Kevin R. Thornton

Will take a look next week--probably Tuesday.

Yes, we don't want anything slipping through the cracks, so more verification is a good idea. On Sat, Aug 7, 2021, 12:14 PM Andrew Kern ***@***.***> wrote: > also i...

> Thinking ahead though @andrewkern, can you see reasons why we would want to specify different trajectories? I would rather put a bit of time in now and get the...

> Would that work? I don't see any issue with this, but it'd take a good validation test to make sure. EDIT: had forgotten that the not storing in RAM...

> 've never figured out the details of how to simulate alleles sweeping through multiple populations simultaneously in the coalescent. Doing this right requires either some diffusion results that I've...

The thing that slowed this down, I think, is #1780. There's some idea that we need to get the right API sorted out "at once". Since we're now at 1.x,...

One thing I forgot: allowing `start == end` along with some concept of "forever", allowing the strong balancing selection scenario of Kaplan, Darden, and Hudson?

Naming is tricky because it is related to what's under the hood, at least a bit: * We could have a single class, `msprime.HudsonStructuredCoalescentWithASingleTrajectory`. This is explicit, but there's a...

> let's do it. want to review some code for me @molpopgen as I pull the bits together? Yeah, definitely. I'm quite interested, because I had a near-complete diploSHIC workflow...

> diploSHIC training is gonna fly once this is done. Yeah, in my little test work flows, it was about 30% faster. There's considerable file I/O + time spend in...