Nested Math.Ceiling Math.Round etc does not work with C-sharp and Java

[mittelmark]

Hi,

our project progresses fast ...

Here an issue we found with Java and C# if the Math statements are nested:

public static class Test {
        public static void Main (string[] argv) {
                int x = Math.Ceiling(1.0); // works on C, C++, C#, D, Java, Javascript, Python (Swift and Typescript untested) 
                List<int>() y;
                y.Add(Math.Ceiling(2.0)); // fail Java, C-sharp missing typecast to int
                Console.WriteLine($"x is {x}");
                Console.WriteLine($"y0 is {y[0]}");
        }
}

same issue with Math.Round etc.

Best,
Detlef

[mittelmark]

Thank you for adding this to the Bug list.

For now I added workarounds in my Makefile to fix files in the javam and csm folders, where I output my Java and C# files, like the following:

java:
    fut Stats.fu Rand.fu -o javam/Stats.java
    perl -i.bak -pe 's/\(( *)Math\.(ceil|rint)/($$1(int)Math.$$2/g' javam/*java
cs: 
    fut Stats.fu Rand.fu -o csm/stats.cs.
    perl -i.bak -pe 's/\(( *)MathF\.(Ceiling|Round)/($$1(int)MathF.$$2/g' csm/*cs

But that list will grow over time ;)

[mittelmark]

If I read the reference it states that:

Conversion from floating-point type to integer must be explicit, by using one of the above methods that return an integer:

double d = 10.5;
int i = Math.Truncate(d); // translated to: (int) d
int j = Math.Ceiling(d);
int k = Math.Floor(d);
int n = Math.Round(d);

If we 'fix' that, we probably can't use these methods, which return a float or double in most languages, to add floats or doubles to a growing list?

[pfusik]

List<int>() y;
y.Add(Math.Ceiling(2.0));

should work.

You can use this workaround:

int tmp = Math.Ceiling(2.0);
List<int>() y;
y.Add(tmp);

[mittelmark]

The temporary int approach does work, the y.Add(Math.Ceiling(2.0)) does not work for C# and Java but for C, C++, D, Javascript and Python.

C++ for instance does a static cast int the fut generated output file even if inline:

Here the input in fusion:

   int x = 0;
   List<int>() y;
   x = Math.Ceiling(2.0);
   y.Add(Math.Ceiling(2.0));

And here the correct generated C++ code:

 int x = 0;
 std::vector<int> y;
 x = static_cast<int>(std::ceil(2.0));
 y.push_back(static_cast<int>(std::ceil(2.0)));

whereas in Java and C# that casting is not done.

int x = 0;
final ArrayList<Integer> y = new ArrayList<Integer>();
x = (int) Math.ceil(2.0);
y.add(Math.ceil(2.0)); // missing (int)

and as well not in C#:

  int x = 0;
  List<int> y = new List<int>();
  x = (int) Math.Ceiling(2.0);
  y.Add(Math.Ceiling(2.0)) // missing (int)

D looks again fine:

int x = 0;
Array!(int)* y = new Array!(int);
x = cast(int) ceil(2.0);
y.insertBack(cast(int) ceil(2.0));