AI-GCS: Create AI-generated code test dataset. Reuse existing dataset where relevant

[pombredanne]

For proper evaluation, we need to create a reusable test data set mixing AI-generated code examples with more traditional code snippets code examples.

The data set should be tagged so that we know the origin of the code snippets -- when possible -- as well as the origin tools used this data set to create and the people or group.

It should be designed for large scale evaluation of the search algos.

Eventually we should publish:

• An open data dataset with its basic documentation, and published on Zenodo.

See some links:

• https://github.com/IBM/Project_CodeNet and https://arxiv.org/abs/2105.12655v2 and https://developer.ibm.com/exchanges/data/all/project-codenet/
• https://github.com/Back3474/AI-Human-Generated-Program-Code-Dataset
• https://github.com/codefuse-ai/Awesome-Code-LLM
• https://ieee-dataport.org/documents/formai-dataset-large-collection-ai-generated-c-programs-and-their-vulnerability

[JonoYang]

We have started a repo to store and run extended matchcode tests at https://github.com/aboutcode-org/matchcode-tests