Good pattern for allowing on_birth functions to have slicing or random access?

[KevinMcCarthyAtIDM]

I wanted to give myself the freedom to have on_birth functions do slicing or random access. Slicing is better when adding agents in a block to the population Laserframe.
There are lots of ways to do constant-population demography. One is "constant on average" - just set birth rate = death rate and have random births and deaths as normal. Another would be deterministic constant population: add your births as normal to the laser frame, and then make sure that exactly that many existing agents are removed. As a way to be memory-efficient at very large populations, I wanted to implement a mode to enable births while maintaining strictly constant population, and one obvious way to do this is to have each "birth" just be re-initializing agents as newborns.
To do this, I needed to be able to randomly index the Laserframe rather than remove agents in slices - doing this via slicing could induce potentially buggy and hard-to-find correlations between position in the laserframe and probability of dying.

I don't want to have to build separate classes for every class with an on-birth function just to make this change. And having multiple "on_birth" methods seems wrong too, since code specifically looks for a method named on_birth, and method overloading didn't seem like the right way. So I did the below pattern - if the input parameter "iend" is none, assume istart is an index list; if istart and iend are present, assume they are the limits of a slice. Good pattern?

def on_birth(self, model, _tick, istart, iend): if iend is not None: model.population.susceptibility[istart:iend] = 1 else: model.population.susceptibility[istart] = 1

[clorton]

Good question. IMO, this relates to our experiment on grouping agents by state - grouping agents by node (in order to efficiently select k from N).

Another option we can try is passing a Boolean array of all the agents of interest in a slice rather than having to call on_birth() once per agent.