fix InexactError on conversion for HiGHS solver

[thchr]

I was trying out the HiGHS solver as an alternative to Cbc via

solver_positions(..., Zarate(ordering_solver=optimizer_with_attributes(HiGHS.Optimizer)))

For certain problems, this threw the following error:

ERROR: InexactError: Bool(-2.220446049250313e-16)
Stacktrace:
  [1] Bool(x::Float64)
    @ Base .\float.jl:158
  [2] (::LayeredLayouts.var"#is_before_func#44"{Vector{Any}})(n1::Int64, n2::Int64)
    @ LayeredLayouts C:\Users\tchr\.julia\dev\LayeredLayouts\src\zarate.jl:202
  [3] #1
    @ .\ordering.jl:133 [inlined]
  [4] lt
    @ .\ordering.jl:120 [inlined]
  [5] sort!(v::Vector{Int64}, lo::Int64, hi::Int64, #unused#::Base.Sort.InsertionSortAlg, o::Base.Order.Lt{Base.Order.var"#1#3"{LayeredLayouts.var"#is_before_func#44"{Vector{Any}}, typeof(identity)}})
    @ Base.Sort .\sort.jl:504
  [6] sort!(v::Vector{Int64}, lo::Int64, hi::Int64, a::Base.Sort.QuickSortAlg, o::Base.Order.Lt{Base.Order.var"#1#3"{LayeredLayouts.var"#is_before_func#44"{Vector{Any}}, typeof(identity)}})
    @ Base.Sort .\sort.jl:570
  [7] sort!
    @ .\sort.jl:660 [inlined]
  [8] #sort!#8
    @ .\sort.jl:721 [inlined]
  [9] order_layers!
    @ C:\Users\tchr\.julia\dev\LayeredLayouts\src\zarate.jl:204 [inlined]
 [10] solve_positions(layout::Zarate, original_graph::Graphs.SimpleGraphs.SimpleDiGraph{Int64}; force_layer::Vector{Pair{Int64, Int64}}, force_order::Vector{Pair{Int64, Int64}})
    @ LayeredLayouts C:\Users\tchr\.julia\dev\LayeredLayouts\src\zarate.jl:93
[...]

I imagine that the assumption that value(isbefore[n1][n2]) is an integer is not necessarily true for all solvers. This is a suggested fix for those cases.

[oxinabox]

thanks

[thchr]

Bump :)

[thchr]

Also a gentle bump on this: would be nice to be able to experiment with other solvers.

[oxinabox]

Oh, I totally forgot about this PR.
And recreated it in #34.
This closes #33

[mtfishman]

Unfortunately round(::Type{Bool}, ::Float64) isn't available in Julia 1.7 and below:

julia> round(Bool, 0.9)
ERROR: MethodError: no method matching trunc(::Type{Bool}, ::Float64)
Closest candidates are:
  trunc(::Type{Signed}, ::Float64) at /Applications/Julia-1.7.3.app/Contents/Resources/julia/share/julia/base/float.jl:358
  trunc(::Type{T}, ::Missing) where T at /Applications/Julia-1.7.3.app/Contents/Resources/julia/share/julia/base/missing.jl:154
  trunc(::Type{T}, ::Rational) where T at /Applications/Julia-1.7.3.app/Contents/Resources/julia/share/julia/base/rational.jl:455
  ...
Stacktrace:
 [1] round(#unused#::Type{Bool}, x::Float64)
   @ Base ./float.jl:369
 [2] top-level scope
   @ REPL[2]:1

julia> versioninfo()
Julia Version 1.7.3
Commit 742b9abb4d (2022-05-06 12:58 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin21.4.0)
  CPU: Apple M1 Max
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-12.0.1 (ORCJIT, westmere)

[mtfishman]

Also it's not supported in Compat.jl but I made an issue for that: JuliaLang/Compat.jl#798.

[thchr]

I guess it could be conditionally added to LayeredLayouts by copying over its definition from Base:

if VERSION < v"1.8"
   round(::Type{Bool}, x::AbstractFloat) = (-0.5 <= x < 1.5) ? 0.5 < x : throw(InexactError(:round, Bool, x))
end

Alternatively, we could change round(Bool, value(is_before[n1][n2]))::Bool to convert(Bool, round(value(is_before[n1][n2])))::Bool.

[oxinabox]

@mtfishman please open new errors rather than commenting on PRs merged months ago.

[mtfishman]

Sorry @oxinabox! It annoys me when people do that as well... Just wanted to comment before I forgot, I was planning on raising an issue/making a PR.