ts: Storing time/generation forward in time

[gregorgorjanc]

msprime counts generations/time backwards in time, and I think tskit too, but we should check (I assume that any node sorting based on time would have to follow this logic).

SLiM is simulating forward in time. Is it storing that information into node table, and if so, in what format (negative values to make it consistent with positive values in tskit)? Or is it storing it separately into the top-level metadata?

AlphaSimR is simulating forward in time, but it is not storing anywhere the notion of time really - apart from the notion of generations in pedigree SP$pedigree. However, users often keep track of time - usually in years. @hannesbecher we had some discussion with @gaynorr, but I can not fine that issue/discussion on AlphaSimR anymore. Do you recall where we brainstormed about this?

• Does tskit assume that time goes backwards in time?
• How is SLiM storing time and where?

@LynxJinyangii can you explore what tskit and SLiM do? For the later check also pyslim package - that enables you to load SLiM tree sequence into Python and work with it.

[LynxJinyangii]

I think the answers are: https://tskit.dev/pyslim/docs/latest/time_units.html#converting-from-slim-time-to-tskit-time-and-back and https://tskit.dev/pyslim/docs/latest/time_units.html#example-calculating-generation-times-in-a-slim-simulation. Just want to add one more thing: the IDs in the tables and the order of their time do not necessarily have to be related; the only limitation here is the time (record in node table) of a parent node (record in the edge table) must > (not >=) that of its child node.

[gregorgorjanc]

@LynxJinyangii perfect find. Yes, this is what we will have to be mindful of too and just to add, we might be needing to keep track in unison of the tskit time, SLiM time, and AlphaSimR time!

The examples at #12 (comment) will be a good testing case for this.