Skip to content
Snippets Groups Projects
Select Git revision
  • main default protected
1 result
Compare
Name Last commit Last update
README.md
personal-site.md
README.md

Professional Skills Brainstorm

Workshop Ideas

  • Tikz MasterClass
  • Personal Academic Website
  • Git (for LaTeX projects)
  • LaTeX installation tour
  • Linux/Unix shell
  • Computer Algebra Systems
  • Publishing Python/SageMath packages

Overall Thoughts

Proposed Format

Given the variety of topics, I think the goal for most sessions should not be a tutorial as such but an exposition of what can be done, and an opportunity to get started trying it.

Talk + Workshop

The format I have in mind for the topics above are a 15-30min presentation, with the purpose of introducing the topic and demonstrating what can be done, as opposed to proper instructions. Followed by 30-60min workshop time, to either work through an (maybe a choice of) exercise(s) for some hands on practice. Ideally even provide a space for people to bring their own projects relating to the topic to work/talk about.

Expository Talk

If the above format doesn't make sense for a given topic, guest speaker, or software setup. Then the odd session could involve an alternative seminar-like format, where a speaker presents on a topic of interest to them, with the goal of inspiring the use of tech in research.

Devcontainers/Codespaces (a logistical solution)

To allow for independent topics each week, with software that people might decide not to keep using, it would be nice to not have attendents install a bunch of software each time (if possible), and have to deal with troubleshooting each time.

Although this may not be possible for all topics, I believe that every session where it is set up, it will be worth it.

Precedence for Codespaces

Following the releases of codespaces from GitHub, the Computational Numerics, and Python Programming courses (by Charlotte Desvages) have adopted it. The main benefit was to set up the students of their course with a uniform environment, which greatly improved how quickly they could get started with the actual workshop content in the first few weeks, as well as restricting the range of technical issues experienced.

Special Considerations for Our Case

Compute Limitations

The course mentioned above has some connections with GitHub which may not be worth pursuing for a small workshop, however even free tier GitHub users enjoy up to 60hrs of CodeSpaces compute time per month, which is hopefully much more than necessary.

One other backup that comes for free when setting up for Codespaces, is that any users with VSCode and Docker (see [corresponding section](VSCode + Devcontainers extension)) can get an identical (superior) experience on their own computer, modulo the time and space to download the associated container (measured in GB). But has the benefits of avoiding software complications and conflicts.

Variety of Software

The above mentioned courses had a single linux container image prepared to run in the codespaces for the whole semester. However this workshop will likely require different software each week, creating a chore to prepare a new container image and setup for each session. I (Luke) do have some experience with it at this point, in fact creating the final container image used in the courses above to cut down in load times. I have also been using devcontainers in my own (LaTeX) work, using existing container images when available, or creating my own when necessary.

DockerHub is a good place to look for existing containers which could be useful. Two notable ones are:

  • TexLive for LaTeX based workshops
  • Micromamba which can be used as a small base for any environment that can be created with conda-forge packages (most OSS scientific computing). Here's a reference guide for building container images from it.

VSCode + Devcontainers extension

Linux Setup

A fairly large proportion of PhD students likely to attend chose to receive Ubuntu laptops from the school. Luckily this is the easiest to install docker on, requiring no deep virtualization to be enabled in the BIOS, just sudo apt install ... a couple packages. From then, the devcontainers extension needs to be installed in VSCode (note not VSCodium), and at that point they're ready to get going.

Windows Setup

The most likely difficulty with the setup on Windows is running Docker, which in turn requires WSL2. Microsoft has been pushing this tech quite hard since 2019 and the installation experience has gotten simpler though. One note is that WSL2 requires a BIOS setting to be enabled, which may-or-may-not be possible to enable without visiting the IT office. Anecdotally, the laptop that I was given from the school in Sept 2020 had this option enabled and there was no difficulty installing the above. The 2021 cohort did get different laptops though.

MacOS Setup

I've done this once on an old MacBook air a while ago, I don't remember any real hurdles.

OpenBSD Setup

I'm aware of a singular PhD student running none of the above. They're also less likely to be keen to run Microsoft software. I'm sure they can manage to do their own thing.