[takeda]

I used code from: https://endler.dev/2017/go-vs-rust/ good point, I overlooked that.

Anyway in Go (ironically because of lack of generics) if you use any numeric type other than int, int64, float64 you will be in the word of hurt. Rust doesn't have that issue.

So in practice you will likely use int, and I suppose you can add an assertion.

BTW: I only see that it would remove 3 lines though, where are the other 3?

[tptacek]

I don't follow. I use unsigned ints in Go all the time. I've never been in a world of hurt with them. Mandatory explicit integer conversions (and the way Go consts work) are something Go gets right.

[takeda]

Ok, so you like that. I myself really hated when I used float32 and had to do this when I was doing calculations:

    result = (float32)Max((float64)a, (float64)b)

I ended up switching the type to float64, and wonder why they even offer float32 if it's practically unusable. I had similar experience when I needed to use int8 or int16 etc.

An alternative was to make own version of Max/Min and other math functions, but this is what generics would solve.

[tptacek]

I'm not sure I follow, because Rust is also (thankfully) fussy about integer types.

[masklinn]

I think what they mean is utility functions (e.g. min/max here) tend to only be implemented for one type, so if you’re using an other you keep casting back and forth.

Rust is very fussy about types and some operations are bound to a single one (e.g. stdlib sequences only index with usize) but utility functions tend to either be genetic or be mass-implemented using macros (or by hand probably for f32::min and f64::min though I did not check).

[takeda]

I don't have problem with being strong typed, I like that as well. The problem is that the stdlib doesn't really support other types and that's mostly due to lack of generics, so using uncommon types becomes quite annoying.

[masklinn]

TBF the float Min/Max issue has nothing to do with generics, there isn't a "floating-point" generic class, and the entire reason why Go has an f64 Min/Max in the first place is that floats are not fully ordered[0], so you "can't" use regular comparisons (which you'd be told to do for integers) and thus you could not have a generic min/max relying on that even if there was one (e.g. in Rust `f32` and `f64` are not `Ord`, which `std::cmp::{min, max}` require, hence both `f32` and `f64` having their own inherent min/max methods).

So what your issue comes down to is Go's designers couldn't be arsed to duplicate the entire `math` package to also work on `float32`. Some members of the community did rise to the challenge[1] tho.

[0] well recent revisions of IEEE-754 have a total ordering predicate but I wouldn't say that it's really useful for such a situation as it positions NaNs on "outside" of the numbers, so a negative NaN is smaller than a negative number and a positive NaN larger than a positive number

[1] https://github.com/chewxy/math32

[takeda]

I'm aware of it, in the example given I used float32. Anyway Max/Min was just example I used other functions from math as well.

Anyway my point was that with generics, they wouldn't need to copy anything, the math package would work with both float32 and float64 and many functions likely would also work on all integers.