conflict with "ScikitLearn" package

[brad-sturt]

There seems to be conflict between master branch of JuMPeR and ScikitLearn. When adding

using ScikitLearn

to an example (such as inventory or portfolio examples), I get the following error during the solve(model):

ERROR: LoadError: JuMPeR.BasicUncertaintySet has not implemented setup_set!
 in error at /Applications/Julia-0.4.3.app/Contents/Resources/julia/lib/julia/sys.dylib
 in _solve_robust at /Users/brad-sturt/.julia/v0.4/JuMPeR/src/solve.jl:65
 in solve at /Users/brad-sturt/.julia/v0.4/JuMP/src/solvers.jl:114
 in include at /Applications/Julia-0.4.3.app/Contents/Resources/julia/lib/julia/sys.dylib
 in include_from_node1 at /Applications/Julia-0.4.3.app/Contents/Resources/julia/lib/julia/sys.dylib
 in process_options at /Applications/Julia-0.4.3.app/Contents/Resources/julia/lib/julia/sys.dylib
 in _start at /Applications/Julia-0.4.3.app/Contents/Resources/julia/lib/julia/sys.dylib
while loading /Users/brad-sturt/Code/Experiments/Jumper/inventory.jl, in expression starting on line 68

The examples work correctly when using ScikitLearn is removed. My package versions:

• ScikitLearnBase: 0.0.5
• ScikitLearn: 0.0.4
• JuMPeR: 0.3.1+ (master)
• JuMP: 0.13.1

Is this a known bug / is there a workaround? Sorry in advance if I am missing an obvious solution. Thanks alot!

[IainNZ]

I wonder if ScikitLearn defines something called Model...

[IainNZ]

Although it is getting through solve(::Model) so maybe not that.

[IainNZ]

If you are at a REPL with using JuMP, JuMPeR, ScikitLearn, and do methods JuMPeR.setup_set!, what do you see?

[IainNZ]

JuMPeR hardly exports anything so I'm not sure the root of this really...

[brad-sturt]

In the REPL, I ran the following sequences of commands:

using JuMP, JuMPeR
model = RobustModel()
@variable(model,x)
@objective(model,Min,x)
@constraint(model, x>= 0)
@uncertain(model,u >= 0)
@constraint(model, x == u)
solve(model) # Solves correctly

Then, I tried again with ScikitLearn

using JuMP, JuMPeR, Scikitlearn
model = RobustModel()
@variable(model,x)
@objective(model,Min,x)
@constraint(model, x>= 0)
solve(model)  
# :Optimal

@uncertain(model,u >= 0)
@constraint(model, x == u)
solve(model) 
# ERROR: JuMPeR.BasicUncertaintySet has not implemented setup_set!
 in error at /Applications/Julia-0.4.3.app/Contents/Resources/julia/lib/julia/sys.dylib
 in _solve_robust at /Users/brad-sturt/.julia/v0.4/JuMPeR/src/solve.jl:65
 in solve at /Users/brad-sturt/.julia/v0.4/JuMP/src/solvers.jl:114

In both cases, JuMPeR.setup_set is a defined method, but JuMPeR.setup_set! is undefined.

[brad-sturt]

(sorry, in the previous post, it should be ScikitLearn, not Scikitlearn)

[IainNZ]

JuMPeR.jl/src/uncsets.jl

Line 35 in ea9222a

error("$(typeof(us)) has not implemented setup_set!")

is the cause of setup_set! appearing - its just an incorrect error message.

So I guess what I meant before was methods(setup_set)

[brad-sturt]

Sorry! Here are the methods.

julia> methods(JuMPeR.setup_set)
# 3 methods for generic function "setup_set":
setup_set(us::JuMPeR.BasicUncertaintySet, rm::JuMP.Model, idxs::Array{Int64,1}, scens_requested::Bool, other_prefs::Dict{Symbol,Any}) at /Users/brad-sturt/.julia/v0.4/JuMPeR/src/uncsets_basic.jl:82
setup_set(us::JuMPeR.BudgetUncertaintySet, rm::JuMP.Model, idxs::Array{Int64,1}, scens_requested::Bool, other_prefs::Dict{Symbol,Any}) at /Users/brad-sturt/.julia/v0.4/JuMPeR/src/uncsets_budget.jl:58
setup_set(us::JuMPeR.AbstractUncertaintySet, rm::JuMP.Model, idxs::Array{Int64,1}, scens_requested::Bool, other_prefs) at /Users/brad-sturt/.julia/v0.4/JuMPeR/src/uncsets.jl:33

[IainNZ]

OK, good. So its falling through to the abstract error case, but the only reason it should do that is if other_prefs made it - I just noticed that the abstract one doesn't have that restriction.
The thing I don't get is,

• That dictionary is defined here:

  JuMPeR.jl/src/solve.jl

  Line 46 in ea9222a

  prefs = Dict{Symbol,Any}([name => value for (name,value) in kwargs])

• Why ScikitLearn being loaded would affect that

[brad-sturt]

Interestingly, I found that the first file works correctly, whereas the second file crashes.

File1.jl

using JuMP, JuMPeR

model = RobustModel()
@variable(model, x)
@uncertain(model, u)
@constraint(model, u >= 0)
@constraint(model, u <= 1)
@constraint(model, x >= u)
@objective(model, Min, x)
@show solve(model)

using ScikitLearn

model = RobustModel()
@variable(model, x)
@uncertain(model, u)
@constraint(model, u >= 0)
@constraint(model, u <= 1)
@constraint(model, x >= u)
@objective(model, Min, x)
@show solve(model)

File2.jl

using JuMP, JuMPeR
using ScikitLearn

model = RobustModel()
@variable(model, x)
@uncertain(model, u)
@constraint(model, u >= 0)
@constraint(model, u <= 1)
@constraint(model, x >= u)
@objective(model, Min, x)
@show solve(model)

[brad-sturt]

The following also works:

using JuMP, JuMPeR

model = RobustModel()
solve(model)

using ScikitLearn

model = RobustModel()
@variable(model, x)
@uncertain(model, u)
@constraint(model, u >= 0)
@constraint(model, u <= 1)
@constraint(model, x >= u)
@objective(model, Min, x)
@show solve(model)

It appears that is we run a solve() prior to loading ScikitLearn, the problem goes away.

[IainNZ]

Can you try Julia 0.4.5?

[brad-sturt]

Same problem

[brad-sturt]

The workaround a couple posts above works for now. I'll keep looking through the code to see if I can figure out the problem.

[yeesian]

I can't reproduce this in #69. Is it still an issue?

[yeesian]

Closing stale issues. Re-open if it's still a problem.