Confusing error message on creation of ADNLPModel when matrix of linear constraints contains integers

[ForceBru]

Code

import SparseArrays: sparse
import ADNLPModels

params0 = [1.0, 0.1, 0.9]
A = sparse([
	0 -1 1 # NOTE: integers here!
])
problem = ADNLPModels.ADNLPModel(
	x -> x' * x, params0, # f, x0
	[0., 0, 0], [Inf, Inf, 1], # lvar, uvar
	A, [0.], [Inf] # A, lcon, ucon
)

Output

julia> A = sparse([
               0 -1 1 # NOTE: integers here!
       ])
1×3 SparseArrays.SparseMatrixCSC{Int64, Int64} with 2 stored entries:
 ⋅  -1  1

julia> ADNLPModels.ADNLPModel(
               x -> x' * x, params0, # f, x0
               [0., 0, 0], [Inf, Inf, 1], # lvar, uvar
               A, [0.], [Inf] # A, lcon, ucon
       )
ERROR: MethodError: no method matching ADNLPModels.ADNLPModel(::var"#1#2", ::Vector{…}, ::Vector{…}, ::Vector{…}, ::Vector{…}, ::Vector{…}, ::Vector{…}, ::Vector{…}, ::Vector{…})
The type `ADNLPModels.ADNLPModel` exists, but no method is defined for this combination of argument types when trying to construct it.

Closest candidates are:
  ADNLPModels.ADNLPModel(::Any, ::S, ::S, ::S, ::Any, ::Any, ::S, ::S, ::S; kwargs...) where S
   @ ADNLPModels ~/.julia/packages/ADNLPModels/bOFzz/src/nlp.jl:330
  ADNLPModels.ADNLPModel(::Any, ::S, ::S, ::S, ::Any, ::Any, ::S, ::Any, ::S, ::S; kwargs...) where S
   @ ADNLPModels ~/.julia/packages/ADNLPModels/bOFzz/src/nlp.jl:427
  ADNLPModels.ADNLPModel(::Any, ::Any, ::Any, ::Any, ::SparseArrays.AbstractSparseMatrix{Tv, Ti}, ::Any, ::Any, ::Any; kwargs...) where {Tv, Ti}
   @ ADNLPModels ~/.julia/packages/ADNLPModels/bOFzz/src/nlp.jl:507
  ...

Stacktrace:
 [1] ADNLPModels.ADNLPModel(f::Function, x0::Vector{…}, lvar::Vector{…}, uvar::Vector{…}, A::SparseArrays.SparseMatrixCSC{…}, lcon::Vector{…}, ucon::Vector{…}; kwargs::@Kwargs{})
   @ ADNLPModels ~/.julia/packages/ADNLPModels/bOFzz/src/nlp.jl:368
 [2] ADNLPModels.ADNLPModel(f::Function, x0::Vector{…}, lvar::Vector{…}, uvar::Vector{…}, A::SparseArrays.SparseMatrixCSC{…}, lcon::Vector{…}, ucon::Vector{…})
   @ ADNLPModels ~/.julia/packages/ADNLPModels/bOFzz/src/nlp.jl:358
 [3] top-level scope
   @ REPL[8]:1
Some type information was truncated. Use `show(err)` to see complete types.

Problem

Turns out, this happens because A is a matrix of integers, but the error message doesn't tell me this.

Also, I'm not calling ADNLPModels.ADNLPModel(::var"#1#2", ::Vector{Float64}, ::Vector{Float64}, ::Vector{Float64}, ::Vector{Int64}, ::Vector{Int64}, ::Vector{Int64}, ::Vector{Float64}, ::Vector{Float64}) - the package tries to call it here:

ADNLPModels.jl/src/nlp.jl

Lines 358 to 369 in 5d917ce

	function ADNLPModel(
	f,
	x0::S,
	lvar::S,
	uvar::S,
	A::AbstractSparseMatrix{Tv, Ti},
	lcon::S,
	ucon::S;
	kwargs...,
	) where {S, Tv, Ti}
	return ADNLPModel(f, x0, lvar, uvar, findnz(A)..., lcon, ucon; kwargs...)
	end

The signature of this method doesn't constrain S or Tv in any way. Why not write:

const AV{T} = AbstractVector{T}
ADNLPModel(
    f, x0::AV{Tv}, lvar::AV{Tv}, uvar::AV{Tv},
    A::AbstractSparseMatrix{Tv, Ti},
    lcon::AV{Tv}, ucon::AV{Tv};
    kwargs...
) where {Tv, Ti} = ...

This way, the error will move to user code, because then I'll be calling this constructor with (Tv = eltype(x0)) === Float64, but the AbstractSparseMatrix would need Tv = eltype(A) == Int64.