Welcome to the Churchill MCR postgraduate-room-ballot code, please read on to find out how to:
- Run the ballot code if you are the computing officer this year.
- Verify that the room ballot was run honestly!
To build the codebase you will need a c compiler supporting c 20, git, make, and a version of cmake greater than or equal to 3.14. Then follow the standard procedure:
git clone https://github.com/ConorWilliams/ballot
cd ballot
mkdir build && cd build
cmake -DCMAKE_BUILD_TYPE=Release ..
make
You will now have your very own copy of the ballot
executable!
Alternatively 64-bit Linux binaries are provided with each release.
First up you're your going to need a csv file containing everyone's room preferences see example/example.csv
to get an idea of the expected format. Now compile the program, navigate to example/
and run the ballot like:
../build/ballot run example.csv
This will generate two files public_ballot.json
and secret_ballot.csv
. The first can be distributed and used by members of the MCR to anonymously verify the ballot was run fairly. The second contains the room assignments and some additional info. You should email each student their result, "id" and "secret_name" (last three fields in secret_ballot.csv
respectively).
If you would like to encourage particular rooms to fill up (e.g. the hostels) then you can pass in a list of prefixes, for example:
../build/ballot run example.csv -h RR CJ
would preferentially fill all rooms beginning with the letters "RR" or "CJ".
Finally you can control the total number of allocated rooms using the -m
or --max-rooms
options.
To verify the MCR computing officer hasn't fiddled your position you need a copy of the public_ballot.json
file they generated, your "id" and "secret_name" which you should have received securely. Now run:
./ballot verify YOUR_ID YOUR_SECRET_NAME
where you can supply the optional flag -i /path/to/public_ballot.json
to specify the location of the public ballot file if it is not in your current working directory.
The ballot code formulates the task as solving the balanced linear assignment problem. We define a cost function which assigns a value to allocating any student to any room. The student-room pairs are then permuted until the global minimum of the cost function (value summed over all pair) is found. This is done using the Jonker-Volgenant algorithm.
In order to allow the possibility that all students get kicked off the ballot the list of rooms is augmented with p (the number of people) "kicked-rooms". To ensure balanced assignment, preference-free null-people are appended to the list of people.
If the total number of allocated rooms needs to be limited the lowest priority students are removed from the ballot.
In summary, the cost function prioritises people getting their first choices but prefers kicking people off the ballot to assigning them to a room they didn't want. Finally, the cost function ensures all houses designated as "hostels" are preferentially filled.