Hi everyone, I get different results using `cos` f...

# kotlin-native

[Omar Mainegra]

Hi everyone, I get different results using cos function in K/N, i.e:

@Test
fun cosine_functions_works_the_same_in_all_platform() {
    assertEquals(0.6441559905471732, cos(0.8708769899641021))
}

This test pass in JVM and JS, but fails in iOS:

kotlin.AssertionError: Expected <0.6441559905471732>, actual <0.6441559905471733>.

[Kris Wong]

that's simply the nature of passing floating point numbers across language boundaries

[Kris Wong]

there's a fuzzy version of assert equals

[Kris Wong]

use a tolerance of something like .000001, or whatever works for you.

[Nico Buescher]

I can't seem to find a fuzzy version of assertEquals .

[Omar Mainegra]

Yes, that's what I'm doing now, but I was actually worried about the implications, since it looks like a Double precision issue that might be related to the implementation of the functions in K/N

[Omar Mainegra]

For instance, this is passing in all platforms:

fun multiply(a: Double, b: Double) = a*b

@Test
fun cosine_functions_works_the_same_in_all_platform() {
    assertEquals(0.49271114849400754, 0.7648941494365021*0.6441559905471732)
    assertEquals(0.49271114849400754, multiply(0.7648941494365021, 0.6441559905471732))
}

[Omar Mainegra]

Does the second case counts as passing floating point numbers across language boundaries? If so I'd expect a loss in the precision as happened with cos and sin functions

[Kris Wong]

no, it doesn't go through a language binding

[Omar Mainegra]

Ahh, I see

[Omar Mainegra]

Just to see if I get it right, If I implement multiply function in C or ObjC and create the bindings, can I expect the loss in the precision too?

[Kris Wong]

most likely

[Omar Mainegra]

Nice, thanks for everything