Value and Indexing-by-rational errors

[kaltsoplyn]

Hello, thank you for making this available.
I use it for large quantum numbers and it's super fast. However, I noticed this today:

julia> WignerFamilies.wigner3j_f(Float64, 1, 1//2, 1, -1//2)[3//2]
-0.2886751345948129  # ok

This is ok.

julia>  WignerFamilies.wigner3j_f(Float64, 3//2, 1, -1//2, 1)[1//2]
0.32025630761017426  # not ok

This doesn't look good. It's a cyclic permutation of the previous one, but yields a different value.
And then the next cyclic permutation yields an error due to indexing with a rational:

julia> WignerFamilies.wigner3j_f(Float64, 1//2, 3//2, -1//2, -1//2)[1]
ERROR: ArgumentError: invalid index: 1//1 of type Rational{Int64}
Stacktrace:
 [1] to_index(i::Rational{Int64})
   @ Base .\indices.jl:300
 [2] to_index(A::WignerSymbolVector{Float64, Rational{Int64}, Vector{Float64}}, i::Rational{Int64})
   @ Base .\indices.jl:277   ... etc.

I'm not somehow blundering spectacularly here, am I?
I haven't tested extensively (i.e., I started writing this issue report 2 minutes after I noticed it), but the indexing error seems to occur when all arguments are half-integers. The value error, I noticed it as I was testing the indexing error! I'll take a look at the source code if I find some time later. Cheers. And thanks again,

[xzackli]

Thanks @kaltsoplyn, I imagine this is a bug. I'll look into this!

I am a cosmologist and I wrote this package mostly for spin-0 and spin-2 fields on the sphere, which do not typically involve half-integer indices.

[xzackli]

In the meantime, I believe the BigInt factorization strategy in https://github.com/Jutho/WignerSymbols.jl will be slow but correct.