Using NativeCall to call the C fn `erf` gets more precise output than `erf` in C

Question

I have written a Raku script to call erf function in C standard library:

use NativeCall;
sub erf(num64) returns num64 is native { * };

say [0.5,1,2,3,4,-0.9].map: {erf($_.Num)};

The output of this script

(0.5204998778130465 0.8427007929497149 0.9953222650189527 0.9999779095030014 0.9999999845827421 -0.7969082124228322)

matches with the output from C for all values [0.5,1,2,3,4,-0.9] except 4.

For 4 the C outputs 1.000000 while Raku gives 0.9999999845827421.

To test the output for 4 in C, run this code:

#include <stdio.h> // Including header file for printf function
#include <math.h>  // Including header file for erf function

int main (){

  double param, result;
  param = 4.0;
  result = erf(param);
  printf("erf (%f) = %f\n", param, result);
  return 0;
}

Any idea what's going on? I need to output 1.0 from Raku too.

Answer 1

You're comparing apples with oranges.

use NativeCall;
sub erf(num64) returns num64 is native { * };

say .fmt("%f") for [0.5,1,2,3,4,-0.9].map: {erf($_.Num)}

0.520500
0.842701
0.995322
0.999978
1.000000
-0.796908

You're using printf in C, if you use .fmt (the easier way to say sprintf in Raku), then you'd also get 1.0.