Why is this code nondeterministic?

General Usage
1

I don’t understand what is going on with the code below (MWE): the first invocation of isequal is false, from then on it is true. Running v0.6.

struct MaybeParsed{T}
    pos::Int
    value::T
    MaybeParsed{T}(pos) where {T} = new{T}(pos)
    MaybeParsed{T}(pos, value::T) where {T} = new{T}(pos, value)
    MaybeParsed(pos, value::T) where {T} = new{T}(pos, value)
end

function Base.isequal(x::MaybeParsed{Tx}, y::MaybeParsed{Ty}) where {Tx,Ty}
    if isbits(Tx) && isbits(Ty)
        (x.pos == y.pos) & ifelse(x.pos > 0, isequal(x.value, y.value), true)
    else
        x.pos == y.pos && (x.pos > 0 ? isequal(x.value, y.value) : true)
    end
end

isequal(MaybeParsed{Int}(1), MaybeParsed{Int}(1)) # false

isequal(MaybeParsed{Int}(1), MaybeParsed{Int}(1)) # true
2

It doesn’t look like you initialize value, yet it is used for computing the result of isequal.

3

Indeed. Sorry, silly mistake after refactoring, forgot to switch signs.