If you don't use zig, you might have missed why that error indicates a stdlib bug. The error in the article is down to a quirk of the type system where literals (like the literal float) have their own unique type (comptime_float) that can coerce to related types (f32, f64) but in this particular case, when atan switches on the type, it doesn't explicitly include comptime_float and fails. Seems like an oversight in the stdlib to me.

Missed all of that, thanks for the explanation. Why doesn't zig try and apply the coercion before the type switch - is this deliberate for safety?

Because "else" technically includes "comptime_float", so that case is handled, just not in the way that's expected.

One of the downsides of comptime over generics is that, because it's low-level and procedural instead of high-level and declarative, things like automatically inserting coercions become harder.

I'm not sure but I would guess it's in the spirit of Zig's "explicit over implicit" philosophy. They are technically different types.

The issue is that atan wasn't implemented for a comptime float.

A "comptime f32" is different from an "f32"