Closed beckermr closed 3 years ago
@esheldon @erykoff This one is ready for review!
By the same token, we only need healpy for getting unit vectors from the positions. That is easily done ourselves too.
ok @erykoff This is ready for another look!
Thanks @beckermr and @erykoff, this looks awesome! I totally agree with pulling out the healpy
and esutil
dependencies.
I know that @erykoff doesn't like to let scipy
spawn its own subprocesses, but I just wanted to call attention to the fact that there is a workers
(formerly n_jobs
) argument to cKDTree.query
that we could expose to the user if anyone feels strongly about it. I know that my big matching jobs are always limited by memory, so I have been interested in see if this actually provides a significant speed-up (I haven't had a chance to do much testing).
This PR adds the matcher from scipy and updates the tests.