Tipsy, a tool to parse, analyze, cluster programming MOOC submissions, in order to tackle the problems such as bias in grading by TAs, helping students by providing relevant tips and suggestions.
- JVM
- sbt
- postgres (optional)
- python2 (optional)
- matplotlib (optional)
- scipy (optional)
- numpy (optional)
Download sbt (Scala Build Tool), and place it in your path. It will automatically download the required version of Scala.
For the Python dependencies, you can use a virtualenv in the scripts folder with the above listed dependencies.
You can run postgres using Docker (follow instructions on their website regarding how to install Docker). To do that, run the following:
docker run -d --name tipsy-db -e POSTGRES_PASSWORD=default -e POSTGRES_USER=default -e POSTGRES_DB=default -p 5432:5432 postgresbash> sbt
sbt> run help
This would display the available commands. For instance, you can run the following to display corrections when comparing two files:
sbt> run exec --files <file1>.c,<file2>.c --corrections
To see the linear representation of multiple files, you can do either of the following. The second one will run the command on all files in the provided directory.
sbt> run exec --linearRep --files <file1>.c,<file2>.c,<file3.c>.....
sbt> run exec --linearRep --dirs <dir1>,<dir2> --limit 40
This work has been accepted at the International Conference on Artificial Intelligence in Education, 2018 as a poster paper. The full text can be found on arXiv.
A partial documentation is available, and can be viewed as follows:
sbt> doc
cd target/scala-2.12/api
python3 -m http.server
This would serve the documentation at localhost:8000.
This documentation will be extended in the future.
The project's source code is structured into the following folders inside src/main/scala/tipsy/:
- actor
- compare
- compiler
- db
- frontend
- lexer
- parser
The actual execution starts from the file src/main/scala/tipsy/Tipsy.scala. The various folders of the project are described here:
Contains the code to allow interaction of this software with the user. We currently allow CLI and a Web-server based interactions. The file frontend/CLI.scala contains the entrypoint of this file, and it decides whether to launch the web server, or whether to execute actions from the CLI.
Contains functions which launch gcc preprocessor, and the lexer/parser provided by Tipsy, to compile input programs to the internal representation of Tipsy.
Contains the lexer tokens used by our parsing process, and the lexer code which outputs those tokens, given an input C program.
Contains the parser code which converts input C programs to parse trees, and later into linear representations.
Contains the program comparision logic, cluster actions, normalization functions, et cetera. Detailed description of these files is incoming.
Contains the database models (using Slick ORM), and helper functions for database actions. Also contains serialization functions which convert internal types to database-friendly formats.
Contains Akka actors which run as separate threads. They execute various asynchronous tasks on the database (for instance, inserting new programs, updating the clusters, et cetera). Some of these tasks have to be executed sequentially to avoid race conditions (for example: Updating distances on inserting a new program). Thus, only one actor is spawned for those tasks, and that actor can only use one thread of execution.
In addition, there is some more code in:
scripts/: Python code facilitating program insertion, testing, and clustering.src/main/scala/forcelayout: Code from scala-force-layout placed here because it has no builds for recent scala versions. Can be ignored.
- Provide a requirements.txt file to facilitate a virtualenv in the scripts folder.