patternsqlModerate
CASE expression returns wrong value when using CEILING
Viewed 0 times
caseexpressionvalueusingwrongreturnswhenceiling
Problem
I've run into an issue where a
As a test, I added a decimal variable and ran the same
Here is the SQL code:
Here is a snippet of the results:
And here is the d
CASE expression does not return what I expect. As a test, I added a decimal variable and ran the same
CASE expression against it and it works fine, returning the results as I would expect (rounding the value up when IsGun=1. But when I run that same CASE expression against another decimal value, it always returns the value with the CEILING() function and never returns the original value.Here is the SQL code:
DECLARE @Num decimal(8,2);
set @Num = 12.54;
WITH PQ AS
(
SELECT
UPC,
Price1,
DBID,
AVG(Price1) OVER (PARTITION BY UPC) AS Price1Avg
FROM
vProducts_PriceQty_Union
)
SELECT
PQ.UPC,
PQ.Price1,
PQ.Price1Avg,
(CASE WHEN p.IsGun = 1 THEN CEILING(@Num) ELSE @Num END) AS UsingVar,
CAST(
(CASE WHEN P.IsGun = 1 THEN CEILING(PQ.Price1Avg) ELSE PQ.Price1 END)
AS NUMERIC(8,2))
AS PriceAdj,
PQ.DBID,
P.IsGun
FROM
PQ
INNER JOIN
products P ON PQ.UPC = P.UPCHere is a snippet of the results:
UPC Price1 Price1Avg UsingVar PriceAdj DBID IsGun
942000899195 14.9900 14.990000 12.54 15.00 1 0
980420671300 29.9900 29.990000 12.54 30.00 1 0
980420671310 29.9900 29.990000 12.54 30.00 1 0
980426713020 29.9900 29.990000 12.54 30.00 1 0
980426713120 29.9900 29.990000 12.54 30.00 1 0
000998622130 319.0000 319.000000 13.00 319.00 1 1
000998624730 314.0000 314.000000 13.00 314.00 1 1
000998624970 419.0000 419.000000 13.00 419.00 1 1
008244284754 1015.0000 1015.000000 13.00 1015.00 2 1
010633012288 267.0000 267.000000 13.00 267.00 6 1
And here is the d
Solution
To reproduce the problem:
What happens here is that
According to the documentation, the output type of
To verify:
As a workaround, you could explicitly convert the output of the
SELECT *, (CASE
WHEN IsGun=1 THEN CEILING(Price1Avg)
ELSE Price1 END)
FROM (
SELECT UPC, IsGun, Price1,
AVG(CAST(Price1 AS numeric(8, 2))) OVER (PARTITION BY UPC) AS Price1Avg
FROM (
VALUES ('A', 0, 14.99),
('B', 0, 29.99),
('C', 1, 319.00),
('D', 1, 314.00)
) AS x(UPC, IsGun, Price1)
) AS sub;What happens here is that
CEILING(PQ.Price1Avg) produces a numeric(38, 0).According to the documentation, the output type of
CEILING() is of the same base datatype as the input, although the scale (the number of decimals) may change, which is what happens here.- The
AVG()function, in my tests, returnsnumeric(38, 6).
- The
CEILING()function on that column, however, outputsnumeric(38, 0):
To verify:
SELECT CEILING(CAST(123.45 AS numeric(38, 6)))As a workaround, you could explicitly convert the output of the
CEILING() function, which should give you the correct results:SELECT *, (CASE
WHEN IsGun=1 THEN CAST(CEILING(Price1Avg) AS numeric(8, 2)) -- Explicit CAST.
ELSE Price1 END)
FROM (
SELECT UPC, IsGun, Price1,
AVG(CAST(Price1 AS numeric(8, 2))) OVER (PARTITION BY UPC) AS Price1Avg
FROM (
VALUES ('A', 0, 14.99),
('B', 0, 29.99),
('C', 1, 319.00),
('D', 1, 314.00)
) AS x(UPC, IsGun, Price1)
) AS sub;Code Snippets
SELECT *, (CASE
WHEN IsGun=1 THEN CEILING(Price1Avg)
ELSE Price1 END)
FROM (
SELECT UPC, IsGun, Price1,
AVG(CAST(Price1 AS numeric(8, 2))) OVER (PARTITION BY UPC) AS Price1Avg
FROM (
VALUES ('A', 0, 14.99),
('B', 0, 29.99),
('C', 1, 319.00),
('D', 1, 314.00)
) AS x(UPC, IsGun, Price1)
) AS sub;SELECT CEILING(CAST(123.45 AS numeric(38, 6)))SELECT *, (CASE
WHEN IsGun=1 THEN CAST(CEILING(Price1Avg) AS numeric(8, 2)) -- Explicit CAST.
ELSE Price1 END)
FROM (
SELECT UPC, IsGun, Price1,
AVG(CAST(Price1 AS numeric(8, 2))) OVER (PARTITION BY UPC) AS Price1Avg
FROM (
VALUES ('A', 0, 14.99),
('B', 0, 29.99),
('C', 1, 319.00),
('D', 1, 314.00)
) AS x(UPC, IsGun, Price1)
) AS sub;Context
StackExchange Database Administrators Q#136379, answer score: 11
Revisions (0)
No revisions yet.