Inflection Points

Studying changepoints in open source software evolution. The team name comes from an initial focus on inflection points, which we broadened to cover changepoints.

Team Members

• Noah Burgin (noah22)
• Kuljit Chahal (kkc-al)
• James Walden (jwalden)

Research Questions

• How common are changepoints in software evolution?
• What are the signs and sizes of changepoints?
• What types of metrics might show changepoints?
  • Commits/month
  • Unique authors/month
  • Users/month
  • Popularity (stars, watches on github)
  • Pull requests
  • Issues
  • Code metrics (SLOC, cyclomatic complexity, etc.)
• What are the causes of inflection points in software evolution?
  • Migration to GitHub: check commit messages
  • New releases: check for release tags near changepoint
  • Security events: check news or CVE database for important dates
  • Bot adoption: check commit messages
  • Users adopt fork or competing project: check forks, compare to other projects
  • core team reorganization (e.g. a project lead leaving the project/ newcomers bringing in fresh ideas)
  • change in team culture e.g. from hostile to more inclusive (may be through sentiment analysis) or vice-versa

Data

• Sample of projects from WoC R and S dataset that have
  • 50 authors
  • 5000 commits
  • Age 48 months or more
  • Earliest commit date (unix timestamp) > 0
• Data files on da0 in /data/play/inflection_points

Tasks

• (James) Test changepoint detection libraries.
• (Kuljit) Obtain sample of projects.
• (Noah) Write script to get time series commit data.
• (Noah) Obtain commit time series data.
• (James) Write changepoint detection script for commits.
• (James) Compute changepoint data for commit time series.
• (James) Check-in changepoint script.
• (Noah) Check-in time series scripts and instructions to use.
• (Kuljit) Check-in project sample script and/or instructions.
• (Noah) Investigate why we could not obtain time series for all projects.
• (Noah) Obtain number of unique authors/month time series for sample.
• (Kuljit, Noah) Obtain files changed/month time series for sample.
• (James) Visualize changepoints for sample of projects.
• (James) Tune changepoint detection.
• (James) Run tuned changepoint detection on commits time series.
• (James) Check-in changepoint data and visualizations.
• (James) Compute changepoint data for unique authors time series.
• (All) Prepare for hackathon final presentation on Dec 5.
• (All) Write hackathon 2-page paper by Jan 19.

MSR 2021 World of Code Hackathon Schedule

Checkpoints are scheduled at 10-11:30am EST.

• Checkpoint 1 (Sun, Nov 15): Present tasks and research questions. Complete WoC tutorial before or after checkpoint.
• Checkpoint 2 (Weds, Nov 18): Refine RQs. Have sample of projects. Determine how to create time series. Plan for next checkpoint.
• Checkpoint 3 (Fri, Nov 20): Obtained project sample based on commits and authors using MongoDB. Investigating clickhouse for time series data. Research segmented regression for modeling time series.
• Checkpoint 4 (Wed, Nov 25): Having problems getting time series data from clickhouse. Audris suggests we need to get commits and authors time series from maps using a shell script instead of using clickhouse, which only has data for files changed per month time series.
• Checkpoint 5 (Wed, Dec 2): Obtained commits/month time series and analyzed with changepoint detection script.
• Final presentations (Sat, Dec 5): 10am-1pm EST
• Hackathon track paper (Tue, Jan 19): 2-page + bibliography paper due.