, 120, 'cat', 11),
(8, 100, 'dog', 20),
(93, 102, 'dog', 20),
(104, 108, 'dog', 29),
(111, 109, 'dog', 31),
(122, 155, 'dog', 20),
(13, 165, 'dog', 20),
(1445, 172, 'dog', 31);

In the above, the ID column is all messed up, perhaps there are other records in between. Also the time no longer increments by 1 and I’ve thrown a dog into the mix.

Answer 1

The query you need is this:

select curr.time  ,
       curr.item  ,
       curr.price 
from myTable curr
join myTable prev on prev.time = ( select max(time)
                                   from myTable p
                                   where p.time < curr.time
                                 )
where curr.price != prev.price
order by t.time

It takes advantage of the fact that time is an alternate key for the table. We take all the rows from myTable (let’s call it current) such that its price is different than that of the previous row based on the the value of the alternate key time (lets call this set previous).

To do this, we need a correlated subquery that identifies the highest value of time less than or equal to the value of of time for the current row. So…we wind up referencing myTable thrice:

• First, for the set of ‘current’ rows,
• Second, for the set of ‘previous’ rows’
• Third, to identify the ‘previous’ row based on the data in the ‘current’ row.

Easy!