Redefine root location

[bcumming]

Currently (root) returns (location 0 0), which is the proximal end of the first segment.

However, multiple segments and branches can be at the root, i.e. have arb::mnpos as their parent, so we should have a (root) location that expresses the common root of all these.

I think that (root) = (mnpos 1) is a good value for this.

Discussion points:

• is (mnpos 1) a good way to represent this
• should it be called (root), or something else? I vote for calling it root.

[halfflat]

I think (location mnpos 1) is a good way to represent it. In terms of implementation, should we make the branch id a signed integer, or special case the comparisons?

On calling it root, I'm a bit worried that this might be confusing: it's not a point on the morphology (you can't stick a current clamp there, for example), and it doesn't correspond to e.g. the first point in an SWC file. I think origin or least or bottom might be safer options.

Regarding the semantics: is this point included in reg::all() or completions of regions that include the proximal points of top-level branches?

• If it is, then all will give a connected cell; otherwise it depends on whether or not we have multiple top-level branches.
• On the other hand, if it is included, then the result of most-proximal on e.g. all will give a locset containing just this point, which isn't a valid location for placement of anything.

[halfflat]

Hmm, the problem of placement can be resolved in the discretization; allowing stuff to be placed there is not the issue I thought it was.

[brenthuisman]

Consensus: can't place anything on (location mnpos 1), and say that (location mnpos x) for x ≠ 1 is simply invalid. @bcumming ping

[thorstenhater]

Hi,

I am a bit late to the party, but wouldn't most issues be resolved by making (root) return { (location k 0) | branch k has parent=mnpos }. Thus, (root) is the dual of (terminal).