I’m very conscious about wanting to share as much of what I do as a research/instructor in as easy a way as possible. One example of this is the slides/decks that I use for talks I give. 3 or 4 years ago I was doing this with a Dropbox link. More recently this has lead to me putting everything on github but this wasn’t ideal as I’ve started to accumulate a large number of repositories for the various talks I give.

One of the great things about using github though is that for html presentations (reveal.js is the one I’ve used a couple of times), you can use the github deployement branch gh-pages to serve the files directly. A while back I put a video together showing how to do this:

So there are positives and negatives. After moving to a jekyll framework for my blog I started thinking of a way of getting all the positives without any negatives…

  • I want to use a git+github workflow;
  • I don’t want to have a bunch of separate repositories anymore;
  • I want to be able to just write my talks and they automatically appear online.

My immediate thought was to just use jekyll but as far as I can tell I’d have to hack it a bit to get blog posts be actual .pdf files (for my beamer slides) (please tell me if I’m wrong!). I could of course write a short blog post with a link to the file but this was one extra step to just writing my talk that I didn’t want to have to do. Instead of hacking jekyll a bit I decided to write a very simple Python script. You can see it here but I’ll use this blog post to just walk through how it works.

I have a Talks repository:

~$ cd Talks
Talks$ ls

2014-07-07-Using-a-flipped-classroom-in-a-large-programming-course-for-mathematicians
2014-07-09-Embedding-entrepreneurial-learning-through-the-teaching-of-programming-in-a-large-flipped-classroom
2014-07-25-Measuring-the-Price-of-Anarchy-in-Critical-Care-Unit-Interactions
README.md
favicon.ico
footer.html
head.html
header.html
index.html
main.css
reveal.js
serve.py

What you can see in there are 3 directories (each starting with a date) and various other files. In each of the talk directories I just have normal files for those talks:

Talks$ cd 2014-07-25-Measuring-the-Price-of-Anarchy-in-Critical-Care-Unit-Interactions/
...$ ls

2014-07-25-Measuring-the-Price-of-Anarchy-in-Critical-Care-Unit-Interactions.nav    2014-07-25-Measuring-the-Price-of-Anarchy-in-Critical-Care-Unit-Interactions.snm
2014-07-25-Measuring-the-Price-of-Anarchy-in-Critical-Care-Unit-Interactions.pdf    2014-07-25-Measuring-the-Price-of-Anarchy-in-Critical-Care-Unit-Interactions.tex

I can work on those slides just as I normally would. Once I’m ready I go back to the root of my Talks directory and run the serve.py script:

Talks$ python serve.py

This file automatically goes through my sub-directories reading the date from the directory names and identifying .html or .pdf files as talks. This creates the index.html file which is an index of all my talks (sorted by date) with a link to the right file. To get the site online you simply need to push it to the gh-pages branch of your github repository.

You can see the site here: drvinceknight.github.io/Talks. Clicking on relevant links brings up the live version of my talk, or at least as live as my latest push to the github gh-pages branch.

The python script itself is just:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47

#!/usr/bin/env python
"""
Script to index the talks in this repository and create an index.html file.
"""
import os
import glob
import re

root = "./"
directories = sorted([name for name in os.listdir(root) if os.path.isdir(os.path.join(root, name))], reverse=True)
talk_formats = ['.pdf', '.html']


index = open('index.html', 'w')
index.write(open('head.html', 'r').read())
index.write(open('header.html', 'r').read())

index.write("""
            <body>
            <div class="page-content">
            <div class="wrap">
            <div class="home">
            <ul class='posts'>""")

for dir in directories:
    if dir not in ['.git', 'reveal.js']:
        talk = [f for f in glob.glob(root + dir + "/" + dir[:10] + '*') if  os.path.splitext(f)[1] in talk_formats][0]
        date = talk[len(root+dir) + 1: len(root + dir) + 11]
        title, extension =  os.path.splitext(talk[len(root+dir)+11:].replace("-", " "))
        index.write("""
                    <li>
                    <span class="post-date">%s [%s]</span>
                    <a class="post-link" href="%s">%s</a>
                    <p>
                    <a href="%s">(Source files)</a>
                    </p>
                    </li>
                    """ % (date, extension[1:], talk, title, 'https://github.com/drvinceknight/Talks/tree/gh-pages/' + dir ))
index.write("""
            </ul>
            </div>
            </div>
            </div>
            """)
index.write(open('footer.html', 'r').read())
index.write("</body>")
index.close()

The main lines that do anything are lines 25-38, everything else just reads in the relevant header and footer files.

So now getting my talks written and online is hardly an effort at all:

# Write awesome talk
Talks$ git commit -am 'Finished talk on proof of finite number of primes'  # This I would do anyway
Talks$ python serve.py  # This is the 1 extra thing I need to do
Talks$ git push origin  # This I would do anyway

There are various things that could be done to improve this (including pushing via serve.py as an option) and I’m not completely convinced I can’t just use jekyll for this but it was quicker to write that script then to figure it out (or at least that was my conclusion after googling twice).

If anyone has any fixes/improvements (including: “You idiot just run jekyll with the build-academic-conference-talk-site flag”) that’d be super appreciated and if you want to see the Talk repository (with python script, css files, header.html etc…) it’s here: github.com/drvinceknight/Talks.

Now to finish writing my talk for ORAHS2014…