Failure on julia-nightly #402

dmbates · 2020-10-05T15:13:04Z

When running the tests by hand on

julia> versioninfo()
Julia Version 1.6.0-DEV.1128
Commit 8519538bb7 (2020-10-05 13:05 UTC)
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: Intel(R) Core(TM) i5-7500 CPU @ 3.40GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-10.0.1 (ORCJIT, skylake)
Environment:
  JULIA_EDITOR = "/usr/share/code/code"
  JULIA_NUM_THREADS = 4

I got a failure on

julia> pca = models(:kb07)[3].PCA.item

Principal components based on correlation matrix
Error showing value of type MixedModels.PCA{Float64}:
ERROR: MethodError: no method matching zero(::String)
Closest candidates are:
  zero(::T) where T<:Dates.TimeType at /home/bates/git/julia/usr/share/julia/stdlib/v1.6/Dates/src/types.jl:423
  zero(::Diagonal{T, StaticArrays.SVector{N, T}}) where {N, T} at /home/bates/.julia/packages/StaticArrays/l7lu2/src/SDiagonal.jl:43
  zero(::FillArrays.Zeros{T, N, Axes} where Axes) where {T, N} at /home/bates/.julia/packages/FillArrays/RM6r2/src/FillArrays.jl:517
  ...
Stacktrace:
  [1] iszero(x::String)
    @ Base ./number.jl:40
  [2] fzeropreserving(bc::Base.Broadcast.Broadcasted{LinearAlgebra.StructuredMatrixStyle{LowerTriangular{Float64, Matrix{Float64}}}, Tuple{Base.OneTo{Int64}, Base.OneTo{Int64}}, typeof(string), Tuple{LowerTriangular{Float64, Matrix{Float64}}}})
    @ LinearAlgebra ~/git/julia/usr/share/julia/stdlib/v1.6/LinearAlgebra/src/structuredbroadcast.jl:107
  [3] similar(bc::Base.Broadcast.Broadcasted{LinearAlgebra.StructuredMatrixStyle{LowerTriangular{Float64, Matrix{Float64}}}, Tuple{Base.OneTo{Int64}, Base.OneTo{Int64}}, typeof(string), Tuple{LowerTriangular{Float64, Matrix{Float64}}}}, #unused#::Type{String})
    @ LinearAlgebra ~/git/julia/usr/share/julia/stdlib/v1.6/LinearAlgebra/src/structuredbroadcast.jl:125
  [4] copy
    @ ./broadcast.jl:862 [inlined]
  [5] materialize
    @ ./broadcast.jl:837 [inlined]
  [6] show(io::IOContext{Base.TTY}, pca::MixedModels.PCA{Float64}; ndigitsmat::Int64, ndigitsvec::Int64, ndigitscum::Int64, covcor::Bool, loadings::Bool, variances::Bool, stddevs::Bool)
    @ MixedModels ~/.julia/dev/MixedModels/src/utilities.jl:248
  [7] show
    @ ~/.julia/dev/MixedModels/src/utilities.jl:238 [inlined]
  [8] show(io::IOContext{Base.TTY}, #unused#::MIME{Symbol("text/plain")}, x::MixedModels.PCA{Float64})
    @ Base.Multimedia ./multimedia.jl:47
  [9] (::REPL.var"#38#39"{REPL.REPLDisplay{REPL.LineEditREPL}, MIME{Symbol("text/plain")}, MixedModels.PCA{Float64}})(io::Any)
    @ REPL ~/git/julia/usr/share/julia/stdlib/v1.6/REPL/src/REPL.jl:219
 [10] with_repl_linfo(f::Any, repl::REPL.LineEditREPL)
    @ REPL ~/git/julia/usr/share/julia/stdlib/v1.6/REPL/src/REPL.jl:461
 [11] display(d::REPL.REPLDisplay, mime::MIME{Symbol("text/plain")}, x::Any)
    @ REPL ~/git/julia/usr/share/julia/stdlib/v1.6/REPL/src/REPL.jl:212
 [12] display(d::REPL.REPLDisplay, x::Any)
    @ REPL ~/git/julia/usr/share/julia/stdlib/v1.6/REPL/src/REPL.jl:224
 [13] display(x::Any)
    @ Base.Multimedia ./multimedia.jl:328
 [14] #invokelatest#2
    @ ./essentials.jl:709 [inlined]
 [15] invokelatest
    @ ./essentials.jl:708 [inlined]
 [16] print_response(errio::IO, response::Any, show_value::Bool, have_color::Bool, specialdisplay::Union{Nothing, AbstractDisplay})
    @ REPL ~/git/julia/usr/share/julia/stdlib/v1.6/REPL/src/REPL.jl:246
 [17] (::REPL.var"#40#41"{REPL.LineEditREPL, Tuple{MixedModels.PCA{Float64}, Bool}, Bool, Bool})(io::Any)
    @ REPL ~/git/julia/usr/share/julia/stdlib/v1.6/REPL/src/REPL.jl:230
 [18] with_repl_linfo(f::Any, repl::REPL.LineEditREPL)
    @ REPL ~/git/julia/usr/share/julia/stdlib/v1.6/REPL/src/REPL.jl:461
 [19] print_response(repl::REPL.AbstractREPL, response::Any, show_value::Bool, have_color::Bool)
    @ REPL ~/git/julia/usr/share/julia/stdlib/v1.6/REPL/src/REPL.jl:228
 [20] (::REPL.var"#do_respond#61"{Bool, Bool, VSCodeServer.var"#36#37"{REPL.LineEditREPL, REPL.LineEdit.Prompt}, REPL.LineEditREPL, REPL.LineEdit.Prompt})(s::REPL.LineEdit.MIState, buf::Any, ok::Bool)
    @ REPL ~/git/julia/usr/share/julia/stdlib/v1.6/REPL/src/REPL.jl:797
 [21] #invokelatest#2
    @ ./essentials.jl:709 [inlined]
 [22] invokelatest
    @ ./essentials.jl:708 [inlined]
 [23] run_interface(terminal::REPL.Terminals.TextTerminal, m::REPL.LineEdit.ModalInterface, s::REPL.LineEdit.MIState)
    @ REPL.LineEdit ~/git/julia/usr/share/julia/stdlib/v1.6/REPL/src/LineEdit.jl:2435
 [24] run_frontend(repl::REPL.LineEditREPL, backend::REPL.REPLBackendRef)
    @ REPL ~/git/julia/usr/share/julia/stdlib/v1.6/REPL/src/REPL.jl:1124
 [25] (::REPL.var"#44#49"{REPL.LineEditREPL, REPL.REPLBackendRef})()
    @ REPL ./task.jl:392

