Github in the classroom

There’s a piece on the GitHub blog about GitHub Classroom. I tried it last year, and I was not really impressed by it. I want to use GitHub in my teaching, but GitHub Classroom is useless, in my opinion, and here is why.

GitHub Classroom

With Classroom you set up a repository that the students can clone. You can set up a number of assignments and then you get URLs that you can give the students.

When they click the URL a new repository is created

But what they get is a clone that is completely decoupled from the repository you set up; worse, it lives in an Organisation space and not in the students’ own list of repositories—which makes the repositories harder for them to find.

Since the new repositories are completely decoupled from the originals, If you make changes to the original repository—for example, you can add some test data or test cases for the students to validate their code after they believe they have solved the problems you give them—these changes are not available for them to merge into their own repositories.

Now, I could see some benefits in being able to amend an assignment by updating a repository. Especially in the—admittedly unlikely—scenario where I have made a mistake in the assignment description or example code. It could also be very useful to be able to add data in a data science projects. Quite often, in real projects, the data you are working on gets updated, which is why it is so important to have automated analysis pipelines.

Just as a learning experience, I would like to be able to tell my students, “right, now you have analysed the data, but I have just changed it, do the analysis again!”. With GitHub Classrooms, I cannot do that.

Even worse, since there is no real connection between the original repository and the students’, you don’t get the review functionality you have for real repositories. You can see a snapshot of what the students have right now, and you can, of course, go back and look over their commits, but if you want to go and comment on what they are doing, you have to do it as individual line comments. You cannot combine a number of comments in a single review.

That is probably what annoyed me the most with using Classroom, and why I gave up on it half-way through the class where I used it. The functionality I’m used to on GitHub simply isn’t there with Classroom.

Using real repositories

I think there are two points to using GitHub in teaching: you teach the students skills they will most likely find use for in the future where they are likely to actually use GitHub, and it makes it a lot easier for me to correct assignments.

First point first. Using a crippled version of GitHub for assignments doesn’t show the students how to actually use GitHub. If you just want to show them how to use version control, it is fine, but for GitHub there is so much more you can do. Much of which you can’t do with Classroom.

You can, of course, still use Issues to have todo-lists of tasks and you have them as a discussion forum if they work in distributed groups—which they didn’t in the class where I used Classroom, but which they often do in other classes I teach. They can also use pull-requests within a group, if they want to, but if everyone in the group can push to the repository there is little need for it.

When you actually use GitHub, pull-requests and reviews are part of the usual workflow. It would only make sense to include it in the workflow used in class.

I realise and appreciate that there is a difference between a real project and a classroom setting, but it takes very little in addition to make the full experience part of class compared to just using GitHub as a version control system.

If, instead of using Classroom, you make a repository that the students fork, you have your repository and the students’ repositories connected. You can make changes to the repository and they can pull them. More importantly, if hand-ins are done as pull-requests, they get experience with that, and you can use the review features to correct the hand-ins.

You have the hand-ins as pull-requests; you can go in and compare changes from the forked version or since your last review; and you have a built-in discussion forum where you can discuss the code, either commenting directly on individual lines or give overall feedback.

This is an active document. If they make changes after a review, they become part of the pull-request, so you can keep track of their progress. And you can directly see what they changed since the last review, so you don’t have to read through the entire code base to see what the modified.

If they get stuck, you can even jump to their pull-request branch, fix an error for them, and push it back to them (for example as a pull-request to their own repository), after which they can continue.

Even better, you can set up automatic checks of their hand-ins using e.g. Travis. If you set up a suite of tests, the students get immediate feedback on how they are doing with solving the problem, and you can check at a glance—by the green tick-marks on the pull-requests—whether they have actually solved the problem they are supposed to.

You cannot set up a complete unit-test suite of their code, of course; you can only check that their code passes a specification you have made. You can, however, set up a code coverage check and keep track of how much they are testing their own code. Again, this is a way for yourself to see how they are doing at a glance and a way for them to get immediate feedback for themselves. coverage report coverage report

Setting up tests and coverage statistics is, of course, also something they could set up themselves for their own Classroom repositories, but that would require them to get familiar with those tools in addition to GitHub. At some point they probably should learn those tools, but if you set it up in the original repository they don’t have to before they earn the benefits of them—which might motivate them more to learn the tools later. coverage report coverage report

You can even set up automatic code reviews using codacy if the students are using a programming language supported there—unfortunately for my R classes, it doesn’t seem to support R.

By and large, having the students fork a repository seems, to me, a much better approach to using GitHub in the classroom that using GitHub Classroom is.

One remaining issue

The only thing that really bugs me with using GitHub and pull-requests in class is that I cannot easily include the pull-request merging in the workflow. At least, I haven’t figured out a good way of doing it yet.

There is nothing inherently wrong with just having an open pull-request per student group, but I would like to be able to merge these pull-requests as the way of accepting the hand-ins. It would be the right kind of closure to finishing an exercise.

It would also make it a lot easier for me to keep track of which groups have finished an assignment and which are still working on it.

The problem with just merging pull-requests is, of course, that if they all make pull-requests to the same branch, then you can’t merge them all.

One solution I’ve been playing with is to have separate branches per group. If they make pull-requests to separate branches, then these can easily be merged and everything is fine. This just requires that I set up branches for each group before they make pull-requests—and that they actually make pull-requests to the right branches—but it is a solution that would work. It just requires a little more work.

It’s not a problem for me to have to set up the branches for each group. What worries me is that there is just that little extra work involved for the students in making the pull-request, especially if they use a GUI like GitHub Desktop. There, you can make pull-requests very easily, but for the branch solution to work, you would have to remember to pick the right branch to make the pull request to.

Not impossible to remember, but a little extra to keep track of while they are learning to use a new tool.

I don’t know if there is any way to change the “to”-branch of a pull-request. I don’t think so. If there were, that would definitely be the way to go.

Author: Thomas Mailund

My name is Thomas Mailund and I am a research associate professor at the Bioinformatics Research Center, Uni Aarhus. Before this I did a postdoc at the Dept of Statistics, Uni Oxford, and got my PhD from the Dept of Computer Science, Uni Aarhus.

Leave a Reply