Johannes U. Lange

Fantastic! I'll check this PR against halotools in the next few days and report back.

I currently don't find agreement with halotools. Take the following code that counts pairs in spheres with radius of 10 in a uniformly sampled, non-cubic box. ```import numpy as np...

Awesome! Just tested the updated version and everything looks good.

The current PR doesn't implement non-cubic boundary conditions for Corrfunc.theory functions other than DD, right? Is it possible to also add it for the other pair counters, especially DDsmu?

> @johannesulf This PR should now support per-axis boxsize and periodicity for the other modules, including `DDsmu`. Would you like to check that `DDsmu` works as expected before we merge?...

I performed some initial tests of `DDsmu` against halotools using per-axis periodicity and the results look good.

I did some more in-depth tests and now I get some errors but only sometimes. Let me see whether I can find out what exactly triggers it.

@lgarrison I get errors when the number of tracers is very small. For example, if I take my code above and set `N=2`, it'll crash for me. I'm not sure...

Works much better. But I'm getting an out-of-bounds error when calculating an auto-correlation function with e.g. `Corrfunc.theory.DD ` using just 1 particle, even if the particle is within the bounds.

Wonderful! Everything looks good now.