Shell Script Automarking System: Building a File-Backed Autograder

This repo implements an automarking workflow with shell scripts. It models the core lifecycle of a university-style assignment system: add an assignment, submit files, summarize submissions, run tests, and calculate marks.

The project is interesting because it uses only shell and the filesystem to model state.

Assignment and Submission Flow

The add script validates assignment names, solution files, autotest specifications, and automarking specifications before storing them under a hidden project directory.

The submit script validates the assignment, student zID format, and submitted file. It creates per-student submission folders and records metadata such as timestamp and file size.

Marking Flow

The marking script reads automarking specs, runs the latest student submission against the reference solution, captures stdout, stderr, and exit status, then compares results.

It supports comparison options for blanks, case, digits, and whitespace. These switches are important because real grading often needs flexible output matching rather than byte-for-byte strictness.

Design Constraints

Writing this in dash keeps the project close to POSIX-style scripting. That means the implementation must be careful with quoting, path validation, process exit codes, and temporary output files.

Knowledge Points

• Shell scripts can model real systems when the data model maps cleanly to files and folders.
• Input validation is essential in CLI tools because every argument can become a path or command input.
• Autograders need to capture stdout, stderr, and exit status separately.
• Flexible diff options make tests fairer for harmless formatting differences.
• File-backed state is simple to inspect but needs consistent naming and cleanup rules.
• POSIX shell encourages small commands and explicit control flow.
• A robust CLI should fail loudly and consistently when inputs are malformed.

Takeaway

This project shows how much system behavior can be built with shell alone. It is a practical exercise in automation, validation, and test infrastructure.