tl;dr: I had to run specs on a lot of different projects and gather results. Here’s how I did it with shell scripts and Ruby.
For a couple of years now I’ve been teaching a Ruby on Rails class for approximatively 100 students. Since I believe in learning by doing, I try to get them to actually write code as much as possible. To do this I usually give them automated tests that they have to get to pass. I’m not a huge believer in grades, but I do have to give them one and it’s always a good incentive for the students to get to work. Therefore all these exercices go this way: 20 tests, 1 point per test passing.
Every week, I would give them a new exercice and I try to get it corrected before the next class. I would then adapt my slides according to their results to focus on pain points. It also allows me to spot students in difficulty. Overall I think it’s a good approach and it gets everyone coding on the first day of class.
The problem with this is that every week I would end up with close to a hundred Ruby on Rails projects on which I would need to run the specs to see how my students performed. Doing this manually would take forever, especially if I wanted insights like “how many students failed the 5th test?”, so obviously I had to automate this.
Running RSpec on a simple Ruby project without Rails is easier because loading a few Ruby files is pretty straightforward and rspec comes with tooling for this situation.
Using a Formatter
First we need to use a JSON formatter so that rspec’s output becomes easily usable:
Running The Specs
Once the formatter is setup, we need to iterate over every projects with
Dir.entries and load the implementation files for every project:
… and actually run the specs right after:
The problem here is that we don’t unload the previously loaded files, which can lead to issues. To fix this, the simplest solution is to run this in a new process and the memory encapsulation will take care of itself:
Reading the Results
Once the specs are done running, you can just access the results via the formatter with
formatter.output_hash. For instance if you want the number of failures:
Once you managed to read the information you need, you can store it in a file somewhere and manipulate it easily in another process.
Getting a project ready to run rspec can get tricky as a lot of new factors are introduced (gems, databases…). Loading rails to run rspec is more complex than just looping throught the project’s files and doing a
require. I tried doing it by manually loading Rails from a fork like before, but it turned out to be a mess and there was some cases where it would simply not work. Remember, I have to work with hundreds of different student projects, and it has to run for all of them.
Shell Scripts To The Rescue!
After a few hours stuck on weird Rails issues, I figured it would be simpler and faster to just introduce bash and brute force my problem. Since it’s script issue, why not write shell scripts?
First I created a Ruby script that would generate a shell script for each project looking like this:
cd FULL_PATH bundle install rake db:drop RAILS_ENV=test rake db:create RAILS_ENV=test rake db:migrate RAILS_ENV=test bundle exec rspec spec > ../rspec_results.log
Running this script would setup the environment and run the specs while outputting the results in a log file that I will be able to process later on.
I don’t think that sharing this piece of code is as interesting since it’s a bit messy and depends on your file structure. The idea is to create a new
.sh file for each project and don’t forget to set it to the appropriate access rights:
If you really want to take a look at what I’ve done here, I created a gist. Also note that you could have done the same thing with a bigger and more complex shell script, but I figured it would be simpler to it this way with a bunch of really small and independent files.
Running The Scripts
Once all the scripts are created, we just need to run them. So, for every project folder run:
Note that I run once again the code in a new process. This is mainly because I want to take advantage of the multiple cores on my machine. Of course this will result in my computer to really slow down, as running a hundred rails processes and bundler installation is very memory heavy, but from my experience it always turns out to be faster - at least on my computer. Using
exec is nice here as it completely replace the ruby process by the shell script.
Reading The Results
Once all the scripts are done running, you can just parse the multiple
rspec_results.log files you ended up with. In my experience it involves a few regular expressions but it’s quite easy. You could also make the rspec output more to your taste with the
--format option if you’d like.