Слава Україні! Героям слава!

@alfonsogarciacaro Perhaps you need to get ceiling or round on the number of steps after you divide? Idk, I haven't looked at the dotnet implementation. Floating point representations are the...

@alfonsogarciacaro Looks like the salient part is [here](https://github.com/dotnet/fsharp/blob/main/src/FSharp.Core/prim-types.fs#L5541-L5552): - do a division to get number of steps (as you do above) - if number of `steps` is a 32-bit integer...

@daz10000 Technically there are ways to increase precision through [compensated summation](https://en.wikipedia.org/wiki/Kahan_summation_algorithm). ```js //TODO: handle NaNs, infinity //TODO: handle negative steps let start = 0.5; let stop = 1.5; let step...

Should be possible to relax the identifier sanitization to allow unicode characters. Unicode identifiers already work fine when compiling to Rust, but it needs to be fixed for the other...

Perhaps later we can move all the date stuff into `Date.ts`, since `Util.ts` is getting quite big. A few tests to cover the new options will go a long way....

@alfonsogarciacaro Kind of depends on the timeline for v3.0 release. IMO, given that we're already in RC territory, and the stated goal of stability and backwards compatibility of v3.0, we...

> Seems some test failing because this is overruling StringEnum or Erase attribute. Is that intended? I guess Erased and StringEnum unions should behave the same as without the optimization...

@alfonsogarciacaro Sure, perhaps that solves some corner cases, but would you elaborate on what would be the benefits vs. just a string for the case name in the array? Also,...

@alfonsogarciacaro Doing it before the Fable AST phase makes this PR a lot smaller. But perhaps what you suggest makes sense, if we assume plugins want to know if something...

@alfonsogarciacaro IMO using a compiler switch (like it is right now in the PR) is cleaner, as you can change the generated code output without making changes to the source...