Improve GitHub tutorials

[gpsaggese]

Specs are: https://docs.google.com/document/d/1xPgQ2tWXQuVWKkGVONjOGd5j14mXSmGeY_4d1_sGzAE/edit?tab=t.0#heading=h.1phuo9h5z68n

Let's make small steps to make sure we are in agreement with the design and with the specs. You can make a proposal in terms of API, without implementing.

Let me know if it makes sense

[Shaunak01]

Read the following files and improve:

• Tutorial writing guide
• GitHub API notebook
• GitHub example notebook

Tasks:

• Stats to compute

Tasks :

• Stats to compute

  • P0, How many lines of code people write
  • P0, How many PRs
  • Compute the time it takes for someone to close an issue
  • E.g., from the first activity to when it's closed
  • Measure how much interaction there is on a PR
  • E.g., how many reviews, how many comments per line

• We want to have ability to compute stats

  • Per person
  • Per period
  • Per set of repos

• The main problem is that every query is very slow

• The idea is to cache the queries

• The problem is that we don't want to "replicate" queries, in the sense that a query for 2 persons can be covered by a 2 query for 1 person

• Suggestion is to break the queries into "chunks", e.g., querying for a period of time gets split into queries by month and then cache at that level

• In practice the GitHhub call / caching happens for only one value (e.g., one person, one month, one repo), have our internal API do the loops and put things together

• Make a design proposal

[Shaunak01]

@gpsaggese

• Now the queries are wayy faster and take few seconds
• We retrieve all the data we want in cache:
  • Multiple users, repo for commits, PRs and LOC (additions and deletions) for any time period
  • Then we query over cached data for any needed insights and comparisons
  • FYI I also found Github insights pretty useful
  • Helper funcs added in github_utils.py for insights

PR: #543

• I have proposed the design in the PR itself as its simpler. I can work with feedback from this if that works
• Lets address the P0 first, i.e Prs, commits and LOC to evaluate performance and then I can work on stats of time and other insights

[Shaunak01]

closing with: #543