Skip to content

Latest commit

 

History

History
28 lines (20 loc) · 2.74 KB

README.md

File metadata and controls

28 lines (20 loc) · 2.74 KB
Raw

pycon2015-steve-example

This repo has the contents of the directory where I did work to push pycon2015 videos to the pyvideo pycon2015 category.

Currently steve uses vidscraper to scrape youtube, and that doesn't work with channels or playlists. We've not had time to change steve to rework it yet, and in the meantime I have some one-off youtube script that I use to grab json data. Once I have that, I use parts of steve to save the data to json files so that I can finish the process by using steve's command line interface.

For this example,

  • I used the pyvideo.org admin to add a new category, PyCon US 2015. (this is not in the API)
  • I tweaked things to massage the title, description, and speaker data from the PyCon2015 channel
  • I used steve-cmd webedit to change a few titles (I found that I forgot about Lightning talks)
  • and then I used steve-cmd push to push things to pyvideo.
  • At the admin, I toggled everything from draft to live.
  • I still had some names and titles to edit in pyvideo.org, when I noticed some dedups and some titles that could be improved -- I usually notice these when I'm watching some videos. For example, I corrected Titus Brown to C. Titus Brown, lvh to his full name, Vanderplas to VanderPlas, etc.
  • When I watch videos and see find useful links that relate to the talks, I add those by hand. (this is not in the API)

Observations

When Carl has done events, adding the videos happens a lot quicker and we get more metadata because he has scripts and a webapp that links up the schedule data from a conference with the location of the videos on youtube (or vimeo, etc.) along with downloadable files (because he also pushes them to Rackspace cloudfiles and/or archive.org).

I'd like for people who upload videos to a location to give me data like that because it's neat to include the longer descriptions -- I've managed to hack it up by hand in the past when I'm more deeply involved in a conference. For example for US SciPy 2014 I wrote some scripts to mesh together descriptions from the schedule with the scraped youtube data because I was able to add a json api to the conference site to extract the schedule data in a machine readable format. That was cool because the speakers had written really nice and detailed abstracts with graphs and images. I was also able to do something similar for EuroPycon once because I got similar data from their system (frab.xml from the conference system), and they had also followed a convention in linking up uploaded files with the schedule data so that I could include download links for the conference.

Re-scraping a channel when videos are added later is still painful. there is no graceful way to handle that.