europython2025-recap

My recap of the EuroPython 2025.

From 2025-07-14 to 2025-07-20 in Prague (Czech Republic).

• Monday and Tuesday : tutorials (workshops)
• Wednesday, Thursday and Friday : conferences (talks)
• Saturday and Sunday : sprints (contribution to Open Source)

Additional events :

• Tuesday evening : speakers' dinner
• Wednesday evening : "pyvo" (beer)
• Thursday morning : data leaders breakfast
• Thursday evening : social event

Tutorials

Monday morning : Rambo Python

@Julien : clarify that the repo is the end-goal, that participants should start a new one @Julien : prepare formatting/typechecking/typing exercices @Julien : rename the repo to not have a dash (-) in its name, for the package = true trick

Monday afternoon : PyScript

May be useful in some cases, still not convinced of its potential for global adoption (too many edge cases to not being a first-party Python target, nor a first-party for browsers)

Tuesday morning : CI/CD with Python, K8s, TestContainers, AWS, Jenkins, ...

A total waste of time. I left as soon as possible.

Tuesday morning : Cleaner Code, Better Queries: SQLAlchemy, ORMs and asyncio

Very well prepared, clear, and insightful.

Repo : https://github.com/rhythm-patel/sqlalchemy-workshop

Course : https://aelsayed95.github.io/sqlalchemy-wkshop/

Tuesday afternoon : Packaging Summit

Super interesting ! Too many things going too fast to take notes myself.

Shared notes : https://hackmd.io/-RZmb68eSsyCTk-aTZ8Q4g

PyCon US notes : https://hackmd.io/3AUJd0GkRFKHclfzULowTw

Conferences

Planning :

• Wednesday

  • 09h30 - You don’t have to be a compiler engineer to work on Python
    • interesting, but the talk's message was quite simple : "there are many ways to contribute, and they may be simpler than you expect"
  • 10h45 - Design Pressure: The Invisible Hand That Shapes Your Code
    • @Julien : replay at Kaizen
  • 10h45 - Teamwork makes the dream work
  • 10h45 - Myths and fairy tales around Python performance
    • very interesting, made clear what are the problems caused by Python's extreme dynamicity
    • at the end, Antonio Cuni presented what he thinks is the way towards a more efficient Python, and this bold direction he started to follow by implementing a proof-of-concept language SPy
  • 11h40 - Inside the Black Box: The Anatomy of Virtual Environments
    • interesting, mostly the comparison between CPython's and uv's venvs
    • just a description of what they contain, no real insight on their role in the current ecosystem
  • 12h20 - Building my own (accurate!) Spotify Wrapped
    • great talk ! actually delivers what its title promised : an actually good and insightful Spotify Wrapped, without much work (for European citizens only, through RGPD)
    • @Julien : replay at Kaizen for the Data team ?
  • 12h50 - [poster] Exploring LLM latency
    • intéressant, mais je ne sais pas quoi en tirer
  • 13h50 - Godot & Python: open source in game development
    • disappointing, just a technical demo of the communication (through socket) with a Python server doing OpenCV
  • 13h50 - How to contribute to Python docs in your own language
  • 14h30 - Typing at scale: statically type-checking a large codebase
    • @JULIEN : replay (for myself only ?)
  • 14h30 - Bridging the gap: Viewing your project through a downstream packager's lens
    • a collection of good practices to ease the job of packagers, but they are often contradicting the development good practices, so there is no globally satisfying solution
    • mostly relevant for OSS maintainers who have their projects packaged downstream
  • 14h30 - The Evolution of Advanced Python Monitoring
  • 15h25 - Type Hints in Real-World Projects: Maintenance & Improvement
    • I was late so the room was full, I had to watch the livestream
    • new features in 3.13 : TypeGuard and TypeIs, backported into typing_extensions
    • new syntax for list, dict, Optional, ... (upgrade by Ruff or pyupgrade)
    • avoid raw # type: ignore but add [attr-defined] (or other), and avoid typing.Any
    • see also its talk at EuroPython 2024, for the decorators better use ParamSpec and Concatenate
  • 16h05 - The FastAPI-Postgres stack you've been waiting for
    • https://github.com/geldata/gel-python client for the Gel DB (ex: EdgeDB), actually Postgres
    • no, I was not waiting for it
    • a bit too commercial and trying to hype for it
    • said it is not an ORM, it is actually partly an ORM, and sketchily tried to evade the question about it
  • 16h45 - Building a large SaaS AI product with Python: The tale of three ecosystems
    • slow to start
    • they seem to have over-engineered from the start (not having deadlines ?)
    • the talk's message was not clear to me
  • 17h30 - Python quiz
    • some really hard questions !
  • 17h45 - Lightning talks

