Teaching basic lab skills
for research computing

Software Carpentry at Cambridge University

At the beginning of this week Thomas Kluyver and I taught a Software Carpentry bootcamp at the University of Cambridge. We had around 35 learners of very different experience and took them through Bash, Git and Python over the course of two days.

First of all I want to thank everyone who helped organize and run the bootcamp this week, especially Thomas for sharing the teaching duties and our excellent helpers Jelena Aleksic and David Molnar. We also had a lot of help from our host Gabriella Rustici along with Paul Judge and David Judge. We held the bootcamp in a university teaching room, so we used a suite of desktop computers and had them all loaded with the Software Carpentry virtual machine, which eliminated much of the extra time spent solving setup issues on the first morning.

Since we also wanted people to go away with a working setup that they could use on their own, I sent round the standard bootcamp installation instructions to all the learners the week before the bootcamp and asked them to have a go installing all the software on their own machines. The idea was that we could then help people with individual installation issues during the coffee/lunch breaks if necessary. In the end, only two people came to us having had problems with installation, which probably reflects a mixture of improvements to the SWC windows installer, a relatively high number of more experienced learners and a number of people not attempting to install things before the bootcamp.

The pre-bootcamp questionnaire revealed a very wide spread of previous experience. We had a number of absolute beginners, but nearly one third of the group reported that they currently programmed once a week or more. Almost all of the respondants also indicated that they had already had some experience with the command line. In order not to frustrate those who had more advanced knowledge, we decided to add links to some of the intermediate Software Carpentry lessons on the bootcamp website. At the beginning of the first day we let everyone know that we were aware of the wide range in experience but we would be working at a pace suitable for novices, so anyone who felt comfortable with the material should feel free to work at their own pace and move on to the intermediate material if they finished early. Walking around during the challenges I had the impression that the vast majority of the class was working at the instructor's pace, but perhaps knowing that they weren't being forced to slow down helped the more advanced people not to get annoyed. Even so, we had a few pink stickies at the end of the day from people who felt that the Bash lessons were too slow for more experienced users, and that the course website had not been clear enough about the expected entry level.

Before the bootcamp I invested some time in organizing the files for the shell lessons. Since we were working from virtual machines, it made sense to have all the example files in a consistent file tree. I think this helped a lot with the lesson flow, and we had a few learners comment that the unix session was especially easy to follow. One person did say that we could have been clearer about which directory everyone was supposed to be in at the beginning of each new section. Thomas took the afternoon's Git session which seemed to go really well. Instead of having the learners collaborate from two repositories on their local machines, we had them pair up and work on each other's repositories. This approach seemed to motivate people really well and we got a lot of positive comments about that session at the end of the day.

Our biggest technical hurdle came on Wednesday morning, when it transpired that I had forgotten to check that the iPython notebook was installed on the virtual machines. After some copying and pasting of "sudo apt" commands, we got the notebook working only to find that numpy and matplotlib were not installed either. In the end it only took around 15 minutes to get everyone set up and people seemed quite quick to forgive me my oversight! Our audience were keen to learn about debugging and defensive programming, so Thomas and I decided to skip SQL in order to give ourselves enough time to cover everything. Despite having a whole day for python, we had still only gotten through the first three lessons by the afternoon coffee break and decided to skip over if/else to get to the material on assertions and test driven development. We got quite a few positive comments about this material at the end, but there was also a little frustration as some of the challenges in lesson five require if/else statements to solve.

One final thing I'd recommend for anyone planning to run a bootcamp in the future - I took a small notebook with me and wrote down lots of small issues that cropped up as we were going through the material. If I hadn't written these all down immediately I would have forgotten them by the end of the day, but I now have a little stockpile of material for issues and pull requests I can send back upstream, hopefully making things easier for myself (and everyone else) next time around!

Dialogue & Discussion

You can review our commenting policy here.