InexactError: converting to Int64 using kmeans

[aborzunov]

Julia version: Version 1.11.1 (2024-10-16)
Package version:

(@v1.11) pkg> status Clustering
Status `~/.julia/environments/v1.11/Project.toml`
  [aaaa29a8] Clustering v0.15.7

Error: ERROR: InexactError: Int64(1.5143524864540896e6)

Steps to reproduce:

julia> @show fs
fs = Real[2530.6228, 3588.6382, 3734.6614, 3743.5217, 3787.343, 3781.5833, 3786.8713, 3786.8774, 3795.4502, 3797.0173, 3792.176, 3798.5337, 3802.8047, 3798.4106, 3797.0593, 3801.621, 3793.6523, 3797.8796, 3788.598, 3782.3643, 3783.711, 3782.1304, 3775.1907, 3775.5623, 3781.2214, 3778.0942, 3784.7712, 3783.2368, 3787.496, 3783.89, 3779.4907, 3778.8562, 3781.099, 3781.0989, 3783.989, 3785.749, 3779.4814, 3786.7463, 3776.925, 3786.4768, 3784.4246, 3781.518, 3786.1404, 3778.7288, 3770.2327, 3769.5112, 3769.3184, 3772.6284, 3779.833, 3778.3235, 3777.4395, 3777.688, 3772.608, 3774.4868, 3765.0747, 3774.5864, 3768.8337, 3766.881, 3763.6145, 3765.7983, 3763.7893, 3768.7952, 3771.589, 3771.647, 3770.689, 3770.2998, 3773.4404, 3760.1555, 3768.2588, 3764.1575, 3766.0757, 3766.8535, 3763.753, 3765.077, 3756.2734, 3759.2808, 3761.2131, 3767.3147, 3770.9436, 3769.1045, 3774.5337, 3763.52, 3761.4167, 3765.7832, 3765.9082, 3761.6826, 3757.2097, 3755.8606, 3758.1658, 3760.027, 3762.6748, 3758.6973, 3759.4707, 3763.619, 3757.0476, 3761.4084, 3750.9497, 3744.2056, 3746.2617, 3755.1316, 3770.0051, 3772.382, 3764.9985, 3763.3247, 3756.66, 3739.35, 3740.5583, 3748.689, 3760.9043, 3766.0525, 3761.701, 3769.4585, 3765.6973, 3769.515, 3772.6848, 3774.4402, 3774.8489, 3771.528, 3773.139, 3762.212, 3758.393, 3753.7148, 3755.2974, 3753.2249, 3751.801, 3756.4004, 3751.262, 3740.1104, 3744.5264, 3738.4866, 3736.9583, 3741.925, 3744.4048, 3745.2512, 3740.0276, 3735.7217, 3734.1936, 3731.8965, 3722.8894, 3727.5469, 3729.073, 3734, 3740.857, 3741.8394, 3742.6428, 3733.0754, 3728.5334, 3731.9055, 3740.8435, 3745.121, 3725.6047, 3728.026, 3726.7766, 3726.7058, 3726.6072, 3725.531, 3725.107, 2732.9595, 951.42993, 608.17145, 607.59, 606.88995, 605.2483, 705.0368, 613.60144, 630.2143, 673.90076, 647.07733, 611.26227, 656.975, 669.02594, 585.1286]

julia> kmeans(fs', 2)
ERROR: InexactError: Int64(1.6037550296302496e6)
Stacktrace:
  [1] Int64
    @ ./float.jl:994 [inlined]
  [2] convert
    @ ./number.jl:7 [inlined]
  [3] setindex!(A::Vector{Int64}, x::Float64, i::Int64)
    @ Base ./array.jl:976
  [4] colwise!(metric::Distances.SqEuclidean, r::Vector{Int64}, a::LinearAlgebra.Adjoint{Real, Vector{…}}, b::SubArray{Real, 1, LinearAlgebra.Adjoint{…}, Tuple{…}, true})
    @ Distances ~/.julia/packages/Distances/mlz7p/src/generic.jl:106
  [5] colwise
    @ ~/.julia/packages/Distances/mlz7p/src/generic.jl:186 [inlined]
  [6] initseeds!(iseeds::Vector{Int64}, alg::KmppAlg, X::LinearAlgebra.Adjoint{Real, Vector{Real}}, metric::Distances.SqEuclidean; rng::Random.TaskLocalRNG)
    @ Clustering ~/.julia/packages/Clustering/JwhfU/src/seeding.jl:173
  [7] initseeds! (repeats 2 times)
    @ ~/.julia/packages/Clustering/JwhfU/src/seeding.jl:160 [inlined]
  [8] initseeds
    @ ~/.julia/packages/Clustering/JwhfU/src/seeding.jl:43 [inlined]
  [9] initseeds(algname::Symbol, X::LinearAlgebra.Adjoint{Real, Vector{Real}}, k::Int64; kwargs::@Kwargs{rng::Random.TaskLocalRNG})
    @ Clustering ~/.julia/packages/Clustering/JwhfU/src/seeding.jl:75
 [10] initseeds
    @ ~/.julia/packages/Clustering/JwhfU/src/seeding.jl:75 [inlined]
 [11] kmeans(X::LinearAlgebra.Adjoint{…}, k::Int64; weights::Nothing, init::Symbol, maxiter::Int64, tol::Float64, display::Symbol, distance::Distances.SqEuclidean, rng::Random.TaskLocalRNG)
    @ Clustering ~/.julia/packages/Clustering/JwhfU/src/kmeans.jl:111
 [12] kmeans(X::LinearAlgebra.Adjoint{Real, Vector{Real}}, k::Int64)
    @ Clustering ~/.julia/packages/Clustering/JwhfU/src/kmeans.jl:95
 [13] top-level scope
    @ REPL[337]:1
Some type information was truncated. Use `show(err)` to see complete types.