This is just an FYI issue as I will assign it to myself.

palday · 2020-10-05T15:15:44Z

Interesting. I'm curious about whether the underlying issue is some implicit assumption I made in the new show method.

dmbates · 2020-10-05T15:21:21Z

Appears to be. The highlighted line is utilities.jl:248. I'll keep digging unless you want to take over.

palday · 2020-10-05T15:23:25Z

I won't get around to it until much later today, so if this is blocking your work on other issues, please keep digging. Otherwise, I'll take a look later tonight.

dmbates · 2020-10-05T15:38:59Z

It seems to be that 1.6.0 doesn't like string.(m) where m is LowerTriangular

julia> LowerTriangular(Float64.(I(4)))
4×4 LowerTriangular{Float64, Diagonal{Float64, Vector{Float64}}}:
 1.0   ⋅    ⋅    ⋅ 
 0.0  1.0   ⋅    ⋅ 
 0.0  0.0  1.0   ⋅ 
 0.0  0.0  0.0  1.0

julia> string.(LowerTriangular(Float64.(I(4))))
ERROR: MethodError: no method matching zero(::String)
Closest candidates are:
  zero(::T) where T<:Dates.TimeType at /home/bates/git/julia/usr/share/julia/stdlib/v1.6/Dates/src/types.jl:423
  zero(::Diagonal{T, StaticArrays.SVector{N, T}}) where {N, T} at /home/bates/.julia/packages/StaticArrays/l7lu2/src/SDiagonal.jl:43
  zero(::FillArrays.Zeros{T, N, Axes} where Axes) where {T, N} at /home/bates/.julia/packages/FillArrays/RM6r2/src/FillArrays.jl:517
  ...
Stacktrace:
 [1] iszero(x::String)
   @ Base ./number.jl:40
 [2] fzeropreserving(bc::Base.Broadcast.Broadcasted{LinearAlgebra.StructuredMatrixStyle{LowerTriangular{Float64, Diagonal{Float64, Vector{Float64}}}}, Tuple{Base.OneTo{Int64}, Base.OneTo{Int64}}, typeof(string), Tuple{LowerTriangular{Float64, Diagonal{Float64, Vector{Float64}}}}})
   @ LinearAlgebra ~/git/julia/usr/share/julia/stdlib/v1.6/LinearAlgebra/src/structuredbroadcast.jl:107
 [3] similar(bc::Base.Broadcast.Broadcasted{LinearAlgebra.StructuredMatrixStyle{LowerTriangular{Float64, Diagonal{Float64, Vector{Float64}}}}, Tuple{Base.OneTo{Int64}, Base.OneTo{Int64}}, typeof(string), Tuple{LowerTriangular{Float64, Diagonal{Float64, Vector{Float64}}}}}, #unused#::Type{String})
   @ LinearAlgebra ~/git/julia/usr/share/julia/stdlib/v1.6/LinearAlgebra/src/structuredbroadcast.jl:125
 [4] copy
   @ ./broadcast.jl:862 [inlined]
 [5] materialize(bc::Base.Broadcast.Broadcasted{LinearAlgebra.StructuredMatrixStyle{LowerTriangular{Float64, Diagonal{Float64, Vector{Float64}}}}, Nothing, typeof(string), Tuple{LowerTriangular{Float64, Diagonal{Float64, Vector{Float64}}}}})
   @ Base.Broadcast ./broadcast.jl:837
 [6] top-level scope
   @ REPL[27]:1
 [7] eval
   @ ./boot.jl:360 [inlined]
 [8] include_string(mapexpr::typeof(identity), mod::Module, code::String, filename::String)
   @ Base ./loading.jl:1051

