GraphPlot.jl Incompatibility

[andreaperin]

Hi,

I am trying to run this simple code:

using BayesNets

CPDa = StaticCPD(:a, Normal(1.0))
CPDb = LinearGaussianCPD(:b, [:a], [2.0], 3.0, 1.0)
CPDb(:a => 0.5)
rand(CPDb, :a => 0.5) # condition and then sample
pdf(CPDb, :a => 4.0, :b => 3.0)

bn = BayesNet([CPDa, CPDb])

and I got this error:

ERROR: MethodError: no method matching stressmajorize_layout(::Graphs.WARNING: both ProbabilisticGraphicalModels and IterTools export "groupby"; uses of it in module BayesNets must be qualified
SimpleGraphs.SimpleDiGraph{Int64})
Closest candidates are:
  stressmajorize_layout(::LightGraphs.AbstractGraph) at ~/.julia/packages/GraphPlot/vrIy8/src/stress.jl:58
  stressmajorize_layout(::LightGraphs.AbstractGraph, ::Int64) at ~/.julia/packages/GraphPlot/vrIy8/src/stress.jl:58
  stressmajorize_layout(::LightGraphs.AbstractGraph, ::Int64, ::Any) at ~/.julia/packages/GraphPlot/vrIy8/src/stress.jl:58

Downgrading GrapPlot to 0.4.1 works, but is a very outdated version (more than 2 y.o.).

Graphplot is now at version 0.5.2 and since 0.4.4(allowed in your last version) migrates fromLightGraphs to Graphs. Maybe the usage of an outdated GraphPlot version could be the problem.

Best regards,

Andrea

[dylan-asmar]

I think this was fixed when BayesNets was moved to using GraphPlot 0.5. However, now GraphPlot is at 0.6 with very few references to 0.5. I think BayesNets can move to GraphPlot 0.6 without any issues.

[dylan-asmar]

Closing this issue based on the above comment and PR #154.