RE: Rounding up to nearest 1000

["Jonathan Mason" <jonathan.mason@xxxxxxxxxxxxxxxx>]

Hi Greg

> Did I do something wrong ?

I'm afraid so.  As Adam said, you have to use "integer division" for it to
work, so in your example of:

        Say I want to round 1461 up to the nearest thousand, which would be
        2000.
        So N is 1461 and S = 1000

        Temp = 1461 + 1000/2
        Temp = 1461 + 500
        Temp = 1961
        Result = (1961/1000) * 1000
        Result = 1961 ?  

You need to have Result calculated as:

        Result = Int(1961/1000) * 1000
        Result = 1000

If N had been 1641 then you would have had:

        Temp = 1641 + 1000/2
        Temp = 1641 + 500
        Temp = 2141
        Result = Int(2141/1000) * 1000
        Result = 2000

The half adjustment works on the mid boundary of the scale so anything below
gets rounded down while anything equal to or above gets rounded up.

All the best

Jonathan

-----Original Message-----
From: rpg400-l-bounces@xxxxxxxxxxxx [mailto:rpg400-l-bounces@xxxxxxxxxxxx]
On Behalf Of Fleming, Greg (ED)
Sent: 10 November 2006 19:47
To: RPG programming on the AS400 / iSeries
Subject: RE: Rounding up to nearest 1000


|The rounding algorithm goes something like this:
|
|Let N be the number to round, S be the scale of the rounding (in this
|example S = 1000)
|1)  temp = N + S/2
|2) result = (temp / S) * S
|
|I think this has to use 'integer division' to work.
|
|This is essentially what the H extender is about - hence the name -
|"(H)alf add".
|
|HTH,
|Adam

Adam,

I'm trying that out using my trusty calculator, and I'm not ending up
with a round number.

Say I want to round 1461 up to the nearest thousand, which would be
2000.
So N is 1461 and S = 1000

Temp = 1461 + 1000/2
Temp = 1461 + 500
Temp = 1961
Result = (1961/1000) * 1000
Result = 1961 ?  

Did I do something wrong ? 

Greg