Selected root id order changes on refresh

[ceesem]

In a segmentation layer, the order of root ids changes on refresh based on the checkbox state (selected root ids come to the top) and is resorted. This makes keeping track of a list of root ids very difficult if you need to refresh (which one does sometimes for a variety of reasons, e.g. image downloading gets stuck).

[mmc46]

Just to add that addressing this would make some things much easier for us. In addition to the case above, we often generate neuroglancer links in R, ordering and colouring ids in a specific way, so being able to have that order maintained would be really helpful.

[schlegelp]

The same applies if a selected neuron is edited (merge or split): it gets bumped to the bottom of the selection table.

I sometimes find myself quickly checking a list of a couple dozen neurons for obvious errors/omissions. Currently, any neuron that I touch ends up at the bottom at the list which makes it harder to keep track of what I have already looked at. It would be super useful if neurons stayed put on edit. I do appreciate that the current behaviour could also be useful in certain situations. Could perhaps be a toggle?

[jefferis]

Overlapping, if you construct a scene with a set of neurons in a particular order (either by listing the ids or by specifying colours for them) the order is scrambled when the scene is opened. This is really annoying for another couple of use cases:

1. providing a list of upstream or downstream partner neurons for proof-reading. You would really like to keep this order based on the number of synaptic connections. Especially if you were reviewing the targets of a pair of left and right homologous neurons at the same time. Right now I am pasting ids manually into flywire because this seems the only way to retain the order.

2. If you take some tracing in another environment (e.g. CATMAID) and you would like to create a list of all the overlapping flywire pieces you would like to do so by size.

[ceesem]

As a workaround, my workflow for working through a sorted list of ids is to make an annotation layer with a list of points with linked segmentations. The points retain their order, but unfortunately there is no automatic tool for "select neuron under point" that would let the associated ids update with proofreading (see #505), but at least you still retain order and the ability to select the neuron. Moreover, you can't lose a neuron forever by accidentally deselecting it (see #493).

[jefferis]

Thanks @ceesem. This can help. But it's not as useful if you want to be able to toggle multiple neurons simultaneously. Also I guess we can programmatically update the ids on annotation layers, but not ideal.