• Thursday

  • 09h20 - Why it took 4 years to get a lock files specification
    • very interesting, but more about the process and the struggles than actionnable ideas
  • 10h30 - A new safe external debugger interface for CPython
    • interesting
    • difficult to follow the talk due to the talking speed, that we warned about
    • when we will migrate to 3.14 I think I'll start using it
  • 10h30 - Intuition vs. Reality: Surprising Truths in Python Performance
  • 11h25 - Choosing Between Free Threading and Async
  • 11h25 - Testing the Tests - Assess and Improve Your Python Testing Code
    • OK
    • the examples were too simple, so the advices did not seem to match actual problems
    • 5 Es : tests exist, are executed, enough for coverage, efficient and effective
    • differences between 2 mutation testing libraries : mutmut and Cosmic Ray
  • 12h05 - Fixtures meets parameters: Writing efficient, elegant and reusable tests
    • one key hack in order to be able to parametrize fixture
    • applying pytest.parametrize on a function requesting the pytest.RequestFixture allow to do introspection and parametrize fixtures
    • did not know about hexagonal architecture nor contract testing
  • 12h45 - Turbocharge your Python test suite today!
    • interesting
    • some actual useful advices on how to speed up tests (apart from fixture scope) : pytest -n auto (for local) and pytest-split (for distributed)
  • 13h15 - [poster] Preserving Culture with Python: AI plays Ayo, a Traditional Nigerian Game
    • meh
  • 14h15 - Pydantic, Everywhere, All at Once
    • a very good sequel to Hynek's "Design pressure" talk
    • either use a shared Model (with caveats), one for each interface (verbose), or use inheritance (difficult)
  • 14h15 - Snapshot Testing: A New Era of Reliability
  • 14h55 - Good Practices for Testing Web User Interfaces
    • trainwreck : no rythm, no charisma, low on content, and managed to only last 15 minutes
    • just : test only the part that you want to test, for the rest either mock or do with an API (setup/teardown)
    • also, take care of fixtures' scopes
  • 14h55 - Continuous Documentation: basics and advanced techniques
  • 15h50 - End to End with Testcontainers and Playwright
    • very good
    • @Julien : replay at Kaizen ?
  • 16h30 - Behind the scenes of FastAPI and friends for developers and builders
    • good overview of (design) principles to have a good experience creating and maintaining open-source products
    • @Julien : replay at Kaizen ?
  • 17h15 - Lightning talks

Friday:

• 09h00 - The Boom of Generative AI: realities, promises and awkward situations
  • boring, just a description of the landscape, anti-predictions (already happened)
  • not a good keynote
• 10h10 - What comes after Rust in the Python ecosystem?
• 10h10 - It's All About Time ⏰ - What developers need to know about time
  • very good !
  • used Marp (with JS+CSS hackery) for dynamic slides (time passing)
• 11h05 - Building a NoGIL Load Balancer in Python in 30 minutes
• 11h05 - Efficient Caching in Python: From Local to External Solutions
  • @Julien : replay at Kaizen ?
• 11h05 - Your Cookiecutter Is Wrong
  • a good explanation of the problems caused by the PYTHONPATH, a an explanation of the usual 2 solutions : flat and src-layout
• 11h05 - Explaining the 5 types of database and how to choose between them
  • @Julien : replay at Kaizen ?
• 11h45 - What does = do?
• 11h45 - Django and HTMX: Patterns to Success
• 11h45 - Navigating the mesh: Istio for Python applications
  • was not "advanced" at all (as indicated by the talk metadata), just an introduction to Mesh tooling and use-cases
  • the examples were trivial
  • there was too much information density, which was not explained, so it felt rushed
• 12h25 - Beyond the Code: Manipulating Bytecode and Building Community
• 12h25 - Python Config That Will Give You Chills (In a Good Way, I Promise!)
  • very good talk, about how to use pydantic-setting and pydantic.Annotated for type-safe, polymorphic, dependency injection
  • @Julien : replay at Kaizen
• 13h55 - When in practice is Python performance an issue? Facts and myths.
• 13h55 - The art of yield
  • @Julien : replay at Kaizen
• 14h35 - You Came to a Python Conference. Now, Go Do a PR Review!
• 14h35 - Broken __slots__ are a silent performance killer—Let's fix them!
  • did not attend, but looked at the slides, the presentation seemed very good
  • https://programme.europython.eu/media/europython-2025/submissions/DCZCJP/resources/europython_l5KqgBd.pdf
• 14h35 - Zero-Drama Upgrades: Real-World Tactics for Major Python Migrations
  • disappointed, was actually only about Pydantic, no generalisation
• 15h15 - "Python: the Documentary" screening and Q&A
  • interesting to have perspective on Python 2 to 3 transition (cf my own talk), but mostly on female inclusion in the community
  • the full version will drop on August 28th
• 17h00 - Lightning talks
• 17h30 - Conference Closing
• 17h50 - Sprint Orientation

Sprints

TODO