Checking LinearAlgebra with JET

[goerch]

Just checked LinearAlgebra with JET on 1.8.0. Result:

═════ 70 possible errors found ═════
[...]

There are some real beauties like this one

┌ @ C:\Users\Win10\AppData\Local\Programs\Julia-1.8.0-DEV\share\julia\stdlib\v1.8\LinearAlgebra\src\adjtrans.jl:314 #self#(v, 0)
│┌ @ C:\Users\Win10\AppData\Local\Programs\Julia-1.8.0-DEV\share\julia\stdlib\v1.8\LinearAlgebra\src\adjtrans.jl:314 LinearAlgebra.pinv(Base.getproperty(v, :parent), tol) 
││┌ @ C:\Users\Win10\AppData\Local\Programs\Julia-1.8.0-DEV\share\julia\stdlib\v1.8\LinearAlgebra\src\generic.jl:1078 LinearAlgebra._vectorpinv(LinearAlgebra.adjoint, v, tol)
│││┌ @ C:\Users\Win10\AppData\Local\Programs\Julia-1.8.0-DEV\share\julia\stdlib\v1.8\LinearAlgebra\src\generic.jl:1082 den = LinearAlgebra.sum(LinearAlgebra.abs2, v)      
││││┌ @ C:\buildbot\worker\package_win64\build\usr\share\julia\stdlib\v1.8\SparseArrays\src\sparsevector.jl:1377 SparseArrays.sum(f, SparseArrays.nonzeros(x))
│││││┌ @ reducedim.jl:890 Base.#sum#748(Base.:, Base.pairs(Core.NamedTuple()), #self#, f, a)
││││││┌ @ reducedim.jl:890 Base._sum(f, a, dims)
│││││││┌ @ reducedim.jl:894 Base.#_sum#750(Base.pairs(Core.NamedTuple()), #self#, f, a, _4)
││││││││┌ @ reducedim.jl:894 Base.mapreduce(f, Base.add_sum, a)
│││││││││┌ @ reducedim.jl:322 Base.#mapreduce#740(Base.:, Base._InitialValue(), #self#, f, op, A)
││││││││││┌ @ reducedim.jl:322 Base._mapreduce_dim(f, op, init, A, dims)
│││││││││││┌ @ reducedim.jl:330 Base._mapreduce(f, op, Base.IndexStyle(A), A)
││││││││││││┌ @ reduce.jl:419 Base.mapreduce_empty_iter(f, op, A, Base.IteratorEltype(A))
│││││││││││││┌ @ reduce.jl:367 Base.reduce_empty_iter(Base.MappingRF(f, op), itr, ItrEltype)
││││││││││││││┌ @ reduce.jl:371 Base.reduce_empty(op, Base.eltype(itr))
│││││││││││││││┌ @ reduce.jl:348 Base.mapreduce_empty(Base.getproperty(op, :f), Base.getproperty(op, :rf), _)
││││││││││││││││┌ @ reduce.jl:361 Base.reduce_empty(op, T)
│││││││││││││││││┌ @ reduce.jl:339 Base.reduce_empty(Base.+, _)
││││││││││││││││││┌ @ reduce.jl:330 Base.zero(_)
│││││││││││││││││││┌ @ missing.jl:106 Base.throw(Base.MethodError(Base.zero, Core.tuple(Base.Any)))
││││││││││││││││││││ MethodError: no method matching zero(::Type{Any})
│││││││││││││││││││└──────────────────

with code like

function sum(f, x::AbstractSparseVector)
    n = length(x)
    n > 0 || return sum(f, nonzeros(x)) # <-- ???
    m = nnz(x)
    (m == 0 ? n * f(zero(eltype(x))) :
     m == n ? sum(f, nonzeros(x)) :
     Base.add_sum((n - m) * f(zero(eltype(x))), sum(f, nonzeros(x))))
end

I'm wondering: why is that?

Edit: attached JET output.

LinearAlgebra.txt

[aviatesk]

The error accounts for a possibility of runtime error when nonzeros(x) is empty.
sum(f, xs) throws when xs is empty and its element type is unknown:

julia> sum(identity, [])
ERROR: MethodError: no method matching zero(::Type{Any})
Closest candidates are:
  zero(::Type{Union{Missing, T}}) where T at ~/julia/julia/usr/share/julia/base/missing.jl:105
  zero(::Union{Type{P}, P}) where P<:Dates.Period at ~/julia/julia/stdlib/Dates/src/periods.jl:53
  zero(::T) where T<:Dates.TimeType at ~/julia/julia/stdlib/Dates/src/types.jl:450
  ...
Stacktrace:
  [1] zero(#unused#::Type{Any})
    @ Base ./missing.jl:106

[aviatesk]

Reading the attached trace, I'd say we definitely want to exclude those top-level throw calls appropriately,
since those methods are just defined to throw.

═════ 70 possible errors found ═════
┌ @ C:\Users\Win10\AppData\Local\Programs\Julia-1.8.0-DEV\share\julia\stdlib\v1.8\LinearAlgebra\src\LinearAlgebra.jl:265 LinearAlgebra.throw_uplo()
│ may throw: LinearAlgebra.throw(LinearAlgebra.ArgumentError("uplo argument must be either :U (upper) or :L (lower)")::ArgumentError)
└────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
┌ @ C:\Users\Win10\AppData\Local\Programs\Julia-1.8.0-DEV\share\julia\stdlib\v1.8\LinearAlgebra\src\adjtrans.jl:303 *(::LinearAlgebra.Adjoint{T, <:AbstractVector} where T, ::LinearAlgebra.Adjoint{T, <:AbstractVector} where T)
│ may throw: LinearAlgebra.throw(LinearAlgebra.MethodError(LinearAlgebra.*, Core.tuple(u::LinearAlgebra.Adjoint{T, <:AbstractVector} where T, v::LinearAlgebra.Adjoint{T, <:AbstractVector} where T)::Tuple{LinearAlgebra.Adjoint{T, <:AbstractVector} where T, LinearAlgebra.Adjoint{T, <:AbstractVector} where T})::MethodError)
└────────────────────────────────────────────────────────────────────────────────────────────────────

[goerch]

The error accounts for a possibility of runtime error when nonzeros(x) is empty. sum(f, xs) throws when xs is empty and its element type is unknown:

julia> sum(identity, [])
ERROR: MethodError: no method matching zero(::Type{Any})
Closest candidates are:
  zero(::Type{Union{Missing, T}}) where T at ~/julia/julia/usr/share/julia/base/missing.jl:105
  zero(::Union{Type{P}, P}) where P<:Dates.Period at ~/julia/julia/stdlib/Dates/src/periods.jl:53
  zero(::T) where T<:Dates.TimeType at ~/julia/julia/stdlib/Dates/src/types.jl:450
  ...
Stacktrace:
  [1] zero(#unused#::Type{Any})
    @ Base ./missing.jl:106

Thanks, I would have expected f(zero(eltype(x))) instead, but didn't think about missing eltype.

[goerch]

Update to reflect @aviatesk latest commit:

LinearAlgebra.txt

[ViralBShah]

Perhaps this should be reopened with a more recent analysis (or even be a new issue).