TreeSequence.preorder(u) function

[jeromekelleher]

This should equivalent to

all_descendants = []
for tree in ts.trees():
    all_descendants = np.union1d(
        all_descendants, np.sort(tree.preorder(alpha_origin["node_id"])))

But in graph preorder.

I think the algorithm is here: #2869 (reply in thread)

We can add other traversal orders if they seem useful, but this seems like a good place to start. Return value is a numpy array.

[jeromekelleher]

Ah, it's not quite the same, as the algorithm referenced above will return anything descending even if there is no inherited genetic material from the original sample. It's not clear to me whether it's worth worrying about this, as I think the cases where we actually want to do this will be where we have very limited recombination.

[hyanwong]

Both all-descendants and only-genetic-descendants are nice to have. It would be great to have an efficient way to do these. I can envisage using them even in high-recombination cases (e.g. within human pedigrees)

[hyanwong]

Also, I'm not sure about the name. It would be very easy to get confused between Tree.preorder and TreeSequence.preorder(). I wonder if TreeSequence.arg_preorder() or TreeSequence.graph_preorder() would lead to less confusion?

[jeromekelleher]

I think it's more helpful to have ts.preorder() produce the thing that's analogous to tree.preorder() rather than having to remember a different name.