compareConstructs is an evaluation function for Lambda Feedback that aims to check and provide feedback on Python code. The intended use case is in teaching the various Mathematics and Computing modules across Imperial College, where the advanced equation entry features already present on the platform can be combined with code to provide a more cohesive experience for students.
Currently, compareConstructs focusses on the following aspects of code checking:
- Syntax and style:
- Aims to provide helpful feedback in the case of a syntax error.
- Checks for correct and consistent indentation style.
- Console output:
- Verifies that
print'ed values match between the response and answer.
- Verifies that
- Variable content:
- Compares the contents of specified global variables between the response and answer.
- Special feedback cases for NumPy arrays.
- Functions:
- Functions written by students can be subjected to a battery of tests to ensure that they work correctly for any input.
If these tests are inconclusive, as an experimental feature the checking can be handled by GPT-4, which can provide reasonably accurate feedback.
Further documentation on the exact implementation of these features can be found in the 'docs' directory. Start by reading dev.md, which contains lots of information on this function from a developer's perspective.