float(::Type) width inconsistency

[nsajko]

float(Union{Int,BigInt}) being narrower than BigFloat seems like a bug.

julia> float(Union{Int,BigInt})
Float64

julia> float(BigInt)
BigFloat

julia> versioninfo()
Julia Version 1.11.0-DEV.1276
Commit fc6295df630 (2024-01-15 05:47 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 8 × AMD Ryzen 3 5300U with Radeon Graphics
  WORD_SIZE: 64
  LLVM: libLLVM-16.0.6 (ORCJIT, znver2)
Threads: 1 default, 0 interactive, 1 GC (on 8 virtual cores)
Environment:
  JULIA_NUM_PRECOMPILE_TASKS = 3
  JULIA_PKG_PRECOMPILE_AUTO = 0

julia> @which float(Union{Bool,BigInt})
float(::Type{T}) where T<:Number
     @ Base float.jl:373

The method looks like this:

float(::Type{T}) where {T<:Number} = typeof(float(zero(T)))

The problem is that zero(T) for T that is an union doesn't behave as we would like here (xref. #34003):

julia> typeof(zero(Union{Int,BigInt}))
Int64

[mikmoore]

Very tangentially related to #52528 (comment), where the insufficient generality of Base.zero(::Union) also causes a hiccup.

It looks like zero (and maybe several other functions like one, float, etc) should attempt to Base.promote_union on union types. E.g.,

julia> zero(Base.IEEEFloat)
ERROR: MethodError: no method matching Union{Float16, Float32, Float64}(::Int64)

julia> zero(Base.promote_union(Base.IEEEFloat)) |> typeof
Float64

This won't work for a Union that doesn't promote to a unique usable type (like ones that promote to Any), but it should work for many "sane" instances like the one here and in that other issue.

[nsajko]

It looks like zero (and maybe several other functions like one, float, etc) should attempt to Base.promote_union on union types.

FTR this conflicts with the design proposed for zero in #34003. And lots of people put a 👍 on that proposal.

But I guess we could fix this issue using Base.promote_union!