tsql determinate price trends - tsql

Question

I have a table with 3 column
+---------------+-------------------------+-------+
| InstrumentId | Date | Price |
+---------------+-------------------------+-------+
| 39 | 2012-10-31 00:00:00.000 | 150 |
| 39 | 2012-11-01 00:00:00.000 | 160 |
| 39 | 2012-11-01 00:00:00.000 | 200 |
| 40 | 2012-10-31 00:00:00.000 | 150 |
| 40 | 2012-11-01 00:00:00.000 | 140 |
| 40 | 2012-11-01 00:00:00.000 | 200 |
| 50 | 2012-10-31 00:00:00.000 | 150 |
| 50 | 2012-11-01 00:00:00.000 | 150 |
| 50 | 2012-11-01 00:00:00.000 | 150 |
+---------------+-------------------------+-------+
I need to recive next result:
+--------------+-------+
| InstrumentId | Price |
+--------------+-------+
| 39 | 200 |
| 40 | 0 |
| 50 | 150 |
+--------------+-------+
rules:
if price for same InstrumentId is growing or is equal => return last price (that means every next price greater or equal to a previous price.
For instance Id 39: 150 <= 160 <= 200 => return 200)
if any price for same InstrumentId is less than previous => return 0 (see instrumentId 40)
I can do that with a cursor... but I think that exist a simply workaround to do this.
Any ideas?
test data:
DECLARE #table TABLE(
instrumentId INT NOT NULL,
priceListDate DATETIME NOT NULL,
price DECIMAL NOT NULL
)
INSERT INTO #table
(
instrumentId,
priceListDate,
price
)
VALUES( 39, '2012-10-31 00:00:00.000', 150),
(39,'2012-11-01 00:00:00.000', 160),
(39,'2012-11-01 00:00:00.000', 200),
(40,'2012-10-31 00:00:00.000', 150),
(40,'2012-11-01 00:00:00.000', 140),
(40,'2012-11-01 00:00:00.000', 200),
(50,'2012-10-31 00:00:00.000', 150),
(50,'2012-11-01 00:00:00.000', 150),
(50,'2012-11-01 00:00:00.000', 150)

Answer 1

Let me know if this works ok. I'm guessing you won't ever have a price = -1 in your table, I think that would cause problems for the current solution.
WITH CTE
AS ( SELECT RN = ROW_NUMBER() OVER ( ORDER BY instrumentId ) ,
*
FROM #table
)
SELECT CASE WHEN MIN(X.xPrice) = -1 THEN 0
ELSE MAX(X.xPrice)
END 'price' ,
X.instrumentId
FROM ( SELECT CASE WHEN [Current Row].instrumentId = [Next Row].instrumentId
THEN CASE WHEN [Current Row].price > [Next Row].price
THEN -1
ELSE [Current Row].price
END
ELSE CASE WHEN [Previous Row].instrumentId = [Current Row].instrumentId
THEN CASE WHEN [Previous Row].price <= [Current Row].price
THEN [Current Row].price
ELSE -1
END
ELSE [Current Row].price
END
END 'xPrice' ,
[Current Row].RN ,
[Current Row].instrumentId
FROM CTE [Current Row]
LEFT JOIN CTE [Previous Row] ON [Previous Row].RN = [Current Row].RN
- 1
LEFT JOIN CTE [Next Row] ON [Next Row].RN = [Current Row].RN
+ 1
) X
GROUP BY X.instrumentId
It might seem a bit convoluted but the basic idea is to determine the next and previous row of the current on in order to test the value of the price column for that row.