Multi Stage Assignments - the good and the bad - Advent of Code

# COMMENTS

In the real world you're usually not building your own projects from scratch. Much more frequently you're working on a team and you and other players come and go over time.

This is in stark contrast to most CS educational experiences where students typically complete relatively small assignments from beginning to end. One of my biggest fears way back when as I was about to graduate college was when I woke up one night in a cold sweath "oh crap - in a couple of weeks I'm going to have to actually produce good code. Not just code good enough to get by the grader."

So, as a general rule, having students build a project and then later extend it or, even better, having them extend another student's project can be a very good experience. On the other hand, it can also be a very bad one.

It's a great feeling when you leverage a previous project to create something new. On the other hand, there can be tremendous frustration if the original project forms a faulty foundation and you have to go back and rebuild and even then you're never 100% sure that you're building on a solid base.

Building on a previous project has been big in this years Advent of Code. It started with day 2 when we were asked to write a simple computer that ran "intcode programs." Not a bad little challenge. It then returned in day 5 where we had to add a number of features to our intcode machines. This too was a nice assignment. It had to be built on a correct and complete day 2 solution but as far as I could tell, regardless of how you completed day 2 there was a straightforward way to extend it to integrate the requirements for day 5.

Things got interesting in day 7 and this is where a teacher, or competition designer for that matter has to start being careful. Day 5 had us modify and build on day 2 but day 7 had us use what we created in days 2 and 5. From a teacher who's building assignments point of view, this changes the equation. Day 2 and 5 had you build something that took some input and produced some output. Now, in day 7 you had to do something on top of your intcode machine, feed it into it (or them as it turned out) and get an output. Depending on the assumptions you made in building your intcode machine you would either have an easy time with day 7 or you'd have to go back and redo day 2 and/or 5 - much less fun. As part 2 of day 7 had us running multiple intcode machines "simultaneously," my set up didn't work. I still have to go back and finish that problem.

Herein lies a big problem with multi part assignments. Unless you really lay out the parameters of the early assignment students can, and frequently will make assumptions you never dreamed of. Those assumptions can make their lives easier in parts 2 and beyond but it can also leave them with an impossible task. This on top of any student that doesn't finish the early version to begin with.

Moving on, day 9 added some more features to the intcode machine and since then, days 11, 13, 15, 17, and 19 have all used our intcode machine. I've had issues with my machine for days 11 and 13 and have to fix my intcode machine's input/output handling to finish those challenges but my machine worked well enough to complete days 15, 17, and 19.

So, at the end of the day building on a previous project can be very cool and for a student very rewarding but it can also be really challenging. On top of the issue I've already talked about this can also leave students in a very uneasy state when extending a project by using it rather than building on it. Did they write the early parts fully correctly or just enough to get by the early test cases? It can be very unsettling to work on an uncertain foundation. What's more, while a teacher can even provide base code, this might only be useful to help check results as now having a student use provided code adds yet another challenge (albeit one that provides many useful lesosns).

As to doing this in Advent of Code? I doubt Eric Wastl and the other organizers will read this but in case they do, this isn't a complaint about Advent of Code. I love the event and we can all quibble about details we like or don't like. Advent of Code also isn't a class with a grade and if I never finish, let's say, day 11, there are no consequences - unlike if a student doesn't finish part 2 of a graded project.

I'm writing this because I think there are some interesting parallels between creating an event like Advent of Code and designing projects and lessons for students and more interestingly , where those //s diverge.

Now to get to grading. Hoping to get done to spend at least a few minutes on day 20.