This project provides a language for testing adversaries in the context of UCSD's CSE 107 course.
See setup instructions to get started.
The interpreter is written in Python and uses PLY to lex and parse the language. The current implementation runs multiple interpreters in Docker containers. By using Docker, we can implement memory constraints and limit damage to the server in case of a vulnerability. As each code run is stateless, interpreter instances can run on different machines and round robinned. The interpreter server runs a thrift service, currently implemented with ThriftPy since Apache Thrift's Python bindings do not support Python 3.
The front end is currently a rudimentary interface for a thrift client. Thanks to the language's similarity to JavaScript, the code editor uses CodeMirror's JavaScript syntax support for easy editing. Since there are some differences in syntax, it can occasionally be misleading, but it is often more convenient than problematic.
On top of simply running code, there is a way to 'grade' code. We can pass the target code to be graded (typically the adversary code) and the grader code. By separating the target and grader code, we can guarantee that the target code cannot interfere with a well-written grader code. The grader code must provide a function called grader, which receives the main function from the target code. Presumably, the grader code runs a series of test inputs and checks against expected outputs and return its verdict. The grader environment can also be extended to include special functions that may be useful only for the grader.