Round Up, Round Down and Half adjust

[gregwga50]

Does anyone know of a method, routine, procedure, bif, etc. to perform round up or round down (not to be confused with half adjust) . The chart below shoes the difference between round up, round down and half adjust:

[JonBoy]

It has been about 25 years since I had to code this but something like this?

**free
dcl-s rounded packed(5:2);
dcl-s base packed(5:3) inz(12.999);
dcl-s i int(5);
for i = 1 to 100;
   eval(h) rounded = base + 0.004;
   dsply ( 'Base = ' + %char(base) + ' rounded = ' + %char(rounded) );
   base += 0.001;
endfor;
​

[B.Hauser]

Use SQL:

With x (Number) as (Values(430,123), (430,125), (430,128))
Select Number, Round(Number, 2)           "Round",
               Ceiling(Number * 100) /100 "Round Up",
               Floor(Number * 100) / 100  "Round Down"
  from X;​

[vazymimil]

I think these in fact are round ceiling and round floor, and when input values are negative, round up is equivalent to Ceiling(abs(Number) * 100) /100 * Sign(Number)

[gregwga50]

Use SQL:

With x (Number) as (Values(430,123), (430,125), (430,128))
Select Number, Round(Number, 2) "Round",
Ceiling(Number * 100) /100 "Round Up",
Floor(Number * 100) / 100 "Round Down"
from X;​

That will only work if the the numbers to be rounded are integers, right? X has 3 decimal places in the examples I posted.

[vazymimil]

I think these in fact are round ceiling and round floor, and when input values are negative, round up is equivalent to Ceiling(abs(Number) * 100) /100 * Sign(Number)

[UserName10]

Rounding down is truncation. To round up compare the truncated value to the original value and adjust if needed.

dcl-ds;
  x zoned(5:3) pos(1);
  x_trunc zoned(5:2) pos(1);
end-ds;

dcl-s result zoned(5:2);


// rounding down
x = 430.123;
result = x_trunc;    // result = 430.12

x = 430.125;
result = x_trunc;    // result = 430.12

x = 430.128;
result = x_trunc;    // result = 430.12


// rounding up
x = 430.123;
if x > x_trunc;
  result = x_trunc + .01;    // result = 430.13
endif;

x = 430.125;
if x > x_trunc;
  result = x_trunc + .01;    // result = 430.13
endif;

x = 430.128;
if x > x_trunc;
  result = x_trunc + .01;    // result = 430.13
endif;​

[UserName10]

correction:

x_trunc zoned(4:2) pos(1);

[Peder Udesen]

In RPG you could code like this:

dcl-pr ceil float(8) extproc(*cwiden:*dclcase);
           val float(8) const;
end-pr;​


y=ceil( x *100) / 100;

[B.Hauser]

That will only work if the the numbers to be rounded are integers, right? X has 3 decimal places in the examples I posted.

If you had run my example you would have seen it is rounded (in all 3 cases) to 2 decimal positions.

[gregwga50]

If you had run my example you would have seen it is rounded (in all 3 cases) to 2 decimal positions.

Sorry about that, you are usually always right on the money. I saw one of your posts from several years ago about rounding and made my conclusion from that, my bad.