Why does rSPR generate multifurcating trees sometimes?

[teng-gao]

Reproducible example:

run_chain = function(tree_init, max_iter = 100, propose_tree = TreeSearch::NNI, seed = 0) {
    set.seed(seed)
    trees = list()
    tree = tree_init
    for (i in 1:max_iter) {
        message(i)
        trees[[i]] = tree
        tree = propose_tree(tree)
    }
    class(trees) = 'multiPhylo'
    return(trees)
}

set.seed(0)
ntips = 4
tree_init = rtree(ntips)

trees = run_chain(tree_init, max_iter = 16, propose_tree = phangorn::rSPR)

output:

1

2

3

Error in oneOf4(tree, ind[2], ind[1], sample(c(1, 2), 1), sample(c(1, : trees must be binary
Traceback:

1. propose_tree(tree)
2. kSPR(trees, k = k)
3. oneOf4(tree, ind[2], ind[1], sample(c(1, 2), 1), sample(c(1, 
 .     2), 1), root)
4. stop("trees must be binary")
5. .handleSimpleError(function (cnd) 
 . {
 .     watcher$capture_plot_and_output()
 .     cnd <- sanitize_call(cnd)
 .     watcher$push(cnd)
 .     switch(on_error, continue = invokeRestart("eval_continue"), 
 .         stop = invokeRestart("eval_stop"), error = invokeRestart("eval_error", 
 .             cnd))
 . }, "trees must be binary", base::quote(oneOf4(tree, ind[2], ind[1], 
 .     sample(c(1, 2), 1), sample(c(1, 2), 1), root)))

[KlausVigo]

Hi @teng-gao,
sorry it took me so long. I looked into the code and the function was originally intended for unrooted trees. So for now I will add a test for this and add a version for rooted trees later on.
Kind regards,
Klaus