SkyLink

Code for efficiently matching sky catalogs using KDTrees and graphs. This includes internal matching via friends-of-friends algorithm. Even in its serial mode, SkyLink performs faster than many other approaches I came across (e.g. the FoFCatalogMatching+fast3tree packages hosted here and here, respectively).

Example

A minimal usage with three catalogs and multiple values for the linking length¹ looks something like the following. This has been made similar in use to the FoFCatalogMatching package to provide an effortless way to switch to this new package and speed up your matching codes. Note that the range of the sky coordinates in the parallel mode should allow for a gnomonic projection which will be used in mosaicking.

from numpy.random import uniform, normal
from astropy.table import Table
import skylink as sl

n = 1000
ra = uniform(0, 179, n)
dec = uniform(-90, 90, n)

# Create catalogs
cat_a = Table({'ra': ra, 'dec': dec})
cat_b = Table({'ra': ra + normal(0, 0.0002, n),
               'dec': dec + normal(0, 0.0002, n)})
cat_c = Table({'ra': ra + normal(0, 0.0003, n),
               'dec': dec + normal(0, 0.0003, n)})

# Run an FoF match with 8 processors in parallel
res = sl.match({'a': cat_a, 'b':cat_b, 'c':cat_c},
               {3.0: 3, 2.0: 3, 1: None}, nprocs = 8)

¹ See FoFCatalogMatching

Installation

The latest version of SkyLink can be installed as follows:

pip install git+https://github.com/enourbakhsh/skylink

I strongly recommend doing this in a conda virtual environment.

Common Installation Issues

This package takes advantage of f-strings which were introduced with Python 3.6. In older python versions, an f-string will result in a syntax error.

Make sure you have an updated version of setuptools before installing SkyLink. A simple pip install setuptools -U does it for you.

For most users, all the dependencies will install automatically. However, some might encounter issues with the networkit package. If you are one of those users, try conda install networkit and if that fails follow the instructions here. networkit requires CMake 3.5 or higher. Make sure you check this by running the command cmake --version in your terminal.

In case pip gives you a gcc error while installing psutil in a conda environment, try conda install psutil before installing SkyLink (source).

Testing and maintainability

In the root directory, you can run the unit tests available in the tests directory using the pytest command (assuming you have pytest installed):

pytest .

While in the root directory, if you only want to check the code base against coding style (PEP8), programming errors and to check cyclomatic complexity, you can run this command (assuming you have flake8 installed):

flake8 .

Citing SkyLink

You can cite SkyLink using the following BibTex reference format:

@software{Nourbakhsh2020,
title={A parallel Python code for efficiently matching sky catalogs using KDTrees and graphs},
url={https://github.com/enourbakhsh/skylink},
author={Erfan Nourbakhsh},
version={1.0.0},
year={2020}}