Making packages and packaging "just work"

on YouTube ↗40m

" summary

Shows exact fixes and working tools: a new conda-build 3 that renders Jinja templates, records compiler and variant info into a core dependency tree, and emits reproducible package hashes; prebuilt compiler toolchains via cross‑tool‑ng to avoid ABI breakage; an automated, cross‑platform CI built on Concourse that scales workers and lights builds green or red. Claims are quantified — about 700 maintained packages versus ~80,000 on PyPI and ~2,600 in Conda Forge — and the build system demonstrably produces installable, test‑verified conda packages.

AI-written summary · about

" chapters (13)

1. 0:00Packaging and Conda intro
2. 2:06Python package anatomy
3. 5:28Conda Build basics
4. 9:14Recipe generation
5. 11:21Packaging challenges
6. 13:26Anaconda.org channels
7. 16:35API, ABI, constraints
8. 19:47Compiler compatibility
9. 23:17Core Dependency Tree
10. 25:10Conda Build 3 pinning
11. 31:59Dynamic compilers
12. 33:50CI automation
13. 36:52Q&A binary compatibility

" related talks

• 37mEmbrace conda packages.Juan Luis Cano Rodríguez · PyData Madrid 2016
• 45mConda: a cross-platform package manager for any binary distributionIlan Schnell · PyData Berlin 2014
• 30mConda: a cross-platform, python-agnostic binary package manager.Lex Hider · PyCon Australia 2015
• 39mConda-press, or Reinventing the WheelAnthony Scopatz · PyData New York City 2019
• 37mAutomated deployment of Python packages for developmentAndrew MacDonald · PyCon AU 2016
• 01h 01mLessons Learned from Packaging 10,000+ C++ ProjectsBret Brown, Daniel Ruoso · CppCon 2021
• 50mQ&A with the CEO of AnacondaPeter Wang · PyData Austin 2019
• 41mTaming the toxic python environment on your laptopDharhas Pothina · PyData NYC 2023