Skip to content

Latest commit

 

History

History
33 lines (26 loc) · 1.82 KB

README.md

File metadata and controls

33 lines (26 loc) · 1.82 KB

SkyviewBot

SkyviewBot: base repo to go alongside "Good Coding Practice" (ASTERICS-OBELICS 2019).
See also GCP talk slides: https://drive.google.com/open?id=1vdV290w_2hsvVmsoXKdgTo1hQuGxMhlU

Overview

The main goal of this repository is to introduce three primary concepts in good coding practices: 1) argument parsing, 2) modularisation, and 3) docstrings. The code will "run" as it is (once the setup is done), but there are many ways to improve on what it does with better structure and more customised options.

The overall flow:

  1. Download a FITS cutout from any of the 100+ Skyview surveys (or use pre-existing FITS)
  2. Use APLpy to make an image of the FITS file with custom settings
  3. Upload the resulting image to Google Drive using PyDrive wrapper around REST API
  4. Attach the web-ready image to a Slack post and send to the #gcp channel

The best image post will win a box of world-famous Dutch stroopwafels.

Dependencies:

Things you need to do to get this running:

  1. git pull origin master in School 2019, conda env update -f environment.yml (from T. Dijkema)
  2. Preferred: fork this repository, then git clone your version so you can push changes back
    Alternative: git clone https://github.com/cosmicpudding/skyviewbot.git to a sensible location
  3. Download skyview.jar: https://skyview.gsfc.nasa.gov/current/jar/skyview.jar
  4. Move skyview.jar to the main code folder (or change the path in call_skyview()
  5. When first running code: authenticate using autoskyview@gmail.com login details for Google Drive API
  6. Join the #gcp channel on http://obelics-school.slack.com (GCP = Good Coding Practices)

If you get stuck