Using type Real in a .mos-script gives non precise values

[davbr]

If an array of Reals, for example {23.2,12.7,1.1} is sent as input to a
function, in this case a bubblesort algorithm (see func_bubble.mo), the
resulting array will be {1.10000002384186,12.6999998092651,23.2000007629395}.

[Adrian Pop]

I think this is because in read_write.c is used float instead of double!!!???
int read_modelica_real(FILE* file, modelica_real* data)
{

float f;
...
if (fscanf(file,"%e",&f) != 1)
{ cleanup_description(&desc); in_report("rs parse"); return 1; }
*data = f;
read_to_eol(file);
cleanup_description(&desc);
return 0;

}

Peter, can you tell us why you used float here?
I know that float has less precision than double.

I will do some tests with everything as double and get back to you.

Cheers,
Adrian Pop/

[Adrian Pop]

Fixed in revision: 2641
Before:

bubblesort({23.2,12.7,11.1})

{23.2000007629395,12.6999998092651,11.1000003814697}
After changing all floats to double and use
fscanf(%le",...)
now:

loadFile("C:/OpenModelica1.4.3/testmodels/bubblesort.mo")

true

bubblesort({23.2,12.7,11.1})

{23.2,12.7,11.1}