Root lengths in the unrooted layout?

[fedarko]

Following up on a point raised in #363.

There is some code in the unrooted layout that accesses the length of the root node:

empress/empress/support_files/js/layouts-util.js

Lines 470 to 478 in 5d19f0d

	var n = tree.postorderselect(tree.size);
	var x1 = 0,
	y1 = 0,
	a = 0,
	da = angle;
	// NOTE: x2 will always be 0, since sin(0) = 0.
	var rootLen = ignoreLengths ? 1 : tree.length(n);
	var x2 = x1 + rootLen * Math.sin(a);
	var y2 = y1 + rootLen * Math.cos(a);

As far as I can tell, the root node's length is actually used in the layout: if you replace tree.length(n) with 1000 on this line for the moving pictures dataset, the unrooted layout looks like:

We currently do not do any sort of validation (in validate_tree()) that the root node's length is valid (e.g. the length is specified and is a nonnegative number). Even in the moving pictures dataset, the root length is unspecified, so it looks like it defaults to 0:

I think it would be best to just ignore the root branch length in the unrooted layout (and maybe just set it to 0 or something), but I am not sure what you all think.

One of the reasons for bringing this up is that as of last week Empress' UI says the following:

...so we should try to be precise about when / if the root length is used :)

[wasade]

The edge associated with a node in the BP or TreeNode model is from that node to its parent. The root by definition does not have a parent. If a length value is associated with the root, it should be ignored.

[fedarko]

Cool, thanks. It sounds like just setting the root length to 0 in the python code, and adjusting the unrooted layout to always keep the root length as 0 (even if "ignore lengths" is set to true), would be the way to go here.