I'll see if I can conjure a workaround.

palday · 2020-10-05T15:41:48Z

The crude, inefficient way would be to convert LowerTriangular to just Matrix.

dmbates · 2020-10-05T15:46:28Z

It may be possible to avoid the LowerTriangular conversion in the construction of printmat completely because the upper triangle is later overwritten with dotpad.

palday · 2020-10-05T15:50:33Z

Of course: just create printmat as a Matrix of empty strings and populate the lower triangle with the actual values when populating the upper triangle with dotpad. (I'm sure you already figured this out; just writing this down for my own notes.)

dmbates · 2020-10-05T15:55:38Z

I was going to do that but then decided it was easier to just use the existing logic but with pca.covcor.data instead of LowerTriangular(pca.covcor). Just running the tests now then will commit. Feel free to change the logic if you prefer.

palday · 2020-10-05T15:56:55Z

Perfect. I'm a fan of easy. 😄

dmbates · 2020-10-05T16:17:09Z

Something is still wrong. On my computer using julia-1.5.2 it takes 28 seconds to run models(:kb07) from a cold start. With julia-nightly it has been running for over 10 minutes with no sign of completing. I'll do some more digging.

palday · 2020-10-05T16:19:44Z

I've noticed that the Future builds fail after several hours, which is probably timing out on this.

dmbates · 2020-10-05T16:27:54Z

It is not a matter of fitting the models as those can be fit in a comparable length of time to 1.5.2. It seems to have something to do with the modelcache. When I interrupt it manually it seems to be in a compilation loop around modelcache.jl:40.

dmbates · 2020-10-05T16:50:24Z

It seems to be a problem with the dot mapping of fit although I don't know why. I will replace it with a comprehension.

palday · 2020-10-05T16:51:35Z

So the try...catch in there should probably be changed to use isdefined or @isdefined. I suspect it's catching too much.

dmbates · 2020-10-05T17:01:27Z

Unfortunately, using the comprehension worked in the REPL but not inside the models function. This construction still seems to throw the JIT compiler into a loop.

dmbates · 2020-10-05T17:43:13Z

@palday Feel free to modify #404 to use @isdefined or isdefined

dmbates self-assigned this Oct 5, 2020

dmbates mentioned this issue Oct 5, 2020

Testing on julia-v1.6.0-DEV #387

Closed

palday mentioned this issue Oct 5, 2020

PCA show method fixes for Julia 1.6 #403

Closed

dmbates mentioned this issue Oct 5, 2020

Fixes for running tests on 1.6.0-dev #404

Merged

palday closed this as completed in #404 Oct 5, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Failure on julia-nightly #402

Failure on julia-nightly #402

dmbates commented Oct 5, 2020

palday commented Oct 5, 2020

dmbates commented Oct 5, 2020

palday commented Oct 5, 2020

dmbates commented Oct 5, 2020

palday commented Oct 5, 2020

dmbates commented Oct 5, 2020

palday commented Oct 5, 2020

dmbates commented Oct 5, 2020

palday commented Oct 5, 2020

dmbates commented Oct 5, 2020

palday commented Oct 5, 2020

dmbates commented Oct 5, 2020

dmbates commented Oct 5, 2020

palday commented Oct 5, 2020

dmbates commented Oct 5, 2020

dmbates commented Oct 5, 2020

Failure on julia-nightly #402

Failure on julia-nightly #402

Comments

dmbates commented Oct 5, 2020

palday commented Oct 5, 2020

dmbates commented Oct 5, 2020

palday commented Oct 5, 2020

dmbates commented Oct 5, 2020

palday commented Oct 5, 2020

dmbates commented Oct 5, 2020

palday commented Oct 5, 2020

dmbates commented Oct 5, 2020

palday commented Oct 5, 2020

dmbates commented Oct 5, 2020

palday commented Oct 5, 2020

dmbates commented Oct 5, 2020

dmbates commented Oct 5, 2020

palday commented Oct 5, 2020

dmbates commented Oct 5, 2020

dmbates commented Oct 5, 2020