Auto Plotting Thesis Word Count

Procrastination time! By inspiration from my supervisor, who told me he did a similar thing to procrastinate his thesis writing, I’ve decided to write a tiny thing that automatically plots the word count of my thesis on a nightly basis. In addition to being a fun little exercise, it’ll probably also serve as a good motivator / guilt mechanism to keep me writing ^^;;


  • Wowchemy/HugoBlox w plotly json
  • bc easy
  • could have done PNG from some other thing, but that’s less easy + nice to store


  • Edwin: cron job

  • Don’t have a server or permanently on machine which can write to my GH, so GH actions it is

  • Considered triggering from push to thesis repo

  • then looked into actions and discovered scheduled

    • well that seems convenient!
  1. schedule nightly
         - cron: '55 23 * * 1-5'
  1. clone (curr.y private) thesis repo
  • Some stuff w. Personal Access Tokens (PATs)
  1. run the xwcount script to get current word count
  2. subtract previous word count from today’s
  3. load data into python
  4. plot
  5. dump the plot into json


  • plotly
    • saves me dumping the data into the right JSON format by hand
  • setup-python action
    • caches things
  • install dependencies
    • plotly for graphs and its json output
    • pandas for data reading, of course
  • can we plot the words?
    • yes (assuming we’ve remembered to have a header line in the csv)
  • what about the word-diff?
    • need the diff between each day, i.e. each row/entry
    • pandas has a function for this: diff! fantastic! ^^
    • now we can also plot that
  • okay, can we combine the two plots?
    • quick ddg search later: yes
    • the magic word is “subplots”
    • from plotly.subplots import make_subplots
  • subplotting
    • dates are the same: shared_xaxes=True — actually, scratch that; not as easily readable as I’d hoped / worth duplicating that data imo
    • one plot above the other: rows=2, cols=1
    • probably good to have some vertical spacing (remembering to fit the labels): vertical_spacing=0.05
    • need to “graph objects” instead of ‘raw’ express plots, it seems: import plotly.graph_objects as go
    • graph objects take an x and y series or dictionary, not a dataframe with x/y being keys into the dataframe
    • and there is a whole bunch of shenanigans to label the axes
    • At least not as bad as matplotlib – eyoo

Adding plot to website

  • It is apparently easy:
  • It is never just easy.
    • Plotly docs: print(fig)
    • Error: cannot access "data", chart is null
  • First look at the JSON: okay, the Figure(...) bit is not JSON, but I assumed it was a Plotly thing. Clearly not, rm that.
    • Still errors. Shoot.
  • Opening the JSON in vim again: “Huh. The fields are single-quoted, resulting in them being highlighted in red. And it stops highlighting them when I change it to double-quotes. That’s got to be it. And there’s got to be a way to output ‘correct’ JSON”
  • Plotly docs: “Figures also support fig.to_json()
    • ohreally.png
    • print(fig.to_json())
    • well that looks a lot better!
  • And now it works!
Thomas Ekström Hansen
Thomas Ekström Hansen
PhD student in Computer Science

My interests include information visualisation, formal methods, and low-level programming.