CLuster And RepAir tool for introductory programming assignments

This is an implementation of the clustering and repair tool for introductory programming assignments, described in the following paper: Automated Clustering and Program Repair forIntroductory Programming Assignments (https://dl.acm.org/doi/10.1145/3192366.3192387 and https://arxiv.org/abs/1603.03165).

The package is currently supported/tested on the following operating systems:

• Fedora
• Debian
• Ubuntu

The package is currently supported/tested on the following Python versions:

• 3.6
• 3.7
• 3.8
• 3.9

Any other version might or might not work as expected. If you encounter any error with the above mentioned combination, please do file an issue.

• C compiler
• Cython
  • $ sudo aptitude install cython (Debian)
  • # dnf install Cython (Fedora)
• lpsolve 5.5 (development files and library)
  • $ sudo aptitude install lp-solve liblpsolve55-dev (Debian)
  • # dnf install lpsolve-devel (Fedora)

• make (in this directory) installs a binary file called clara
• Run clara help or see examples below.

• Create a new virtual environment (using virtualenv)
• Install Cython (using pip install Cython inside the new virtual enviroment)
• Run python setup.py develop

On Debian system the following is required before running the tool: export LD_LIBRARY_PATH=/usr/lib/lp_solve/

The examples/ directory contains some example programs:

• c1.py and c2.py are the correct examples from the paper
• i1.py and i2.py are the incorrect example from the paper
• c3.py is a constructed example.

To test matching between examples/c1.py and examples/c2.py on inputs [4.5] and [1.0,3.0,5.5] use:

clara match examples/c1.py examples/c2.py --entryfnc computeDeriv --args "[[[4.5]], [[1.0,3.0,5.5]]]" --ignoreio 1

This should output Match!.

To test matching between examples/c1.py and examples/c3.py on inputs [4.5] and [1.0,3.0,5.5] use:

clara match examples/c1.py examples/c3.py --entryfnc computeDeriv --args "[[[4.5]], [[1.0,3.0,5.5]]]" --ignoreio 1

This should output No match!.

To repair examples/i1.py using examples/c1.py on the same inputs as above, use:

clara repair examples/c1.py examples/i1.py --entryfnc computeDeriv --args "[[[4.5]], [[1.0,3.0,5.5]]]" --ingoreio 1

To cluster correct programs on the same inputs as above use:

mkdir clusters
clara cluster examples/c*.py --clusterdir clusters --entryfnc computeDeriv --args "[[[4.5]], [[1.0,3.0,5.5]]]" --ignoreio 1

This should produce two clusters in the directroy clusters/ and two .json files with additional experssion extracted from the clusters.

To produce feedback from the above clusters for an incorrect program, for example examples/i1.py, use:

clara feedback clusters/c*.py examples/i1.py --entryfnc computeDeriv --args "[[[4.5]], [[1.0,3.0,5.5]]]" --ignoreio 1 --feedtype python

You can add --verbose 1 to any of the examples to obtain a more verbose output.