ERROR: convert has no method matching convert(::Type{Int64...}, ::Uint64)

[Jutho]

I seem to be getting this quite a few times.

ERROR: `convert` has no method matching convert(::Type{Int64...}, ::Uint64)
 in convert at base.jl:9
 in convert at base.jl:17
 in read at /Users/jutho/.julia/v0.4/HDF5/src/JLD.jl:301
 in read at /Users/jutho/.julia/v0.4/HDF5/src/JLD.jl:284

My variables are normal Arrays and both saved and opened with HDF5.jl version 4.3.

[timholy]

Can you provide an explicit test case?

[Jutho]

Hold on, there seems to be some interference with PyPlot. The error does not occur if I don't use using PyPlot at the same time, at least that seems to be the case right now. I will investigate further...

[Jutho]

Ok, so I am on latest master

Julia Version 0.4.0-dev+1012
Commit 139fa47* (2014-10-08 19:49 UTC)

with latest HDF5.jl (0.4.3) and PyPlot.jl (1.4.0).

In a freshly opened Julia REPL, doing this

using PyPlot
using HDF5, JLD
f=jldopen("newspectrum_32_1.30_0.45.jld","r")
spectrum1=read(f,"spectrum1")
close(f)

fails with the aforementioned error. Exiting and restarting Julia, this then works

using HDF5, JLD
f=jldopen("newspectrum_32_1.30_0.45.jld","r")
spectrum1=read(f,"spectrum1")
close(f)

If I do using PyPlot afterwards, I can reopen the previous file and read more data, however, opening a different file and trying to read data gives the aforementioned error.

Not sure if I need to report this here, at PyPlot.jl or at Julia itself. I can try to share the dataset.

[Jutho]

All this data is on my Dropbox folder, so I can just share this:
https://www.dropbox.com/s/3tabil14bxxrfkk/newspectrum_32_1.30_0.45.jld?dl=0

[simonster]

Somehow, loading PyPlot breaks:

convert((Int...), (uint(1),))

[Jutho]

That's a surprisingly simple and yet very surprising explanation. Thanks.
I'll report it at PyPlot.jl.

[timholy]

Note that convert seems to have a number of issues, though: JuliaLang/julia#8618
JuliaLang/julia#8128
It's quite possible it's PyPlot's problem, but also possible this is exposing some other issue. convert has more methods than just about anything.

[simonster]

Definitely a Julia bug, and actually triggered by Color.jl:

julia> using Color

[deprecation warnings...]

julia> convert((Int...), (uint(1),))
ERROR: `convert` has no method matching convert(::Type{Int64...}, ::Uint64)
 in convert at base.jl:9
 in convert at base.jl:17

julia> f() = convert((Int...), (uint(1),));

julia> f()
(1,)

[Jutho]

Ok thanks. I'll hold off reporting this.

[timholy]

See JuliaLang/julia#8631. Nice detective work, folks.

[CCWRoy]

I just ran into this issue, and found this ticket via web search.

I see Issue #8631 references this issue, so I'll post here to let others know this issue between JLD.load(), convert, and PyPlot also exists in v0.3.1. To reproduce the error, start a new Julia REPL session, and type:

import PyPlot
import HDF5,JLD 
a=[1:5];
JLD.save("hi.jld","a",a)
z=JLD.load("hi.jld");
b=z["a"];

or

import PyPlot
import HDF5,JLD 
a=[1:5];
JLD.save("hi.jld","a",a)
z=JLD.load("hi.jld");
b=z["a"];

both give:

ERROR: `convert` has no method matching convert(::Type{Int64...}, ::Uint64)
 in convert at base.jl:13
 in convert at base.jl:21
 in read at /home/roy/.julia/v0.3/HDF5/src/JLD.jl:301
 in read at /home/roy/.julia/v0.3/HDF5/src/JLD.jl:284
 in anonymous at /home/roy/.julia/v0.3/HDF5/src/JLD.jl:932
 in jldopen at /home/roy/.julia/v0.3/HDF5/src/JLD.jl:205
 in load at /home/roy/.julia/v0.3/HDF5/src/JLD.jl:931

import PyCall works though.

[timholy]

Does pinning your version of Color work? See JuliaAttic/Color.jl#68.

It does seem there is some long-lurking bug that this has just exposed.

[CCWRoy]

Ah hah! Thanks! I verified pinning Color.jl to v0.3.9 would not cause the JLD.load() and convert problem I posted here.

[timholy]

Update: hopefully this is fixed with the latest release of Color, so do a Pkg.update().

[timholy]

Updated Color fixes some problems, so it's worth trying, but not the specific test case here. See JuliaLang/julia#8631

[GunnarFarneback]

Maybe it's already known but the commit in Color triggering this was 9367c4ef754580693cac5258859256717d812791

[timholy]

Do you have any clue why that commit causes this problem?

[GunnarFarneback]

None at all. That was the point when I gave up debugging and started searching for issues.

[GunnarFarneback]

This issue should be safe to close now.