Brainstorming the next NumPy comic for 2025

[MarsBarLee]

Did you enjoy the NumPy comics from my Lightning Talk? Have a bold new vision or idea of the next edition? Let me know, let's sprint on it!

Some ideas

• Making your first pull request + how to understand a codebase
• Making full use of community calls, slack, Mailing list (especially in making a pull request)
• preventing and dealing with contributor burnout
• informative vs narrative comics

After the sprint: I'm looking into applying for the Google Season of Docs 2025 with NumPy with our sprint idea

[MarsBarLee]

Notes from my sprint session with Ross.

[MarsBarLee]

Next Comic: Making your first PR in NumPy

Logical starting point: Making a sequel to the first comic, using the same characters

• Ties to the topic 'Understanding the Codebase' and 'Making full use of the Community Call, Slack, Mailing list'

Understanding the Codebase

• There are some existing guides to understanding a big codebase eg this guide.
• My goal would be to take this general advice and make specific to NumPy and open source.
• The strength of making resource specific to NumPy: NumPy is a foundational package made nearly 2 decades ago- it's very worthwhile to contribute but since the codebase is so big, it can be intimidating and hard to start
• The comic would got through a specific part of the codebase eg .npz filetype change in more detail ie a case study
• Ross will ask NumPy team if there's any IRL examples eg past PRs we can use for the story

Making full use of the Community Call, Slack, Mailing list

• Showing the communication channels the NumPy community uses would highlight the parts that makes open source unique. How the process may differ from propriety software and internal teams eg what is etiquette on volunteer basis
• Can close the gap, intimidation of now know who the NumPy community is, how they will respond to your issue
• How Community Call is great human starting point
• 'the value of a triage call to focus specifically discussing the implementation of a PR,
• how communication can move over to Slack
• the value of the mailing list for broadly reaching entire NumPy community
• The subtitles of what communication channel is good for what may not be obvious to those outside of NumPy

A logical story progression combining all these concepts together

• Student A is using NumPy for their cell imaging research
• However, they face Problem X and want to write their own solution
• They eagerly start looking into the NumPy codebase, but get intimidated by the large codebase
• Eventually they find a part that matches, maybe they could change
• However, they're not entirely sure
• They remember that the NumPy team meets at the Community Call- that could a good starting point
• Student A goes to a NumPy Community Call to ask 'I want to solve Problem X by making Solution Y. Is this a step in the right direction?'
• NumPy team member, Contributer B provides guidance, big picture overview, and possible
• Student A feels affirmed of moving in right direction, got some correction and most of all, a human connection
• Contributor B says 'Let's continue this conversation on Slack, I'll walk you through it in more detail"
• Student A and Contribute B talk on Slack, and then decide to meet at Triage call
• Student A and Contributor B get in Triage call to go through codebase,
• Student A works on Solution Y
• Student A submits their first PR, with context of both the full NumPy codebase and guidance of NumPy team
• Student A's PR is reviewed by Contributor B and a NumPy maintainer
• Student A PR is merged! Problem X has been solved with Solution Y!
• Solution Y is announced on the mailing list

Scope problem

• This story is logical and follows the actual experience of contribution
• However, this story is too long to fit within a 12-16 page comic
• I would know, based on my experience, the first grant was overscoped

Possible Solution

• Split up the grant and this story over 2 separate grants and 2 separate comics over 2 years

Other discussion points

Changing NumPy on the C level

• Some of the biggest, worthwhile changes are at a C level
• However, the C level is more complicated, and not a common starting point for a newcomer such as Student A in the comic
• Maybe a future comic could talk about NumPy at a C level

Promote best practice: test-driven development

• Ross mentioned the importance of test-driven development- a comic could talk about why that's a worthwhile best practice and why do that in NumPy
• Test behavior and constraints by implementation
• See what you want it to do and not to do, experience in CI, trial and error with helpful errors you can build off of
• Learn with confidence - see Red X into Green ✓
• Method: How do I teach myself?

Selecting a good first PR: Reality vs Story

• Docstring changes are approachable contribution a newcomer could make, but is often isolated from the rest of the codebase and may be bikeshedded.
• For a story that teaches how NumPy codebase works, more worthwhile to tackle changing the codebase, even if that's more difficult to do

[MarsBarLee]

Next steps after this Summit Sprint:

• I'll bring this conversation to the next NumPy Community Call
• Decide on story scope
• Ask NumPy team about if there's a previous example/real PR that we could use for the story.
• Write grant application for GSOD 2025