

Grasshopper
Group: General Forum Members
Last Login: Monday, September 15, 2014 9:57 AM
Points: 19,
Visits: 19


Hi, I want to perform the following steps 1. Convert the real number into a character. 2. Convert the resulting character to a decimal. 3. Round the value at the N+1 th place.
My requirement is When n=2 I want to round 18.005 to 18.01 or 18.0049 to 18.00(at 100th place)
when i try with
Declare @real real Set @real=18.005 print Round(Convert(Decimal(15,7),Convert(varchar(30),@real)),2)
i can able to get the result until i get some indefinite real value like 4.8892562E08. so when i try the same logic, i am getting a conversion error(Error converting data type varchar to numeric.). Please help me to get the problem resolved. thanks in advance




Old Hand
Group: General Forum Members
Last Login: Thursday, December 11, 2014 12:53 AM
Points: 317,
Visits: 409


You're going to run into that problem because of the rule for the floating point. How about converting to decimal(38,17) from real before trying to cast it out to a varchar?
Alternatively, do this in code.




SSCrazy
Group: General Forum Members
Last Login: Tuesday, October 28, 2014 11:02 AM
Points: 2,397,
Visits: 3,419


DECLARE @n TINYINT
SET @n = 2
DECLARE @Sample TABLE ( Value DECIMAL(15, 7) )
INSERT @Sample SELECT 18.0049 UNION ALL SELECT 18.005
SELECT Value, ROUND(Value + CAST(0.5E * POWER(0.1E, @n) AS DECIMAL(15, 7)), @n, 1) FROM @Sample
N 56°04'39.16" E 12°55'05.25"




Old Hand
Group: General Forum Members
Last Login: Thursday, December 11, 2014 12:53 AM
Points: 317,
Visits: 409


Or do it the hardcore way like Peso's example




Grasshopper
Group: General Forum Members
Last Login: Monday, September 15, 2014 9:57 AM
Points: 19,
Visits: 19


Thanks for the Reply. I have already tried with converting the real into decimal before converting to Varchar
Declare @real real Set @real=Input print Round(Convert(varchar(30),Convert(Decimal(15,7),@real)),2)
If Input = 19.005, expected Result is 19.01. but i cant able to reproduce the same. Actual result is 19 i can able to achieve the result when Input = 19.0051( Result is 19.01).




SSCDedicated
Group: General Forum Members
Last Login: Yesterday @ 8:47 PM
Points: 35,770,
Visits: 32,436


idrisgani (7/2/2009) Thanks for the Reply. I have already tried with converting the real into decimal before converting to Varchar
Declare @real real Set @real=Input print Round(Convert(varchar(30),Convert(Decimal(15,7),@real)),2)
If Input = 19.005, expected Result is 19.01. but i cant able to reproduce the same. Actual result is 19 i can able to achieve the result when Input = 19.0051( Result is 19.01).
Based on that reply, I'm thinking that you didn't try Peso's code which actually does work. Don't assume that just reading the code will be your answer. Try it.
Jeff Moden "RBAR is pronounced "reebar" and is a "Modenism" for "RowByAgonizingRow".
First step towards the paradigm shift of writing Set Based code: Stop thinking about what you want to do to a row... think, instead, of what you want to do to a column."
(play on words) "Just because you CAN do something in TSQL, doesn't mean you SHOULDN'T." 22 Aug 2013
Helpful Links: How to post code problems How to post performance problems




Grasshopper
Group: General Forum Members
Last Login: Monday, September 15, 2014 9:57 AM
Points: 19,
Visits: 19


Hi Peso, Thanks for the code. Your logic works well when the input type is of decimal. but my input type is real.it is rounded to 19 and not 19.01, when the input type is of real.




SSCrazy
Group: General Forum Members
Last Login: Tuesday, October 28, 2014 11:02 AM
Points: 2,397,
Visits: 3,419


And that's the culprit. Due to the binary interpretation of REAL and FLOAT datatypes
1 / 2^n + 1 / 2^(n+1) + 1 / 2^(n+2) + 1 / 2^(n+3) + ...
the value may not be stored exactly as you want. So you are left with two choices
1. Keep REAL and learn to live with it 2. Change to decimal and get what you want, all the times.
N 56°04'39.16" E 12°55'05.25"




Grasshopper
Group: General Forum Members
Last Login: Monday, September 15, 2014 9:57 AM
Points: 19,
Visits: 19


Hi. i can able to get what i am expected when i use the below code
Declare @real Decimal(15,7) Set @real=19.005 print ROUND(@real + CAST(0.5E * POWER(0.1E, 2) AS DECIMAL(15, 3)), 2, 1) Output = 19.01
i am not able to predict the length of the digit after dot, it may vary. so the length of the decimal used while casting should not be hardcoded.
i cant able to supply the length at runtime like decimal(15,@n+1).
Thanks, Idris Gani R




SSCrazy
Group: General Forum Members
Last Login: Tuesday, October 28, 2014 11:02 AM
Points: 2,397,
Visits: 3,419


print cast( ROUND(@real + CAST(0.5E * POWER(0.1E, 2) AS DECIMAL(15, 3)), 2, 1) AS decimal(15, 2))
N 56°04'39.16" E 12°55'05.25"



