An automated testing tool for graph-processing systems via Graph-cutting. The codebase for the paper "Finding Logic Bugs in Graph-processing Systems via Graph-cutting"

We plan to actively maintain GSlicer and extend support for more algorithms in NetworkX (see /graphs/networks/algs). We warmly welcome contributions—feel free to open a pull request if you’d like to be part of the project!

The code has been tested on a Linux (Ubuntu 22.04 LTS) workstation with Python 3.10. To set up the environment, follow these steps:

docker compose up
pip install neo4j
pip install networkx
pip install kuzu
pip install numpy
pip install pandas

• Neo4j GDBMS: To test Neo4j, simply run:

  python ./databases/neo4j/test.py

• Neo4j-GDS (Graph Data Science Library): To test specific algorithms, navigate to the corresponding directory. For example, to test triangle counting:

  cd ./graphs/neo4j/algorithms/triangle_counting
  python triangle_counting.py

• NetworkX: To test using NetworkX, run:

  python ./graphs/networkx/entrance.py

  Note that for NetworkX, we do not provide instances for applying graph-cutting oracles other than those identified by Algorithm 1. Users may implement them and add to ./graphs/networkx/output.json file before running the above command. You can get the basic ./graphs/networkx/output.json file by reproducing the task coverage results (see below).

• Kuzu: To test using Kuzu, run:

  python ./graphs/kuzu/launcher.py

To reproduce the task coverage results, run:

python ./graphs/networkx/sample.py

To reproduce the code coverage results:

1. Manually compile Kuzu v0.4.2 and install the LCOV tool.

   • See Kuzu Developer Guide and LCOV Documentation.

2. Once Kuzu and LCOV are set up, run:

   python ./graphs/kuzu/launcher.py

To keep the double-blind rules, we have removed the original links for bug reports in the artifacts submission.

1. GSlicer detected 39 unique and previously unknown bugs, of which 34 have been fixed and confirmed by developers. The overall bug information can be found in ./found-bugs/overall.csv.

2. Triggering test cases for logic bugs can be found in ./found-bugs/cases.