Kingman prototype

[JereKoskela]

Two new functions:

• sim_setup_arg uses 1 child, 1 parent jumps to move all lineages to a single pixel. Spatial parameters and a virtual event location are overwritten so that lineages can only intersect with a single pixel, and are unable to migrate out.
• sim_simulate_arg simulates the arg phase using the two types of transitions we discussed: 2 children, one parent for a merger and 1 child, two parents for a recombination. The different recombination mechanism in the ARG has led to some changes to the sim_generate_parents function, and spatial locations are still set so that every lineage intersects with one single pixel only.

I've tried some trial simulations with small numbers of lineages and both few and many loci, and everything looks good. Valgrind also reports no memory leaks. But I haven't done extensive testing, which would probably be easier with a Python wrapper.

[jeromekelleher]

Minor style nit: there should be a space between the if and the bracket.

[jeromekelleher]

This is superb, thanks @JereKoskela! I think this is a great framework, and we'll be able to build the remaining functionality pretty easily on top of this. There are some minor changes I'd like to see:

1. Some binary files have been included in the git commit. These should be removed.
2. I think we're using some pretty recent C features which might cause trouble for portability. C99 is the version I'd like to target.
3. I've made some superficial style comments also, just to try to keep the style consistent through the codebase.

[JereKoskela]

I think I caught all the style discrepancies and removed the binary files. I also split up the z[] and r inits like we discussed. Self->num_parents still gets modified as before. Let me know if you spot anything else.

[jeromekelleher]

This file should be removed as well (KDE eh?).

[jeromekelleher]

Also need to remove 'main' from the commit.

[JereKoskela]

Both done

[jeromekelleher]

space needed here - if (..)

[jeromekelleher]

Indentation.

[jeromekelleher]

OK, I've had a more detailed look through this now, and it's clearer what's going on. You've done a great job integrating in with the existing code, but I wonder if perhaps you've gone a little too far in this. I think it would simplify things a lot if we made top level functions to perform the two types of event we want in the Kingman phase. It would be a lot easier to follow what's happening then, and we wouldn't need to trace through a bunch of functions using flags to determine what code path to take.

Does this make sense?

[jeromekelleher]

Need to check if this is null here:

if (recomb_rates == NULL) {
    ret = ERR_NO_MEMORY;
    goto out;
}