Same error using kmeans(reshape(fs, 1, length(fs)), 2), but works using inline data passing:

julia> kmeans([2530.6228, 3588.6382, 3734.6614, 3743.5217, 3787.343, 3781.5833, 3786.8713, 3786.8774, 3795.4502, 3797.0173, 3792.176, 3798.5337, 3802.8047, 3798.4106, 3797.0593, 3801.621, 3793.6523, 3797.8796, 3788.598, 3782.3643, 3783.711, 3782.1304, 3775.1907, 3775.5623, 3781.2214, 3778.0942, 3784.7712, 3783.2368, 3787.496, 3783.89, 3779.4907, 3778.8562, 3781.099, 3781.0989, 3783.989, 3785.749, 3779.4814, 3786.7463, 3776.925, 3786.4768, 3784.4246, 3781.518, 3786.1404, 3778.7288, 3770.2327, 3769.5112, 3769.3184, 3772.6284, 3779.833, 3778.3235, 3777.4395, 3777.688, 3772.608, 3774.4868, 3765.0747, 3774.5864, 3768.8337, 3766.881, 3763.6145, 3765.7983, 3763.7893, 3768.7952, 3771.589, 3771.647, 3770.689, 3770.2998, 3773.4404, 3760.1555, 3768.2588, 3764.1575, 3766.0757, 3766.8535, 3763.753, 3765.077, 3756.2734, 3759.2808, 3761.2131, 3767.3147, 3770.9436, 3769.1045, 3774.5337, 3763.52, 3761.4167, 3765.7832, 3765.9082, 3761.6826, 3757.2097, 3755.8606, 3758.1658, 3760.027, 3762.6748, 3758.6973, 3759.4707, 3763.619, 3757.0476, 3761.4084, 3750.9497, 3744.2056, 3746.2617, 3755.1316, 3770.0051, 3772.382, 3764.9985, 3763.3247, 3756.66, 3739.35, 3740.5583, 3748.689, 3760.9043, 3766.0525, 3761.701, 3769.4585, 3765.6973, 3769.515, 3772.6848, 3774.4402, 3774.8489, 3771.528, 3773.139, 3762.212, 3758.393, 3753.7148, 3755.2974, 3753.2249, 3751.801, 3756.4004, 3751.262, 3740.1104, 3744.5264, 3738.4866, 3736.9583, 3741.925, 3744.4048, 3745.2512, 3740.0276, 3735.7217, 3734.1936, 3731.8965, 3722.8894, 3727.5469, 3729.073, 3734, 3740.857, 3741.8394, 3742.6428, 3733.0754, 3728.5334, 3731.9055, 3740.8435, 3745.121, 3725.6047, 3728.026, 3726.7766, 3726.7058, 3726.6072, 3725.531, 3725.107, 2732.9595, 951.42993, 608.17145, 607.59, 606.88995, 605.2483, 705.0368, 613.60144, 630.2143, 673.90076, 647.07733, 611.26227, 656.975, 669.02594, 585.1286]', 2)
KmeansResult{Matrix{Float64}, Float64, Int64}([3748.343334177213 655.1108621428573], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1  …  2, 2, 2, 2, 2, 2, 2, 2, 2, 2], [1.4828432993568368e6, 25505.729882564396, 187.1953228339553, 23.248156137764454, 1520.973934289068, 1104.895327899605, 1484.404150441289, 1484.8742288313806, 2219.0568076446652, 2369.1549489200115  …  2486.2751034501707, 2492.599270915147, 1723.032126633916, 619.8388065331383, 353.0602614817908, 64.53763869032264, 1922.6990329106338, 3.4750099503435194, 193.62939177034423, 4897.517014631536], [158, 14], [158, 14], 2.7425830743179293e6, 2, true)

[alyst]

Thanks for the report!
The cause of the issue is that fs elements are of type Real, not Float64.
The issue itself is in Distances.jl, not in Clustering.jl; in the first colwise() call, to be precise.
There it incorrectly estimates the numeric type of the distance as Int64, not Float64 or Real:

julia> result_type(SqEuclidean(), Matrix{Real}(undef, 1, 1), Vector{Real}(undef, 1))
Int64

So I suggest you submit the bug report to Distances.jl.

Meanwhile, you can workaround it by converting fs to Vector{Float64} (inline data passing works because that vector has Float64 eltype).