ethand605/GitHub_Owner

Getting Started

Running the Project

1. Load the dependencies by running the following commands:

   git clone git@github.com:golang/go.git
   

   pip install -r requirements.txt
   

2. Run the main Python script:

   python main.py
   # or
   python3 main.py

When asked to "Please enter the subdirectory:", you may enter subdirectory paths, eg. ./.github or ./api. (It is recommended to cd into ./go and run ls first to see what directories are available.)

It is also recommended to query subdirectories instead of root directory for shorter wait time. There are definitely room for improvements for execution time/performance, but due to time constraint I will not be implementing them.

GitHub Owners: Expert Ranking System Summary

Our ranking system identifies the "owners" of a GitHub directory based on four key metrics:

• Commit Contribution (%): Measures the percentage of commits in a directory made by a contributor (git log).

• File Touch (%): Measures the percentage of files in a directory modified by a contributor (git log).

• Line Touch (%): Estimates the percentage of lines in a directory modified by a contributor (git blame).

• Time Commitment (%): Evaluates the time a contributor has spent committing relative to the repository's age (git log).

These metrics contribute to an overall "expert score" for each contributor, ranking them based on their engagement with the directory.

TODOs

• Unify the commit traversal logic
  • Currently, the commit history is traversed multiple times
• Use tqdm for progress bars
  • Progress bar for overall progress
  • Progress bar for each scoring function
• Multithreading for each subdirectory
  • Each subdirectory is independent of each other and can be processed in parallel
  • most of the tasks are I/O bound

Ideas for Future Work

1. Break it down to smaller units - file/line level scoring
2. Pull requests/issues support
3. (Wild idea) If queried frequently, translate commit graph to a graph database, e.g., neo4j. No need for aggregation/parsing at code level
4. Assuming this is a long-running server, implement re-pull and cache invalidation
   • invalidate cache when new commits are pushed
   • invalidating subdirectory cache also invalidates parent directory cache
5. Different weights for different files, e.g., README.md is more important than .gitignore
6. Support recursive query: querying a parent directory should also calculate and cache the experts for its subdirectories
   • https://imgur.com/rr7gWz4
   • this would require reworking the current git log logic to exclude subdirectories, not implmemented due to time constraint
   • will also need to take into considertaion of weights of subdirectories