PyCon 2025 Friday - Keynote and Talks
Friday at PyCon 2025 brought the first keynote speaker and many talks! Talks run concurrently in four rooms so you do not get to see everything. In general for a conference like this I try to read the talks description and the speaker bios when deciding which sessions to go to. Its tough to really know what a talk will be about though until you go into it. Give yourself permission to walk away from a talk if you're getting anything form it. At the same time try and give talks a chance because they might suprise you!
Here are some highlights.
Cory Doctorow Keynote:
Absolutely amazing. Cannot wait for the recording to go online so I can share it with everyone.
Update: recording is live!
Diving into the Deep End: A Python Journey in Water Management:
I was looking forward to learning more about water management as a part of this. It was mostly a lessons learned presentation, but I got quite a bit out of it. The presenter Jack had a really cool life story and it was nice to see someone on stage talking about navigating ADHD as a part of running a software company.
The PyArrow revolution in Pandas:
This struck a good balance of how deep into the weeds to go. Started with a nice history of how Pandas was built on top of numpy since numpy provided many things Pandas needed. Then went on to nicely summarize why numpy's design is not aligned with the type of workloads Pandas is used in.
Finally they dug into the details of how Apache Arrow fits their usecase better and how they are replacing numpy bit by bit Ship of Theseus style.
Processing Large Geospatial Datasets with Dask & Xarray:
Similar to the Water Management talk I was hoping to learn a bit about geospacial datasets along the way on this talk. The presentation was framed around an example 252 TiB data set from NOAA. However the only details about the dataset that the speaker went into was describing the shape of it: time series + coordinates + measurement. That did nicely setup the discussion to show how xarray's data labeling is useful.
503 days working full-time on FOSS: lessons learned:
There is a recording for the talk here.
I really enjoyed this talk. You could tell Rodrigo is passionate about this topic and able to explain his thoughts clearly. I'm going to include a summary of my favorite takeaways but when the video goes online I really encourage giving it a watch.
Interacting Online:
- Everything you do online is publicity. Its there and others may find it.
- Give the first reply to a issue quickly. The first interaction is important for making someone feel like their contribution is being considered.
- If you cannot get to the issue soon let them know! "Hey thanks you for submitting this, I will take a look as soon as I can"
- Give very clear instruction to bug reporters. Ideally written down or baked into the bug reporting form.
- If they are missing details let them know at the start. You can point to the written guidance.
- If you have to close a PR, take anything you can salvage, put it in a new PR and mark them as a co author so they don't feel like you are stealing their contributions.
Contribution Guides:
- If you run an open source project CREATE A CONTRIBUTING GUIDE.
- People will not read your guide but the doc will have your back.
- You can point contributors towards your guide. Then people are less likely to view feedback as a personal attack.
- Include a style guide in your guide, then its just a thing the project does and not something to be argued.
Unlocking Python's Power: A Practical Guide to Metaprogramming with Decorators, Metaclasses, and Dynamic Code Generation:
I left five minutes into this talk...
Talks Detecting Honeybee Swarms Using the Integration of OpenCV, Pandas, AI, and PyTorch:
... and walked across the hall to this one! The speaker is very passionate about beekeeping and it turns out bees will seasonally leave the hive in mass. The talk focused on using consumer hardware and a model to detect when bees are swarming.