2-cycle (parallel edges) proposed in SNaQ

[cecileane]

Here is a reproducible example of the error reported here.

For now, reproducing the error requires to use the snaq2cfix branch of PhyloNetwork. Do this in pkg mode within julia to use that branch:

pkg> add PhyloNetworks#snaq2cfix

The code below reproduces the error on julia v1.10.2, using the lower-level function optTopLevel! instead than snaq! to control the type of topology updates being proposed.

using Random, PhyloNetworks

dcf = readTableCF(joinpath(dirname(pathof(PhyloNetworks)), "..", "examples", "tableCF.txt"))
net3c_newick = "(1,2,(((3,4))#H1:::0.6,(#H1:::0.4,(5,6))));" # 3-cycle adjacent to 3 cherries
Random.seed!(5)
PhyloNetworks.optTopLevel!(readTopologyLevel1(net3c_newick),0.01,75, dcf, 2, # hmax=2
  1e-5,1e-6,1e-3,1e-4, false,true, [0,0,10,0,0,0], # to propose move 3 only: MVtarget
  stdout,false)

ERROR: cycle with only 2 nodes: parallel edges

The error occurs because the current network has a 3-cycle (whose 3 nodes are each adjacent to a group with 2+ taxa), and the proposed topology modifies this 3-cycle into a 2-cycle of 2 parallel edges. We have a diagnostic and a small reproducible example: a great start to find a fix 😄

[cecileane]

After the simplified random integer generation in commit 8cf9b85, and in julia v1.10.3, we need a different seed to reproduce the error, for example:

Random.seed!(3)
PhyloNetworks.optTopLevel!(readTopologyLevel1(net3c_newick),0.01,75, dcf, 2, # hmax=2
  1e-5,1e-6,1e-3,1e-4, false,true, [0,0,10,0,0,0], # to propose move 3 only: MVtarget
  stdout,false)

Also, the error is fixed with commit e890cc2.
To reproduce an analysis with the fix but without the change in random integer generation, then, we can pin the package at that commit, doing this in package mode within julia:

pkg> add PhyloNetworks#e890cc2be3d3068